Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Marital quality, gender, and biomarkers of disease risk in the MIDUS cohort
(USC Thesis Other)
Marital quality, gender, and biomarkers of disease risk in the MIDUS cohort
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
MARITAL QUALITY, GENDER, AND BIOMARKERS OF DISEASE RISK IN THE MIDUS COHORT by Carrie Joy Donoho ________________________________________________________________________ A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (GERONTOLOGY) December 2012 Copyright 2012 Carrie Joy Donoho ii DEDICATION This work is dedicated to my daughter, Isabelle, and my husband, Rob. Isabelle, thank you for being so understanding when I missed so many of your weekend adventures with Daddy. I hope you accomplish everything you want to in life. Rob, thank you for your patience and encouragement through this seemingly endless endeavor. Your constant strive to be the best in your own work inspires me. iii ACKNOWLEGEMENTS This work would not have been possible without the dedication, time, and effort put forth by my committee members, Dr. Eileen Crimmins, Dr. Tara Gruenewald, Dr. Cleopatra Abdou, and Dr. Stanley Azen. Thank you to Dr. Crimmins for her guidance and mentorship. She set a standard that I will always strive to live up to. Her dedication to my success was unwavering. Thank you to Dr. Gruenewald for her guidance on the methodological complexities of studying cortisol, and for providing me with the resources to pursue the methods used in this study. Thank you to Dr. Abdou for her guidance on social relationships theory, and for continually providing encouragement when I needed it. Thank you to Dr. Azen for pushing me to keep an epidemiologist’s perspective, and for the jovial conversations between USC and HSC. I also want to thank Dr. Jennifer Ailshire, Dr. Teresa Seeman, and Dr. Chih-Ping Chou for contributing their time and energy to advising me on this work. Lastly, I want to thank Dr. Glenn Ehresmann for keeping me well during my studies. Without his dedication, encouragement, and understanding, this dissertation would not have been possible. This work was supported by the National Institute on Aging T32AG0037. iv TABLE OF CONTENTS DEDICATION .................................................................................................................... ii ACKNOWLEGEMENTS .................................................................................................. iii INTRODUCTION .............................................................................................................. 1 Specific Aims .................................................................................................................. 3 Dissertation Format ......................................................................................................... 5 CHAPTER 1: MARRIAGE, HEALTH, AND THE MIDUS STUDY .............................. 6 Marriage and Health ....................................................................................................... 6 The Survey of Midlife in the United States .................................................................. 13 CHAPTER 2: MARITAL QUALITY, GENDER, AND MARKERS OF INFLAMMATION IN THE MIDUS COHORT .............................................................. 15 Method .......................................................................................................................... 25 Results ........................................................................................................................... 30 Discussion ..................................................................................................................... 32 CHAPTER 3: THE PRICE THE HEART PAYS IN A BAD MARRIAGE: MARITAL STATUS, MARITAL QUALITY, GENDER, AND HEART RATE VARIABILITY IN THE MIDUS COHORT ................................................................................................... 42 Method .......................................................................................................................... 48 Results ........................................................................................................................... 53 Discussion ..................................................................................................................... 55 CHAPTER 4: MARITAL QUALITY, GENDER, AND DIURNAL CORTISOL RHYTHMS IN THE MIDUS COHORT .......................................................................... 63 Method .......................................................................................................................... 67 Results ........................................................................................................................... 75 Discussion ..................................................................................................................... 77 CHAPTER 5: CONCLUSION ......................................................................................... 94 Summary ....................................................................................................................... 94 Findings and Implications ............................................................................................. 94 Limitations .................................................................................................................... 96 Future Directions .......................................................................................................... 97 REFERENCES ................................................................................................................. 99 APPENDIX A: METHODOLOGICAL APPROACHES TO ASSESSING DIURNAL CORTISOL RHYTHMS IN EPIDEMIOLOGICAL STUDIES: HOW MANY SALIVARY SAMPLES ARE NECESSARY? .............................................................. 116 APPENDIX B: SENSITIVITY ANALYSES FOR CORTISOL ................................... 140 1 INTRODUCTION A marital relationship is one of the most important social relationships established in adulthood, and the better health experienced by married adults is well documented (Waite & Gallagher, 2000). Married adults live longer (Johnson, Backlund, Sorlie, & Loveless, 2000), rate their health better (Williams & Umberson, 2004), and report fewer chronic conditions and functional limitations (Hughes & Waite, 2009), compared to their non-married counterparts. While prior research has established that marriage confers health benefits, there has been less focus on how the health of married individuals may vary as a function of the quality of their marital relationship and the duration of time spent in a marriage. Larger studies have generally focused on the association between marital quality and broadly defined health outcomes, whereas smaller laboratory-based studies have focused on experimentally induced marital conflict or collaboration and physiological processes (Carr & Springer, 2010). Thus it is important to examine the association between marital quality and physiological processes related to health in a population study of adults. Greater understanding of physiological processes among couples in long-term marriages could enhance our understanding of the health benefits associated with long, high-quality marriages, and the health risks associated with long, low-quality marriages. This dissertation aims to elucidate the psychosocial and physiological mechanisms through which marriage affects health outcomes and disease processes by linking measures of marital quality (satisfaction, support, strain, disagreement) to physiological mechanisms that have been associated with psychosocial factors and 2 adverse health outcomes: inflammation, neural regulation of the heart, and hypothalamic- pituitary-adrenal axis (HPA) activity. Biomarkers of inflammation, interleukin-6 and c-reactive protein, have been associated with cardiovascular disease (Danesh et al., 2008), diabetes (Festa, 2002), cognitive decline (Gimeno et al., 2009; Weaver et al., 2002), disability (Ferrucci et al., 1999), and mortality (Hamer, Chida, Stamatakis, 2010), and have also been observed under unfavorable social conditions such as social isolation (Cole et al., 2007), perceived stress (McDade, Hawkley, & Cacioppo, 2006), and low socioeconomic status (Alley et al., 2006; McDade, Lindau, & Wroblewski, 2010). Neural regulation of the heart, measured using heart-rate variability (HRV), is beat-to-beat variation in the heart rate that reflects the resilience, flexibility, and adaptability of the autonomic nervous system in regulating the heart. HRV has been associated with social isolation and emotion, as well as cardiovascular events and all-cause mortality (Thayer, Yamamoto, & Brosschot, 2010). The circadian cortisol rhythm, an indicator of HPA activity, may also be an important link between marriage and health. Flattened circadian rhythms have been associated with major depression and post-traumatic stress disorder (Yeheuda, 1996), and a recent meta- analysis found that higher cortisol awakening responses are observed in individuals experiencing general life stress and that lower cortisol awakening responses are observed in individuals with post-traumatic stress disorder and among individuals with positive psychological characteristics such as happiness, optimism, and self esteem (Chida & Steptoe, 2009). These aberrations in cortisol production have been linked to downstream 3 health effects such as obesity, diabetes, and infection (Cutolo et al., 2008; Donoho, Weigensberg, Emken, Hsu, & Spruijt-Metz, 2010). This study is important because it will be the first study to examine associations between positive and negative aspects of marital quality and multiple biomarkers of disease risk in a national sample of US adults. This will also be the first study to compare the health status of married individuals in high quality marriages versus married individuals in low quality marriages. Findings from this study will contribute to our understanding of the physiological mechanisms linking marriage to health, and will answer questions regarding how individual perceptions of the quality of their marriage ‘get under the skin’ to cause disease in the US. Specific Aims Aim 1: To investigate the association between four indices of marital quality (satisfaction, support, strain, disagreement) and variation in inflammation among individuals in long- term (>10y) marriages, hypothesizing positive marital characteristics will be associated with lower levels of CRP and IL-6, while negative marital characteristics will be associated with higher levels. Aim 2a: To test whether marital status is associated with autonomic nervous system control of the heart (i.e., heart rate variability), hypothesizing married individuals will have higher HRV than unmarried individuals, and that continuously married individuals will have higher HRV than remarried individuals. Aim 2b: To test whether marital quality is associated with autonomic nervous system control of the heart (i.e., heart rate variability) among individuals in long-term marriages, 4 hypothesizing that positive marital characteristics will be associated with higher HRV and negative marital characteristics will be associated with lower HRV. Aim 3: To link marital quality to aberrations in diurnal cortisol production among individuals in long-term marriages, hypothesizing that individuals in marriages characterized by strain and conflict will have flat rhythms, while those in marriages characterized by support and satisfaction will have normal rhythms. This study will make a significant contribution to the scientific study of how marital quality influences disease risk. First, this study takes a novel approach to examining the association between marriage and health by examining individuals in long-term, stable marriages to test whether the health benefits of marriage are conditional on the quality of marriage. Examining long-term relationships is important because these couples are more likely to remain married, therefore increasing their exposure to the support and/or stress of their relationship. Second, the use of multiple measures of marital quality (satisfaction, support, strain, disagreement) among married individuals in long-term relationships is novel and has the potential to address which aspects marriage are associated with health among individuals in relatively stable relationships. This has the potential to guide further research aimed at developing interventions to increase the quality of marriages, and perhaps increase the health of the married. Third, the use of multiple biomarkers indicating dysregulation of multiple physiological systems will provide evidence for the physiological pathways through which marriage affects health. This study will be the first population study to use detailed measures of marital quality with a novel set of biomarkers. 5 Dissertation Format This dissertation opens with a chapter providing a broad overview of marriage and health, discussing previous research findings and the theoretical framework that guides this research, and details gaps in the literature that this study fills. This chapter ends with an overview of the Survey of Midlife in the United States (MIDUS), which is the study cohort from which these arise. The first chapter is followed by three research chapters that address a specific aim. Each research chapter contains its own literature review relevant to the specific aim of the empirical study. Because of the various sub- studies included in MIDUS, each research chapter also contains its own method section that clearly states the study method and study sample that were used for the analyses. The first two chapters are manuscripts that have been prepared for publication in peer-review journals; therefore, these chapters were left with the content requested from the editors and reviewers, and in the format requested by the journal. Thus, these chapters may contain some brief redundancy from Chapter 1. 6 CHAPTER 1: MARRIAGE, HEALTH, AND THE MIDUS STUDY Marriage and Health Social relationships play a significant role in health outcomes ranging from catching a cold (Cohen, Doyle, Skoner, Rabin, & Gwaltney, 1997) to mortality (Johnson, Backlund, Sorlie, & Loveless, 2000). Marriage is undoubtedly one of the most important social relationships established in adulthood, and the health benefits experienced by married adults are well documented. Married adults rate their health better, (Williams & Umberson, 2004) and report fewer chronic conditions and functional limitations (Hughes & Waite, 2009) compared to their non-married counterparts. However, despite the significance of marriage, the precise psychosocial and physiological mechanisms relating marriage to better health are unclear. Marriage is thought to reflect stable access to social, emotional, and tangible support (Ross, 1995), but marital status is a crude measure for understanding marriage because it does not account for the length of time that an individual has spent in the married state, and it does not account for the quality of the marital relationship. Long, high-quality relationships may be salutary for health, but long, low-quality relationships may be pernicious. The present study aims to evaluate three hypothesized physiological pathways through which marriage may influence health, by linking marital quality to biomarkers indicating inflammatory processes, autonomic nervous system function, and neuroendocrine function. Conceptual Framework Several theoretical frameworks have been key in understanding how marital relationships influence health. For the purpose of this study, we predominantly draw from 7 social relationships theories, examining marriage as a social relationship that may confer both social support and social stress. Social relationships theories (i.e., stress-buffering and social integration) posit that marriage is a source of social support (emotional, tangible, informational) that buffers the effects of life stress (Cohen & Wills, 1985; Ross, 1995), and that marriage can be a form of social integration that encourages salutary behaviors through social control, and results in positive psychological characteristics (e.g., self-esteem, identity) that increase health (House, Landis, & Umberson, 1988; Seeman, 1996). Marital Status & Health Martial status has been consistently related to all-cause (Johnson et al., 2000; Kaplan & Kronick, 2006) and cardiovascular disease mortality (Molloy, Stamatakis, Randall, & Hamer, 2009). These health differentials can partially be attributed to better health behaviors among the married, such as diet, exercise, and smoking. For example, married individuals have lower rates of smoking and alcohol consumption (Kaplan & Kronick, 2006; Lee et al., 2005; Malloy et al., 2009), are more likely to adhere to medical recommendations (Kulkarni, Alexander, Lytle, Heiss, & Peterson, 2006), and consume greater amounts of fruits and vegetables. On the other hand, the unmarried exercise the same (Malloy et al., 2009) or more than the married (Kaplan & Kronick, 2006; Lee et al., 2005), and also tend to maintain healthier weights (Kaplan & Kronick, 2006; Lee et al., 2005). Carr and Springer (2011) point out that couples may reinforce negative health behaviors when both spouses engage in similar behavior (i.e., smoking, overeating, 8 sedentary behavior). Thus, marriage may not result in better health if spouses do not inhibit unhealthy behaviors and promote healthy ones. Spouses may attempt to influence or control the health behaviors of their partner out of concern; however, these attempts at control may cause distress in the partner, especially if the tactics used are insensitive or unsympathetic (Hughes & Grove, 1981). Positive control behaviors (e.g., motivating, complimenting) have been shown to be more successful for behavior change than negative control behaviors (e.g., nagging and pressuring; Campbell, 2003; Fekete, Stephens, Druley, & Greene, 2006), although this may depend on whether the spouse is emotionally supportive (e.g., affectionate, points out strengths) or emotionally avoidant (e.g., changes the subject, dismisses feelings). For example, a study of patients recovering from knee surgery found that when spouses used negative control behaviors (e.g., nagging) and were emotionally avoidant (e.g., dismissing feelings), they elicited angry feelings from their spouse. Furthermore, when spouses used positive control behaviors (e.g., pointing out strengths), but were emotionally avoidant (e.g., dismissing feelings), patients adhered less to their treatment compared those with spouses that were not emotionally avoidant. Thus, although the health behaviors of the married may be better in general, they may depend on the health behaviors of both spouses, the positive or negative tactics used to influence behavior, and the type of emotional support provided. Marital History & Health Marital status and marital history (i.e., number of divorces or widowhoods, length of time spent in marriage, age of first marriage) have been related to increased mortality 9 (Dupre, Beck, & Meadows, 2009; Molloy, Stamatakis, Randall, & Hamer, 2009b), chronic conditions, decreased functional status, and self-rated health (Dupre & Meadows, 2007; Hughes & Waite, 2009). Although research on marital history considers a variety of factors, much of this research suggests that increased marital stability (i.e., greater duration) is linked with better health. Marital quality is related to couples remaining married (Gottman & Levenson, 1992), and is also likely to be important for understanding how marriage affects health over time. To date, however, the specific biological and psychosocial pathways through which marital history confers health have not been fully explored. Intermediary biological pathways and development of pre- clinical disease have not been examined. In addition, the relationship between marital duration and health has not considered marital quality – a contextual factor that is likely to be important for understanding how marriage affects health. If the duration of time one is married is an important factor in predicting health, it is likely contextualized by the quality of the marriage(s). Thus, long, supportive marriages should be salutary for one’s health, while long, conflict-ridden marriages should be pernicious. Marital Quality & Health. Marital quality is important for understanding how marriage influences health (Kiecolt-Glaser & Newton, 2001). Historically, research on marital quality has focused on a single dimension of marital quality - marital satisfaction. Marital satisfaction (i.e., How satisfied are you with your marriage?) provides a global assessment of one’s marriage; however, it disregards the multidimensionality of the marital relationship (Fincham & Linfield, 1997). Examining positive and negative evaluations of a marital 10 relationship can provide more nuanced information regarding various dimensions of marital quality and how they are differentially associated with health. As a form of social support, a marital relationship is thought to reduce the appraisal of an event as being stressful, promote positive coping strategies, and reduce negative emotional responses to stress (stress buffering hypothesis, Cohen & Wills, 1985). However, a marital relationship does not always confer social support, integration, or identity, nor does it guarantee that the support provided by one spouse will meet the needs of the other (Uchino, Vaughn, & Matwin, 2008). If a marital relationship is not a source of support, it may not reduce the impact of other life stressors, and individuals in marital relationships that are unsupportive or are not a source of stability or satisfaction may not experience the expected health benefits of marriage. An experiment using fMRI to examine neurological responses to physical stress/pain (an electric shock to the subjects ankle) found that participants subjected to the shock while alone had greater neurological stress responses than participants subjected to the same stress while holding a romantic partners’ hand, and the stress response varied as a function of relationship quality (Coan et al., 2006). A subsequent study indicated that this phenomenon also occurs when a subject views a picture of their romantic partner (Master et al., 2009). Thus, we might expect that the presence of a supportive spouse will reduce neurological responses to stress, therefore reducing the subsequent physiological responses to stress. Intimacy, companionship, and shared leisure activities in marriage are also factors that are important for maintaining hedonic tone and preventing feelings of isolation 11 (Rook, 1987). Hawkley and colleagues (2008) found that marriage prevented loneliness, but only if a spouse was considered a confidant. On the other hand, if relationships are unsupportive or are a source of strain and conflict, then marital relationships may be independent sources of stress, and may also exacerbate the effects of other life stressors on health (Seeman, 1996; Uchino, 2009). Conflict, disagreement, and hostility among marital partners have shown deleterious effects on the cardiovascular, neuroendocrine, and immune systems in laboratory experiments (Robles & Kiecolt-Glaser, 2003; Smith et al., 2009). Marital conflicts in laboratory settings have been associated with higher blood pressure reactivity, increased cardiac output, increased pulse, and decreased heart rate variability compared to collaborative interactions (Smith et al., 2009). In another laboratory study, marital conflict was associated with increases in IL-6 production over a 24-hour period and these increases were not observed when couples participated in the control task and did not engage in hostile behaviors (Gouin et al., 2009). Taken together, these studies indicate that in laboratory settings where conflict is required, married couples experience physiological stress responses that may have long-term health implications (Manuck, 1994). Thus, we expect that over time, couples with high conflict, disagreement, and hostility will have higher levels of inflammation and greater levels of cardiovascular disease risk because they engage in these behaviors frequently. Thus, assessing marital quality is important for understanding heterogeneity in the health of the married, and examining multiple domains of marital quality (e.g., satisfaction, support, strain, conflict) will provide a more detailed examination of how the 12 various facets of marital quality differentially predict proximal markers of downstream heath outcomes (Kiecolt-Glaser et al., 2011). Gender, Marriage, & Health Marriage may confer health differently for men and women. Men and women differ in their social networks and their sensitivity to relationship characteristics (Baumeister & Sommer, 1997), and therefore different facets of marital quality may differentially confer health benefits among men and women. Men have larger social networks and are more inclusive of others, whereas women tend to have smaller social networks focusing on developing stronger bonds among relatively few social ties (Baumeister & Sommer, 1997). Women also tend to become more emotionally upset than men during disagreements (Almeida & Kessler, 1998), and are more sensitive to threats associated with the disruption of relationship harmony; whereas men tend to be more sensitive to threats associated with loss of dominance, power, and control (Smith, Gallo, Goble, Ngu, & Stark, 1998). Thus, women may be sensitive to the negative qualities and the (absence of) positive qualities in their relationships, and men may be sensitive to changes in marital status and to marital conflict over issues of power and control. Biological Pathways Research has examined marital status and marital history with important health indicators and outcomes such as functional status, self-rated health, and mortality (Dupre, Beck, & Medows, 2009; Hughes & Waite, 2009; Sbarra & Nietert, 2009); however, the examination of intermediary biological pathways such as change in biological risk and development of pre-clinical disease have not been thoroughly examined. Additionally, 13 individuals experiencing differing levels of marital quality may differ in physiological stress responses and health-related risk factors and outcomes. Inflammation, autonomic balance, and neuroendocrine function are three important physiological pathways through which marriage may influence health, and the examination of biomarkers of these systems may provide more exact information on the specific physiological mechanisms linking marital quality and health. The Survey of Midlife in the United States To address the specific aims of the current research, publicly available data from the Survey of Midlife in the United States (MIDUS) will be used. MIDUS is study of health and well-being among US adults, which includes: (1) a phone interview and postal survey with detailed self-report measures of marital timing, status, duration, and quality (satisfaction, support, strain, and disagreement); (2) an overnight hospital stay with a detailed health interview and a physical health examination, including an extensive set of biomarkers (Biomarker Study); and (3) a daily diary study that includes salivary cortisol at four time points during the day for four days (National Survey of Daily Experiences [NSDE]). The MIDUS study was initially conducted by the MacArthur Foundation Research Network on Successful Midlife Development, taking place from 1995-1996 (MIDUS 1). The original sample was a national probability sample of non- institutionalized, English speaking midlife adults (age range 25-74) residing in the 48 contiguous states, with an additional sample of twins/triplets identified from an ongoing database of twins. An average of 9.2 years after completing MIDUS 1 (range = 7.8 to 14 10.4 years), participants were asked to participate in a telephone interview and subsequent postal survey (MIDUS 2) similar in content to MIDUS 1. Participants who completed the MIDUS 2 phone interview were recontacted and asked to participate in the Biomarker Study and NSDE. A detailed description of the sampling and participation rates across the MIDUS studies is described in detail elsewhere (Almeida, McGonagle, & King, 2009; Dienberg Love, Seeman, Weinstein, & Ryff, 2010). The Biomarker Study involved an overnight hospital stay at one of the three general clinical research centers in the US that included a detailed health interview, a physical health examination, the collection of biological specimens (e.g., blood, urine, saliva), and an additional psychosocial questionnaire to assess various physiological states (Dienberg-Love et al., 2010). Participants completed the biomarker study an average of 1.91 years after completing the MIDUS 2 phone interview. The age range of the sample at the time of the biomarker collection was 35 to 86. The Biomarker Study contains data from 1,054 respondents from the original MIDUS cohort. Biomarker data collection was carried out at three General Clinical Research Centers (at UCLA, University of Wisconsin, and Georgetown University). The NSDE, a daily diary study, was conducted from 2004 to 2005 and included short (10-20 minute) telephone interviews about daily experiences on 8 consecutive days. On four of the days participants provided salivary samples using home saliva collection kits, which were mailed to participants prior to the start of the study. The NSDE includes 1,842 respondents from the original MIDUS cohort that also participated in the MIDUS 2 survey. The age range of participants at the time of the study was 34 to 84. 