| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Department of Family and Consumer Sciences, University of Hawai'i at Manoa.
2 Department of Human Development and Family Sciences, Oregon State University, Corvallis.
3 Department of Epidemiology, Boston University School of Public Health, Massachusetts.
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
TAYLOR, Repetti, and Seeman (1997)
detailed the ways that unhealthy environments "get under the skin" by describing the physiological pathways through which social environments influence health. In this article, we explore further along this vein and ask how coping gets under the skin. Research suggests that coping is at least modestly associated with physical health (Penley, Tomaka, & Weibe, 2002). However, the results are decidedly mixed across studies, and findings about the relationship between coping and health outcomes are much more consistent for laboratory than for naturalistic stressors (Aldwin & Yancura, 2004). The research setting is another reason for inconsistencies in findings among studies. Laboratory studies often look at stress reactivity (the change in physiological outcomes before and after exposure to a stressor), whereas field studies generally examine direct relationships between coping and outcomes—a less controlled design, but one that is necessary to examine stressors that occur in real life.
An alternative strategy for field studies might be to examine affect as a proxy for stress reactivity, assuming that greater negative affect is associated with greater stress reactivity. It is likely that the relationship between coping and health is mediated by affect (Zautra, 2003). Positive and negative affect are directly related to health outcomes (Krantz & McCeney, 2002; Salovey, Rothman, Detweiler, & Steward, 2000). Affect has also been found to mediate between coping and health in one sample of young men (Billings, Folkman, Acree, & Moskowitz, 2000).
This problem might also be characterized as one of effect magnitude. If the relationship between coping and physical health outcomes is weak, one must have large sample sizes to detect it. Alternatively, one could create a summary index of multiple outcomes to aggregate the effect across variables. The metabolic syndrome is a good candidate for such an indicator, because it is a combination of a variety of factors such as blood pressure and cholesterol levels, both of which previous research has shown to be influenced by stress and coping processes (Bedi, Varshney, & Babbar, 2000; Niaura, Stoney, & Herbert, 1992).
There are relatively few studies on coping and physical health in samples of older individuals (Aldwin & Yancura, 2004). We find this especially surprising, given the assumed greater vulnerability of older individuals to stress and the large individual differences in health functioning in later life. Nevertheless, to our knowledge, a model proposing that affect mediates between coping and health outcomes has not been explicitly tested in an older sample. In the present study we seek to clarify these issues by examining affect as a mediator between coping and the metabolic syndrome in older men.
Does Coping Matter?
The adverse effects of stress on a variety of physical health outcomes are well documented (Cohen & Herbert, 1996; Strike & Steptoe, 2004); however, some have questioned whether coping influences these outcomes (Coyne & Racioppo, 2000). Penley and colleagues (2002) reviewed this literature and concluded that there are associations between coping and health. Unfortunately, the generalizations to older populations that can be made from this review are limited because it focused on psychological health, primarily in samples of young adults.
A limited number of studies have reported effects of coping on some health outcomes in samples of older individuals. Relationships between anger control and avoidance coping and some types of cholesterol have been found in a sample of spousal caregivers for Alzheimer's patients (Vitaliano, Russo, & Niaura, 1995). A Stress x Coping interaction on proliferative immune responses has also been reported (Stowell, Kiecolt-Glaser, & Glaser, 2001). Avoidance coping and anxiety have also been associated with slower cardiovascular recovery to laboratory stressors (Vitaliano, Russo, Paulsen, & Bailey, 1995). Taken together, findings from these few studies of coping in samples of older individuals indicate that coping might influence health among older adults. Clearly, however, more work is needed for researchers to understand how coping gets under the skin.
