Changes in regression coefficients of psychosocial measures after adjustment for each potential mediating variable (behavior, body mass index [BMI], and diabetes mellitus). The base model controls for age, sex, race/ethnicity, site, use of medication, and recent infection. Subsequent models are adjustments of the base model, one variable at a time. All variables are categorical variables except age and BMI. Graphs show change in standard deviation (SD) units of each outcome associated with a 1-unit change in psychosocial predictor. CRP indicates C-reactive protein.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Ranjit N, Diez-Roux AV, Shea S, et al. Psychosocial Factors and Inflammation in the Multi-Ethnic Study of Atherosclerosis. Arch Intern Med. 2007;167(2):174–181. doi:10.1001/archinte.167.2.174
Copyright 2007 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2007
Psychosocial factors are associated with the development and progress of cardiovascular disease, but the pathological mechanisms remain unclear. We examined the associations of psychosocial risk factors for cardiovascular disease with concentrations of inflammatory markers among healthy adults and assessed the extent to which these associations are mediated by behaviors, body mass index (BMI), and diabetes mellitus.
This cross-sectional study used data from the baseline examination of the Multi-Ethnic Study of Atherosclerosis, a multisite study of 6814 men and women aged 45 to 84 years. Regression analyses were used to estimate associations of cynical distrust, chronic stress, and depression with serum levels of C-reactive protein, IL-6, and fibrinogen before and after adjustment for socioeconomic position, behaviors, BMI, and diabetes.
Higher levels of cynical distrust were associated with higher levels of inflammatory markers. The percentage differences (95% confidence intervals [CIs]) comparing the 80th and 20th percentiles of the scale were 7% (3%-11%) for IL-6; 9% (2%-16%) for C-reactive protein; and 1.3% (0.1%-2.4%) for fibrinogen. Higher levels of chronic stress were associated with higher concentrations of IL-6 and C-reactive protein. The percentage differences (95% CIs) comparing 2 and 0 ongoing stressful circumstances were 4% (1%-8%) for IL-6 and 5% (1%-11%) for C-reactive protein. Depression was positively associated with the level of IL-6 (percentage difference [95% CI] comparing the Center for Epidemiologic Studies–Depression Scale scores of ≥21 vs <21 was 7% [1%-14%]). Associations of psychosocial factors with inflammatory markers were reduced by 20% to 55% after adjustment for behavioral factors and by 45% to 100% after adjustment for BMI and diabetes, mostly owing to the effect of BMI. No associations remained after controlling for socioeconomic position, behaviors, BMI, and diabetes.
Psychosocial factors are associated with higher levels of inflammatory markers, most consistently for cynical distrust. Results are compatible with a mediating role of BMI, behaviors, and diabetes.
Numerous studies have documented associations of psychosocial characteristics with cardiovascular disease (CVD) incidence and mortality.1,2 Attention is turning increasingly to the investigation of the possible effects of psychosocial factors on the biological precursors of CVD in an attempt to understand the mechanisms through which these factors may be related to the development of CVD.3
Research has highlighted the importance of inflammation in the initiation and development of atherosclerosis and in the precipitation of cardiovascular events.4 Inflammation may be a mechanistic pathway linking psychosocial factors to CVD. Psychosocial factors may be associated with health behaviors predictive of inflammation, including smoking, physical activity, and alcohol intake.5-7 In addition, psychosocial factors are associated with obesity and insulin resistance, which are established correlates of systemic inflammation.8,9 Psychosocial factors may also be associated with proinflammatory changes in the autonomic nervous system and the hypothalamic-pituitary-adrenal axis. Cytokines and glucocorticoid hormones mediate the acute-phase response,10 and limited experimental evidence suggests that, like cortisol, inflammatory markers respond to acute psychological stressors.11
Depression, hostility, chronic stress, and job stress have been investigated in relation to inflammatory markers. Depression has been associated with elevations in IL-612,13 and C-reactive protein (CRP) levels,14 although these associations have not always been consistent.15 Hostility and chronic stress have been linked to higher IL-6 levels.16,17 High job stress has been associated with increased concentrations of fibrinogen.18 These studies have varied widely in size and sample characteristics and in their ability to control for confounding factors or to examine the factors mediating these associations.