15 CHAPTER 2: MARITAL QUALITY, GENDER, AND MARKERS OF INFLAMMATION IN THE MIDUS COHORT Social relationships play a significant role in health outcomes ranging from catching a cold (Cohen, Doyle, Skoner, Rabin, & Gwaltney, 1997) to mortality (Holt- Lunstad, Smith, & Layton, 2010). A marital relationship is one of the most important social relationships established in adulthood, and the better health experienced by married adults is well documented (Waite & Gallagher, 2000). Married adults live longer (Johnson, Backlund, Sorlie, & Loveless, 2000), rate their health better (Williams & Umberson, 2004), and report fewer chronic conditions and functional limitations (Hughes & Waite, 2009), compared to their non-married counterparts. While prior research has established that marriage confers health benefits, there has been less focus on how the health of married individuals may vary as a function of the quality of a marriage and the duration of time spent in a marriage. Larger studies have generally focused on the association between marital quality and broadly defined health outcomes, whereas smaller laboratory-based studies have focused on experimentally induced marital conflict or collaboration and physiological processes (Carr & Springer, 2010). Thus it is important to examine the association between marital quality and physiological processes related to health in a population-based study of adults. Greater understanding of physiological processes among individuals in long-term marriages could enhance our understanding of the health benefits associated with long, high-quality marriages, and the health risks associated with long, low-quality marriages. 16 The present study aims to evaluate one hypothesized physiological pathway through which marriage may influence health, by linking the quality of marital relationships to two biomarkers indicating inflammatory processes: Interleukin-6 (IL-6), a proinflammatory cytokine, and C-reactive protein (CRP), an acute phase protein. Social relationships have shown strong associations with many diseases also associated with inflammatory processes, such as cardiovascular disease and cancer, indicating that inflammation may be an important pathway linking the social environment to health outcomes (Kiecolt-Glaser, Gouin, & Hantsoo, 2010). It is hypothesized that psychological stress activates a cascade of physiological processes resulting in the production of proinflammatory cytokines (e.g., IL-6, CRP), contributing to reductions in cell division and accelerated cellular aging (Epel, 2009). Marital relationships may provide social and emotional support that can buffer the effects of other life stressors, leading to decreased inflammation and better health outcomes (Cohen & Wills, 1985; Ross, 1995). Conversely, marriage may be a source of stress in its own right and the strain associated with marital conflict and/or increased demands and obligations may lead to increased inflammation and subsequently poorer health outcomes (Berkman, Glass, Brissette, & Seeman, 2000). Thus, inflammation may be one key pathway linking the quality of a marital relationship to health outcomes. Conceptual Framework Several theoretical frameworks have been key in understanding how marital relationships influence health. For the purpose of this study, we predominantly draw from social relationships theories, examining marriage as a social relationship that may confer 17 both social support and social stress. Social relationships theories (i.e., stress-buffering and social integration) posit that marriage is a source of social support (emotional, tangible, informational) that buffers the effects of life stress (Cohen & Wills, 1985; Ross, 1995), and that marriage can be a form of social integration that encourages salutary behaviors through social control, and results in positive psychological characteristics (e.g., self-esteem, identity) that increase health (House, Landis, & Umberson, 1988; Seeman, 1996). Marital status and marital history (i.e., number of divorces or widowhoods, length of time spent in marriage, age of first marriage) have been related to increased mortality (Dupre, Beck, & Meadows, 2009; Molloy, Stamatakis, Randall, & Hamer, 2009), chronic conditions, decreased functional status, and self-rated health (Dupre & Meadows, 2007; Hughes & Waite, 2009). Although research on marital history considers a variety of factors, much of this research suggests that increased marital stability (i.e., greater duration) is linked with better health. If the duration of marriage is an important factor in predicting health, it is likely contextualized by the quality of the marriage(s). Marital quality is related to couples remaining married (Gottman & Levenson, 1992), and is also likely to be important for understanding how marriage affects health over time. Thus, long, supportive marriages should be salutary for one’s health, and long, conflict-ridden marriages should be pernicious. Marital Quality and Health Marital quality is important for understanding how marriage influences health (Kiecolt-Glaser & Newton, 2001). Historically, research on marital quality has focused 18 on a single dimension of marital quality - marital satisfaction. Marital satisfaction provides a global assessment of one’s marriage; however, it disregards the multidimensionality of the marital relationship (Fincham & Linfield, 1997). Examining positive and negative evaluations of a marital relationship can provide more nuanced information regarding various dimensions of marital quality and how they are differentially associated with health. As a form of social support, a marital relationship is thought to reduce the appraisal of an event as being stressful, promote positive coping strategies, and reduce negative emotional responses to stress (Cohen & Wills, 1985). An experiment examining neurological responses to physical stress/pain among women found that wives subjected to a shock while alone had greater neurological stress responses than wives subjected to the same stress while holding their husband’s hand. Furthermore, these stress responses varied as a function of marital satisfaction, with less response observed among individuals reporting higher levels of satisfaction (Coan, Schaefer, & Davidson, 2006). However, a marital relationship does not always provide social support, nor does it guarantee that the support provided by one spouse will meet the needs of the other spouse (Uchino, Vaughn, & Matwin, 2008). If a marital relationship is not a source of support, it may not reduce the impact of other life stressors, and individuals in marital relationships that are unsupportive or are not a source of stability or satisfaction may not experience the expected health benefits of marriage. Thus, we expect that the presence of a supportive spouse will reduce neurological responses to stress, therefore reducing the 19 subsequent cascade of inflammatory cytokines and acute phase proteins, such as IL-6 and CRP. On the other hand, if relationships are unsupportive or are a source of strain and conflict, they may be independent sources of stress, and may exacerbate the effects of other life stressors on health (Seeman, 1996; Uchino, 2009). Conflict, disagreement, and hostility among marital partners have shown deleterious effects on the cardiovascular, neuroendocrine, and immune systems in laboratory experiments (Robles & Kiecolt- Glaser, 2003; Smith et al., 2009). In laboratory settings, marital conflict has also been associated with increases in IL-6, and these increases were not observed when couples participated in the control task and did not engage in hostile behaviors (Gouin et al., 2009). Taken together, these studies indicate that in laboratory settings where conflict is induced, married couples experience physiological stress responses. One implication of such findings is that, if these responses are repeated over many years outside the lab, they may have long-term health implications (Manuck, 1994; McEwen & Seeman, 1999). Thus, we expect that over time, couples with high conflict, disagreement, and hostility will have higher levels of inflammation and greater levels of cardiovascular disease risk because they engage in these behaviors frequently. Thus, assessing marital quality is important for understanding heterogeneity in the health of the married, and examining both positive and negative aspects of marital quality (e.g., satisfaction, support, strain) can provide a more detailed examination of how the various facets of marital quality may differentially predict downstream heath outcomes. 20 Inflammation, Health, and Marriage Inflammation plays a substantial role in aging, and is considered to be a key process linking environmental stressors to longevity and age-related diseases such as cardiovascular disease, diabetes, and cancer (Finch, 2011). Interleukin-6 (IL-6) and C- reactive protein (CRP) are two indicators of general systemic inflammation that can provide information regarding the links between social relationships and disease etiology (Kiecolt-Glaser, Gouin, & Hantsoo, 2010). IL-6 and CRP are part of both the adaptive and innate immune systems (Finch, 2011). IL-6 is an inflammatory cytokine that stimulates the production of several acute phase proteins, including CRP, which is created in the liver and is responsible for reduced endothelial vasodilation and increased platelet aggregation (Kishimoto, 2005). IL-6 is expressed in many cells, including macrophages and adipose (fat) cells, with higher levels of expression in visceral fat cells, or the abdominal fat cells that surround internal organs. IL-6 and CRP are seen as both causes and consequences of health change with aging, and have been associated with cardiovascular disease (Danesh et al., 2008), diabetes (Festa, D'Agostino, Tracy, Haffner, & Insulin Resistance Atherosclerosis Study, 2002), cognitive decline (Gimeno et al., 2009), disability (Ferrucci et al., 2005), and mortality (Alley, Crimmins, Bandeen-Roche, Guralnik, & Ferrucci, 2007). Basal levels of IL-6 and CRP are good proximal markers of disease risk and may become dysregulated with chronic exposure to social stress (Alley, Crimmins, Bandeen- Roche, Guralnik, & Ferrucci, 2007) ). IL-6 and CRP are always present in the body at low levels and can become rapidly elevated by acute illness and psychological stress, 21 returning to normal basal levels shortly after recovery from infection or stress. Research indicates that chronic stressors such as poverty and caregiving and chronic health conditions such as obesity and cardiovascular disease are related to elevated basal levels of IL-6 and CRP (Alley et al., 2006; Kiecolt-Glaser, Bane, Glaser, & Malarkey, 2003). Thus, inflammation is a plausible physiological pathway linking marriage to health. It has been shown to be associated with marital quality in laboratory studies, and therefore it makes sense to study the association between marital quality and inflammation in a population-based study of individuals in long-term marriages. Gender, Marital Quality, and Health Marriage may confer health differently for men and women. Men and women differ in their social networks and their sensitivity to relationship characteristics (Baumeister & Sommer, 1997). Therefore, different facets of marital quality may differentially confer health benefits or costs among men and women. Men have larger social networks and are more inclusive of others, whereas women tend to have smaller social networks focusing on developing stronger bonds among relatively few social ties (Baumeister & Sommer, 1997). Thus, men may experience health benefits from the social tie associated with marriage, as evidenced by previous research indicating that marriage is more beneficial for men than women, but women may be more sensitive to the quality of their marital relationship. Women tend to become more emotionally upset than men during disagreements (Almeida & Kessler, 1998), and are more sensitive to threats associated with the disruption of relationship harmony; whereas men tend to be more sensitive to threats associated with loss of dominance, power, and control (Smith, 22 Gallo, Goble, Ngu, & Stark, 1998). Thus, women may be more sensitive to the absence of positive qualities such as support, affection, and satisfaction, compared to men; whereas men may be more sensitive to changes in marital status and marital strain or conflict over issues of power and control. To better understand how marital quality influences inflammation in men and women, examining both positive and negative marital characteristics is important and may aid in our understanding of various mechanisms leading to gender differences in associations between marriage and inflammation (Kiecolt-Glaser & Newton, 2001). Gender differences in immune response have been found among couples participating in laboratory conflict tasks, but the evidence is conflicting. Individuals that report more negative behaviors during arguments in general, as well as those engaging in hostile and negative behaviors in laboratory conflict resolution tasks, were found to have reduced blastogenic responses to mitogens, reduced NK cell lysis, and increased Epstein- Barr virus antibody titers, compared to those reporting or engaging in fewer of these behaviors - and these effects were greater among women than men (Kiecolt-Glaser et al., 2005). Other laboratory studies have shown acute increases in IL-6 among individuals that display negative behaviors while participating in a conflict task, and among couples displaying hostile behaviors during marital interactions, but have not found gender differences in these associations (Gouin et al., 2009). These conflicting results may be due to several reasons: First, individuals engaging in hostile behaviors during induced laboratory conflict engage in the conflict task regardless of whether they would normally engage in conflict. Second, previous research has indicated a substantial percentage of 23 married participants refuse to participate in laboratory studies of marital conflict after learning a study includes discussion about a current marital problem (Miller, Dopp, Myers, Stevens, & Fahey, 1999). Thus, using a population survey may better capture chronic conflict and feelings of support, and may have better participation rates for all couples. These methodological differences may increase the ability to detect gender differences in the association between marital quality and health. The Present Study The aim of this study is to examine links between marital quality - defined by marital support and marital strain - and inflammatory processes as one potential pathway by which marriage may influence health over the life cycle. To understand the effects of marital quality among couples in long-term relationships, we examine only individuals married to the same person for 10 or more years. We examine associations between positive and negative marital characteristics – support and strain – and IL-6 and CRP, hypothesizing that higher levels of marital strain and lower levels of marital support will be associated with higher levels of inflammatory markers IL-6 and CRP. We also examine gender differences in these associations, hypothesizing that women’s inflammatory levels will be more strongly associated with marital support and strain compared to men’s, and that, for men, inflammatory levels will be associated with marital strain and conflict, due to men’s greater sensitivity to threats of dominance and control, but not support. 24 Demographic, Behavioral, and Psychosocial Correlates of Inflammation Several demographic, behavioral, and psychosocial factors are established corollaries of both inflammation and social relationships. Age is associated with higher levels of IL-6 and CRP and is associated with many diseases of aging (Crimmins, Vasunilashorn, Kim, & Alley, 2008). Socioeconomic status gradients in IL-6 and CRP have been observed among British Civil Servants (Gimeno et al., 2009) and in US populations (Friedman & Herd, 2010), although some studies have only observed social differences at very high levels of CRP (Alley et al., 2006). Behavioral factors have also been associated with inflammation and social relationships. Smoking has been consistently associated with higher levels of CRP and IL-6 (Hamer & Chida, 2009). Regular physical activity and exercise have been associated with lower levels of IL-6 and CRP in population studies (Colbert et al., 2004; Singh & Newman, 2010). Obesity – particularly abdominal obesity – has been associated with increased inflammation, due to the release of IL-6 by visceral (abdominal) adipose tissue cells in close proximity to the liver, which causes greater production of CRP (Fontana, Eagon, Trujillo, Scherer, & Klein, 2007). Abdominal adiposity has also been related to higher levels of social stress in adults (Brunner, Chandola, & Marmot, 2007) and children (Donoho, Weigensberg, Emken, Hsu, & Spruijt-Metz, 2010). The use of statin medications, prescribed for lowering cholesterol, has been associated with lower levels of inflammation, including CRP and Il-6 (März, Winkler, Nauck, Böhm, & Winkelmann, 2003). Poor sleep quality has been associated with higher levels of CRP (Matthews et al., 2010), and has also been associated with marital quality (Friedman et al., 2005). Lastly, 25 depression has shown consistent associations with elevated IL-6 and CRP (Howren, Lamkin, & Suls, 2009), and has also been associated with higher levels of marital conflict (Choi & Marks, 2008) and lower levels of marital satisfaction (Renshaw, Blais, & Smith, 2010). Thus, it is critical to examine the effects of these potential confounders of the association between marital quality and inflammation. Method We analyzed data from the Survey of Midlife in the US (MIDUS) Biomarker Study. The MIDUS study was initially conducted by the MacArthur Foundation Research Network on Successful Midlife Development in 1995-1996 (MIDUS 1). The original sample was a national probability sample of non-institutionalized, English speaking midlife adults (age range 25-74) residing in the 48 contiguous states, with an additional sample of twins/triplets identified from an ongoing database of twins. An average of 9.2 years after completing MIDUS 1 (range = 7.8 to 10.4 years), participants were asked to participate in a telephone interview and subsequent postal survey (MIDUS 2) similar in content to MIDUS 1. Participants who completed the MIDUS 2 phone interview were subsequently asked to participate in the Biomarker Study. The Biomarker Study involved an overnight hospital stay at one of the three general clinical research centers in the US that included a detailed health interview, a physical health examination, the collection of biological specimens, and an additional psychosocial questionnaire to assess various physiological states (Dienberg-Love, Seeman, Weinstein, & Ryff, 2010). Participants completed the biomarker study an average of 1.91 years after completing the MIDUS 2 phone interview. The age range of the sample at the time of the biomarker collection was 26 35 to 86. A detailed description of the sampling and participation rates across the MIDUS studies is provided elsewhere (Dienberg-Love, Seeman, Weinstein, & Ryff, 2010). The analytic sample was restricted to individuals who were married to the same person at the time of the biomarker study and at MIDUS 1. This sample allowed for the examination of health differences associated with marital quality for individuals in long- term relationships. We also restricted our analyses to non-Hispanic White race/ethnicity because of the small number of sample members with other ethnicities: (Hispanic White (n = 6), Native American (n = 3), African American (n = 12), Multiracial/Other (n = 11)). This resulted in an analytic sample of 553 participants (264 female), including 70 monozygotic twins, 55 dizygotic twins (1 set of triplets), and 2 siblings. The MIDUS study did not include any spousal dyads, and therefore there are no individuals reporting on the same marriage in this study. Descriptive characteristics of the final analytic sample are presented in Table 1. Measures Marital Quality. Marital quality was measured with two scales indicating spousal support and spousal strain (Schuster, Kessler, & Aseltine, 1990; Turner, Frankel, & Levin, 1983). Support (α = .90) was assessed using the mean of six items rated on a 4-point Likert scale from not at all to a lot. Questions included feelings of being cared for, understood, and appreciated, as well as being able to rely on, open up to, and relax around one’s spouse. Strain (α = .87) was assessed using the mean of six items tapping perceptions of criticism, demands, tension, feeling let-down, irritability, and arguments, 27 using a 4-point Likert scale from not at all to a lot. Marital support and strain scores from MIDUS 2 were used. Interleukin-6 and C-Reactive Protein (CRP). After overnight fasting, blood samples were obtained from all participants in the morning following a standard study protocol (Ryff, Seeman, & Weinstein, 2010). Samples were frozen at -60 to -80°C and shipped on dry ice to the MIDUS Biocore Laboratory where they were stored at -65°C for monthly batch analysis to ensure consistency across the three collection sites. IL-6 was assayed using Quantikine® High-sensitivity ELISA kits (R&D Systems, Minneapolis, MN). The assay sensitivity, or minimum amount of IL-6 that could be accurately be measured using this assay, was 0.16 pg/ml. The inter-assay coefficient of variation (CV), or variation in the control substrate run on each plate used to measure plate-to-plate consistency, was 13%. The intra-assay CV, or variation observed when many duplicate samples are run, was 4.1%. CRP was assayed using particle-enhanced immunonephelometry (BN II nephelometer, Dade Behring Inc., Deerfield, IL). The assay sensitivity was 0.18 ug/ml, inter-assay CV was 5.7%, and intra-assay CV was 4.4%, varying slightly between the different batches run. The intra- and inter-assay CVs for these samples were well below 20%, an established acceptable range (DeSilva et al., 2003). Anthropometric, Behavioral and Psychosocial Variables. Waist and hip circumferences were measured to the nearest millimeter by a trained clinician, using a Gulick II tape measure. Waist circumference was measured at the narrowest point between the hips and the iliac crest. Hip circumference was measured at the iliac crest. 