Affect as a Potential Mediator Between Stress and Coping Processes and Health Outcomes
Recent theories link physical health and affective states through several channels, including direct effects on physiology (Hall & Irwin, 2001; Salovey et al., 2000). Emotions directly influence cardiac (Krantz & McCeney, 2002), endocrine (McEwen, 2003) and immune systems (Cohen & Herbert, 1996; Kiecolt-Glaser, McGuire, Robles, & Glaser, 2002). They have also been shown to influence frailty (Ostir, Ottenbacher, & Markides, 2004), and longevity (Danner, Snowdon, & Friesen, 2001). Interestingly, positive and negative emotions appear to have discrete neural substrates (LeDoux, 1995) as well as independent effects on health (Cacioppo & Gardner, 1999). Results from studies examining the influence of stress, coping, and affect on the metabolic syndrome suggest that both positive and negative affect might mediate between them (Vitaliano, Scanlan, Siegler, McCormick, & Knopp, 1998; Vitaliano et al., 2002).
Affect and Stress and Coping Processes
Although a great deal of empirical work has been done on affect in social contexts in older individuals (Antonnucci, 2001; Carstensen, Fung, & Charles, 2003), relatively few studies have explored both positive and negative affect as part of the stress and coping process (Folkman & Moskowitz, 2004). The model by Lazarus and Folkman (1984) provides a framework to study this because it is adaptable to the range of stressors faced by older individuals (Aldwin, 1991). Its process view of coping is also theoretically appealing because it concurs with the prevalent view of stress and coping as a dynamic transaction between person and environment.
In this model, an individual's first response to a stressor is to make an appraisal of its possible impact and of the tools she or he has for dealing with it. This involves subjective interpretation of the situation in a complex cognitive-affective response and enables the choice of coping strategy (Lazarus, 1999). There are many conceptual approaches to the classification of coping strategies, including a problem–emotion focus, approach–avoidant dichotomies, and meaning making (Folkman & Moskowitz, 2004). In addition to these traditional approaches, coping strategies may also be conceptualized as being valenced toward positive or negative affect (Aldwin, Sutton, & Lachman, 1996). It is this affective valence that may influence physiological outcomes such as the metabolic syndrome.
The Metabolic Syndrome
The metabolic syndrome is one term used to describe a cluster of physiological symptoms linked to cardiovascular disease (Wilson & Grundy, 2003). Because it is a relatively new concept, there is a fair amount of disagreement between researchers on its exact clinical definition (Aguilar-Salinas et al., 2001; Ford, Giles, & Dietz, 2002; Reilly & Radar, 2003). The clinical diagnosis most commonly used in the United States was put forth in the third report of the NCEP-ATPIII, or National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol (Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults, 2001; Wilson & Grundy). Abnormalities in the metabolism of adipose tissue, as reflected by insulin resistance, are believed to be the mechanism by which the metabolic syndrome leads to the development of cardiovascular disease (Einhorn et al., 2003; Grundy et al., 2005). Accepted standards of measurement use both insulin and glucose measurements. Measurements of glucose do not typically deliver as much information about peripheral tissue or whole-body sensitivity to insulin (Matsuda and DeFronzo, 1999), but they are the most commonly used indicators of insulin sensitivity in the United States (Einhorn et al., 2003; Kahn, Buse, Ferrannini, & Stern, 2005).
Because of its close association with the development of diabetes and coronary heart disease, the metabolic syndrome is of great interest to researchers and clinicians (Pladevall et al., 2006; Wilson & Grundy 2003). Those individuals diagnosed with it are two to three times more likely to develop coronary artery disease than those who have been diagnosed with only one of its factors (Meigs, 2002; Scott, 2003). Recent research suggests that it is best represented by a second-order model with four factors: hypertension, lipids, obesity, and glucose (Grundy et al., 2005). We chose it as the outcome for this study because its factors are closely related to the physiological stress response (Chandola, Brunner, & Marmot, 2006) and are relatively common in older men (Ford et al., 2002).