Using data from the Multi-Ethnic Study of Atherosclerosis (MESA), we examined the extent to which a range of commonly assessed inflammatory markers are associated with known psychosocial risk factors for CVD in a large sample of healthy adults. We investigated the following 2 questions: (1) Are higher levels of cynical distrust, depression, or chronic stress associated with higher concentrations of inflammatory markers? (2) If so, to what extent are these associations independent of socioeconomic position (SEP), health behaviors, body mass index (BMI), and diabetes mellitus? We hypothesized that psychosocial risk factors would be positively associated with concentrations of inflammatory markers; that these associations would be partly reduced after adjustment for health behaviors, BMI, and diabetes, indicating that the associations of psychosocial factors with inflammation may operate through (or be mediated by) these factors; and that residual associations would persist after adjustment, indicating the possible role of other stress-related biological processes. The large and diverse MESA subject population, the range of psychosocial measures and inflammatory markers included in the study, and the detailed information on covariates make the MESA findings particularly suited to such an investigation.
The MESA is a 10-year longitudinal study with the goal of identifying risk factors for subclinical atherosclerosis. The cohort includes 6814 men and women aged 45 to 84 years at baseline who were recruited from 6 field centers using a variety of population-based approaches. Only persons free of clinical CVD at baseline were eligible. Details of the study design have been described elsewhere.19 The cohort is 38.5% white, 27.8% black, 21.9% Hispanic, and 11.8% Chinese. The baseline visit (on which these analyses are based) took place between July 2000 and September 2002. Cynical distrust was measured at a follow-up visit that occurred between September 2002 and February 2004.
Cynical distrust, depression, and chronic stress, the measures used herein, correspond to 3 distinct psychosocial domains that were measured in the MESA because of their links to CVD in prior work.12 Psychosocial risk factors were assessed using standardized questionnaires in English, Spanish, or Chinese; in all cases, higher scores reflect reports of more negative psychosocial characteristics (ie, characteristics that are expected to be associated with higher levels of inflammatory markers). Cynical distrust was based on an 8-item subset20 of the full Cook-Medley Hostility Scale that has been linked to atherosclerosis. Depression was assessed using the Center for Epidemiologic Studies–Depression Scale (CES-D) score.21 Chronic psychological stress was assessed using the Chronic Stress Scale,22 which consists of 4 items on ongoing difficulties in the domains of health of others, job or ability to work, finances, and relationships. Participants were coded as having a difficulty in the domain in question if they reported a moderately stressful or severely stressful ongoing difficulty; accordingly, the Chronic Stress Scale score ranges from 0 to 4. To investigate threshold effects, cynical distrust and depression were categorized into 4 groups based on the nearest approximate quartiles of the full distribution. The CES-D score was additionally dichotomized into higher than 21 and 21 or lower. A cutoff score of 21 on the CES-D has been proposed as a screening instrument for major depression.23 The Chronic Stress Scale was categorized into 4 groups based on the number of domains for which severe or moderate ongoing difficulties were reported.
Information on demographic characteristics (age, race/ethnicity, income, and education) and behaviors (physical activity, smoking, and alcohol use) was collected by means of a questionnaire. Physical activity was assessed using a questionnaire adapted from the Cross-Cultural Activity Participation Study.24 A composite measure of physical activity based on minutes and intensity of weekly activities was constructed and categorized into 4 levels. Smoking and alcohol use were categorized into current, past, or never. These measures were supplemented by information on pack-years of cigarette use. The BMI was calculated as the weight in kilograms divided by the square of height in meters and was modeled as a continuous variable. Diabetes (fasting glucose level of >125 mg/dL [>6.99 mmol/L] or use of hypoglycemic medication) and impaired glucose tolerance (fasting glucose level of 100-125 mg/dL [5.60-6.99 mmol/L ]) were defined according to the 2003 American Diabetes Association fasting criteria. Participants were categorized as having a recent infection if they reported a cold or flu, sinus infection, urinary tract infection, tooth infection, bronchitis, or pneumonia in the preceding 2 weeks. Medication use was defined as current use of lipid-lowering medications, nonsteroidal anti-inflammatory drugs, or, in the case of women, hormone therapy. All of these medications are known to alter the levels of inflammatory markers.
The inflammatory markers examined include IL-6, CRP, and fibrinogen. Standardized methods were used for collection, processing, and shipping of blood samples, with fasting morning phlebotomy and use of a central laboratory (Laboratory for Clinical Biochemistry Research, University of Vermont, Burlington). The IL-6 level was measured by an ultrasensitive enzyme-linked immunosorbent assay (R&D Systems Minneapolis, Minn); CRP and fibrinogen antigen levels were measured using the BNII nephelometer (N high sensitivity to CRP and N antiserum to human fibrinogen; Dade-Behring, San Mateo, Calif).