28 Age was calculated from self-reported date of birth. Exercise was coded as a dummy variable, with 1 indicating 20 minutes of light, moderate, or vigorous activity three times/week or more, and 0 indicating fewer than three times/week. Current smoking status was self-reported and coded as a dummy variable (0 = nonsmoker, 1 = smoker). Sleep quality was measured using the subjective sleep item from the Pittsburgh Sleep Quality Index (Buysse, Reynolds, Monk, Berman, & Kupfer, 1989), “During the past month, how would you rate your sleep quality overall?” Response options included: very good, fairly good, fairly bad, and very bad. Depressive symptoms were measured using the Center for Epidemiologic Studies Depression Scale [CES-D (Radloff, 1977)]. This scale is well validated and has been widely used to characterize depressive symptomatology in large epidemiological studies. The scale contains 20 items tapping feelings of depression (e.g., sad, lonely) absence of positive affect (e.g., happy, joyful), somatic symptoms (e.g., restless, poor appetite), and interpersonal difficulties (e.g., feeling disliked, people were unfriendly) over the past week. Items were rated on a 4-point Likert scale ranging from rarely or none of the time to most or all of the time, with higher scores indicating greater symptoms and the mean of the 20 items was used. Internal consistency of the scale was high in this sample (α = .89). Data Analysis To contextualize the effect of each marital quality characteristic (support, strain), the two indices were examined separately with serum IL-6 and CRP. We used OLS regression with clustered standard errors to account for the non-independence of the twin and sibling pairs. The initial distributions of IL-6 and CRP were positively skewed; 29 therefore, natural-log transformations were used to normalize these distributions and eliminate issues of heteroscedasticity in multivariate models. Linear models with log- transformed dependent variables yield estimates, that when multiplied by 100, are interpreted as percent change in the dependent variable with a one-unit increase in the independent variable, when all other covariates are held constant. We used a Type 1 error rate of .10 to indicate statistical significance, and placed a stronger emphasis on the effect size estimates and their confidence intervals. We interpreted the results using a traditional Type 1 error rate of less than .05 to indicate ‘significant’, and interpreted rates between .05 and .10 as being ‘marginally significant’. We used a hierarchical modeling strategy (Aiken & West, 1991; Cohen, Cohen, West, & Aiken, 2003), where Model 1 examined the association of each marital quality indicator (support, strain) with each inflammatory marker (Il-6, CRP), controlling for age, and gender, and education. In Model 2, the interaction of the quality indicator with gender was added, to examine whether the interactions account for greater variance than gender and marital quality alone. In Model 3, smoking, exercise, waist-hip ratio, statin use, sleep quality, and depressive symptoms were added as behavioral and psychosocial covariates that may confound associations between marital quality and inflammation. The interaction was only retained in Model 3 if significant (p < .10) in Model 2. All interaction terms were created with centered, continuous independent variables. When significant, interactions were probed using simple slope analysis (Aiken & West, 1991). Unstandardized coefficients are presented because standardized coefficients are inappropriate in models 30 containing interactions with centered predictors (Aiken & West, 1991). All statistical analyses were performed using Stata 11.1 (StataCorp., College Station, TX). Results Marital support was negatively correlated with marital strain (r = -.63), and IL-6 was positively associated with CRP (r = .53). Descriptive characteristics of the sample are presented in Table 2.1. There were significant gender differences in levels of marital support, t(540) = 3.24, p < .001, with men reporting greater levels of support. There were marginal gender differences in marital strain, t(540) = 1.56, p = .06, with men reporting lower levels of marital strain compared to women. Marital Support and Inflammatory Markers Table 2.2 presents the results of regression analyses of IL-6 and CRP on support. In the main effects model including age and education (Model 1), support was marginally and negatively associated with IL-6 (b = -.06, 95% CI = -.11 − .001, p = .06). In Model 2, the addition of the interaction of support and gender was significant (b = -.15, 95% CI = - .27 − -.03, p = .01). The interaction was probed using methods described by Aiken and West (1991), and is shown in Figure 2.1. Tests of simple slopes indicate that for every one unit increase in support, women have 12% less IL-6, on average (b = -.12, 95% CI = -.20 − -.05, p = .001), but not men (b =.03, 95% CI = -.06 − .12, p = .53). This interaction is illustrated in Figure 2.1. Education was not associated with IL-6, but age was significantly associated with higher IL-6 (b = .01, 95% CI = .00 − .01, p < .001). In Model 3, the addition of behavioral covariates did not attenuate the interaction between gender and support. Exercise and the use of statin medications were associated with 31 lower IL-6 (b = -.08, 95% CI = -.14 − -.01, p = .02, and b =-.06, 95% CI = -.12 − .00, p = .06, respectively), although statin drugs were only marginally associated with IL-6. Waist-to-hip ratio was associated with higher IL-6 (b = .77, 95% CI = .41 − 1.13, p < .001), and depressive symptoms were marginally associated with higher IL-6 (b = .004, 95% CI = .00 − .008, p = .08). Sleep and smoking were not significantly associated with IL-6. The results of regression analyses of CRP on support are presented in Table 2.2. In the main effects model including age and education (Model 1), support was not significantly associated with CRP (b = -.06, 95% CI = -.16 − .03, p = .17). In Model 2, the addition of the interaction of support and gender was marginally significant (b = -.17, 95% CI = -.35 − .01, p = .07). The interaction was probed using methods described by Aiken and West (1991), and tests of simple slopes indicated higher support was associated with lower CRP for women, with each unit increase in support corresponding to 14% less CRP, on average (b = -.14, 95% CI = -.27 − -.02, p = .03), but not for men (b = .03, 95% CI = -.10 − .16, p = .42). This interaction is illustrated in Figure 2.2. Education was not significantly associated with CRP, but age was associated with higher CRP (b = .01, 95% CI = .00 − .01, p = .003). In Model 3, the addition of behavioral covariates did not attenuate the interaction between gender and support. Exercise was associated with lower CRP (b = -.17, 95% CI = -.27 − -.07, p = .001), and waist-to-hip ratio was associated with higher CRP (b =1.44, 95% CI = .68 – 2.21, p < .001). Other behavioral covariates, smoking, use of statin drugs, and depressive symptoms, were not significantly associated with CRP. 32 Marital Strain and Inflammatory Markers Table 2.3 presents the regression of IL-6 on marital strain. Strain was associated with higher IL-6 in the main effects model, with a one-unit increase in strain being associated with a 5% higher level of IL-6 among men and women, on average (b = .05, 95% CI = .004 – .10, p = .03) controlling for age and gender. The interaction of gender and support was not significant (b = .06, 95% CI = -.03 – .16, p = .20), and was omitted from Model 3. The addition of behavioral covariates in Model 3 attenuated the association between strain and IL-6 (b = .03, 95% CI = -.02 – .08, p = .19). Marital strain was not associated with CRP in main effects models or in models including interactions with gender (Table 2.3). Discussion Prior research has indicated that being married is salutary for health outcomes (Johnson, Backlund, Sorlie, & Loveless, 2000; Kaplan & Kronick, 2006), but additional studies of women have shown marriage to be beneficial only to the happily married (Gallo et al., 2003; Gallo, Troxel, Matthews, & Kuller, 2003). Findings from the current study provide further evidence consistent with the latter point, but provide more nuanced measures of marital quality showing that for women, levels of inflammation are associated with levels of spousal support and are weakly associated with levels of spousal strain; and for men, levels of inflammation are only weakly associated with marital strain. These findings illustrate that differences in marital quality, broadly defined, are important for inflammation among women but may not be as important for men. It also answers important questions concerning two aspects of marital quality, marital support and 33 marital strain, and their associations with two physiological markers of inflammation, IL- 6 and CRP (Kiecolt-Glaser & Newton, 2001). In the present study marital support was associated with lower levels of CRP and IL-6 among women, but not men. Among women in this sample, a one-unit increase in spousal support was associated with 12-14% reductions in IL-6 and CRP, respectively – an effect that was stronger than the observed effect of regular exercise, being a non- smoker, or taking a statin medication. This finding is consistent with prior research indicating that long-term health trajectories are influenced by marital satisfaction among women (Gallo et al., 2003; Gallo, Troxel, Matthews, & Kuller, 2003; Troxel, Matthews, Gallo, & Kuller, 2005). Although we did not examine broader social ties among men, previous research indicates that marital status is associated with lower CRP among men but not women (Sbarra, 2009), and broader measures of social support and integration have typically been related to reductions in CRP among men, but not women. For example, general social support was related to lower CRP in adult men (Mezuk, Diez Roux, & Seeman, 2010), and having a greater number of social ties was related to lower CRP for men, but not for women (Ford, Loucks, & Berkman, 2006). Thus men may experience reduced inflammation in any marriage, and women may only experience the health benefits of marriage in a good marriage. Marital strain was associated with higher levels of IL-6, but not CRP, among both men and women; however, this effect was no longer significant in fully adjusted models. This finding indicates that strain is not as strongly or consistently associated with inflammation as support, and that associations between marital strain and IL-6 may be 34 accounted for by other factors such as obesity, medication use, and lack of physical activity, and depressive symptoms. Further studies with longitudinal measures of IL-6 and CRP could examine the potential mediating effects of these behavioral and psychosocial variables in the association between marital strain and inflammation. Although our findings are relatively consistent across IL-6 and CRP, the estimates were slightly different. The reasons for these differences cannot be directly assessed; however, these differences are likely a result of differing physiological pathways and functions. Interleukin 6 is an inflammatory cytokine, and CRP is an acute phase protein. While both are increased during infection and inflammation, they are produced differently. IL-6 is temporally more proximal than CRP in the inflammatory response, and IL-6 has a broader range of physiological functions (Kishimoto, 2005). When a macrophage comes in contact with bacteria, this causes greater expression of IL-6. Some of this IL-6 is carried to the liver to make CRP, which binds to bacteria and is removed from the system by CRP receptors. However, some IL-6 is transported to macrophages and is involved in the production of IL-1 receptor agonist in order to inhibit the pro- inflammatory effects of IL-1 beta. IL6 is also expressed in many cells, including macrophages and fat cells, with higher levels of expression in visceral fat cells, or fat surrounding the internal organs. The closer proximity of visceral fat cells to the liver results in increased production of CRP. Thus, there are several reasons for subtle differences in IL-6 and CRP, which is why it is useful to measure both. We hypothesized that men would be sensitive to issues of dominance, power, and control (Smith, Gallo, Goble, Ngu, & Stark, 1998), and thus their levels of inflammation 35 would be more sensitive to marital strain than marital support. Our findings provide some evidence indicating that men are more sensitive to marital strain than marital support. Among men, marital support was not associated with IL-6 or CRP; whereas, marital strain was associated with higher levels of IL-6, but was not associated with CRP. Our findings do not support our hypothesis that men are more sensitive to strain than women, as there was no interaction between gender and strain in predicting levels of inflammation. These findings are somewhat different than the stronger findings from laboratory studies, where conflict has consistently been associated with acute increases in inflammatory markers, including IL-6 (Kiecolt-Glaser et al., 2005). This could be due to methodological differences, as the participants in laboratory studies are engaging in a conflict task, whereas in this study, participants were reporting on general levels of marital strain. These differences could also be due to the sample, as the individuals in this study were considerably older and were in long-term marriages. This study was strengthened by the use of a large sample of US residents, well- validated indicators of immune function, and multiple measures of marital quality, capturing both positive and negative dimensions of marriage. However, there are also several notable limitations. First, the study population was Caucasian adults married for ten or more years, with relatively high educational attainment. This limits the generalizability of these findings to the broader population of US adults. There is considerable evidence indicating that marriage may be different for individuals of various ethnic, racial, and socioeconomic backgrounds (McLoyd, Cauce, Takeuchi, & Wilson, 2000). Although this is the reason we chose to restrict this study to Caucasians, larger 36 studies that can examine these associations among more diverse populations would be helpful. A second limitation is that declines in health may contribute to declines in marital quality, and there could also be reciprocal effects of marital quality and health over time. Longitudinal evidence indicates that low-quality marriages contribute to declines in self-rated health (Umberson, Williams, Powers, Liu, & Needham, 2006), but there have been no studies using measures of both biological indicators and marital quality indicators, therefore we cannot assume the directionality of this association. Lastly, like most studies, this study examined only one spouse in a spousal dyad, and therefore we were unable to examine the family unit. This is an important limitation in research on marriage because the marital quality rating of the one spouse could influence the health of the other spouse. Future research in this area should examine longitudinal changes in health indicators as a function of marital support in order to better understand the effects of different facets of marital quality over the life course. Marital quality should also be examined in more complex models of social relationships and as a potential moderator of the association between stress and inflammation (Uchino, Vaughn, & Matwin, 2008). Marital strain may exacerbate the effects of stress on inflammation or may have a direct effect of levels of inflammation, and marital support may decrease the effects of stress on levels of inflammation. There is some indication that marriage is good for men regardless of the positive qualities of the marriage (i.e., a main effect), whereas marriage may be good for women only when they are satisfied in their marriage. 37 In sum, these findings suggest that marital support and strain are important factors to consider when examining the relationship between marriage and health, and also suggest that inflammation may be one mechanism through which marital quality influences health outcomes. However, there are likely many biological pathways through which marital support and strain may influence health, and these other pathways are important to consider in future research. 38 Table 2.1. Descriptive Characteristics of the Analytic Sample by Gender (n = 542). Men Women (n = 283) (n = 259) p a Mean (SD) Mean (SD) t-test Age (years) 60.9 (11.3) 57.3 (1.6) <.001 Body mass index (kg/m 2 ) 29.6 (4.7) 28.3 (6.4) .01 Waist-hip ratio .97 (.08) .83 (.07) <.001 CES-D b 6.67 (6.54) 7.28 (7.13) .20 Sleep .86 (.59) 1.03 (.69) .001 Marital quality Support (1-4 c ) 3.72 (.43) 3.60 (.50) .002 Strain (1-4 c ) 2.11 (.57) 2.19 (.60) .16 Median Median t-test d Interleukin-6 (pg/mL) 1.91 1.87 .53 C-reactive protein (mg/L) 1.12 1.28 .009 Percent Percent χ 2 Education ≥ college degree 52% 46% .32 < college degree 48% 54% Current smoker 10% 5.4% .05 Exercise ≥ 3 times/week 80% 82% .43 Use of statin drugs 40.5% 26.2% <.001 Note: p-values are for gender differences. a t-tests, df = 540; χ 2 tests, df = 1. b Center for Epidemiological Studies Depression Scale. c 1 = not at all, 4 = a lot. d Test based on natural log-transformed values. Table 2.2. Regression Coefficients (95% Confidence Intervals) for Marital Support on Interleukin-6 a and C-Reactive Protein a (n = 542). Gender is coded 0 = male, 1 = female; CES-D = Center for Epidemiological Studies – Depression Scale a natural log transformed. + p < .10. * p < .05. ** p < .01. *** p < .001. Interleukin-6 a C-Reactive Protein a Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 Variable b (SE) b SE b SE b SE b SE b SE Support -.06 (.03) + .03 (.04) .06 (.04) -.06 (.05) .03 (.07) .06 (.07) Gender .00 (.03) .00 (.03) .12 (.04) ** .12 (.04) ** .12 (.04) ** .33 (.06) *** Support*Gender -.15 (.06) * -.15 (.06) ** -.17 (.09) + -.17 (.10) + Age .01 (.00) *** .01 (.00) *** .01 (.00) *** .01 (.00) ** .01 (.00) ** .00 (.00) * Education .00 (.01) .00 (.01) .00 (.01) -.01 (.01) -.01 (.01) .00 (.01) Smoker -.01 (.05) .02 (.07) Exercise -.08 (.03) * -.17 (.05) *** Waist/hip ratio .77 (.18) *** 1.44 (.37) *** Use of statins -.06 (.03) + .02 (.05) Sleep quality .00 (.02) -.03 (.03) CES-D .00 (.00) + .00 (.00) Constant .32 (.05) *** .32 (.05) *** .38 (.08) *** .15 (.07) * .15 (.07) * .09 (.12) Adusted R 2 .07 .08 .13 .03 .04 .10 39 Table 2.3 Regression Coefficients (95% Confidence Intervals) for Marital Strain Interleukin-6 a and C-Reactive Protein (n = 542). Gender is coded 0 = male, 1 = female; CES-D = Center for Epidemiological Studies - Depression a natural log transformed. + p < .10. * p < .05. ** p < .01. *** p < .001. Interleukin-6 a C-Reactive Protein a Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 Variable b SE b SE b SE b SE b SE b SE Strain .05 (.02) * .02 (.04) .03 (.03) .05 (.04) .00 (.05) .03 (.04) Gender .01 (.03) .01 (.03) .12 (.04) ** .13 (.04) ** .13 (.04) ** .33 (.06) *** Strain*Gender .06 (.05) .09 (.07) Age .01 (.00) *** .01 (.00) *** .01 (.00) *** .01 (.00) ** .01 (.00) ** .01 (.00) * Education .00 (.01) .00 (.01) .00 (.01) -.01 (.01) -.01 (.01) .00 (.01) Smoker .00 (.05) .02 (.07) Exercise -.08 (.03) * -.17 (.05) *** Waist/hip ratio .77 (.20) *** 1.44 (.40) *** Use of statins -.06 (.03) + .02 (.05) Sleep quality .00 (.02) -.03 (.03) CES-D .00 (.00) .00 (.00) Constant .31 (.05) *** .31 (.05) *** .37 (.08) *** .15 (.07) * .15 (.07) * .08 (.12) Adusted R 2 .07 .08 .13 .03 .03 .10 40 41 Figure 2.1 Figure 2.2 42 CHAPTER 3: THE PRICE THE HEART PAYS IN A BAD MARRIAGE: MARITAL STATUS, MARITAL QUALITY, GENDER, AND HEART RATE VARIABILITY IN THE MIDUS COHORT There is clear evidence linking marital status to cardiovascular disease (CVD) and mortality from CVD and all-causes (Johnson, Backlund, Sorlie, & Loveless, 2000; Kaplan & Kronick, 2006; Molloy, Stamatakis, Randall, & Hamer, 2009). Married individuals are consistently found to have better mental and physical health than their unmarried counterparts (Kiecolt-Glaser & Newton, 2001; Marks & Lambert, 1998). Martial status has been consistently related to all-cause (Kaplan & Kronick, 2006; Manzoli, Villari, Pirone, & Boccia, 2007; Rendall, Weden, Favreault, & Waldron, 2011) and cardiovascular disease mortality (Molloy, Stamatakis, Randall, & Hamer, 2009), with married individuals living longer than their non-married counterparts. A meta-analysis of over 50 studies linking marriage to mortality among older adults found that married individuals were 12% less likely to die compared to unmarried individuals, indicating that the married state is associated with better survival (Manzoli, Villari, Pirone, & Boccia, 2007). However, despite the apparent health benefits of marriage, the state of being married may not confer health unless one has a supportive and satisfying marriage (Seeman, 1996; Uchino, Cacioppo, & Kiecolt-Glaser, 1996). Although high conflict marriages may be expected to be short-lived, many people remain in low-quality marriages for many years, which could contribute to stress-related wear-and-tear on physiological systems (McEwen & Seeman, 1999). 43 Married individuals live longer (Rendall, Weden, Favreault, & Waldron, 2011), have fewer chronic conditions and mobility limitations (Hughes & Waite, 2009), and rate their health better (Williams & Umberson, 2004). Further evidence suggests that marital quality is also important, as low marital quality has been found to be a risk factor for myocardial infarction and heart failure (Coyne et al., 2001; De Vogli, Chandola, & Marmot, 2007; Orth-Gomer et al., 2000). However, despite marital status and marital quality having strong associations with these disease endpoints, the specific mechanisms linking marital status and marital quality to coronary events are largely unknown. Many risk factors have been identified as important indicators of health (for an extensive review, see Crimmins et al., 2008), and many of these have been previously associated with social relationships including marriage (Robles & Kiecolt-Glaser, 2003; Seeman, Singer, Ryff, Dienberg Love, & Levy-Storms, 2002). Research to date has examined resting parameters such as blood lipids and blood pressure (Gallo et al., 2003; Gallo, Troxel, Matthews, & Kuller, 2003; Troxel, Matthews, Gallo, & Kuller, 2005), and experimentally manipulated (stress induced) dynamic parameters of cardiovascular reactivity and recovery such as blood pressure, heart rate, and heart rate variability (Smith et al., 2009). However, to date, the association between marital status, marital quality, and resting heart rate variability has not been examined. Heart rate variability (HRV) is a measure of neural regulation of the heart that has been prospectively associated with atherosclerosis, congestive heart failure, myocardial infarction, cardiac death, and mortality (Bigger, Fleiss, Rolnitzky, & Steinman, 1993; de Bruyne et al., 1999; Dekker et al., 1997; Huikuri et al., 1999; La Rovere, Specchia, 44 Mortara, & Schwartz, 1988; Liao et al., 1997; Tsuji et al., 1996), and thus is a valuable indicator of cardiovascular health. The present study examines the association between marital status, marital quality, and heart rate variability - an important indicator of cardiovascular health (Kleiger, Stein, & Bigger, 2005; Thayer, Yamamoto, & Brosschot, 2010). HRV is beat-to-beat variation in the heart rate that reflects the resilience, flexibility, and adaptability of the autonomic nervous system in regulating the heart (Thayer & Lane, 2009). The sympathetic nervous system (known for ‘fight or flight’) drives the heart to beat faster, while the parasympathetic nervous system (known for ‘rest and restore’) simultaneously inhibits it through the direct innervation from the vagus nerve (Porges, 2007). The vagus nerve’s ability to rapidly inhibit (or disinhibit) sympathetic nervous system control of the heart has led to this process being referred to as ‘the vagal brake’ (Porges, 2007). The vagal break allows for rapid physiological adaptation to changing environmental cues and results in variability in the intervals between heartbeats (Task Force, 1996; Thayer & Sternberg, 2006). Thus, higher HRV during rest indicates greater vagal control and a greater degree of physiological adaptability, whereas lower HRV reflects slower reaction, indicating a lesser degree of adaptability. HRV may be an important physiological link between marital quality and health due to shared neurological mechanisms that regulate both HRV and social behavior and emotion (Thayer et al., 2009; Porges, 2007). Resting HRV has been has been described as a measure of self-regulatory capacity, or an indicator of one’s ability to socially engage 45 and attend to stimuli. For example, adults with higher HRV report greater emotion regulation and decreased negative emotional arousal when exposed to stress (Fabes & Eisenberg, 1997). Resting HRV is malleable as it decreases with age and may be influenced by long-term disruptions to social relationships and chronic stress. For example, a study of socially isolated prairie voles found sustained decreases in resting HRV after four weeks, compared to socially housed prairie voles (Grippo, Trahanas, Zimmerman, Porges, & Carter, 2009). In a study of married couples, decreases in resting HRV were observed 10 minutes after a negative marital interaction, and increases were observed after a positive interaction – although these effects were only observed among women (Smith et al., 2011). In humans, work stress and chronic worry have both shown associations with lower resting indices of HRV (Brosschot, Van Dijk, & Thayer, 2007; Hintsanen et al., 2007; Loerbroks et al., 2010; Uusitalo et al., 2011), although in some cases associations with work strain were only observed among women (Uusitalo et al., 2011), or within specific age groups (Loerbroks et al., 2010). Taken together, HRV appears to be affected by social stressors, and when these stressors are chronic - as they are in bad marriages - they may have long-term implications for regulation of the heart. Understanding marital quality and how it contributes to health is complex and is likely dependent on several factors such as individual perceptions of relationship satisfaction and support from a spouse, as well as the amount of conflict and stress that results from the relationship. Positive marital characteristics, such as satisfaction and support, may promote feelings of relationship security, positive affect, and well-being, and reduce feelings of loneliness (Hawkley et al., 2008; House, Landis, & Umberson, 46 1988; Rook, 1987; Seeman, 1996). Support from a marital partner may also buffer the effects of other life stressors, such as job stress (Cohen, Janicki-Deverts, & Miller, 2007; S. Cohen & Wills, 1985). Negative marital characteristics, such as conflict and strain, may cause stress and negative affect, and thus may have deleterious effects on health (Seeman, 1996). Epidemiological studies of marital quality and health outcomes have not focused on examining multiple dimensions of marital quality, and these may differentially contribute to disease. Feelings of satisfaction and support are likely to increase vagal control and promote social affiliation and calming. Conversely, conflict and stress in a marriage may decrease vagal control, resulting in active avoidance and excitability. Several studies indicate married women unsatisfied in their marriages have worse cardiovascular and metabolic profiles compared to women who are satisfied in their marriages (Gallo et al., 2003; Gallo, Troxel, Matthews, & Kuller, 2003; Troxel, Matthews, Gallo, & Kuller, 2005). Disagreements during marital interactions in laboratory settings have been associated with higher cardiovascular reactivity, as indicated by increases in blood pressure, cardiac output, and pulse during a laboratory conflict as compared to a collaborative interaction among older adults (Smith et al., 2009). In younger adults, findings were similar. Negative discussions were associated with increases in blood pressure, heart rate, and cardiac output; and decreases in