The Present Study
In the present study we test a model within the framework of the Lazarus and Folkman (1984) process model of stress and coping (see Figure 1). We hypothesize that appraisal and coping will influence affect along pathways differentiated by emotional valence. In the positive pathway, stress leads to challenge appraisals, which leads to positive coping, and then positive affect. In the negative pathway, stress leads to threat appraisal, which leads to negative coping, and then negative affect. Affect, in turn, will influence the metabolic syndrome. We hypothesize negative affect to be a risk factor, whereas we hypothesize positive affect to be a protective factor.
|
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
), a longitudinal study of aging in men. The study has followed 2,280 men since the 1960s. The men enrolled in the study for the 1994–1997 wave of data collection were mailed a packet with the Health and Social Behavior Survey (HSB) the month before they reported for their triennial physical examinations. We used psychosocial data from the HSB; all questions prompted for responses within the past month. The physical data we used were from measurements taken during the physical examination. The 1994–1997 wave of the HSB was completed by 801 men, most of whom completed it the day prior to their exam. We excluded data from 93 men who reported a problem but did not report stress and coping data, and 154 men with missing data on one or more metabolic syndrome variables. Because diabetes is associated with abnormal glucose levels, we also excluded 36 men with diabetes, resulting in a final sample of 518 individuals. We made mean comparisons of the model variables between men included in our sample and those excluded for missing data by using the Bonferroni correction for multiple t tests. As we expected, men who were dropped for having diabetes had significantly lower HDL cholesterol levels t(610) = 2.59, p <.01, and they had significantly higher triglyceride levels, t(30.7) = –4.40, p <.05, fasting glucose levels, t(30.5) = –10.57, p <.001, postchallenge glucose levels, t(13.1) = –3.18, p <.001, waist-to-hip ratios, t(674) = –2.08, p <.05, and body mass indices, t(30.5) = –3.40, p <.001. However, the 100 men who were dropped for missing data on the lipid variables had significantly lower values on blood pressure, both systolic, t(660) = 2.17, p <.05, and diastolic, t(660) = 2.28, p <.05. The 16 men who were dropped for missing data on obesity had lower postchallenge glucose levels, t(15.4) = 1.91, p <.01. The ages of the included men ranged from 51–89 years (M = 68.87, SD = 6.87). Thus, the excluded men were both higher and lower on the physiological variables, depending on the exclusion criteria.
Measures
Stress
Participants were asked to identify the most serious problem or concern they had experienced in the past month. We assessed specific problems that had occurred during the past month because reporting was not likely to be influenced by memory problems. Comparable studies of coping in older adults use the specific problem approach (Aldwin, 1994). After the problem was identified, participants rated its stressfulness on a 7-point scale; 1 was "not troubled at all," whereas 7 was "the most troubled I've ever been."
Appraisal
After each participant had identified and rated his most serious problem, he was asked to endorse relevant appraisals from a list extended from the original list by Lazarus and Folkman (1984). There were five appraisals: threat, harm–loss, challenge, at a loss for what to do next, and annoyed or worried about others.
Coping
We used the California Coping Inventory (CCI) to assess coping. It lists 50 coping strategies that were specifically valenced toward positive and negative items. Respondents were asked to think of the problem they had reported and indicate whether they used each strategy to cope with it (0 = not at all to 3 = used a lot). Factor analyses in other samples have identified five factors (Aldwin, Shiraishi, & Cupertino, 2001). We used only the positive action coping and negative action coping factors in this study. These factors are characterized by emotional tone, and they are not equivalent to traditional approach–avoidance dichotomies. Appendix A lists the questions comprising these factors. The CCI subscales demonstrate high internal consistency, with Cronbach's alpha values of 
= 0.88 for the positive action and = 0.83 for the negative action factors (Aldwin et al.).
|
).
Obesity indicators
We computed body mass index as the ratio of weight to height in kg/m2. We measured abdominal circumference at the height of the umbilicus, and we measured hip circumference at the greatest protrusion of the buttocks. We computed the waist-to-hip ratio as the ratio of the measurements of abdomen to hip circumference in centimeters.
Lipids
We obtained triglycerides (TRG) and high-density lipoprotein cholesterol (HDL-C) values from serum samples drawn after the participants had fasted overnight. We assayed serum cholesterol enzymatically (SCALVO Diagnostics, Wayne, NJ). We precipitated the low-density lipoprotein cholesterol fraction with dextran sulfate and magnesium with the Abbot Biochromatic Analyzer 100 (Abbot Laboratories, South Pasadena, CA). We measured the HDL-C fraction in the remaining supernatant. We used the Dupont ACE discrete clinical analyzer (Dupont Company, Biomedical Products Department, Wilmington, DE) to measure TRG concentration.