Inflammatory markers were log-transformed for analyses and back-transformed to the original metric (as geometric means) for purposes of presentation, for selected analyses. Correlation coefficients were estimated among the inflammatory markers and among psychosocial measures. Linear regression was used to estimate associations of logged inflammatory markers with each psychosocial measure after adjustment for potential confounders or mediators. Exponentiated coefficients from these models can be interpreted as relative differences in the amount of a given inflammatory measure associated with the variable of interest. For example, a relative difference of 1.10 implies that the outcome increases by 10% for each additional unit of the predictor. Interactions of race/ethnicity and sex with the psychosocial measures were tested. The associations of psychosocial factors with inflammatory markers (hypothesis 1) were investigated in a base model (model 1) adjusted for demographic characteristics (age, sex, race/ethnicity, and field center), recent infections, and medication use. The role of SEP, health behaviors, BMI, and diabetes (hypotheses 2 and 3) were investigated by adding different sets of covariates to the base model. Model 2 added measures for SEP (income and education categories). Model 3 adjusted the base model for health behaviors (smoking, alcohol use, and physical activity), whereas model 4 did the same with BMI and diabetes status. Model 5 included covariates in the base model, SEP, health behaviors, BMI, and diabetes. These models were supplemented by analyses examining the effect of adjusting for each variable separately. All reported P values correspond to 2-tailed tests and were considered significant at the .05 level. The study was approved by institutional review boards at all participating institutions. All subjects gave written informed consent.
A total of 6814 participants underwent the MESA baseline examination; 6035 to 6736 participants (depending on the psychosocial measure and the inflammatory marker examined) were included in analyses of psychosocial factors and inflammatory markers. Participants excluded from the analyses did not differ substantially from those included on psychosocial factors or inflammatory markers. Pearson correlations among the 3 log-transformed inflammatory markers ranged from 0.43 to 0.53. Psychosocial measures were weakly or moderately correlated with each other (Spearman correlations of ≤0.21 for cynical distrust and the other measures and 0.34 for the CES-D score and chronic stress). The internal consistency of the cynical distrust and scales was acceptably high across sex and race/ethnicity (Cronbach α, 0.71-0.79 for cynical distrust and 0.72-0.80 for CES-D score across race/ethnicity and sex groups). Patterning of psychosocial measures by age and sex was as similar across race/ethnicity groups. Means and ranges were comparable to those reported in other population samples, wherever such comparison was possible.
Selected characteristics of the sample at baseline are shown in Table 1 and Table 2. The mean age was 62.2 years; 52.9% were female and 38.5% were white. Mean (SD) values for the psychosocial variables were 2.7 (2.3) for cynical distrust, 7.6 (7.6) for the CES-D score, and 0.8 (1.0) for chronic stress. Cynical distrust and depression were inversely associated with income and education, and chronic stress was inversely associated with income. Physical activity was negatively associated with cynical distrust. The proportion of current smokers increased across levels of all psychosocial measures, and the proportion of current alcohol users was lower at higher levels of cynical distrust and CES-D score. Higher levels of psychosocial factors were associated with higher mean BMI and diabetes prevalence.
Psychosocial factors were positively associated with concentrations of inflammatory markers (Table 3). Higher levels of cynical distrust were associated with progressively higher levels of each of the inflammatory markers. Similar patterns were seen for chronic stress, with significant associations apparent for IL-6 and CRP levels and a weaker nonsignificant association for fibrinogen level. Having a CES-D score higher than 21 was associated with higher levels of all 3 inflammatory markers, although differences were only statistically significant for the IL-6 level. There was no clear evidence of a threshold effect in the association of the psychosocial measures with inflammatory markers, except possibly for IL-6 level at a CES-D score of about 20. Accordingly, subsequent analyses of the CES-D score dichotomized the measure at 21. There was no consistent evidence of differences in these associations by race/ethnicity or by sex. In exploratory analyses stratified by race/ethnicity, the effects were generally weaker or even reversed among Chinese subjects. The sample size in this group was small, however, and sensitivity analyses showed that the results were robust to the inclusion or exclusion of Chinese subjects in the analyses. Because patterns were largely similar across the sex and racial/ethnic groups, subsequent analyses are presented for the full sample and adjusted for race/ethnicity and sex.