peripheral resistance and pre-ejection period (Nealey-Moore, Smith, Uchino, Hawkins, & Olson-Cerny, 2007). Only one study has evaluated the association between marital quality and resting HRV, and this study found marital quality to be associated with HRV 47 for both women (Smith et al., 2011). For women, positive marital characteristics were associated with higher HRV; for men, negative characteristics were associated with lower HRV and positive characteristics were associated with higher HRV. This study consisted of 114 young married couples, most of which were married 1-3 years, recruited to participate in a laboratory disagreement with their spouse. Previous research has indicated a substantial percentage of married participants refuse to participate after learning a study includes discussion about a current marital problem (Miller, Dopp, Myers, Stevens, & Fahey, 1999). Thus, it is important to understand if marital quality is associated with HRV in a population-based sample of individuals in long-term marriages. Several demographic, behavioral and psychosocial factors may be associated with HRV and may alter the associations between marital quality and HRV. Age has been associated with decreased HRV (Antelmi et. al., 2004; Bonnemeier et al., 2003), and gender differences in marital strain across the life course have been observed, contributing to variation in the effect of marital quality on self-rated health (Umberson & Williams, 2005). Thus, age and gender may moderate the association between marital quality and health. Behavioral factors have been associated with HRV and may also explain associations between marital quality and HRV. Physical activity has been associated with higher resting HRV, whereas obesity and smoking have been associated with lower HRV. Depression has shown associations with decreased HRV in several studies (Carney et al., 2001), and other studies have indicated that marital quality may contribute to depression and that depression may contribute to poor marital quality 48 (Gotlib & Whiffen, 1989; Kouros, Papp, & Cummings, 2008). Thus, depression could mediate the association between marital quality and HRV. The present study examines marital status and positive and negative marital characteristics - support, strain, satisfaction, and disagreement - and their associations with HRV, in a large prospective study of US adults. We examine the contributions of demographic, behavioral, and psychosocial variables that are known corollaries of HRV, in order to understand whether these factors influence associations between marriage and HRV. We examine marital status among all participants; however, to understand the long-term effects of marital quality, we examine associations between marital quality and HRV among individuals married to the same person for 10 or more years. Method We analyzed data from the Survey of Midlife in the US (MIDUS) Biomarker Study. The MIDUS study was initially conducted by the MacArthur Foundation Research Network on Successful Midlife Development in 1995-1996 (MIDUS 1). The original sample was a national probability sample of non-institutionalized, English-speaking midlife adults (age range = 25-74) residing in the 48 contiguous states. It consisted of a randomly selected respondent from each household, along with siblings of a subset of respondents, and an additional sample of twins identified from an ongoing database of twins (see Brim et al., 2000). An average of 9.2 years after MIDUS 1 (range = 7.8 to 10.4), participants were asked to participate in a telephone interview and subsequent postal survey (MIDUS 2) similar in content to MIDUS 1. Participants who completed the MIDUS 2 phone interview were asked to participate in the Biomarker Study. The 49 Biomarker Study involved an overnight hospital stay that included a detailed health interview, a physical health examination, the collection of biological specimens (e.g., blood, urine, saliva), and an additional psychosocial questionnaire to assess various physiological states (Dienberg Love, Seeman, Weinstein, & Ryff, 2010). Morning blood samples were obtained and participants completed the psychophysiology assessment that included electrocardiograph recording (Dienberg Love, Seeman, Weinstein, & Ryff, 2010). Participants completed the biomarker study an average of 1.91 years after completing the MIDUS 2 phone interview, and were between 35 and 86 years of age at this time. Marital status was assessed at all three interviews (MIDUS 1, MIDUS 2, Biomarker Study). We examined differences in HRV across various marital status categories, and then restricted to individuals who were married at the time of the biomarker study and who had been married to the same person at MIDUS 1. This approach allowed for the examination of health differences associated with marital quality for individuals in long-term marriages. We also restricted our analyses to participants reporting non-Hispanic White ethnicity because there were too few participants reporting other ethnicities to include measures of ethnicity. Descriptive characteristics of the final analytic sample (N = 834) are presented in Table 3.1. Measures Marital Status. Marital status was self-reported in all analyses. Marital status categories indicate the participant’s self-reported marital status during the Biomarker Study. Marital status was assessed at all three interviews (MIDUS 1, MIDUS 2, 50 Biomarker Study), and classifications included married, separated/divorced, widowed, and never married. Using self-reported data on current marital status and the number of times an individual was married, two sub-categories were created to indicate whether a participant was married continuously to his/her first spouse (always married), or if the participant had divorced or widowed from a previous relationship and was currently remarried. Marital Quality. Marital quality was measured with a set of measures indicating marital satisfaction, support, strain, and disagreement. Global marital satisfaction was tapped with a single item: (1) How would you rate your marriage these days (a scale of 0 [worst] to 10 [best])? Spousal support (α = .88) was assessed using the mean of six items rated on a 4-point Likert scale from not at all to a lot. Questions included feelings of being cared for, understood, and appreciated, as well as being able to rely on and relax around one’s spouse. Spousal strain (α = .86) was assessed using the mean of six items tapping feelings of criticism, demands, tension, and arguments, using a 4-point Likert scale from not at all to a lot. Spousal disagreement α = .72) was assessed using the sum of responses to the question, “How much do you and your spouse disagree on the following issues?” The three issues included money matters (how to spend, save, or invest), household tasks (what needs to be done and who does it), and leisure time activities (what to do and with whom). Response options were rated on a 4-point Likert scale from not at all to a lot, and scores were combined to give a final scale range of 3-12. Marital quality measures from MIDUS 2 were used. 51 Heart Rate Variability (HRV). We present a commonly used measure of HRV, high frequency HRV (HF-HRV) also known as respiratory sinus arrhythmia (RSA). HF- HRV captures variance in the normal RR intervals within a specific frequency band (.15 - .50 Hz) that included the typical resting respiratory rate. Pharmacological blockade studies indicate HF-HRV is driven almost exclusively by parasympathetically driven cardiac vagal control (Cacioppo et al., 1994). HRV was obtained using electrocardiograph records, analyzed according to established guidelines (Task Force, 1996). During an 11-minute seated resting period, the analog ECG signal was digitized at 500 Hz and collected by a microcomputer. R waves were detected using event detection software, and were visually inspected to correct for any errors and files of RR intervals were generated. These files were submitted to Fourier based spectral analysis and power in the HF band (0.15-0.50 Hz) was computed in 300 sec epochs. If a 300-second duration was not obtainable, 60-second decreases were made until a continuous epoch could be examined. If the duration of the epoch was shorter than 180-seconds, the observation was omitted. If two measurements were obtained, the two measurements were averaged. More detailed information is provided elsewhere (Ryff et al., 2006). Anthropometric, Behavioral and Psychosocial Variables. Age was calculated from self-reported date of birth. Exercise was coded as a dummy variable, with 1 indicating 20 minutes of light, moderate, or vigorous activity three times/week or more, and 0 indicating fewer than three times/week. Current smoking status was self-reported and coded as a dummy variable (0 = nonsmoker, 1 = smoker). Height and weight were 52 both measured by clinical staff. Body mass index (BMI), was calculated using weight in kilograms divided by squared height in meters. Education was dichotomized into a two categories indicating whether a person graduated from a 4-5 year college or not. Participants were asked to bring in all current prescription, over-the-counter, and alternative medications in their original bottles. Two dummy variables were created to identify individuals using medications determined to increase HF-HRV or decrease HF- HRV in at least 10% of individuals taking these medications, based on previous literature. Medications determined to increase HF-HRV included beta-blockers and dioxin drugs. Medications determined to decrease HF-HRV included anticholinergic drugs and drugs with anticholinergic side-effects, sedatives, selective serotonin reuptake inhibitors, antipsychotics, and GABA-acting analgesics. Depressive symptoms were measured using the Center for Epidemiologic Studies Depression Scale (Radloff, 1977). This scale is well validated and has been widely used to characterize depressive symptomatology in large epidemiological studies. The scale contains 20 items tapping feelings of depression (e.g., sad, lonely) absence of positive affect (e.g., happy, joyful), somatic symptoms (e.g., restless, poor appetite), and interpersonal difficulties (e.g., feeling disliked, people were unfriendly) over the past week. Items were rated on a 4-point Likert scale ranging from rarely or none of the time to most or all of the time, with higher scores indicating greater symptoms and the mean of the 20 items was used. Internal consistency of the scale was high in this sample (α = .89). 53 Data Analysis To contextualize the effect of each marital characteristic, the four indicators of marital quality were examined separately. We used OLS regression with clustered standard errors to account for the non-independence of the twin and sibling pairs. HF- HRV was natural-log transformed to normalize the distribution, a standard practice in the analysis of HRV (Task Force, 1996). Linear models with log-transformed dependent variables yield estimates, that when multiplied by 100, are interpreted as percent change in the dependent variable with a one-unit increase in the independent variable, when all other covariates are held constant. Model 1 examines the association of HRV with each marital quality indicator, controlling for age and gender. Model 2 adds additional covariates: education, smoking status, physical activity, body mass index (BMI), medication use and depressive symptoms. There were no significant two- or three-way interactions between the marital quality indicators, age, or gender. All statistical analyses were performed using Stata 11.1 (StataCorp., College Station, TX). Results Descriptive characteristics of the analytic sample are presented in Table 3.1. Marital quality was generally high, as was expected in couples in long-term marriages. Bivariate correlations of major study variables are presented in Table 3.2. The four marital quality indicators were correlated between .44 and .75, with expected negative correlations between positive marital characteristics (i.e., support, strain) and negative marital characteristics (i.e., strain and disagreement). Age was moderately correlated with 54 marital satisfaction and support, and inversely correlated with marital strain and disagreement. To examine gender differences in the association between marital status and HF- HRV, we first tested interactions between gender and traditional marital status categories (married, divorced/separated, widowed, never married) as well as marital status categories differentiating always married and remarried individuals. These interactions were not statistically significant; therefore men and women were combined in the same models. Table 3.3 presents the regression of HF-HRV on marital status. The results of the analysis for traditional marital status categories indicated that there were no significant differences between married and unmarried individuals, controlling for all covariates. However, after separating always married and remarried, the results indicated that the remarried have lower HF-HRV compared to the always married after controlling for gender, age, and education (b = -.21, p = .07), and this effect remained even after controlling for all covariates (b = -.23, p = .05). Next, we examined gender differences in the association between marital quality and HF-HRV among individuals married for 10 or more years. Interactions of each marital quality variable and gender were tested, and none of the interactions were significant; therefore men and women were combined. Table 3.4 presents the regression of HF-HRV on the four marital quality indicators. Significant associations between marital quality and HF-HRV were observed. As expected, positive associations between HF-HRV and marital satisfaction and spousal support were observed, and negative associations between HF-HRV and marital strain and disagreement were observed. 55 Satisfaction was significantly associated with higher HF-HRV (b = .07, p = .02), and the addition of covariates did not attenuate this relationship. Spousal support was modestly associated with HF-HRV (b = .21, p = .09), although this association was not significant with the addition of covariates (b = .19, p = .13). Marital strain was only modestly associated with decreased HF-HRV (b = -.18, p = .06), and this association was attenuated by the addition of covariates (b = -.13, p = .20). Disagreement was associated with HF-HRV (b = -.05, p = .04), and was slightly attenuated by the addition of covariates (b = -.04, p = .08). Depressive symptoms were not associated with HF-HRV in any of the multivariate models. Medication use was a strong predictor of HF-HRV, and many of these medications included drugs used to treat depression (e.g., selective serotonin reuptake inhibitors). When individuals taking medication were excluded from the analyses, the results were nearly identical; therefore, analyses using depression as a mediator were not conducted because depressive symptoms were not associated with HF- HRV in this sample. Interestingly, education, smoking, and exercise were not significantly associated with HF-HRV in multivariate models, but BMI was associated with lower HF-HRV in all of the models. Discussion This is the first study to examine the association between marital quality and HF- HRV in a national sample of midlife adults in long-term marriages. Surprisingly, significant differences in HF-HRV between married and unmarried adults were not observed. However, when differentiating between always married and remarried adults, 56 the results suggested that remarried individuals have lower HF-HRV than always married individuals. To put this in context, the effect of being remarried, compared to being continuously married, was equivalent to being a smoker or being 8 years older. We also observed modest associations between marital quality and HF-HRV, with marital satisfaction and support associated with higher HF-HRV, and marital strain and disagreements associated with lower HF-HRV. Controlling for behavioral factors, medication use, and depressive symptoms attenuated these relationships but did not eliminate associations with marital satisfaction or disagreement. These findings suggest that there is heterogeneity in the health of the married, and that having a low quality relationship, or being remarried, may have implications for cardiovascular health independently of depressive symptoms and negative emotional states. Although the associations between marital quality and HF-HRV reported here are modest, they illustrate that even in relatively stable, long-term marriages, small variations in the quality of one’s marriage can affect the heart. HRV is a valuable indicator of cardiovascular health that has previously been associated with disease endpoints and mortality (Bigger, Fleiss, Rolnitzky, & Steinman, 1993; de Bruyne et al., 1999; Dekker et al., 1997; Huikuri et al., 1999; La Rovere, Specchia, Mortara, & Schwartz, 1988; Liao et al., 1997; Tsuji et al., 1996). Marital status has shown consistent associations with cardiovascular disease and mortality (Johnson, Backlund, Sorlie, & Loveless, 2000; Kaplan & Kronick, 2006), and marital quality has been shown to be a predictor of cardiac events (Coyne et al., 2001; Orth-Gomer et al., 2000). The significant associations observed between marital quality and HF-HRV in this 57 study provide a plausible physiological mechanism linking the quality of one’s marriage to a proximal indicator of cardiovascular health. However, the lack of difference in HRV between the married and unmarried suggests that HRV may not be a mechanism linking marital status to downstream health outcomes such as mortality. The lower heart rate variability observed among the remarried compared to the always married suggests that previous marital disruptions may have lasting effects on cardiovascular health, or that there may be inherent differences (e.g., personality, socioeconomic status) between individuals that are continuously married versus those who remain married to the same person. Surprisingly, age and gender did not modify the associations between the marital quality indicators and HRV. Bivariate associations showed age was moderately correlated with marital satisfaction and support, and inversely correlated with marital strain and disagreement. This was surprising, as these were all individuals married to the same person for 10 or more years. This could be due the positivity effect with aging (Mather & Carstensen, 2005), where events may be recalled more positively in older adults compared to younger adults. This study has a number of strengths, including the use of a large sample of US residents and multiple measures of marital quality, capturing both positive and negative dimensions of marriage. However, there are also several notable limitations. First, the study population was Caucasian adults married for ten or more years, with relatively high educational attainment, limiting the generalizability of these findings to the broader population. Additionally, declines in vagal control (indexed by HF-HRV) could 58 contribute to poor emotion regulation and thus result in poor marital quality, although this seems unlikely given that age was associated with both lower HF-HRV and higher ratings of marital quality. Future research in this area should examine longitudinal changes in HRV and marital quality, in order to better understand whether longitudinal changes in marital quality coincide or with or precede changes in HRV. Marital quality should also be examined in the context of other social relationships, such as friendships and other family members. There is some evidence that although marriage may be an important relationship for adults, it may not be a primary source of emotional support. If emotional support is met from other social relationships, then changes in marital quality may not have as strong of an effect on more distal outcomes (Barger, Donoho, & Wayment, 2009). In sum, these findings suggest that marital quality is an important factor to consider when examining the relationship between marriage and health. Interventions aimed at increasing marital quality should examine whether increasing marital quality could also increase individual health. 59 Table 3.1. Descriptive characteristics of the MIDUS analytic sample by gender. Male Female (n = 379) (n = 455 ) Mean (SD) Mean (SD) Age (years) 58.25 (11.7) 56.5 (9.9) Body Mass Index (kg/m 2 ) 29.7 (5.0) 28.4 (6.2) Heart Rate Variability High Frequency (.15-.50 Hz) 4.70 (1.2) 4.79 (1.2) Marital Quality a Satisfaction (0-10) 8.33 (1.72) 8.18 (1.58) Support (1-4) 3.71 (.44) 3.59 (.49) Strain (1-4) 2.14 (.58) 2.19 (.58) Disagreement (3-12) 5.75 (2.04) 5.74 (2.11) Percent Percent Marital Status Always Married 58% 45% Remarried 20% 19% Divorced/Separated 12% 19% Widowed 1% 11% Never Married 9% 6% Current Smoker 11% 11% Physical Activity ≥ 3 times/week 77% 81% Medications Activating 13% 13% Deactivating 15% 18% Education ≥ College Degree 53% 46% < College Degree 47% 54% a For individuals married 10 or more years, based on 259 men and 244 women. Table 3.2. Bivariate correlations of all major study variables 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 1. HF-HRV - 2. Satisfaction .11 - 3. Support .08 .75 - 4. Strain -.09 -.63 -.60 - 5. Disagreement -.09 -.49 -.44 .60 - 6. Female .09 -.02 -.10 .03 .00 - 7. Age -.29 .31 .18 -.17 -.23 -.17 - 8. BMI -.11 -.06 -.07 .03 .02 -.10 .06 - 9. Smoking .06 .01 .03 .03 .08 -.04 -.12 -.04 - 10. Exercise .05 -.04 .02 -.01 .00 .07 -.09 -.13 -.01 - 11. Depressive Symptoms .00 -.23 -.19 .23 .20 .04 -.17 .09 .11 -.02 - 12. Drugs (up) .01 .11 .01 -.08 -.05 -.02 .21 .08 -.02 -.12 .04 - 13. Drugs (down) -.13 .04 .01 .09 .09 .06 .00 .06 .05 -.10 .21 .09 Note. Shaded region indicates age-adjusted partial correlations. HF-HRV = High frequency heart rate variability (.15 - .50Hz) 60 61 Table 3.3. Regression coefficients (95% confidence intervals) high frequency heart rate variability on marital status (N = 844) Model 1 Model 2 Model 1 Model 2 Marital Status Married ref. ref. ref. ref. Remarried - - -.21* -.23** (-.44 , .01) (-.46 , .00) Divorced/ Separated .05 .04 -.01 -.03 (-.19 , .28) (-.17, .29) (-.26 , .23) (-.24 , .23) Widowed .02 .02 -.05 -.05 (-.36 , .38) (-.33, .42) (-.43 , .33) (-.41 , .36) Never Married -.03 -.00 -.08 -.06 (-.40 , .34) (-.38 , .36) (-.46 , .29) (-.44 , .30) Age -.03*** -.04*** -.03*** -.03*** (-.04 , -.02) (-.04 , -.03) (-.04 , -.02) (-.04 , -.03) Gender .07 .04 .07 .04 (-.11 , .24) (-.14 , .21) (-.10 , .24) (-.13 , .21) College .06 .01 .03 .01 (-.11 , .23) (-.14 , .22) (-.14 , .20) (-.16 , .19) Smoker .21 .24* (-.08 , .16) (-.07 , .17) Exercise .11 .11 (-.10 , .32) (-.10 , .31) BMI -.02** -.02** (kg/m 2 ) (-.04 , -.01) (-.04 , -.01) Medication .29* .31* (increase) (.01 , .59) (.00 , .61) Medication -.39** -.41** (decrease) (-.65 , -.17) (-.67 , -.15) Depressive Symptoms -.01 -.01 (-.02 , .00) (-.02 , .00) R 2 .08 .10 .09 .10 + p < .10 * p < .05 ** p < .01 *** p < .001 Table 3.4. Regression coefficients (95% confidence intervals) for high frequency heart rate variability on marital quality measures among married individuals (N =503). Satisfaction Support Strain Disagreement Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Marital Quality .07* .07* .21+ .19 -.18+ -.13 -.05* -.04+ (.01 , .14) (.00 , .14) (-.03 , .45) (-.06 , .43) (-.38 , .01) (-.33 , .07) (-.10 , -.00) (-.09 , .01) Age -.03*** -.04*** -.03*** -.03*** -.03*** -.03*** -.03*** -.03*** (-.04 , -.03) (-.05 , -.03) (-.04 , -.02) (-.04 , -.02) (-.04 , -.02) (-.04 , -.02) (-.04 , -.02) (-.05 , -.02) Gender .04 .05 .07 .07 .05 .05 .04 .04 (-.17 , .25) (-.17 , .26) (-.14 , .28) (-.15 , .28) (-.16 , .26) (-.16 , .27) (-.17 , .25) (-.17 , .26) Education .01 .01 .01 .01 (-.03 , .06) (-.03 , .06) (-.03 , .06) (-.03 , .06) Smoking .17 .17 .18 .20 (-.24 , .57) (-.24 , .58) (-.22 , .59) (-.20 , .60) Exercise .02 .01 .02 .02 (-.23 , .27) (-.24 , .27) (-.23 , .27) (-.23 , .27) BMI -.02+ -.02+ -.02+ -.02+ (kg/m 2 ) (-.04 , .00) (-.04 , .00) (-.04 , .00) (-.04 , .00) Medication .31* .32* .30+ .31* (increase) (.00 , .61) (.02 , .62) (-.01 , .61) (.01 , .62) Medication -.41** -.40** -.37** -.37** (decrease) (-.67 , -.15) (-.66 , -.13) (-.63 , -.11) (-.63 , -.11) Depressive Symptoms .00 .00 .00 .00 (-.02 , .01) (-.02 , .01) (-.02 , .01) (-.02 , .01) R 2 .11 .12 .11 .12 .10 .12 .10 .12 + p < .10 * p < .05 ** p < .01 *** p < .001 62 63 CHAPTER 4: MARITAL QUALITY, GENDER, AND DIURNAL CORTISOL RHYTHMS IN THE MIDUS COHORT Social relationships play a significant role in health outcomes ranging from catching a cold (Cohen, Doyle, Skoner, Rabin, & Gwaltney, 1997) to mortality (Johnson, Backlund, Sorlie, & Loveless, 2000). Marriage is one of the most important adult relationships, and the better health of the married is well established (Waite, 1995). One theory regarding the better health of individuals with large social networks and strong social ties is that these relationships help to buffer the negative health effects of chronic and acute stressors (Cohen & Wills, 1985). There is support for this theory in the epidemiological literature, with evidence showing that high numbers of life stressors increase mortality risk, but that this excess mortality risk is eliminated among individuals reporting high levels of emotional support (Rosengren, Ornth-Gomer, Wedell, & Wilhelmson, 1993). The hypothalamic-pituitary-adrenal (HPA) axis is thought to be the central physiological pathway linking chronic stress to disease endpoints, and thus may be an important link between marital quality and health (McEwen & Seeman, 1999). The current study aims to examine the links between marital quality and diurnal cortisol rhythms – the most widely accepted physiological measure of HPA activity – in order to better understand how marital quality influences health. The circadian cortisol rhythm, an indicator of HPA activity, may be one important link between marriage and health. Cortisol has been indicated as a key pathway linking social factors to age-related diseases (McEwen & Seeman, 1999; Sapolsky, Krey, & McEwen, 1986; Epel, 2010). It is thought that psychosocial stress activates certain 64 regions of the brain (e.g., the hypothalamus), which results in increased production of corticotropin releasing hormone (CRH). Increases in CRH signal for the release of adrenocorticotropin hormone (ACTH) from the anterior pituitary, resulting in increased production of glucocorticoids (cortisol in humans, corticosterone in non-human animals) from the zona fasciculate of the adrenal cortex. Chronic exposure to glucocorticoids has been shown to cause damage and dysfunction in the central nervous system, as well as peripheral tissue damage (Sapolsky, Krey, & McEwen, 1986). Aberrations in cortisol production in humans have been linked to downstream health effects such as obesity, diabetes, and infection (Cutolo et al., 2008; Donoho, Weigensberg, Emken, Hsu, & Spruijt-Metz, 2010). Understanding social factors that influence cortisol secretory patterns in human populations has the potential to make significant contributions to research on social relationships, stress, and aging. Several epidemiological studies have indicated that psychological stress leads to increased mortality, but that this excess mortality risk is buffered by individuals with high perceptions of emotional support (Rosengren, Ornth- Gomer, Wedell, &Wilhelmson, 1993). For example, men reporting high numbers of stressful life events had a substantially greater risk of dying compared to men reporting low numbers of stressful life events (Rosengren, Ornth-Gomer, Wedell, &Wilhelmson, 1993). Among those reporting higher numbers of events and lower levels of emotional support were at the greatest risk, and those reporting higher numbers of events and higher levels of emotional support had mortality rates that were equivalent to men reporting lower numbers of stressful life events. 65 As a form of social support, a marital relationship is thought to reduce the appraisal of an event as being stressful and reduce negative emotional responses to stress (stress buffering hypothesis, Cohen & Wills, 1985). However, if a marital relationship is unsupportive or is a source of strain and conflict, then a marital relationship may be an independent source of stress, and may also exacerbate the effects of other life stressors on health (Seeman, 1996; Uchino, 2009). Conflict, disagreement, and hostility among marital partners have shown deleterious effects on the cardiovascular, neuroendocrine, and immune systems in laboratory experiments (Robles & Kiecolt-Glaser, 2003; Smith et al., 2009). Therefore, if a high quality marriage (i.e., high in support) buffers the effects of general life stressors, then we may expect healthier cortisol profiles among those higher in marital support. Conversely, if a marriage is high in strain or conflict, and is an independent source of stress, then we may expect less healthy cortisol profiles among these individuals. Marital satisfaction has been associated with diurnal cortisol rhythms in small community samples, showing that women satisfied in their marriages have higher awakening values and steeper daily declines than women lower in marital satisfaction (Saxbe, Repetti, & Nishina, 2008). This indicates that, for women, marital satisfaction is associated with a more dynamic, healthy diurnal profile. Additionally, another small study found that married couples influence each other’s cortisol levels across the day, with the negative affect of one spouse influencing the cortisol levels of the other throughout the day (Saxbe & Repetti, 2010). Interestingly, these researchers found that higher levels of marital satisfaction buffered partner increases in cortisol associated with 66 their spouses’ negative affect. Although these studies provide considerable detail regarding associations between marital satisfaction and diurnal cortisol rhythms, they are based on the examination of small, non-representative samples with fewer than 60 participants. Furthermore, these studies only examine marital satisfaction, which is only one aspect of marital quality. Thus, it is necessary to examine associations between marital quality and diurnal cortisol rhythms in a large population, in order to understand these relationships in broader samples. Additionally, examining other domains of marital quality is necessary in order to understand how the negative attributes of marriage (e.g., strain, disagreement) may further explain variation in diurnal cortisol. The study of cortisol in human populations is complex because the definition of a healthy diurnal cortisol profile is unclear and is highly dependent on the groups being compared (Adam & Kumari, 2009). A normal, healthy cortisol rhythm is thought to have a sharp morning increase that peaks within an hour of waking, followed by a decline throughout the remainder of the day, with increases around mealtimes, with a nadir in preparation for evening sleep (Kirshbaum & Hellhammer, 1989; Pruessner et al., 1997). There are numerous studies identifying what appear to be pathological aberrations from normal diurnal cortisol rhythms, and many of these studies suggest that hypoactive and hyperactive HPA function are both signs of poor health (Yehuda, 1996). Blunted responses – indicated by lower awakening responses and less decline in cortisol over the day – have been observed among individuals with major depression and post-traumatic stress disorder (Yeheuda, 1996), as well among individuals with metastatic breast cancer (Abercrombie et al., 2004). A recent meta-analysis found that higher cortisol awakening 67 responses are observed in individuals experiencing general life stress and that lower cortisol awakening responses are observed in individuals with post-traumatic stress disorder and among individuals with positive psychological characteristics such as happiness, optimism, and self-esteem (Chida & Steptoe, 2009). These findings illustrate the fact that being either over-reactive or non-reactive (blunted) may be symptomatic of underlying (Adam & Kumari, 2009). The present study examines the association between positive and negative aspects of marital quality, and diurnal cortisol rhythms in a large cohort of US adults. To understand the effects of marital quality among couples in long-term relationships, we examine only individuals married to the same person for 10 or more years. We hypothesize that higher levels of negative martial characteristics, strain and disagreement, will be associated with greater increases in cortisol within the first hour of waking, and less decline over the day, with higher cortisol levels before bed. We expect the opposite pattern for positive marital characteristics, satisfaction and support, with higher levels being associated with less of an increase in cortisol within the first hour of waking, and steeper declines over the day with lower evening levels. Method The NSDE is a telephone daily diary study was conducted as part of the MacArthur Foundation National Survey of Midlife in the United States (MIDUS), a nationally representative sample of adults in the United States. For a detailed description of the study, see Almeida, McGonagle, and King (2009). For the NSDE, a random subsample of 3,600 MIDUS respondents were asked to participate, and a total of 2,022 68 agreed to participate, 1842 of which had data from MIDUS 1. The age range of participants at the time of the study was 35 to 84. Compared to MIDUS, the NSDE has a greater percentage of non-Hispanic Whites, females, and individuals with a college education. The NSDE contains data from two waves of the study, the first wave (NSDE 1) was conducted in 1996, and the second (NSDE II) was conducted from 2004 to 2006. For the purpose of the current study, we use only data from the 2004-2006 wave, as this was the only wave in which salivary cortisol samples were collected. The NSDE II included short (10-20 minute) telephone interviews about daily experiences on 8 consecutive days. On four of the days participants provided four salivary samples (a total of 16 samples). Home collection kits were sent to participants the week prior to the study start date. The collection kit included directions instructing participants to take the samples at waking, 30 minutes after getting out of bed, before lunch, and before bed. Participants were instructed to refrain from eating a large meal 1 hour before sampling, and were told to refrain from dairy products for 20 minutes before sampling. Less than 3% of the samples were missed or unusable due to contamination. The exact time of each sample was taken from telephone interviews by staff, and from a paper-pencil log sent with the collection kit. Approximately 25% of the participants received a “smart box” containing an electronic recording device that recorded the time of the box opening and closing. Self-reported collection time was highly correlated (r = .90) with electronic recorded time (Almeida, McGonagle, & King, 2009). Salivary samples were taken four times per day (waking, 30 minutes after waking, before lunch, and in the evening before bedtime) for four days. Participants used Starstet 69 salivettes (Nümbrecht, Germany), which are cotton rolls placed in the cheek until saturated and subsequently stored in plastic tubes. Samples were returned to the MIDUS Biological Core center and were frozen at -60°C for shipping and storage. For analysis, salivettes were thawed and centrifuged at 3,000 rpm for 5 minutes. Concentrations of cortisol were measured using a luminescence immunoassay (Hamburg, Germany). Marital Quality. Marital quality was measured with a set of measures indicating marital satisfaction, support, strain, and disagreement. Global marital satisfaction was tapped with a single item: (1) How would you rate your marriage these days (a scale of 0 [worst] to 10 [best])? Spousal support (α = .88) was assessed using the mean of six items rated on a 4-point Likert scale from not at all to a lot. Questions included feelings of being cared for, understood, and appreciated, as well as being able to rely on and relax around one’s spouse. Spousal strain (α = .86) was assessed using the mean of six items tapping feelings of criticism, demands, tension, and arguments, using a 4-point Likert scale from not at all to a lot. Spousal disagreement α = .72) was assessed using the sum of responses to the question, “How much do you and your spouse disagree on the following issues?” The three issues included money matters (how to spend, save, or invest), household tasks (what needs to be done and who does it), and leisure time activities (what to do and with whom). Response options were rated on a 4-point Likert scale from not at all to a lot, and scores were combined to give a final scale range of 3-12. Marital quality measures from MIDUS 2 were used. 70 Covariates Demographic Covariates. Age was calculated by subtracting birthdate from interview date. Education was self-reported and was dummy coded to indicate whether or not the individual completed a 4- or 5-year college degree or bachelor’s degree or less (0 = no college degree; 1= bachelor’s degree or beyond). Medication Use. Medication use was assessed by self-reported use of over-the- counter and prescription medications. Medications known to alter cortisol production were recorded and include steroid inhaler, steroid medications, medications containing cortisone, hormone medications (including birth control pills), anti-depressants and anti- anxiety medications. A dichotomous indicator was created to indicate whether or not a person was taking one or more of these medications (0 = no medication use; 1= current medication use). Other Cortisol-related Covariates. We also examined individual self-reported wake-time, as individuals waking earlier in the day typically have heightened responses to waking (Kudielka & Kirschbaum, 2003), as well as whether it was a weekday or weekend day, as individuals typically experience larger cortisol awakening responses on weekdays compared to weekends (Kunz-Ebrecht, Kirschbaum, Marmot, & Steptoe, 2004). Analytic Strategy Cortisol rhythms are primarily driven by time since waking, and not standard clock time. Thus, time since waking (i.e., wake time = 0) was used as the measure of time. Because previous studies indicate that cortisol values over the day are partially dependent 71 on a person falling asleep (when cortisol should be at its lowest value), and waking (when cortisol will rise for 30-60 minutes), we eliminated individuals from the analyses for the following reasons: (1) slept fewer than 4 hours the night before or stayed awake longer than 20 hours during the day of sampling, (2) slept more than 12 hours the night before or were awake for less than 12 hours the day of sampling, (3) woke up before 4 a.m., after 11 a.m., or were shift-workers, (4) did not take their morning sample within 1-hour of waking, as this would limit the ability to detect the morning cortisol rise, or if their cortisol levels were above 60 nm/dL. We also eliminated individuals reporting race/ethnicity that was non-white because these individuals comprised less than 4% of the sample with data from MIDUS 1, and given the differences in cortisol profiles that are observed between racial/ethnic groups, controlling for race in these analyses would likely be inadequate (Hajat et al., 2010). Distributions of cortisol are typically skewed which was also observed in this sample; therefore we added 1 to each cortisol value (to eliminate problems with values close to 0), and natural-log transformations were used to successfully approximate a normal distribution. Descriptive characteristics of the final analytic sample of married individuals are presented in Table 4.1. Three-level piecewise growth curve models for repeated measures were used to describe the change in cortisol across the day, across multiple days, as well as variation across individuals. Three cortisol samples were used per individual on any given day: waking, 30-60 minutes after waking, and bedtime. We eliminated the third cortisol sample (before lunch) because we were interested in assessing the slope from peak to bedtime, and because of the strong influence of food anticipation on cortisol that may 72 bias this sample (see Appendix A for a discussion of this). Strong evidence indicates that diurnal cortisol rhythms are curvilinear with a strong rise within the first hour of waking, and a decline throughout the day with a nadir before sleep. Piecewise models were used to model the morning rise and the daily decline with an inflection point during the average time for the second cortisol sample, when cortisol levels should peak. The growth models also model curvilinear growth in the second piece (daily decline) in order to capture the deceleration of the daily decline and increase in the late evening, which usually occurs when sleep has been delayed. In this 3-Level model, we can think of the occasions as Level-1, the days as Level-2, and the between-individual as Level-3 (Figure 4.1). The Level-1 investigates the variation of repeatedly measured cortisol within an individual day with time as the predictor. The Level-2 models the variation within an individual between days. The Level-3 examines variation between individuals. Equations of Level-1, Level-2, and Level-3, and composite models are as follows: Level-1 Model: Cortisol ijk = π 0ik + π 1ik S1 ijk + π 2ik S2 ijk + π 3ik S2 2 +ε ijk Level-2 Model: π 0ik = δ 00i + δ 01i WT ik – WT i + u 0ik π 1ik = δ 10i + δ 11i WT ik – WT i + u 1ik π 2ik = δ 20i + δ 21i WT ik – WT i + u 2ik π 3ik = δ 30i + δ 31i WT ik – WT i + u 3ik 73 Level-3 Model: δ 00i = γ 000 + γ 001 Age i + γ 002 College i + γ 003 Medication i + γ 004 WT i + γ 005 Weekend i + γ 006 MaritalQuality i + ζ 00i δ 10i = γ 100 + γ 101 Age i + γ 102 College i + γ 103 Medication i + γ 104 WT i + γ 105 Weekend i + γ 106 MaritalQuality i + ζ 10i δ 20i = γ 200 + γ 201 Age i + γ 202 College i + γ 203 Medication i + γ 204 WT i + γ 205 Weekend i + γ 206 MaritalQuality i + ζ 20i δ 30i = γ 300 + γ 301 Age i + γ 302 College i + γ 303 Medication i + γ 304 WT i + γ 305 Weekend i + γ 306 MaritalQuality i + ζ 30i Composite Model: Cortisol ijk = [γ 000 + γ 001 Age i + γ 002 College i + γ 003 Medication i + γ 004 WT i + γ 005 Weekend i + γ 006 MaritalQuality i + ζ 00i + δ 01i WT ik – WT i + γ 100 S1 ijk + γ 101 Age i S1 ijk + γ 102 College i S1 ijk + γ 103 Medication i S1 ijk + γ 104 WT i S1 ijk + γ 105 Weekend i S1 ijk + γ 106 MaritalQuality i S1 ijk + δ 11i WT ik – WT i S1 ijk + γ 200 S2 ijk + γ 201 Age i S2 ijk + γ 202 College i S2 ijk + γ 203 Medication i S2 ijk + γ 204 WT i S2 ijk + γ 205 Weekend i S2 ijk + γ 206 MaritalQuality i S2 ijk + δ 21i WT ik – WT i S2 ijk + γ 300 S2 2 ijk + γ 301 Age i S2 2 ijk + γ 302 College i S2 2 ijk + γ 303 Medication i S2 2 ijk + γ 304 WT i S2 2 ijk + γ 305 Weekend i S2 2 ijk + γ 306 MaritalQuality i S2 2 ijk + δ 31i WT ik – WT i S2 2 ijk ] + [u 0ik + u 1ik S1 ijk + u 2ik S2 ijk + u 3ik S2 2 ijk + ζ 10i S1 ijk + ζ 20i S2 ijk + ζ 30i S2 2 ijk + ε ijk ] In the Level-1 model, Cortisol ijk is individual i’s value of cortisol on occasion j and day k. π 0ik is individual i’s level of cortisol upon waking. S1 ijk is the amount of time elapsed since waking for individual i on occasion j during day k for the first slope capturing the morning rise (π 1ik ). S2 ijk is the amount of time elapsed since waking for individual i on occasion j during day k for the second slope capturing the daily decline 74 (π 2ik ), with the slope of S2 2 ijk (π 3ik ) capturing the slowing rate of the daily decline and increase during the late evening. The amount of individual i’s cortisol that is unexplained on occasion j during day k is ε ijk . In Level 2 the intercept and slope parameters from Level 1 become the outcomes, where π 0ik, π 1ik, π 2ik, and π 3ik are functions of individual i’s average waking value (δ 00i ), and average slopes of the morning rise (δ 10i ), daily decline (δ 20i ), and slowing of the daily decline (δ 20i ), while accounting for differences in these slopes as a function of individual variability in wake time, which was captured by adding wake time as a covariate to each equation. The level of the intercept and slopes of cortisol that are not explained by the averages of their daily levels are represented by variance terms u 0ik, u 1ik, and u 2ik , and u 3ik . In Level 3, the intercept and slope parameters from Level 2 become the outcomes. Here, δ 00i represents individual i’s true mean cortisol value upon waking and ζ 00i represents the amount that individual i differs from the population mean (γ 000 ). δ 10i represents individual i’s true average rate of change in cortisol during the morning rise and ζ 10i represents the amount that individual i’s rate of change differs from the population average rate of change (γ 100 ); δ 20i represents individual i’s true average rate of change in cortisol during the daily decline and ζ 20i represents the amount that individual i’s rate of change differs from the population average rate of change (γ 200 ); δ 30i represents individual i’s quadratic rate of change in cortisol, or the slowing of the decline, and ζ 30i represents the amount that individual i’s slowing of the decline or late evening increase differs from the population average rate (γ 300 ). Also included in the intercept and slope equations are other predictors that are thought to account for between-person variability 75 in cortisol levels. These include time-independent covariates and independent variables that are the focus of our current analyses – marital quality – with separate models for each marital quality variable. This results in four separate analyses for marital satisfaction, support, strain, and disagreement. Covariates included age (centered), wake time (centered), education (college degree vs. no-college degree), medication use, and weekend (vs. weekday). These covariates were not of substantive interest, but were entered into models to ensure that the final estimates of time were not substantially changed, indicating confounding. Results Descriptive characteristics of the analytic sample are presented in 4.1. The average age of men in the sample was 57.7 years, while the average age for women was 54.6 years. As was expected, men reported higher levels of satisfaction and support compared to women, and slightly lower levels of strain and disagreement. One-third of men and 44.3% of women were on a medication known to influence cortisol. Overall, men had slightly higher levels of cortisol upon waking and at bedtime, but had similar levels to women 30 minutes after waking. Two-thirds of men and three-quarters of women showed evidence of the cortisol awakening response. The raw cortisol values were plotted against time in figure 4.2, with a local second-degree polynomial smoothing function fitted to the data. The distribution of cortisol across the day showed an increase between waking and one hour, and lower values observed in near bedtime, but appeared to become elevated among individuals remained awake for more than 17 or 18 hours. 76 Results from growth models are displayed with four separate panels, indicating the fixed effects estimates for the intercept (waking value), Slope 1 (morning rise), Slope 2 (daily decline), and the quadratic portion of Slope 2. Table 4.2 shows the results of the growth model for marital satisfaction among men, controlling for covariates, age, education, medication, wake time, and weekday/weekend. For men, marital satisfaction was not associated with any part of the diurnal rhythm. Spousal support was marginally associated with higher waking cortisol values among men (b = .10, p = .07), but was not associated with significant differences in the slope of the morning rise, the daily decline, or the quadratic part of the decline. These results are shown in Table 4.3, and differences in the growth curves for individuals high in support (1 standard deviation above the mean) versus low in support (1 standard deviation below the mean) are graphically depicted in Figure 4.3. Among men, negative marital characteristics had stronger effects on diurnal cortisol rhythms. Table 4.4 shows that spousal strain was associated with lower awakening cortisol levels (b = -.08, p = .04), steeper daily declines (b = -.04, p = .07), and a greater deceleration of the daily decline (b = .003, p = .06) leading to higher evening cortisol values among those that were awake for more than 17 hours. Differences in the growth curves for individuals high in spousal strain (1 standard deviation above the mean) versus low in strain (1 standard deviation below the mean) are graphically depicted in Figure 4.4. Similarly, Table 4.5. shows that marital disagreement was associated with lower awakening levels of cortisol among men (b = -.04, p = .001) and a steeper increased slope for the morning rise (b = .03, p = .07); however, marital disagreement was not associated with a steeper daily decline or curve in the daily decline. 77 Differences in the growth curves for individuals high in disagreement (1 standard deviation above the mean) versus low in disagreement (1 standard deviation below the mean) are graphically depicted in Figure 4.5. For women, none of the marital quality indicators were associated with any part of the diurnal rhythm. The results of the piecewise growth curves for marital satisfaction, spousal support, spousal strain, and marital disagreement are shown in tables 4.6, 4.7, 4.8, and 4.9, respectively. Discussion This study aimed to examine how diurnal cortisol rhythms vary as a function of marital quality in a large sample of U.S. men and women participating in a multi-day daily diary study. The current study is the first study to examine the association between cortisol and multiple measures of both positive and negative marital characteristics. Findings suggest that men who perceive higher levels of marital strain and have greater frequency of marital disagreement have lower waking values of cortisol and greater cortisol awakening responses, with some indication of greater decreases in cortisol in the early evening, but higher values in the later evening. For women, on the other hand, marital quality was not associated with differences in cortisol secretory patterns over the day. Our findings are somewhat inconsistent with previous research where studies have found that women satisfied in their marriages have higher awakening values and steeper daily declines than women lower in marital satisfaction (Saxbe, Repetti, & Nishina, 2008). However, consistent with stress-related research, we expected greater cortisol 78 awakening responses among those in lower quality marriages, which our findings supported, at least among men. We also expected individuals in lower quality marriages to have less of a decline across the day, with higher levels in the evening, which was not observed in these data except in the later hours of the evening where predicted cortisol levels were higher among men in relationships high in strain and/or low in support. Given that there is relatively little research examining associations between social relationships and diurnal cortisol rhythms and social relationships - especially marriage - the inconsistencies in the literature are difficult to fully interpret. This study contributes to the growing literature showing that social relationships are important for neuroendocrine functioning. Most of the previous research has been conducted in laboratory settings, examining neuroendocrine responses to acute stress from a conflict(Robles, 2006), or in smaller, non-representative samples(Saxbe, Repetti, & Nishina, 2008). This study has numerous strengths. This was the first study to examine marital quality in population-based cohort of adults in the US, and was also the first study to examine associations between positive and negative marital characteristics and diurnal cortisol rhythms. This study used a validated high-intensity sampling protocol, utilizing three cortisol samples per day over four days. Therefore these data provide a strong test of the association between marital quality and diurnal cortisol rhythms. Despite this studies strengths, however, there are notable limitations. First, this is a cross-sectional study and therefore it is unknown whether there is a causal relationship between marital quality and diurnal cortisol rhythms. Second, this study only trait-like attributes of a marriage (e.g., disagreement in general) and does not examine state-like 79 attributes of marital quality (e.g., day-to-day reports of disagreements), which would be helpful to understand how marital quality influences cortisol profiles over time, as well as how marital quality influences day-to-day fluctuations in cortisol. Figure 4.1. Illustration of 3-Level design. 