Blood pressure
We averaged both systolic and fifth-phase diastolic blood pressure variables from measurements taken in the right and left arms of participants as they were in supine, sitting, and standing positions. We took measurements with a standard mercury sphygmomanometer with a 14-cm cuff, and we checked auscultatory systolic readings by using the palpatory method.
Glucose
We assessed fasting glucose when the men first came into the clinic after an overnight fast. We administered a 100-g challenge dose of glucose, and we drew a sample of blood 2 hr later to assess postchallenge serum glucose. We analyzed glucose levels on an autoanalyzer by using the hexokinase method (see Shen et al., 2003 for more details).
The TRG, HDL-C, postchallenge, and fasting glucose levels were highly skewed, so we used log transformations of these variables in the model.
Medication and smoking status covariates
Clinical interviewers gathered information on the use of antihypertensives, cholesterol-lowering medications, diuretics, and beta blockers on the day of the health examination and coded it by using the American Hospital Formulary Service Codes. They gathered information on self-report smoking status at the same time.
Analysis Plan
The analysis of field-level stress and coping data presents certain difficulties, because many people do not report problems (Aldwin, 1999; Coyne & Racioppo, 2000). Therefore, we created a dichotomous indicator for "have problem" and covaried it in all analyses. We included the 115 men with neither problem nor stress and coping data, but with complete data on other measures, in the analyses; we gave them a score of 0 for the stress and coping variables.
We conducted the main analyses in three stages. In the first stage we made sure that the second-order measurement model of the metabolic syndrome previously confirmed with NAS data collected at a different time point by Shen and colleagues (2003) adequately fit our data. In the second stage, we used the values of the weights from the eight observed variables to the four latent variables in that model to calculate four second-order metabolic syndrome variables (see Appendix B). This resulted in a weighted-sums model with four second-order metabolic syndrome variables, instead of eight observed variables loading onto four latent variables. We did this in order to maximize power in the mediation model. In the third stage, we linked the stress and coping variables to the metabolic syndrome to test if affect mediated between coping and the metabolic syndrome.
|
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
The men reported more positive than negative coping, paired t(517) = 26.56, p <.001. They also reported much more positive than negative affect, paired t(517) = 18.89, p <.001. See Table 2 for mean scores for the stress, coping, and affect variables.
|
). Both the average values of HDL-C and of systolic blood pressure were higher than the clinical cut off for diagnosis of the metabolic syndrome. Average values of the other metabolic syndrome variables in this sample approached, but did not meet, the clinical problem range. Table 3 lists descriptive statistics for the metabolic outcome variables and clinical criteria for diagnosis, where applicable.
|
Correlations Among Model Variables
Bivariate correlations among the psychosocial variables were largely as we expected (see Table 4). Having a problem and stress were both positively associated with all appraisal and coping variables. Stress was also positively associated with negative affect. Threat appraisal was associated with both positive and negative coping, whereas challenge appraisal was associated with positive coping and positive affect. Positive coping was strongly associated with negative coping and both positive and negative affect. Negative coping was associated with negative affect. As we expected, the metabolic syndrome variables were also intercorrelated. The only significant relationship between the psychosocial variables and the individual variables of the metabolic syndrome was the relationship between positive affect and glucose.
|
, 
2(16, N = 518) = 46.48, p <.00, Minimum Sample Discrepancy divided by the Degrees of Freedom (CMIN/DF) = 2.90, Comparative Fit Index (CFI) = 0.96, Root Mean Square Error Of Approximation (RMSEA) = 0.060. The weighted-sums model also provided adequate fit, 2(2, N = 518) = 5.41, p <.07, CMIN/DF = 2.70, CFI = 0.97, RMSEA = 0.057. Adding covariates to this model (age, hypertension medicine, diuretics, cholesterol-lowering medicine, beta blockers, medicine for diabetes, and current smoking status) yielded only two significant relationships, between the metabolic syndrome and blood pressure medication and cholesterol medications. Therefore, we included these medications as covariates in further analyses.
Mediation model
We then tested the hypothesized model. This model was not a good fit to the data, 2(81, N = 518), p <.00, CMIN/DF = 5.18, CFI = 0.74, RMSEA = 0.09. We deleted nonsignificant paths and added paths between positive and negative coping and from stress to negative affect to obtain adequate model fit, 2(78, N = 518) = 166.04, p <.00, CMIN/DF = 2.12, CFI = 0.93, RMSEA = 0.04. We then tested a final model that did not include nonsignificant paths from negative coping, negative affect, and positive affect to the metabolic syndrome. This model also fit the data adequately, 2(81, N = 518) = 164.05, p <.00, CMIN/DF = 2.02, CFI = 0.94, RMSEA = 0.04. The difference between the second and third models was not significant, 
2(3) = 1.9, ns. However, the third model was more parsimonious (Figure 2). Age was negatively related to the metabolic syndrome, ß = –.12, p <.01, and affect did not mediate between coping and the metabolic syndrome; the only significant path to the metabolic syndrome was the path from positive coping, ß = –.12, p <.05.
|
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
; Billings et al., 2000). In other words, coping would "get under the skin" by means of physiological responses associated with positive and negative affect. A first, it would appear that the results of the structural equation models did not support the hypothesis that affect would mediate between stress and coping processes and metabolic outcomes, as neither positive nor negative affect showed direct relationships with the metabolic syndrome. Rather, the relationship between the psychosocial variables and the metabolic syndrome was mediated by coping, which had a direct effect on the metabolic syndrome. Specifically, positive coping was inversely related to the metabolic syndrome, suggesting that it has protective effects. Further reflection suggests that these results do not rule out the possibility of affective mediation. The factors of the CCI, the measure of coping used in this study, were positively and negatively valenced for theoretical reasons. In particular, the positive action scale includes self-regulatory strategies, including taking time outs when needed, taking things one step at a time, being careful not to get overextended, and reminders to self not to get too upset about the problem. In contrast, the positive affect scale of the PANAS included items such as alert, inspired, active, strong, enthusiastic, active, attentive and excited. To the extent that positive coping is more reflective of emotion regulation than the PANAS measure is, then it is not surprising that there was a direct effect of positive coping on the metabolic syndrome rather than being mediated through these indicators of positive affect, which indicate arousal rather than calm. In this instance, a more appropriate measure of positive affect would be one focused on being calm and less excitable.
This fits in very nicely with earlier work in the NAS men showing that emotional stability is one of the strongest protective factors against the development of hypertension (Spiro, Aldwin, Ward, & Mroczek, 1995) and is associated with greater longevity (Aldwin et al., 2001). This would also reflect the calming effect of meditation practices on components of the metabolic syndrome such as blood pressure (see Levenson & Aldwin, in press, for a review).
The lack of direct links between general affect and physical health outcomes in this study might also be due to the chronic nature of the metabolic syndrome. The strongest links in the literature on psychosocial influences on health are found with very specific and proximal outcomes (Biondi & Picardi, 1999). Past studies of the influence of emotions on health outcomes tend to use measures, such as cortisol, cardiovascular activity, immune system variables, or self-report measures. Studies of the effect of coping on health have tended to use acute health outcomes, such as AIDS (Reed, Kemeny, Taylor, Wang, & Visscher, 1994) and cancer (Fawzy et al., 1990). However, the onset of these outcomes is rapid, relative to that of the metabolic syndrome. The fact that coping with a particular situation did affect this chronic condition suggests that the indicators of the metabolic syndrome may well fluctuate and reflect psychological states. To our knowledge, there are no studies examining the stability of the metabolic syndrome over time.
Positive and Negative Affective Pathways
The separate positive and negative pathways between stress and health highlight the importance of appraisal as put forth in Lazarus's process model of coping (Lazarus & Folkman, 1984). Challenge appraisals led to positive action coping, whereas threat appraisals led to negative action coping. Furthermore, coping mediated between appraisal and affect. The strong relationship between positive and negative coping indicates that the men were using multiple coping strategies to resolve their problems; it is well documented that increased stress leads to increased coping efforts (Aldwin, 1999). Studies reporting that challenge appraisals are associated with greater cardiac output than threat appraisals (Tomaka, Blascovich, Kelsey, & Leitten, 1993) also support this interpretation. It is also interesting that the influence of affect on health was through positive, and not negative, pathways. In studies of rheumatoid arthritis patients, Zautra (2003) has found that positive and negative affect influence health along separate pathways under conditions of low stress, but are highly inversely correlated in conditions of high stress. Thus, it is possible that negative pathways might also influence the metabolic syndrome in individuals experiencing severe stressors, such as caregiving.
The Issue of Age
Interrelationships among age and the stress, coping, and affect variables in this study are complex and are further complicated by the fact that this was a cross-sectional study, so relationships with age are confounded by cohort. Age was negatively related to the positively valenced variables in the model. This is in direct contrast to studies reporting positive relationships between age and positive affect (Mroczek & Kolarz, 1998). It has also been demonstrated that older adults report less affect in general (Aldwin, Sutton, Chiara, & Spiro, 1996). The negative relationships found between these variables in the present study may simply be due to a reporting bias on behalf of the older men in the sample toward less positive affect. However, this does not explain why the relationships among age and the negatively valenced variables were not significant. Negative emotions have been demonstrated to have a curvilinear relationship with age, declining in frequency until a person reaches the age of 60 (Carstensen, Pasupathi, Mayr, & Nesselroade, 2000). The nonsignificant relationships between age and the variables with negative emotional valence in the study may be attributable to the fact that we looked only for linear relationships.
Limitations of the Study
The main limitations of this study are related to time. The metabolic syndrome is a slowly developing outcome. More proximal measures of the stress response, such as endocrine or immune measures, might show a direct relationship to affect. Alternatively, a longitudinal study of stress, coping, and affective processes over time might be able to establish direct relationships between affect and the metabolic syndrome. The theoretical effect of stress is cumulative; disease results from the repeated inability to recover from stress (Dienstbier, 1989). Assessment of stress, coping, and affect variables within a one-month period might not accurately tap their long-term influences on health.
Another limitation concerns the homogeneity of the sample. The NAS sample is composed of White men who were selected for study participation for their geographic stability. There are race and gender differences in the prevalence of the metabolic syndrome (Cossrow & Falkner, 2004; Florez et al., 2005). Examing coping and affect as mediators between psychosocial processes and health outcomes in samples characterized by more ethnic diversity might yield different results. In addition to this, because there were some differences on the outcome variables between men included and excluded in the study, the data may have been systematically skewed. However, it is difficult to say in which direction, because the excluded men had lower blood pressure but higher glucose levels.
Suggestions for Future Research
Several promising areas of inquiry arise from the results of the present investigation. The first concerns the factor structure of the metabolic syndrome. It is not surprising that the measurement model of Shen and colleagues (2003) was replicated in this study. It should be noted that their analyses were also done with NAS data, albeit that those were collected at a different time point. Further work is needed on the structure of the metabolic syndrome in other samples. Another promising area arises from our confirmation of a model testing separate positive and negative mediators of stress and affect. The results of this study demonstrate that pathways among stress, coping, and affect variables are emotionally valenced, and that age influences these pathways. This approach to the study of stress and coping processes might be explored with longitudinal designs, which would allow for consideration of age and cohort effects on the development of chronic disease. Additional questions are related to the influence of positive affect on health. The finding that the influence of affect on health was through positive, and not negative, pathways was among the most noteworthy findings of this study. The study of positive influences on health is a promising area of research, yet very little attention has been devoted to how these influences might differ between clinical and healthy samples. Further examination of the differential effects of positive influences on health in various health contexts would provide interesting insights into the relationships among psychosocial variables and health outcomes.
Perhaps the most intriguing finding of this study was that positive coping mediated between stress and physical outcomes in healthy men. It may be that the ability to regulate emotions may be more important than the presence of positive emotions per se. Future studies are needed to clarify the direct effects of positive coping on health, as well as to examine the relative stability of the metabolic syndrome.
|
Footnotes |
|---|
Received for publication September 12, 2005. Accepted for publication March 23, 2006.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