Table 4 shows relative differences in inflammatory markers associated with high vs low levels of psychosocial characteristics. For cynical distrust, these levels correspond approximately to the 80th and 20th percentiles. For chronic burden, corresponding percentiles could not be obtained with the given distribution; we therefore report estimates for persons who reported 2 or more vs those who reported 0 ongoing difficulties, which approximately correspond to the 90th and 50th percentiles. The CES-D score was dichotomized into 21 or higher (5.3%) and lower than 21. Cynical distrust and chronic stress were positively associated with the IL-6 and CRP levels after adjustment for demographic factors, recent infection, and medication use (model 1 in Table 4), although the association was nonsignificant in the case of chronic stress and CRP. Cynical distrust was also positively associated with the fibrinogen level. Depression showed a positive association with the IL-6 level only. The impact of adjustment for income and education (model 2) was comparable to that of the adjustment for behaviors (model 3) (5%-45% reduction of coefficient size with adjustment for SEP and 20%-55% reduction with adjustment for behaviors). Associations were reduced by 45% to 100% after adjustment for BMI and diabetes mellitus (model 4). Adjustment for obesity and diabetes (model 4) was associated with greater attenuation of psychosocial associations than was adjustment for SEP or behavioral risk factors. No associations remained when SEP, behavioral risk factors, obesity, and diabetes were included (model 5).
The Figure shows the extent to which each potential mediator accounts for associations of psychosocial factors with inflammatory markers. Each behavior, BMI, and diabetes are added separately to the base model, which is adjusted for age, sex, site, race/ethnicity, medication use, and recent infection. Standardized scores (obtained by subtracting the mean from the observed value and dividing by the standard deviation) are used to represent inflammatory outcomes in these models to facilitate comparison of the models across outcomes. Overall, adjustment for behavioral risk factors did not lead to large changes in the coefficients, although adjustment for smoking led to a greater attenuation in the coefficient than did adjustment for alcohol use or physical activity. The largest reduction of the coefficients occurred with the adjustment for BMI. Adjustment for diabetes resulted in a change comparable to adjustment for the behavioral factors. These results were generally consistent across all of the psychosocial factors and outcomes examined.
In this cross-sectional analysis of adults aged 45 to 84 years and free of CVD, we found graded associations of psychosocial factors with inflammatory markers. The strongest and most consistent associations were observed for cynical distrust, which was positively associated with all 3 inflammatory markers. Chronic stress was positively associated with the IL-6 level and had a moderately positiveassociation with the CRP level, whereas depression was associated only with the IL-6 level. Although effect sizes appear small (eg, a 7% difference in the level of IL-6 and a 9% difference in the level of CRP, between the 80th and 20th percentiles of cynical distrust, after adjusting for demographic factors, recent infection, and medication use), the effect sizes are not negligible in comparison with other recognized correlates of inflammation. For example, a 7% relative difference in the IL-6 level was associated with a 1.5 difference in BMI, and a 9% relative difference in the CRP level was associated with a 1.1 difference in BMI, in models adjusted for demographic and lifestyle factors. Statins have been shown to reduce CRP concentrations by 13% to 15%.25
Our results are consistent with prior work linking hostility (as assessed using the full Cook-Medley Hostility Scale) and a variety of stress measures, including acute (eg, academic examinations) and chronic (eg, caregiving for patients with chronic illness) stressors, with higher concentrations of IL-6,16,26,27 although much of the previous work has used considerably smaller samples. The full Cook-Medley Hostility Scale was not available in the MESA, but a version of the cynical distrust subscale we used has been associated with subclinical CVD in prior work.20,28 Although the cynical distrust measure was obtained after testing for inflammatory markers, it is believed to be stable, and test-retest reliability coefficients in various studies exceed 0.80.29 The chronic stress measure we used has only a limited range, but its association with IL-6 and CRP levels is suggestive and deserves further investigation with more detailed measures of stress. We also observed a positive association of depression with the IL-6 level, but the association emerged only at CES-D scores higher than 21. The absence of associations of depression with CRP is consistent with a recent meta-analysis of 6 large studies of depression and CRP that concluded that the evidence of a relationship is weak.30 Our sample was limited to mostly healthy adults, and it is possible that this limited our ability to examine associations of severe or major depression with inflammation.