81 Figure 4.2. Scatterplot of raw data and a local polynomial non-weighted smoothing function with a second-degree polynomial function for diurnal cortisol 82 Figure 4.3. Predicted cortisol trajectories over the day for men high (+1 SD) and low (- 1SD) in marital support .00 .50 1.00 1.50 2.00 2.50 3.00 3.50 0 .5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 ln(Cortisol+1) Hours Since Waking High Support Low Support 83 Figure 4.4. Predicted cortisol trajectories over the day for men high (+1 SD) and low (- 1SD) in marital strain .00 .50 1.00 1.50 2.00 2.50 3.00 3.50 0 .5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 ln(Cortisol+1) Hours Since Waking High Strain Low Strain 84 Figure 4.5. Predicted cortisol trajectories over the day for men high (+1 SD) and low (-1SD) in marital disagreement .00 .50 1.00 1.50 2.00 2.50 3.00 3.50 0 .5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 ln(Cortisol+1) Hours Since Waking High Disagreement Low Disagreement 85 Table 4.1. Characteristics of the analytic sample for married NSDE participants. Male Female (n = 408) (n = 433) Mean (SD) Range Mean (SD) Range Age (years) 57.7 (11.5) 34 - 83 54.6 (11.6) 34 - 83 Marital Quality Satisfaction (0-10) 8.47 (1.62) 2 - 10 8.17 (1.84) 0 - 10 Support (1-4) 3.72 (.43) 1.7 - 4 3.58(.55) 1.3 - 4 Strain (1-4) 2.09 (.55) 1 - 4 2.19 (.63) 1 - 4 Disagreement (3-12) 5.69 (1.95) 3 - 12 5.77 (2.11) 3 - 12 Wake Time (hr:min) 6:25 (1:19) 6:41 (1:14) 4:00 – 10:30 Cortisol (µg/dL) a Awakening 17.64 (9.9) .04 – 58.85 15.42 (8.9) .39 – 51.98 30 minutes 22.31 (11.4) .60 – 59.06 22.52 (11.7) .91 – 59.58 Bedtime 3.39 (4.4) .07 – 34.01 2.66 (3.64) .01 – 36.02 Percent Percent Awakening Response > 0 µg/dL 66.3% 74.4% ≤ 0 µg/dL 33.7% 23.3% College Graduate 48.2% 39.2% Currently on Medication b Yes 33.8% 44.3% No 66.2% 55.7% a These are values of cortisol are raw values, but are transformed in subsequent analyses b Medications included steroid inhalers, steroid containing medications, cortisone containing medications, birth control pills, other hormonal medications, and anti- depressant and anti-anxiety medications. Table 4.2. Piecewise growth curve models for repeated measures predicting daily cortisol levels for men with marital satisfaction Waking Value Slope 1 (Morning Rise) Coefficient SE p Coefficient SE p Sample Average 2.72 .04 .00 .44 .05 .00 Marital Satisfaction .02 .01 .18 -‐.01 .02 .60 Age .00 .00 .94 .01 .00 .00 College Graduate .11 .04 .01 -‐.09 .06 .10 Medication User -‐.07 .05 .14 .06 .06 .34 Wake-‐time .00 .01 .70 -‐.10 .02 .00 Weekend -‐.06 .03 .04 -‐.03 .07 .65 Slope 2 (Daily Decline) Slope 2 Quadratic (Deceleration) Coefficient SE p Coefficient SE p Sample Average -‐.31 .02 .00 .013 .001 .00 Marital Satisfaction .01 .01 .37 .000 .000 .42 Age .00 .00 .21 .000 .000 .25 College Graduate -‐.05 .03 .08 .002 .002 .18 Medication User .04 .03 .09 -‐.003 .002 .11 Wake-‐time .02 .01 .07 -‐.001 .001 .35 Weekend .02 .02 .30 -‐.002 .002 .32 Note. Age, wake time, and all marital quality variables were centered on their mean. + p ≤ .10 * p ≤ .05, **p ≤ .01, *** p ≤ .001 86 Table 4.3. Piecewise growth curve models for repeated measure predicting daily cortisol levels for men with spousal support Waking Value Slope 1 (Morning Rise) Coefficient SE p Coefficient SE p Sample Average 2.72 .04 .00 .42 .05 .00 Spouse Support .10 .05 .07 .02 .07 .78 Age .00 .00 .68 .01 .00 .00 College Graduate .11 .04 .01 -‐.08 .06 .15 Medication User -‐.07 .05 .11 .08 .06 .17 Wake-‐time .01 .01 .67 -‐.10 .02 .00 Weekend -‐.06 .03 .03 -‐.02 .07 .80 Slope 2 (Daily Decline) Slope 2 Quadratic (Deceleration) Coefficient SE p Coefficient SE p Sample Average -‐.32 .02 .00 .013 .002 .00 Spouse Support .02 .03 .56 -‐.001 .002 .44 Age .00 .00 .26 .000 .000 .33 College Graduate -‐.05 .03 .06 .003 .002 .14 Medication User .05 .03 .06 -‐.003 .002 .07 Wake-‐time .02 .01 .06 -‐.001 .001 .32 Weekend .03 .02 .27 -‐.002 .002 .28 Note. Age, wake time, and all marital quality variables were centered on their mean. + p ≤ .10 * p ≤ .05, **p ≤ .01, *** p ≤ .001 87 Table 4.4. Piecewise growth curve models for repeated measure predicting daily cortisol levels for men with spousal strain Waking Value Slope 1 (Morning Rise) Coefficient SE p Coefficient SE p Sample Average 2.73 .04 .00 .42 .05 .00 Spouse Strain -‐.08 .04 .04 .04 .05 .50 Age .00 .00 .90 .01 .00 .00 College Graduate .11 .04 .01 -‐.08 .06 .17 Medication User -‐.07 .05 .12 .08 .06 .18 Wake-‐time .01 .01 .63 -‐.11 .02 .00 Weekend -‐.06 .03 .03 -‐.02 .07 .81 Slope 2 (Daily Decline) Slope 2 Quadratic (Deceleration) Coefficient SE p Coefficient SE p Sample Average -‐.32 .02 .00 .013 .001 .00 Spouse Strain -‐.04 .02 .07 .003 .001 .06 Age .00 .00 .42 .000 .000 .49 College Graduate -‐.05 .03 .05 .003 .002 .12 Medication User .05 .03 .05 -‐.003 .002 .05 Wake-‐time .02 .01 .06 -‐.001 .001 .29 Weekend .03 .02 .24 -‐.002 .002 .25 Note. Age, wake time, and all marital quality variables were centered on their mean. + p ≤ .10 * p ≤ .05, **p ≤ .01, *** p ≤ .001 88 Table 4.5. Piecewise growth curve models for repeated measure predicting daily cortisol levels for men with spousal disagreement. Waking Value Slope 1 (Morning Rise) Coefficient SE p Coefficient SE p Sample Average 2.73 .04 .00 .42 .05 .00 Disagreement -‐.04 .01 .00 .03 .01 .07 Age .00 .00 .89 .01 .00 .00 College Graduate .11 .04 .01 -‐.08 .06 .18 Medication User -‐.07 .05 .15 .07 .06 .21 Wake-‐time .01 .01 .66 -‐.11 .02 .00 Weekend -‐.06 .03 .03 -‐.02 .07 .80 Slope 2 (Daily Decline) Slope 2 Quadratic (Deceleration) Coefficient SE p Coefficient SE p Sample Average -‐.32 .02 .00 .013 .002 .00 Disagreement -‐.01 .01 .25 .001 .000 .21 Age .00 .00 .33 .000 .000 .40 College Graduate -‐.05 .03 .05 .003 .002 .13 Medication User .05 .03 .06 -‐.003 .002 .06 Wake-‐time .02 .01 .07 -‐.001 .001 .35 Weekend .03 .02 .28 -‐.002 .002 .30 Note. Age, wake time, and all marital quality variables were centered on their mean. + p ≤ .10 * p ≤ .05, **p ≤ .01, *** p ≤ .001 89 Table 4.6. Piecewise growth curve models for repeated measures predicting daily cortisol levels for women with marital satisfaction Waking Value Slope 1 (Morning Rise) Coefficient SE p Coefficient SE p Sample Average 2.70 .03 .00 .57 .05 .00 Marital Satisfaction -‐.01 .01 .26 -‐.02 .02 .12 Age .00 .00 .68 .00 .00 .06 College Graduate .05 .04 .26 .06 .06 .30 Medication User -‐.12 .04 .00 .06 .06 .27 Wake-‐time -‐.02 .01 .15 -‐.03 .02 .28 Weekend -‐.07 .03 .01 .02 .07 .77 Slope 2 (Daily Decline) Slope 2 Quadratic (Deceleration) Coefficient SE p Coefficient SE p Sample Average -‐.33 .02 .00 .013 .002 .00 Marital Satisfaction .00 .01 .54 .000 .000 .68 Age .00 .00 .01 .000 .000 .01 College Graduate -‐.01 .02 .78 .000 .002 .94 Medication User .05 .02 .03 -‐.003 .002 .07 Wake-‐time -‐.01 .01 .57 .001 .001 .34 Weekend .04 .02 .12 -‐.002 .002 .18 Note. Age, wake time, and all marital quality variables were centered on their mean. + p ≤ .10 * p ≤ .05, **p ≤ .01, *** p ≤ .001 90 Table 4.7. Piecewise growth curve models for repeated measure predicting daily cortisol levels for women with spousal support Waking Value Slope 1 (Morning Rise) Coefficient SE p Coefficient SE p Sample Average 2.70 .03 .00 .57 .05 .00 Spouse Support -‐.06 .04 .11 -‐.02 .05 .65 Age .00 .00 .62 .00 .00 .08 College Graduate .04 .04 .30 .06 .06 .27 Medication User -‐.12 .04 .00 .05 .06 .33 Wake-‐time -‐.01 .01 .25 -‐.03 .02 .24 Weekend -‐.07 .03 .01 .02 .07 .82 Slope 2 (Daily Decline) Slope 2 Quadratic (Deceleration) Coefficient SE p Coefficient SE p Sample Average -‐.33 .02 .00 .013 .002 .00 Spouse Support -‐.01 .02 .82 .000 .000 .80 Age .00 .00 .00 .000 .000 .00 College Graduate .00 .02 .89 .000 .002 .96 Medication User .04 .02 .06 -‐.002 .002 .11 Wake-‐time .00 .01 .79 .000 .001 .51 Weekend .03 .02 .19 -‐.002 .002 .28 Note. Age, wake time, and all marital quality variables were centered on their mean. + p ≤ .10 * p ≤ .05, **p ≤ .01, *** p ≤ .001 91 Table 4.8. Piecewise growth curve models for repeated measure predicting daily cortisol levels for women with spousal strain Waking Value Slope 1 (Morning Rise) Coefficient SE p Coefficient SE p Sample Average 2.70 .03 .00 .57 .05 .00 Spouse Strain .04 .03 .19 .05 .04 .27 Age .00 .00 .60 .00 .00 .07 College Graduate .04 .04 .30 .06 .06 .28 Medication User -‐.12 .04 .00 .05 .06 .34 Wake-‐time -‐.01 .01 .24 -‐.03 .02 .23 Weekend -‐.07 .03 .01 .02 .07 .80 Slope 2 (Daily Decline) Slope 2 Quadratic (Deceleration) Coefficient SE p Coefficient SE p Sample Average -‐.33 .02 .00 .013 .002 .00 Spouse Strain .01 .02 .64 -‐.001 .001 .63 Age .00 .00 .00 .000 .000 .01 College Graduate .00 .02 .87 .000 .002 .98 Medication User .04 .02 .07 -‐.002 .002 .13 Wake-‐time .00 .01 .79 .000 .001 .51 Weekend .03 .02 .19 -‐.002 .002 .28 Note. Age, wake time, and all marital quality variables were centered on their mean. + p ≤ .10 * p ≤ .05, **p ≤ .01, *** p ≤ .001 92 90 Table 4.9. Piecewise growth curve models for repeated measure predicting daily cortisol levels for women with spousal disagreement. Waking Value Slope 1 (Morning Rise) Coefficient SE p Coefficient SE p Sample Average 2.70 .03 .00 .56 .05 .00 Marital Disagreement .01 .01 .23 .00 .01 .93 Age .00 .00 .54 .00 .00 .08 College Graduate .04 .04 .31 .07 .06 .25 Medication User -‐.12 .04 .00 .06 .06 .31 Wake-‐time -‐.01 .01 .26 -‐.03 .02 .23 Weekend -‐.07 .03 .01 .02 .07 .78 Slope 2 (Daily Decline) Slope 2 Quadratic (Deceleration) Coefficient SE p Coefficient SE p Sample Average -‐.33 .02 .00 .013 .002 .00 Marital Disagreement .01 .01 .25 .000 .000 .23 Age .00 .00 .00 .000 .000 .00 College Graduate .00 .02 .84 .000 .002 1.00 Medication User .04 .02 .08 -‐.002 .002 .14 Wake-‐time .00 .01 .80 .000 .001 .52 Weekend .03 .02 .19 -‐.002 .002 .28 Note. Age, wake time, and all marital quality variables were centered on their mean. + p ≤ .10 * p ≤ .05, **p ≤ .01, *** p ≤ .001 93 91 94 CHAPTER 5: CONCLUSION Summary The three empirical chapters in this dissertation make two major contributions to the scientific literature: First, this research examines multiple dimensions of marital quality to test whether positive and negative marital characteristics differentially affect health. Second, this research focuses on how marital quality influences physiological measures theorized to be pathways leading to poor health outcomes such as cardiovascular disease and mortality. Findings and Implications The first empirical paper examined marital quality and inflammation and found that higher spousal strain was weakly associated with higher inflammation among men and women, and that higher spousal support was associated with lower levels of inflammation among women, but not men. The second empirical paper examined associations between marital status, marital quality, and heart rate variability, a measure of neural regulation of the heart. This study found that traditional categories of marital status were not associated with HRV, but that compared to remarried individuals, continuously married individuals had higher HRV. This study also found that higher levels of spousal strain and disagreement were associated with lower heart rate variability, and that higher levels of marital satisfaction and spousal support were associated with higher HRV. The third empirical paper examined associations between marital quality and diurnal cortisol rhythms. Findings from this study indicate that men with higher levels of marital strain and disagreement have lower waking values of cortisol and greater 95 cortisol awakening responses, with some indication of greater decreases in cortisol in the early evening, but higher values in the later evening. However, for women marital quality was not associated with differences in cortisol secretory patterns over the day. These findings were somewhat surprising because although the heightened awakening response was expected among those in bad marriages, the more pronounced daily decline was not expected, as this is usually an indicator of lower cortisol exposure over the day, and better health. However, there continue to be methodological questions concerning cortisol in epidemiological studies (see Adam & Kumari, 2009). Taken together, these empirical papers show the importance of marital quality for understanding the health of the married. In general, the studies from this dissertation highlight the fact that there is heterogeneity in the health of the married. Individuals in good marriages (i.e., low in negative characteristics and/or high in positive characteristics) tend to appear healthier than those in bad marriages (i.e., high in negative characteristics and/or low in positive characteristics). Additionally, this dissertation shows that marital history also appears to contribute to heterogeneity in the health of the married. Although only explored in the context of heart rate variability, it appears that continuously married individuals have better health than remarried individuals. This could be due to carryover effects from the stress of a marital dissolution (i.e., divorce, widowhood), or from the time spent in a previously bad marriage. In sum, these finding suggest that the health benefits experienced by the married may be limited to only those individuals that are happily and continuously married. 96 Gender differences in the patterns of these associations suggest that while good marriages are better for health for both men and women, marital quality may differentially influence biological systems for men and women. For women, inflammatory responses and neural regulation of the heart are influenced by marital quality, while for men neural regulation of the heart and HPA functioning may be influenced by marital quality. Limitations Although this dissertation had many strengths, it also had several notable limitations. First, the study was entirely Caucasian adults with relatively high levels of educational attainment. There is evidence indicating that marriage may confer different health benefits and hold different meaning for individuals of various ethnic, racial, and socioeconomic backgrounds (McLoyd, Cauce, Takeuchi, & Wilson, 2000). The narrow selection of the study sample eliminated the possibility of confounding by race, but it also placed significant limitations on the ability to generalize findings to broader populations in the US. A second limitation is that although the MIDUS studies are longitudinal and include a 10-year follow-up period, there are limited statistical methods to examine individuals with measurements at only two time points. Therefore, the results of this study are similar to those of cross-sectional studies in that causal direction cannot be inferred. This is problematic, as declines in health could lead to declines in marital quality. However, longitudinal evidence indicates that bad marriages contribute to 97 declines in health, and not vice-versa (Umberson, Williams, Powers, Liu, & Needham, 2006). A third limitation of this research is that it only examined one spouse in a spousal dyad and in an entire family unit. This is an important limitation in research on marriage because evaluations of marital quality by one spouse are likely influenced by evaluations of marital quality by the other spouse. In addition, marital relationships frequently include young children that could result in shared obligations and influence the quality of a marriage. Therefore, future research should attempt to examine spousal dyads and entire families in order to understand how marital quality influences health. Future Directions There are several important avenues of further research that remain unexplored. First, although social relationships theories indicate that social relationships buffer the effects of other life stressors (Cohen, Janicki-Deverts, & Miller, 2007; Cohen & Wills, 1985), the current research does not examine marital quality as a mediator of other life stressors. The reason stress was not included in this dissertation is because mediation analyses are limited to studies with longitudinal data with three or more observations (Cohen, Cohen, West, & Aiken, 2003). Because MIDUS only had biomarkers at one time point, including stress would not have yielded interpretable results. However, future work should examine whether marital quality contributes to health directly, or if it buffers the effects of other life stressors. This research examined how evaluations of marital quality influence biomarkers of health. This approach assumes that evaluations of marital quality are stable across time, 98 and that the biomarkers studied are relatively stable from day-to-day. Although this approach is helpful from an epidemiological perspective, it would also be useful to examine fluctuations in marital quality (i.e., giving or receiving support, disagreement) and fluctuations in these biomarkers in naturalistic settings – especially heart rate variability and cortisol, which we might expect to fluctuate with marital interactions. This would allow for researchers to understand if momentary fluctuations in marital quality influence physiology, and if these fluctuations result in chronic disruption to these physiological systems (McEwen & Seeman, 1999). 99 REFERENCES Aiken, L. S., & West, D. S. G. (1991). Multiple regression: Testing and interpreting interactions Sage Publications, Inc. Alley, D. E., Crimmins, E., Bandeen-Roche, K., Guralnik, J., & Ferrucci, L. (2007). Three-year change in inflammatory markers in elderly people and mortality: The invecchiare in chianti study. Journal of the American Geriatrics Society, 55(11), 1801-1807. Alley, D. E., Seeman, T. E., Ki Kim, J., Karlamangla, A., Hu, P., & Crimmins, E. M. (2006). Socioeconomic status and C-reactive protein levels in the US population: NHANES IV. Brain, Behavior, and Immunity, 20(5), 498-504. Almeida, D. M., & Kessler, R. C. (1998). Everyday stressors and gender differences in daily distress. Journal of Personality and Social Psychology, 75(3), 670-680. Barger, S. D., Donoho, C. J., & Wayment, H. A. (2009). The relative contributions of race/ethnicity, socioeconomic status, health, and social relationships to life satisfaction in the united states. Quality of Life Research : An International Journal of Quality of Life Aspects of Treatment, Care and Rehabilitation, 18(2), 179-189. doi:10.1007/s11136-008-9426-2 Baumeister, R. F., & Sommer, K. L. (1997). What do men want? gender differences and two spheres of belongingness: Comment on cross and madson (1997). Psychological Bulletin, 122(1), 38-44; discussion 51-5. 100 Berkman, L. F., Glass, T., Brissette, I., & Seeman, T. E. (2000). From social integration to health: Durkheim in the new millennium. Social Science & Medicine, 51(6), 843- 857. Bigger, J. T.,Jr, Fleiss, J. L., Rolnitzky, L. M., & Steinman, R. C. (1993). Frequency domain measures of heart period variability to assess risk late after myocardial infarction. Journal of the American College of Cardiology, 21(3), 729-736. Brosschot, J. F., Van Dijk, E., & Thayer, J. F. (2007). Daily worry is related to low heart rate variability during waking and the subsequent nocturnal sleep period. International Journal of Psychophysiology : Official Journal of the International Organization of Psychophysiology, 63(1), 39-47. Brunner, E. J., Chandola, T., & Marmot, M. G. (2007). Prospective effect of job strain on general and central obesity in the whitehall II study. American Journal of Epidemiology, 165(7), 828-837. Buysse, D. J., Reynolds III, C. F., Monk, T. H., Berman, S. R., & Kupfer, D. J. (1989). The pittsburgh sleep quality index: A new instrument for psychiatric practice and research. Psychiatry Research, 28(2), 193-213. Cacioppo, J. T., Berntson, G. G., Binkley, P. F., Quigley, K. S., Uchino, B. N., & Fieldstone, A. (1994). Autonomic cardiac control. II. noninvasive indices and basal response as revealed by autonomic blockades. Psychophysiology, 31(6), 586-598. Carney, R. M., Blumenthal, J. A., Stein, P. K., Watkins, L., Catellier, D., Berkman, L. F., et al. (2001). Depression, heart rate variability, and acute myocardial infarction. Circulation, 104(17), 2024 - 2028. 101 Carr, D., & Springer, K. W. (2010). Advances in families and health research in the 21st century. Journal of Marriage and Family, 72(3), 743-761. Choi, H., & Marks, N. F. (2008). Marital conflict, depressive symptoms, and functional impairment. Journal of Marriage and Family, 70(2), 377-390. Coan, J. A., Schaefer, H. S., & Davidson, R. J. (2006). Lending a hand: Social regulation of the neural response to threat. Psychological Science : A Journal of the American Psychological Society / APS, 17(12), 1032-1039. doi:10.1111/j.1467- 9280.2006.01832.x Cohen, J., Cohen, P., West, S. H., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences. Mahwah, N.J.: L. Erlbaum Associates. Cohen, S., Doyle, W. J., Skoner, D. P., Rabin, B. S., & Gwaltney, J. M.,Jr. (1997). Social ties and susceptibility to the common cold. JAMA : The Journal of the American Medical Association, 277(24), 1940-1944. Cohen, S., Janicki-Deverts, D., & Miller, G. E. (2007). Psychological stress and disease. JAMA : The Journal of the American Medical Association, 298(14), 1685-1687. Cohen, S., & Wills, T. A. (1985). Stress, social support, and the buffering hypothesis. Psychological Bulletin, 98(2), 310-357. Colbert, L. H., Visser, M., Simonsick, E. M., Tracy, R. P., Newman, A. B., Kritchevsky, S. B., et al. (2004). Physical activity, exercise, and inflammatory markers in older adults: Findings from the health, aging and body composition study. Journal of the American Geriatrics Society, 52(7), 1098-1104. 102 Coyne, J. C., Rohrbaugh, M. J., Shoham, V., Sonnega, J. S., Nicklas, J. M., & Cranford, J. A. (2001). Prognostic importance of marital quality for survival of congestive heart failure. The American Journal of Cardiology, 88(5), 526-529. Crimmins, E., Vasunilashorn, S., Kim, J. K., & Alley, D. (2008). Biomarkers related to aging in human populations. Advances in Clinical Chemistry, 46, 161-216. Danesh, J., Kaptoge, S., Mann, A. G., Sarwar, N., Wood, A., Angleman, S. B., et al. (2008). Long-term interleukin-6 levels and subsequent risk of coronary heart disease: Two new prospective studies and a systematic review. PLoS Medicine, 5(4), 78. de Bruyne, M. C., Kors, J. A., Hoes, A. W., Klootwijk, P., Dekker, J. M., Hofman, A., et al. (1999). Both decreased and increased heart rate variability on the standard 10- second electrocardiogram predict cardiac mortality in the elderly: The rotterdam study. American Journal of Epidemiology, 150(12), 1282-1288. De Vogli, R., Chandola, T., & Marmot, M. G. (2007). Negative aspects of close relationships and heart disease. Archives of Internal Medicine, 167(18), 1951-1957. Dekker, J. M., Schouten, E. G., Klootwijk, P., Pool, J., Swenne, C. A., & Kromhout, D. (1997). Heart rate variability from short electrocardiographic recordings predicts mortality from all causes in middle-aged and elderly men. the zutphen study. American Journal of Epidemiology, 145(10), 899-908. 103 DeSilva, B., Smith, W., Weiner, R., Kelley, M., Smolec, J., Lee, B., et al. (2003). Recommendations for the bioanalytical method validation of ligand-binding assays to support pharmacokinetic assessments of macromolecules. Pharmaceutical Research, 20(11), 1885-1900. Dienberg Love, G., Seeman, T. E., Weinstein, M., & Ryff, C. D. (2010). Bioindicators in the MIDUS national study: Protocol, measures, sample, and comparative context. Journal of Aging and Health, 22(8), 1059-1080. Donoho, C. J., Weigensberg, M. J., Emken, B. A., Hsu, J. W., & Spruijt-Metz, D. (2010). Stress and abdominal fat: Preliminary evidence of moderation by the cortisol awakening response in hispanic peripubertal girls. Obesity, 19(5), 946-952. Dupre, M. E., & Meadows, S. O. (2007). Disaggregating the effects of marital trajectories on health. Journal of Family Issues, 28(5), 623-652. Dupre, M. E., Beck, A. N., & Meadows, S. O. (2009). Marital trajectories and mortality among US adults. American Journal of Epidemiology, 170(5), 546-555. doi:10.1093/aje/kwp194 Epel, E. S. (2009). Psychological and metabolic stress: A recipe for accelerated cellular aging. Hormones (Athens), 8(1), 7-22. Fabes, R. A., & Eisenberg, N. (1997). Regulatory control and adults' stress-related responses to daily life events. Journal of Personality and Social Psychology, 73(5), 1107. Ferrucci, L., Corsi, A., Lauretani, F., Bandinelli, S., Bartali, B., Taub, D. D., et al. (2005). The origins of age-related proinflammatory state. Blood, 105(6), 2294-2299. 104 Festa, A., D'Agostino, R.,Jr, Tracy, R. P., Haffner, S. M., & Insulin Resistance Atherosclerosis Study. (2002). Elevated levels of acute-phase proteins and plasminogen activator inhibitor-1 predict the development of type 2 diabetes: The insulin resistance atherosclerosis study. Diabetes, 51(4), 1131-1137. Finch, C. E. (2011). Inflammation in aging processes: An integrative and ecological approach. In E. J. Masoro, & S. N. Austad (Eds.), Handbook of the biology of aging (7th ed., pp. 275-296). London: Academic Press. Fincham, F. D., & Linfield, K. J. (1997). A new look at marital quality: Can spouses feel positive and negative about their marriage? Journal of Family Psychology, 11(4), 489-502. Fontana, L., Eagon, J. C., Trujillo, M. E., Scherer, P. E., & Klein, S. (2007). Visceral fat adipokine secretion is associated with systemic inflammation in obese humans. Diabetes, 56(4), 1010-1013. Ford, E. S., Loucks, E. B., & Berkman, L. F. (2006). Social integration and concentrations of C-reactive protein among US adults. Annals of Epidemiology, 16(2), 78-84. Friedman, E. M., Hayney, M. S., Love, G. D., Urry, H. L., Rosenkranz, M. A., Davidson, R. J., et al. (2005). Social relationships, sleep quality, and interleukin-6 in aging women. Proceedings of the National Academy of Sciences of the United States of America, 102(51), 18757. 105 Friedman, E. M., & Herd, P. (2010). Income, education, and inflammation: Differential associations in a national probability sample (the MIDUS study). Psychosomatic Medicine, 72(3), 290-300. Gallo, L. C., Troxel, W. M., Kuller, L. H., Sutton-Tyrrell, K., Edmundowicz, D., & Matthews, K. A. (2003). Marital status, marital quality, and atherosclerotic burden in postmenopausal women. Psychosomatic Medicine, 65(6), 952-962. Gallo, L. C., Troxel, W. M., Matthews, K. A., & Kuller, L. H. (2003). Marital status and quality in middle-aged women: Associations with levels and trajectories of cardiovascular risk factors. Health Psychology : Official Journal of the Division of Health Psychology, American Psychological Association, 22(5), 453-463. doi:10.1037/0278-6133.22.5.453 Gimeno, D., Kivimaki, M., Brunner, E. J., Elovainio, M., De Vogli, R., Steptoe, A., et al. (2009). Associations of C-reactive protein and interleukin-6 with cognitive symptoms of depression: 12-year follow-up of the whitehall II study. Psychological Medicine, 39(3), 413-423. doi:10.1017/S0033291708003723 Gotlib, I. H., & Whiffen, V. E. (1989). Depression and marital functioning: An examination of specificity and gender differences. Journal of Abnormal Psychology, 98(1), 23. Gottman, J. M., & Levenson, R. W. (1992). Marital processes predictive of later dissolution: Behavior, physiology, and health. Journal of Personality and Social Psychology, 63(2), 221. 106 Gouin, J. P., Glaser, R., Loving, T. J., Malarkey, W. B., Stowell, J., Houts, C., et al. (2009). Attachment avoidance predicts inflammatory responses to marital conflict. Brain, Behavior, and Immunity, 23(7), 898-904. Grippo, A. J., Trahanas, D. M., Zimmerman, R. R.,2nd, Porges, S. W., & Carter, C. S. (2009). Oxytocin protects against negative behavioral and autonomic consequences of long-term social isolation. Psychoneuroendocrinology, 34(10), 1542-1553. Hajat, A., Diez-Roux, A., Franklin, T. G., Seeman, T., Shrager, S., Ranjit, N., et al. (2010). Socioeconomic and race/ethnic differences in daily salivary cortisol profiles: The multi-ethnic study of atherosclerosis. Psychoneuroendocrinology, 35(6), 932- 943. Hamer, M., & Chida, Y. (2009). Associations of very high C-reactive protein concentration with psychosocial and cardiovascular risk factors in an ageing population. Atherosclerosis, 206(2), 599-603. Hawkley, L. C., Hughes, M. E., Waite, L. J., Masi, C. M., Thisted, R. A., & Cacioppo, J. T. (2008). From social structural factors to perceptions of relationship quality and loneliness: The chicago health, aging, and social relations study. Journals of Gerontology: Series B: Psychological Sciences and Social Sciences, 63B(6), S375- S384. Retrieved from PsycINFO database. Hintsanen, M., Elovainio, M., Puttonen, S., Kivimaki, M., Koskinen, T., Raitakari, O. T., et al. (2007). Effort-reward imbalance, heart rate, and heart rate variability: The cardiovascular risk in young finns study. International Journal of Behavioral Medicine, 14(4), 202-212. 107 Holt-Lunstad, J., Smith, T. B., & Layton, J. B. (2010). Social relationships and mortality risk: A meta-analytic review. PLoS Med, 7(7), e1000316. House, J. S., Landis, K. R., & Umberson, D. (1988). Social relationships and health. Science, 241(4865), 540-545. Howren, M. B., Lamkin, D. M., & Suls, J. (2009). Associations of depression with C- reactive protein, IL-1, and IL-6: A meta-analysis. Psychosomatic Medicine, 71(2), 171-186. Hughes, M. E., & Waite, L. J. (2009). Marital biography and health at mid-life. Journal of Health and Social Behavior, 50(3), 344-358. Huikuri, H. V., Jokinen, V., Syvanne, M., Nieminen, M. S., Airaksinen, K. E., Ikaheimo, M. J., et al. (1999). Heart rate variability and progression of coronary atherosclerosis. Arteriosclerosis, Thrombosis, and Vascular Biology, 19(8), 1979- 1985. Johnson, N. J., Backlund, E., Sorlie, P. D., & Loveless, C. A. (2000). Marital status and mortality: The national longitudinal mortality study. Annals of Epidemiology, 10(4), 224-238. Kaplan, R. M., & Kronick, R. G. (2006). Marital status and longevity in the united states population. Journal of Epidemiology and Community Health, 60(9), 760-765. Kiecolt-Glaser, J. K., Bane, C., Glaser, R., & Malarkey, W. B. (2003). Love, marriage, and divorce: Newlyweds' stress hormones foreshadow relationship changes. Journal of Consulting and Clinical Psychology, 71(1), 176-188. 108 Kiecolt-Glaser, J. K., Gouin, J. P., & Hantsoo, L. (2010). Close relationships, inflammation, and health. Neuroscience and Biobehavioral Reviews, 35(1), 33-38. Kiecolt-Glaser, J. K., Loving, T. J., Stowell, J. R., Malarkey, W. B., Lemeshow, S., Dickinson, S. L., et al. (2005). Hostile marital interactions, proinflammatory cytokine production, and wound healing. Archives of General Psychiatry, 62(12), 1377-1384. doi:10.1001/archpsyc.62.12.1377 Kiecolt-Glaser, J. K., & Newton, T. L. (2001). Marriage and health: His and hers. Psychological Bulletin, 127(4), 472-503. Kishimoto, T. (2005). Interleukin-6: From basic science to medicine—40 years in immunology. Immunology, 23(1), 1-21. Kleiger, R. E., Stein, P. K., & Bigger, J. T.,Jr. (2005). Heart rate variability: Measurement and clinical utility. Annals of Noninvasive Electrocardiology : The Official Journal of the International Society for Holter and Noninvasive Electrocardiology, Inc, 10(1), 88-101. Kouros, C. D., Papp, L. M., & Cummings, E. M. (2008). Interrelations and moderators of longitudinal links between marital satisfaction and depressive symptoms among couples in established relationships. Journal of Family Psychology, 22(5), 667. Kudielka, B. M., & Kirschbaum, C. (2003). Awakening cortisol responses are influenced by health status and awakening time but not by menstrual cycle phase. Psychoneuroendocrinology, 28(1), 35-47. 109 Kunz-Ebrecht, S. R., Kirschbaum, C., Marmot, M., & Steptoe, A. (2004). Differences in cortisol awakening response on work days and weekends in women and men from the Whitehall II cohort. Psychoneuroendocrinology, 29(4), 516-528. La Rovere, M. T., Specchia, G., Mortara, A., & Schwartz, P. J. (1988). Baroreflex sensitivity, clinical correlates, and cardiovascular mortality among patients with a first myocardial infarction. A prospective study. Circulation, 78(4), 816-824. Liao, D., Cai, J., Rosamond, W. D., Barnes, R. W., Hutchinson, R. G., Whitsel, E. A., et al. (1997). Cardiac autonomic function and incident coronary heart disease: A population-based case-cohort study. the ARIC study. atherosclerosis risk in communities study. American Journal of Epidemiology, 145(8), 696-706. Loerbroks, A., Schilling, O., Haxsen, V., Jarczok, M. N., Thayer, J. F., & Fischer, J. E. (2010). The fruits of ones labor: Effort-reward imbalance but not job strain is related to heart rate variability across the day in 35-44-year-old workers. Journal of Psychosomatic Research, 69(2), 151-159. doi:10.1016/j.jpsychores.2010.03.004 Manuck, S. B. (1994). Cardiovascular reactivity in cardiovascular disease:“Once more unto the breach”. International Journal of Behavioral Medicine, 1(1), 4-31. Manzoli, L., Villari, P., Pirone, G., & Boccia, A. (2007). Marital status and mortality in the elderly: A systematic review and meta-analysis. Social Science & Medicine (1982), 64(1), 77-94. Marks, N. F., & Lambert, J. D. (1998). Marital status continuity and change among young and midlife adults: Longitudinal effects on psychological well-being. Journal of Family Issues, 19(6), 652 <last_page> 686. 110 März, W., Winkler, K., Nauck, M., Böhm, B. O., & Winkelmann, B. R. (2003). Effects of statins on C-reactive protein and interleukin-6 (the ludwigshafen risk and cardiovascular health study). The American Journal of Cardiology, 92(3), 305-308. Mather, M., & Carstensen, L. L. (2005). Aging and motivated cognition: The positivity effect in attention and memory. Trends in Cognitive Sciences, 9(10), 496-502. Matthews, K. A., Zheng, H., Kravitz, H. M., Sowers, M. F., Bromberger, J. T., Buysse, D. J., et al. (2010). Are inflammatory and coagulation biomarkers related to sleep characteristics in mid-life women?: Study of women's health across the nation sleep study. Sleep, 33(12), 1649-1655. McEwen, B. S., & Seeman, T. (1999). Protective and damaging effects of mediators of stress. elaborating and testing the concepts of allostasis and allostatic load. Annals of the New York Academy of Sciences, 896, 30-47. McLoyd, V. C., Cauce, A. M., Takeuchi, D., & Wilson, L. (2000). Marital processes and parental socialization in families of color: A decade review of research. Journal of Marriage and Family, 62(4), 1070-1093. Miller, G. E., Dopp, J. M., Myers, H. F., Stevens, S. Y., & Fahey, J. L. (1999). Psychosocial predictors of natural killer cell mobilization during marital conflict. Health Psychology : Official Journal of the Division of Health Psychology, American Psychological Association, 18(3), 262-271. Molloy, G. J., Stamatakis, E., Randall, G., & Hamer, M. (2009). Marital status, gender and cardiovascular mortality: Behavioural, psychological distress and metabolic explanations. Social Science & Medicine (1982), 69(2), 223-228. 111 Nealey-Moore, J. B., Smith, T. W., Uchino, B. N., Hawkins, M. W., & Olson-Cerny, C. (2007). Cardiovascular reactivity during positive and negative marital interactions. Journal of Behavioral Medicine, 30(6), 505-519. Orth-Gomer, K., Wamala, S. P., Horsten, M., Schenck-Gustafsson, K., Schneiderman, N., & Mittleman, M. A. (2000). Marital stress worsens prognosis in women with coronary heart disease: The stockholm female coronary risk study. JAMA : The Journal of the American Medical Association, 284(23), 3008-3014. Porges, S. W. (2007). The polyvagal perspective. Biological Psychology, 74(2), 116-143. Radloff, L. S. (1977). The ces-D scale. Applied Psychological Measurement, 1(3), 385. Rendall, M. S., Weden, M. M., Favreault, M. M., & Waldron, H. (2011). The protective effect of marriage for survival: A review and update. Demography, 48(2), 481-506. Renshaw, K. D., Blais, R. K., & Smith, T. W. (2010). Components of negative affectivity and marital satisfaction: The importance of actor and partner anger. Journal of Research in Personality, 44(3), 328-334. Robles, T. F. (2006). Positive behaviors during marital conflict: Influences on stress hormones. Journal of Social and Personal Relationships, 23(2), 305 <last_page> 325. Robles, T. F., & Kiecolt-Glaser, J. K. (2003). The physiology of marriage: Pathways to health. Physiology & Behavior, 79(3), 409-416. Rook, K. S. (1987). Social support versus companionship: Effects on life stress, loneliness, and evaluations by others. Journal of Personality and Social Psychology, 52(6), 1132-1147. 112 Ross, C. E. (1995). Reconceptualizing marital status as a continuum of social attachment. Journal of Marriage and the Family, 57(1), 129-140. doi:10.2307/353822 Ryff, C., Almeida, D. M., Ayanian, J. S., Carr, D. S., Cleary, P. D., Coe, C., et al. (2006). National survey of midlife development in the united states (MIDUS II) No. ICPSR04652-v6). Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]. Saxbe, D. E., Repetti, R. L., & Nishina, A. (2008). Marital satisfaction, recovery from work, and diurnal cortisol among men and women. Health Psychology, 27(1), 15. Sbarra, D. A. (2009). Marriage protects men from clinically meaningful elevations in C- reactive protein: Results from the national social life, health, and aging project (NSHAP). Psychosomatic Medicine, 71(8), 828-835. Schuster, T. L., Kessler, R. C., & Aseltine, R. H.,Jr. (1990). Supportive interactions, negative interactions, and depressed mood. American Journal of Community Psychology, 18(3), 423-438. Seeman, T. E. (1996). Social ties and health: The benefits of social integration. Annals of Epidemiology, 6(5), 442-451. Seeman, T. E., Singer, B. H., Ryff, C. D., Dienberg Love, G., & Levy-Storms, L. (2002). Social relationships, gender, and allostatic load across two age cohorts. Psychosomatic Medicine, 64(3), 395-406. Singh, T., & Newman, A. B. (2010). Inflammatory markers in population studies of aging. Ageing Research Reviews, (3), 319-329. 113 Smith, T. W., Cribbet, M. R., Nealey-Moore, J. B., Uchino, B. N., Williams, P. G., Mackenzie, J., et al. (2011). Matters of the variable heart: Respiratory sinus arrhythmia response to marital interaction and associations with marital quality. Journal of Personality and Social Psychology, 100(1), 103-119. Smith, T. W., Gallo, L. C., Goble, L., Ngu, L. Q., & Stark, K. A. (1998). Agency, communion, and cardiovascular reactivity during marital interaction. Health Psychology : Official Journal of the Division of Health Psychology, American Psychological Association, 17(6), 537-545. Smith, T. W., Uchino, B. N., Berg, C. A., Florsheim, P., Pearce, G., Hawkins, M., et al. (2009). Conflict and collaboration in middle-aged and older couples: II. cardiovascular reactivity during marital interaction. Psychology and Aging, 24(2), 274-286. doi:10.1037/a0016067 Task Force. (1996). Heart rate variability: Standards of measurement, physiological interpretation and clinical use. task force of the european society of cardiology and the north american society of pacing and electrophysiology. Circulation, 93(5), 1043-1065. Thayer, J. F., Yamamoto, S. S., & Brosschot, J. F. (2010). The relationship of autonomic imbalance, heart rate variability and cardiovascular disease risk factors. International Journal of Cardiology, 141(2), 122-131. Thayer, J. F., & Lane, R. D. (2009). Claude bernard and the heart-brain connection: Further elaboration of a model of neurovisceral integration. Neuroscience and Biobehavioral Reviews, 33(2), 81-88. 114 Thayer, J. F., & Sternberg, E. (2006). Beyond heart rate variability: Vagal regulation of allostatic systems. Annals of the New York Academy of Sciences, 1088, 361-372. Troxel, W. M., Matthews, K. A., Gallo, L. C., & Kuller, L. H. (2005). Marital quality and occurrence of the metabolic syndrome in women. Archives of Internal Medicine, 165(9), 1022-1027. Tsuji, H., Larson, M. G., Venditti, F. J.,Jr, Manders, E. S., Evans, J. C., Feldman, C. L., et al. (1996). Impact of reduced heart rate variability on risk for cardiac events. the framingham heart study. Circulation, 94(11), 2850-2855. Turner, R. J., Frankel, B. G., & Levin, D. M. (1983). Social support: Conceptualization, measurement, and implications for mental health. Research in Community & Mental Health, , 67-111. Uchino, B. N. (2009). What a lifespan approach might tell us about why distinct measures of social support have differential links to physical health. Journal of Social and Personal Relationships, 26(1), 53-62. Uchino, B. N., Cacioppo, J. T., & Kiecolt-Glaser, J. K. (1996). The relationship between social support and physiological processes: A review with emphasis on underlying mechanisms and implications for health. Psychological Bulletin, 119(3), 488-531. Uchino, B. N., Vaughn, A. A., & Matwin, S. (2008). Social psychological processes linking personality to physical health: A multilevel analysis with emphasis on hostility and optimism. In F. Rhodewalt (Ed.), Personality and social behavior. (pp. 251-283). New York, NY, US: Psychology Press. 115 Umberson, D., & Williams, K. (2005). Marital quality, health, and aging: Gender equity? The Journals of Gerontology.Series B, Psychological Sciences and Social Sciences, 60, 109-112. Umberson, D., Williams, K., Powers, D. A., Liu, H., & Needham, B. (2006). You make me sick: Marital quality and health over the life course. Journal of Health and Social Behavior, 47(1), 1-16. Uusitalo, A., Mets, T., Martinmaki, K., Mauno, S., Kinnunen, U., & Rusko, H. (2011). Heart rate variability related to effort at work. Applied Ergonomics, doi:10.1016/j.apergo.2011.01.005 Waite, L. J., & Gallagher, M. (2000). The case for marriage: Why married people are happier, healthier, and better off financially. New York, New York: Random House. Williams, K., & Umberson, D. (2004). Marital status, marital transitions, and health: A gendered life course perspective. Journal of Health and Social Behavior, 45(1), 81- 98. 116 APPENDIX A: METHODOLOGICAL APPROACHES TO ASSESSING DIURNAL CORTISOL RHYTHMS IN EPIDEMIOLOGICAL STUDIES: HOW MANY SALIVARY SAMPLES ARE NECESSARY? In the past decade there has been a strong movement to understand biomarkers of aging, with a considerable emphasis placed on biomarkers that may link social, behavioral, and psychological factors to diseases of aging (Harris, Gruenewald, & Seeman, 2008). Because of this, many large-scale community and population-based epidemiological studies have integrated biological measures into their studies. This has offered researchers a wealth of data to examine complex associations between behavioral and social risk factors and pre-clinical disease. However, the addition of these measures has also introduced methodological complexities that are only beginning to be untangled. One biomarker of particular interest is cortisol – an important hormone regulated by the hypothalamic-pituitary-adrenal (HPA) axis. Cortisol has regulatory functions for energy use and metabolism, as well as inflammatory/immune responses. The HPA axis has been indicated as a major pathway linking social factors to aging (McEwen & Seeman, 1999; Sapolsky, Krey, & McEwen, 1986). It is thought that psychosocial stress activates certain regions of the brain (e.g., the hypothalamus), which results in increased production of corticotropin releasing hormone (CRH). Increases in CRH signal for the release of adrenocorticotropin hormone (ACTH) from the anterior pituitary, resulting in increased production of glucocorticoids (cortisol in humans, corticosterone in non-human animals) from the zona fasciculate of the adrenal cortex. Chronic exposure to glucocorticoids has been shown to cause damage and dysfunction in the central nervous 117 system, as well as peripheral tissue damage (Sapolsky, Krey, & McEwen, 1986). Thus, understanding cortisol secretory patterns in human populations has the potential to make significant contributions to research on stress and aging. The study of cortisol in human populations is complex because cortisol follows a circadian rhythm that is based on individual wake and sleep times, as well as normal meal times. A normal, healthy cortisol rhythm is thought to have a sharp morning increase that peaks within an hour of waking, followed by a decline throughout the remainder of the day, with increases around mealtimes, and a nadir in preparation for evening sleep (see Figure 1; Kirshbaum & Hellhammer, 1989; Pruessner et al., 1997). There are numerous studies identifying what appear to be pathological aberrations from normal diurnal cortisol rhythms. Many of these studies suggest that hypoactive and hyperactive HPA function are both signs of poor health. Blunted responses – indicated by lower awakening responses and less decline in cortisol over the day – have been observed among individuals with major depression and post-traumatic stress disorder (Yeheuda, 1996), as well among individuals with metastatic breast cancer (Abercrombie et al., 2004). A recent meta-analysis found that higher cortisol awakening responses are observed in individuals experiencing general life stress and that lower cortisol awakening responses are observed in individuals with post-traumatic stress disorder and among individuals with positive psychological characteristics such as happiness, optimism, and self-esteem (Chida & Steptoe, 2009). These findings illustrate the fact that being either over-reactive or non-reactive (blunted) may be symptomatic of underlying pathology. 118 Previous research findings relating cortisol to health are difficult to interpret because in the past researchers have tended to focus on overall cortisol production over the day, levels of cortisol at certain times of the day, or specific parts of the diurnal rhythm, such as the awakening response or daily decline. Examining average levels over a day typically requires 24-hour urine collection, which is not feasible for ambulatory studies. In addition, averaging cortisol values over the day, or examining differences in levels at one point in the day, is not appropriate for understanding the dynamics of the cortisol rhythm, which have been shown to have important links to health (Kudielka, Gierens, Hellhammer, Wüst, & Schlotz, 2012). An additional complication in studying cortisol is that although cortisol production follows a diurnal rhythm, it is also sensitive to acute changes in the environment. Cortisol levels increase with the anticipation of food (Ott et al., 2011), the experience of a stressful event (Dickerson & Kemeny, 2004; Kirschbaum, Pirke, & Hellhammer, 1993), or injury (Kushner, 1982; Thompson, 2003). These acute responses cause problems for researchers interested in studying cortisol in naturalistic settings because they introduce variability in the diurnal rhythm that is difficult to account for without creating considerable burden to participants by asking them to provide numerous samples over the day, or maintain diaries on food intake, social interactions, and pain. Laboratory studies indicate that hourly measures are desirable in order to fully capture the complete curve and pulsatility patterns of cortisol (Young, Carlson, & Brown, 2001). However, providing numerous samples across the day is not feasible in ambulatory epidemiological studies because it not only increases the burden on 119 participants, but also increases the cost of the study. The cost of obtaining and assaying one salivary cortisol sample, on average, ranges between $5-10 per sample, thus, one additional sample in a study with 1,000 participants can easily increase the study cost by $5,000 to $10,000 (Adam & Kumari, 2009). Therefore it is not feasible to measure cortisol numerous times throughout the day. Theoretically and empirically, the cortisol awakening response (a.k.a. morning rise) and the daily decline are the two most important aspects of the diurnal rhythm that have been associated with health, therefore it is important to capture these two aspects of the curve (Saxbe, 2008). Currently, several large-scale epidemiological studies have obtained salivary cortisol measures; however, there is not currently a consensus on what the best research protocol is to assess the diurnal rhythm. Several studies (reviewed in Adam & Kumari, 2009), such the Chicago Health, Aging, and Social Relationships Study (CHASRS) and the Cebu Longitudinal Health and Nutrition Survey obtain 3 measures per day: upon waking, 30-60 minutes after waking, and at bedtime. Other studies, such as Whitehall II and the Coronary Artery Risk Development in Young Adults Study (CARDIA) obtained 6 samples per day: upon waking, 30-60 minutes, 2.5 hours, 8 hours, and 12 hours after waking, as well as at bedtime. A more recent study, the National Study of Daily Experiences (NSDE), employed a different methodology obtaining saliva samples at waking, 30-60 minutes after waking, before lunch (not at a specific time), and at bedtime. One problem with using mid-day cortisol measures is that they are highly influenced by routine eating patterns and the anticipation of food - not just food ingestion. 120 Among individuals that routinely eat lunch, cortisol begins to rise before lunch, in order to prepare the body for a meal, and continues to rise for several minutes after the meal (Quigley & Yen, 1979). However, when a snack is given before a regular meal, the cortisol rise before lunch is not as pronounced (Follenius et al., 1982). For individuals that have abnormal eating habits (i.e., do not eat lunch regularly), then a cortisol rise is not observed during lunchtime, but is entrained to rise before a normally eaten meal (Honma, Honma, & Hiroshige, 1984). Moreover, for individuals anticipating food ingestion (i.e., told they will be given a meal, or presented a meal), but required to refrain from eating, experience substantial rises in cortisol (Ott et al., 2011). Taken together, this indicates that a substantial amount of variability in mid-day cortisol levels will be influenced by regular eating behavior, food availability, dieting, and snacking – factors that are unlikely to be controlled for by time sampling and are likely to be problematic among middle-aged and older individuals with highly variable eating behaviors and meal times. Thus, it is of interest to understand whether a cortisol measure taken before lunch is meaningful, or if it introduces variability that may mask the true shape of the curve. The purpose of the current study is to examine whether a pre-lunch sample of cortisol adds relevant information to the diurnal cortisol rhythm, and if this measure informs the curve or leads to misspecification of the daily rhythm. To do this, two separate growth curve models (GCM) were developed and compared to investigate the necessity of including pre-luch cortisol measures in the model using empirical data from the NSDE. The first GCM involved three salivary samples (wake, 30-60 minutes after 121 waking, and bedtime); while the second considered all four salivary samples (wake, 30- 60 minutes after waking, before lunch, and bedtime). Method The NSDE is a telephone daily diary study was conducted as part of the MacArthur Foundation National Survey of Midlife in the United States (MIDUS), a nationally representative sample of adults in the United States. For a detailed description of the study, see Almeida, McGonagle, and King (2009). For the NSDE, a random subsample of 3,600 MIDUS respondents were asked to participate, and a total of 2,022 agreed to participate. The age range of participants at the time of the study was 35 to 84. Compared to MIDUS, the NSDE has a greater percentage of non-Hispanic Whites, females, and individuals with a college education. The NSDE contains data from two waves of the study, the first wave (NSDE 1) was conducted in 1996, and the second (NSDE II) was conducted from 2004 to 2006. For the purpose of the current study, we use only data from the 2004-2006 wave, as this was the only wave in which salivary cortisol samples were collected. The NSDE II included short (10-20 minute) telephone interviews about daily experiences on 8 consecutive days. On four of the days participants provided four salivary samples (a total of 16 samples). Home collection kits were sent to participants the week prior to the study start date. The collection kit included directions instructing participants to take the samples at waking, 30 minutes after getting out of bed, before lunch, and before bed. Participants were instructed to refrain from eating a large meal 1 hour before sampling, and were told to refrain from dairy products for 20 minutes before 122 sampling. Less than 3% of the samples were missed or unusable due to contamination. The exact time of each sample was taken from telephone interviews by staff, and from a paper-pencil log sent with the collection kit. Approximately 25% of the participants received a “smart box” containing an electronic recording device that recorded the time of the box opening and closing. Self-reported collection time was highly correlated (r = .90) with electronic recorded time (Almeida, McGonagle, & King, 2009). Salivary samples were taken four times per day (waking, 30 minutes after waking, before lunch, and in the evening before bedtime) for four days. Participants used Starstet salivettes (Nümbrecht, Germany), which are cotton rolls placed in the cheek until saturated and subsequently stored in plastic tubes. Samples were returned to the MIDUS Biological Core center and were frozen at -60°C for shipping and storage. For analysis, salivettes were thawed and centrifuged at 3,000 rpm for 5 minutes. Concentrations of cortisol were measured using a luminescence immunoassay (Hamburg, Germany). Covariates Medication Use. Medication use was assessed by self reported use of over-the- counter and prescription medications. Medications known to alter cortisol production were recorded and include steroid inhaler, steroid medications, medications containing cortisone, hormone medications (including birth control pills), anti-depressants and anti- anxiety medications. A dichotomous indicator was created to indicate whether or not a person was taking one or more of these medications (0 = no medication use; 1= current medication use). 123 Smoking Status. A dichotomous indicator was created to indicate whether or not a person smoked (0 = non-smoker; 1 = current smoker). An individual reporting any smoking on 1 or more days of the 8-day study period, was coded as a smoker. Analytic Strategy Cortisol rhythms are primarily driven by time since waking, and not standard clock time. Thus, time since waking (i.e., wake time = 0) was used as the measure of time. Analyses were restricted to Wednesday because previous research indicates there is a “weekend effect” where cortisol levels differ between weekdays and weekends. Wednesday was the day of the week with the highest levels of participation in this sample (N = 1038). Because previous studies indicate that cortisol values over the day are partially dependent on a person falling asleep (when cortisol should be at its lowest value), and waking (when cortisol will rise for 30-60 minutes), we eliminated individuals from the analyses for the following reasons: (1) slept fewer than 4 hours the night before or stayed awake longer than 20 hours during the day of sampling (n = 38), or (2) did not take their morning sample within 1-hour of waking (n = 88), as this would limit the ability to detect the morning cortisol rise. These restrictions result in a final analytic sample of 912. Distributions of cortisol are typically skewed which was also observed in this sample; therefore we added 1 to each cortisol value (to eliminate negative values), and natural-log transformations were used to normalize these distributions. Descriptive characteristics of the final analytic sample are presented in Table 2. Growth curve models for repeated measures were used to describe the change in cortisol across the day as well as variation across individuals. Models were estimated by 124 gender, without covariates to establish the shape of the cortisol curve across the day. Strong evidence indicates that diurnal cortisol rhythms are curvilinear; therefore polynomial functions of time were explored. The growth models assume that overall growth will follow a quadratic or cubic form, but do not examine person-specific quadratic growth because this is not expected and this assumption also introduces a great deal of complexity into an already complex model. Therefore individual-specific linear trends are explored, and polynomial functions of random effects are not included in the model. Maximum likelihood was used instead of restricted maximum likelihood in order to facilitate significance tests between models using likelihood ratio tests with changes fixed-effects parameters, where maximum likelihood is appropriate. However, all models were also fit using restricted maximum likelihood to ensure the standard errors did not change substantially. A 2-level model was developed to examine variability in cortisol across the day as well as across individuals. The Level-1, or within-individual model investigates the variation of repeatedly measured cortisol within an individual with time as the predictor. Because individuals may vary in the time of their sampling, time in hours since waking is modeled as the independent variable. The Level-2, or between-individual, model examines the variation across individuals. Equations of Level-1, Level-2, and Composite equations are as follows: Level-1 Model: Cortisol ij = π 0i + π 1i Time ij + π 2i Time 2 ij + π 3i Time 3 ij +ε ij Level-2 Model: π 0i = γ 00 + ζ 0i 125 π 1i = γ 10 + ζ 1i π 2i = γ 20 π 3i = γ 20 Composite Model: Cortisol ij = γ 00 + γ 10 Time ij + γ 20 Time 2 ij + γ 20 Time 3 ij + (ε ij + ζ 0i + ζ 1i Time ij ) In the Level-1 model, Cortisol ij is individual i’s value of cortisol on occasion j and Time ij is the amount of time elapsed since waking for individual i on occasion j. The amount of individual i’s cortisol that is unexplained on occasion j is ε ij . In level-2, π 0i represents individual i’s true mean cortisol value upon waking and ζ 0i represents the amount that individual i differs from the population mean (γ 00 ), π 1i represents individual i’s true average rate of change in cortisol and ζ 0i represents the amount that individual i’s rate of change differs from the population average rate of change (γ 10 ), π 2i represents individual i’s quadratic rate of change in cortisol, π 3i represents individual i’s cubic rate of change in cortisol. Finally, time invariant covariates were entered into the model. Covariates included wake time (centered), age (centered), smoking, and medication use. These covariates were not of substantive interest, but were entered into models to ensure that the final estimates of time were not substantially changed with the addition of these covariates. Results Descriptive characteristics of the sample are presented in Table 1. Men had higher awakening values of cortisol compared to women, and had higher lunch values as well. Three quarters of the sample showed evidence of the cortisol awakening response. A 126 large percentage of men (70.5%) and women (47.1%) were taking medications that are known to influence cortisol. Chi-square tests indicate that medication use was not associated with experiencing the awakening response for men (χ 2 = 2.12, p = .14) or women (χ 2 = 0.01, p = .92). To examine variability in the time and level of cortisol for the lunch measure, cortisol values were plotted by time since waking only for the lunch sample of cortisol. Figure 2 shows that for both men and women, there is considerable variability in the number of hours since waking that one chooses to eat lunch, ranging from 2.4 hours after waking to 14.4 hours after waking. For men, the median number of hours since waking for taking the lunch sample was 6 hours, with the 25 th percentile taking their lunch sample 5.1 hours past waking and the 75 th percentile taking their sample 7 hours past waking. For women, the median number of hours since waking for taking the lunch sample was 5.8 hours, with the 25 th percentile taking their sample 5 hours past waking and the 75 th percentile taking their sample 6.9 hours past waking. Figure 3 presents the distribution of cortisol by clock time for the lunch sample of cortisol. Clock time shows a slightly different story, indicating that most individuals eat during socially normative times. For men, the median lunchtime was 12:00 p.m., with the 25 th percentile eating lunch at 11:35 a.m., and the 75 th percentile eating lunch at 1:00 p.m. For women, the median lunch time was 12:15 p.m., with the 25 th percentile eating lunch at 11:45 p.m., and the 75 th percentile eating lunch at 1:00 p.m. For the cortisol values in plotted by both clock and person time (Figures 2 and 3), regression lines and 95% confidence intervals were plotted, and the slopes of all four lines were not significant. 127 To test the different modeling strategies with and without the lunch cortisol sample, the results of the GCMs by gender using Time, Time 2 , and Time 3 are presented in Table2. The models using 3 time points are presented in Panel A of Table 2, and the models using 4 time points are presented in the Panel B of Table 2. In Model 1, the linear term for time is presented, followed by Model 2, which presents the quadratic term for time, with Model 3 showing the cubic function for time. Using three data points, the cubic function for time (Model 3) fit the data best for both men and women. For men, the likelihood ratio test between Model 1 and Model 3 was significant (χ 2 (2) = 77.34, p < .001), as was the likelihood ratio test between Model 2 and Model 3 (χ 2 (1) = 67.66, p < .001). For women, the results were similar. The likelihood ratio test between Model 1 and Model 3 was significant (χ 2 (2) = 239.33, p < .001), as was the likelihood ratio test between Model 2 and 3 (χ 2 (1) = 229.13, p < .001), A graphical depiction of the growth curves from Model 3 for men and women are presented in Figure 4, which shows the quadratic curve as rising sharply within the first hours of morning, decreasing throughout the day and coming to a nadir at the end of the day. It should be noted that there is sparseness of the data between hour 1, and hour 5, which is why the models indicate cortisol would still be rising during this time. Comparing the AIC and BIC fit statistics from Panel A to Panel B shows the fit of the models using 3 data points is consistently higher compared to the models with 4 data points for both men and women. When using 4 data points, Model 2 fit the data best for men, whereas Model 3 fit the data best for women. However, examining the coefficients of the models in Panel B shows that the linear term for time is negative for all models, 128 indicating that cortisol has an immediate negative slope, and therefore does not model the morning rise. Graphical depictions of Model 2 for men and Model 3 for women are shown in Figure 4. These show that cortisol has a consistently negative slope for males, and a negative slope for women until the evening, when the slope appears to flatten out. Discussion The current study examined whether a pre-lunch sample of cortisol adds relevant information to the diurnal cortisol rhythm, or if it leads to misspecification of the daily rhythm. Using two separate growth curve models, with and without the pre-lunch measure, the findings of this study suggest that the use of the pre-lunch measure results in model misspecification. The use of the pre-lunch measure reduced model fit, compared to models without this measure, and resulted in growth models that do not resemble diurnal cortisol rhythms that are observed in well-controlled laboratory studies. The models without the pre-lunch measure had considerable better model fit, and also revealed the theoretically expected cortisol rhythm, with a sharp increase in the morning, followed by a decline throughout the day. The findings from this study indicate that greater numbers of samples across the day do not always lead to greater information. Significant costs are incurred for studies incorporating additional measures of cortisol, and the findings from this study suggest that the use of a pre-lunch measure may not provide relevant information. Thus, studies obtaining only 3 cortisol samples over the day, such as CHASRS and Cebu, may not only reduce costs, but may also result data that better describes normal diurnal cortisol rhythms. It would be worthwhile to replicate the methodological comparisons used in this 129 study on studies that have obtained more than one mid-day measure. Whitehall II or CARDIA, which have obtained cortisol samples at 2.5, 8, and 12 hours after waking, may be useful for these comparisons. There are several limitations to this study. First, the NSDE did not obtain daily food diaries. Although food and eating behaviors are suspected to result in significant measurement error in the pre-lunch measure, there is no way of telling whether this is the reason for our findings. It would be interesting to examine diurnal rhythms that include measures of normal meal times, as well as actual eating behaviors and food availability on the day that cortisol is being sampled. A second limitation is that this study population contained a considerable number of individuals on medications, with nearly half of women, and three-quarters of men taking some type of medication that influences cortisol. Although these medications did not appear to play a significant role in cortisol rhythms as indicated by the modest effect size estimates observed in the models, this could be a result of using a combined category of medications. It may be useful for further studies to examine different medications separately (i.e., allergy medications, steroid inhalers, hormones). 130 REFERENCES Chou, C., Bentler P.M., Pentz M.A. (1998). Comparison of two statistical approaches to study growth curves: The multilevel model and latent curve analysis. Structural Equation Modeling: A Multidisciplinary Journal, 5:247-266. Almeida, D. M., McGonagle, K., & King, H. (2009). Assessing daily stress processes in social surveys by combining stressor exposure and salivary cortisol. Biodemography and Social Biology, 55(2), 219-237. Dickerson, S. S., & Kemeny, M. E. (2004). Acute stressors and cortisol responses: A theoretical integration and synthesis of laboratory research. Psychological Bulletin, 130(3), 355. Harris, J. R., Gruenewald, T. L., & Seeman, T. E. (2008). An overview of biomarker research from community and population-based studies on aging. In M. Weinstein, J. W. Vaupel, K. W. Wachter, J. R. Harris, T. L. Gruenewald & T. E. Seeman (Eds.), Biosocial surveys (pp. 96-135). Washington, DC: The National Academies Press (Committee on Population, Division of Behavioral and Social Sciences and Education): National Academies Press (US). Honma, K. I., Honma, S., & Hiroshige, T. (1984). Feeding-associated corticosterone peak in rats under various feeding cycles. The American Journal of Physiology, 246(5 Pt 2), R721-6. Kirschbaum, C., Pirke, K. M., & Hellhammer, D. H. (1993). The ‘Trier social stress test’—a tool for investigating psychobiological stress responses in a laboratory setting. Neuropsychobiology, 28(1-2), 76-81. 131 Kudielka, B. M., Gierens, A., Hellhammer, D. H., Wüst, S., & Schlotz, W. (2012). Salivary cortisol in ambulatory assessment—some dos, some don’ts, and some open questions. Psychosomatic Medicine, 74(4), 418-431. Kushner, I. (1982). The phenomenon of the acute phase response. Annals of the New York Academy of Sciences, 389(1), 39-48. McEwen, B. S., & Seeman, T. (1999). Protective and damaging effects of mediators of stress. elaborating and testing the concepts of allostasis and allostatic load. Annals of the New York Academy of Sciences, 896, 30-47. Ott, V., Friedrich, M., Prilop, S., Lehnert, H., Jauch-Chara, K., Born, J., et al. (2011). Food anticipation and subsequent food withdrawal increase serum cortisol in healthy men. Physiology & Behavior, Quigley, M. E., & Yen, S. S. (1979). A mid-day surge in cortisol levels. The Journal of Clinical Endocrinology and Metabolism, 49(6), 945-947. Sapolsky, R. M., Krey, L. C., & McEwen, B. S. (1986). The neuroendocrinology of stress and aging: The glucocorticoid cascade hypothesis. Endocrine Reviews, 7(3), 284- 301. Saxbe, D. E. (2008). A field researcher's guide to cortisol: Tracking HPA axis functioning in everyday life. Health Psychology Review, 2(2), 163-190. Thompson, B. T. (2003). Glucocorticoids and acute lung injury. Critical Care Medicine, 31(4), S253. 132 Table 1. Characteristics of the analytic sample for NSDE participants with cortisol measures obtained on Wednesday. Male Female (n = 392) (n = 520) Mean (SD) Range Mean (SD) Range Age (years) 56.6 (11.6) 34 - 83 56.0 (12.1) 33 - 84 Wake Time (hr:min) 6:14 (1:10) 6:32 (1:12) Cortisol (µg/dL) a Awakening 2.79 (.65) .02 – 5.67 2.61 (.57) .09 – 4.37 30 minutes 3.08 (.63) .31 – 5.75 3.02 (.59) .05 – 5.05 Lunch 2.10 (.61) .22 – 5.10 1.92 (.65) .02 – 5.20 Bedtime 1.25 (.74) .02 – 4.74 1.17 (.77) .01 – 6.03 Percent Percent Awakening Response > 0 µg/dL 73.0% 78.8% ≤ 0 µg/dL 27.0% 21.2% Current Smoker 11.4% 12.3 % Currently on Medication b 70.5% 47.1% a These are values of cortisol that have been transformed using the equation ln(cortisol+1) b Medications included steroid inhalers, steroid containing medications, cortisone containing medications, birth control pills, other hormonal medications, and anti- depressant and anti-anxiety medications. Table 2. Growth curve models by gender without covariates comparing models without and with the before lunch cortisol sample. Panel A: Growth curve models without the before lunch cortisol sample (3 data points) Male Female Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 Fixed Effects Time -.13 (.004)*** -.002 (.04) .58 (.08)*** -.13 (.003)*** -.004 (.03) .76 (.05)*** Time 2 -.008 (.002)*** -.09 (.01)*** -.009 (.002)*** -.12 (.007)*** Time 3 .003 (.000)*** .004 (.000)*** Intercept 3.88 (.03)*** 3.85 (.03)*** 3.70 (.04)*** 3.75 (.03)*** 3.72 (.03)*** 3.53 (.03)*** Random Effects Intercept .17 (.03) .19 (.04) .23 (.03) .26 (.03) .23 (.03) .28 (.02) Slope .002 (.000) .002 (.000) .003 (.000) .003 (.000) .004 (.000) .004 (.000) AIC/BIC 2967/2997 2959/2994 2893/2934 3782/3814 3774/3811 3547/3589 Log Likelihood -1477 -1472 -1439 -1885 -1880 -1765 LRT test from previous model - 9.68** 67.66*** - 10.20*** 230.13*** Panel B: Growth curve models with the before lunch cortisol sample (4 data points) Male Female Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 Fixed Effects Time -.13 (.003)*** -.16 (.01)*** -.13 (.03)*** -.14 (.003)*** -.21 (.01)*** -.13 (.02)*** Time 2 .002 (.000)*** -.005 (.005) .004 (.001)*** -.01 (.003)** Time 3 .000 (.000) .0007 (.0002)*** Intercept 3.85 (.03)*** 3.88 (.03)*** 3.87 (.03)*** 3.69 (.03)*** 3.76 (.03)*** 3.73 (.03)*** Random Effects Intercept .18 (.03) .18 (.03) .18 (.03) .21 (.02) .21 (.02) .21 (.02) Slope .003 (.000) .003 (.000) .003 (.000) .003 (.000) .003 (.000) .003 (.000) AIC/BIC 3366/3397 3357/3394 3357/3400 4312/4345 4250/4289 4235/4279 Log Likelihood -1677 -1671 -1670 -2150 -2118 -2109 LRT test from previous model - 10.20*** 2.28 - 64.29*** 16.84*** Note. Model 1: Linear function of time; Model 2: Quadratic function of time; Model 3: Cubic function of time. Each likelihood ratio is a χ 2 with 1 degree of freedom. LRT = likelihood ratio test. * p ≤ .05 , **p ≤ .01, *** p ≤ .001 133 1 134 Table 3. Final multilevel models of change by gender with covariates using 3 data points. Male Female Fixed Effects Time .60 (.08)*** .74 (.06)*** Time 2 -.09 (.01)*** -.12 (.01)*** Time 3 .003 (.000)*** .04 (.000)*** Age .01 (.003)*** .01 (.002)*** Wake time .05 (.03) -.03 (.02) Medication .10 (.08) -.08 (.06) Smoker .15 (.12) .05 (.08) Intercept 3.56 (.12)*** 3.66 (.09)*** Random Effects Intercept .29 (.04) .22 (.03) Slope .003 (.000) .004 (.000) AIC/BIC 2011/2067 2465/2525 Log Likelihood -993 -1220 Note. Age is centered on 56.26 and wake time is centered on 6.40. LRT = likelihood ratio test. Each likelihood ratio is a χ 2 with 5 degrees of freedom. * p ≤ .05 , **p ≤ .01, *** p ≤ .001 135 Figure 1. Theoretical diurnal cortisol curve 136 Figure 2. Distribution of person time for the ‘before lunch’ sample. -5 0 5 10 Natural Log Cortisol (ug/dL) 0 5 10 15 Person Time of Lunch Cortisol Sample (hours) Men -5 0 5 10 Natural Log Cortsiol (ug/dL) 0 5 10 15 Person Time of Lunch Cortisol Sample (hours) Women 137 Figure 3. Distribution of clock time for the ‘before lunch’ sample. -5 0 5 10 15 Natural Log Cortisol (ug/dL) 0 5 10 15 20 Clock Time of Lunch Cortisol Sample (hours) Men -5 0 5 10 15 Natural Log Cortisol (ug/dL) 0 5 10 15 20 Clock Time of Lunch Cortisol Sample (hours) Women 138 Figure 4. Mean cortisol values by sample number. 0 1 2 3 4 5 Natural Log Cortisol +1 1 2 3 4 Sample Number Male Female 95% Confidence Interval Mean Cortisol Levels for Each Sample 139 Figure 5. Growth models of cortisol using three data points (solid lines) at various time points with a cubic function of time for men and women, versus 4 points with a quadratic function for men and a cubic function for women. 1 1.5 2 2.5 3 3.5 4 4.5 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Natural Log Cortisol +1 Time Since Waking Female (3 points) Male (3 points) Female (4 points) Male (4 points) 140 APPENDIX B: SENSITIVITY ANALYSES FOR CORTISOL There have been several methodologies used to examine diurnal cortisol rhythms in population surveys (Friedman, Karlamangla, Almeida, & Seeman, 2012; Stawski et al., 2011). The method chosen for the analyses in this dissertation was chosen for theoretical and methodological reasons that are outlined Appendix A. However, the methods used in this dissertation extend the model in Appendix A, incorporating a 3-level multilevel model for change. This better handles the complicated sampling design of the National Study of Daily Experiences, which included multiple samples per day for multiple days. The results of the analyses associating marital quality with cortisol rhythms are presented in Chapter 4, and are not entirely consistent with previous research. Therefore, to investigate whether these patterns of association with marital strain are a result of the method, or a result of marital quality having different patterns of association with cortisol compared to other relationship domains (e.g., friends, family), we compare marital strain, friendship strain, and family strain. The current document serves to report sensitivity analyses for these data, comparing models of marital strain with other models of social strain (friend, family). Comparing the results of the analyses of marital strain with family strain and friend strain, findings suggest that among men, marital strain influences cortisol differently than family and friend strain. Higher marital strain and family strain were both associated with higher awakening cortisol values, while higher friend strain was associated with lower awakening values of cortisol, for men. Higher marital strain was associated with a steeper decline in cortsiol across the day, but higher family and friend 141 strain were both associated with flatter declines over the day, for men. These differences are shown in Figure 2, below. Comparing the results of the analyses of marital strain with family strain and friend strain among women, the findings suggest that marital strain influences cortisol similarly to family and friend strain, although differences in marital strain were not significant for any part of the cortisol rhythm. For family and friend strain, higher strain was associated with significantly less steep of a decline through the day, but a higher upturn among those staying up later in the evening, where higher friend and family strain was associated with lower cortisol. Graphical depictions of the growth curves for these models are shown below in Figure 3. 142 Figure 2. Growth models of marital strain, family strain, and friend strain among men. 143 Figure 3. Growth models of marital strain, family strain, and friend strain among women.
Abstract (if available)
Abstract
This dissertation aims to elucidate the psychosocial and physiological mechanisms through which marriage affects health outcomes and disease processes by linking measures of marital quality (satisfaction, support, strain, disagreement) to physiological mechanisms that have been associated with psychosocial factors and adverse health outcomes: inflammation, neural regulation of the heart, and hypothalamic-pituitary-adrenal axis (HPA) activity. Data from the National Survey of Midlife in the United States are used, utilizing the main survey, the biomarker substudy, and the daily diary study. Higher spousal strain was weakly associated with higher inflammation among men and women, and that higher spousal support was associated with lower levels of inflammation among women, but not men. Higher levels of spousal strain and disagreement were associated with lower heart rate variability, and that higher levels of marital satisfaction and spousal support were associated with higher HRV. Men with higher levels of marital strain and disagreement have lower waking values of cortisol and greater cortisol awakening responses, with some indication of greater decreases in cortisol in the early evening, but higher values in the later evening. However, for women marital quality was not associated with differences in cortisol secretory patterns over the day. Taken together, these empirical papers show the importance of marital quality for understanding the health of the married. In general, the studies from this dissertation highlight the fact that there is heterogeneity in the health of the married. Individuals in good marriages (i.e., low in negative characteristics and/or high in positive characteristics) tend to appear healthier than those in bad marriages (i.e., high in negative characteristics and/or low in positive characteristics).
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Methodological approaches to assessing diurnal cortisol rythms in epidemiological studies: how many salivary samples are necessary?
PDF
Psychological and physiological pathways from social support exchanges to health: a lifespan perspective
PDF
The effects of wisdom-related personality traits on caregivers’ health: an application of the resilience model
PDF
Internet communication use, psychological functioning and social connectedness at older ages
PDF
Health differences between the elderly in Japan and the United States
PDF
Thinking generatively versus acting generatively: exploring the associations of generative self-concept and generative activity with cognitive function among older adults
PDF
Marital matching in West Africa: an examination of interethnic and interreligious first marriages in Benin
PDF
Social determinants of physiological health and mortality in China
PDF
Biomarkers of age-related health changes: associations with health outcomes and disparities
PDF
Lifting up ourselves by lifting up others: examining cognitive-affective states linking generativity to well-being
PDF
The immediate and long term legacy of relationships with grandparents for the well-being of grandchildren
PDF
An examination of the association between spousal support and type 2 diabetes self-management
PDF
Structure and function of the locus coeruleus across the lifespan
PDF
The dollars and sense of caregiving: exploring the financial financial realities of providing informal care to an older adult
PDF
Statistical algorithms for examining gene and environmental influences on human aging
PDF
Is stress exposure enough? Race/ethnic differences in the exposure and appraisal of chronic stressors among older adults
PDF
A biodemographic approach to understanding sociodemographic disparities in kidney functioning on three dimensions: individual, population, and cross-national
PDF
Examination of the long-term psychosocial and functioning effects of a healthy living, life management behavior intervention for older adults
PDF
Intergenerational transmission of values and behaviors over the family life course
PDF
The effects of health-related spousal influence on couples coping with chronic heart failure: an application of the actor-partner interdependence model
Asset Metadata
Creator
Donoho, Carrie Joy
(author)
Core Title
Marital quality, gender, and biomarkers of disease risk in the MIDUS cohort
School
Leonard Davis School of Gerontology
Degree
Doctor of Philosophy
Degree Program
Gerontology
Publication Date
11/12/2012
Defense Date
10/09/2012
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
cortisol,heart rate variability,Inflammation,Marriage,OAI-PMH Harvest,social support
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Crimmins, Eileen M. (
committee chair
), Abdou, Cleopatra (
committee member
), Azen, Stanley P. (
committee member
), Gruenewald, Tara (
committee member
)
Creator Email
carriedonoho@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-109977
Unique identifier
UC11288254
Identifier
usctheses-c3-109977 (legacy record id)
Legacy Identifier
etd-DonohoCarr-1282.pdf
Dmrecord
109977
Document Type
Dissertation
Rights
Donoho, Carrie Joy
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
cortisol
heart rate variability
social support