Only cynical distrust was associated with fibrinogen. The fibrinogen level was 1.3% higher (equivalent to approximately 4 mg/dL [0.12 μmol/L]) for persons at the 80th compared with the 20th percentile of cynical distrust. Although a 1.3% difference appears small, it is equivalent to the difference in fibrinogen associated with a 1.7–index value increase in BMI in the fully adjusted model and to half of the difference observed between current smokers and never smokers (7.8 mg/dL [0.23 μmol/L]). Almost all studies reporting positive relationships of psychosocial factors with fibrinogen have focused on job-related stressors18 and, even then, associations have not always been consistent.31 The literature regarding depression and fibrinogen is scant and findings are mixed.15,32 To our knowledge, no studies to date have examined the association of fibrinogen with cynical distrust.
In all cases, associations were reduced after adjustment for behavioral factors. Although it is impossible to determine the causal relationships between these factors from cross-sectional analyses, our results are compatible with the hypothesis that at least part of the association of psychosocial factors with inflammatory markers is due to the mediating role of behaviors; ie, psychosocial factors are causally associated with uptake of behaviors such as smoking that result in increased concentrations of inflammatory markers. Significant weakening of associations after adjustment for BMI and diabetes are compatible with 1 or both of 2 possible explanations. Body mass index and diabetes may mediate the associations between psychosocial factors, or they may confound these associations. Evidence consistent with a mediating role comes from studies showing that stress and psychosocial factors may be causally related to obesity and associated metabolic disturbances,33 possibly through mechanisms involving the hypothalamic-pituitary-adrenal axis and the autonomic system.34 An alternative explanation is that obesity and diabetes are the underlying causes of both adverse psychosocial profiles and increased inflammation through separate mechanisms. If this is true, BMI and diabetes are confounders (rather than mediators) of the association between psychosocial factors and inflammation. Both processes may operate at the population level. We found no evidence that associations of psychosocial factors with inflammatory markers persisted after adjustment for behaviors, BMI, and diabetes. Thus, in our data there is no evidence that mechanisms involving other factors play a role in the associations we observed. Although we also report analyses after adjustment for SEP, we do not view SEP as a mediator of psychosocial effects on health. Rather, SEP is likely to be an antecedent to psychosocial characteristics, as well as to other factors (eg, behaviors, BMI, and diabetes) associated with inflammation. The reduction of associations after adjustment for SEP may simply reflect associations of SEP with these factors.
Psychosocial factors are difficult to measure and may change over time. Thus, measurement error (plus the fact that our measures refer to a single point in time) may have resulted in important bias toward the null. Exclusion of persons with clinical CVD (which may be associated with psychosocial factors and inflammation) could also have resulted in underestimates of the associations, although this exclusion also eliminated the possibility of confounding by disease status.
In summary, we found that cynical distrust, depression, and chronic stress are associated with higher concentrations of inflammatory markers in a population-based healthy sample. Our cross-sectional results are compatible with a mediating role of behaviors, BMI, and diabetes mellitus, although definite conclusions regarding mediation can only be drawn from longitudinal designs.
Correspondence: Nalini Ranjit, PhD, Center for Social Epidemiology and Population Health, University of Michigan, 1214 S. University, Ann Arbor, MI 48104 (firstname.lastname@example.org).
Accepted for Publication: October 11, 2006.
Author Contributions:Study concept and design: Diez-Roux, Shea, and Ni. Acquisition of data: Diez-Roux, Shea, Cushman, and Ni. Analysis and interpretation of data: Ranjit, Diez-Roux, Shea, Seeman, Jackson, and Ni. Drafting of the manuscript: Ranjit and Diez-Roux. Critical revision of the manuscript for important intellectual content: Diez-Roux, Shea, Cushman, Seeman, Jackson, and Ni. Statistical analysis: Ranjit, Diez-Roux, Shea, and Ni. Obtained funding: Diez-Roux and Shea. Administrative, technical, and material support: Diez-Roux, Shea, Cushman, and Ni. Study supervision: Diez-Roux and Shea. Substantive content interpretation: Seeman.
Financial Disclosure: None reported.
Funding/Support: This study was supported by grants 1R01HL076831 from the National Heart, Lung, and Blood Institute and R24HD047861 from the National Institute of Child Health and Human Development. MESA was supported by contracts N01-HC-95159 through N01-HC-95165 and N01-HC-95169 from the National Heart, Lung, and Blood Institute.
Acknowledgment: We thank the other investigators, the staff, and the participants of the MESA study for their valuable contributions.
Create a personal account or sign in to: