Trait Anger, Cynical Hostility and Inflammation in Latinas: Variations by Anger Type?

Cardiovascular disease (CVD) is the leading cause of mortality within the U.S., and among Latinos, the largest U.S. ethnic minority group (American Heart Association, 2010). Mexican-American women in particular have rates of CVD only slightly lower than non-Latino white women (30.9% versus 33.8%), and rates of coronary heart disease (CHD) that are comparable [5.6% versus 5.8%; (Roger et al., 2011)]. Although inroads have been made [e.g., INTERHEART (Rosengren et al., 2004), Multi-Ethnic Study of Atherosclerosis (Bild et al., 2002)], knowledge about psychosocial factors related to CVD in Latinos remains limited. Anger and hostility appear to play a salient role in CVD risk and progression (Smith et al., in press; Smith et al., 2004); however, their association with CVD in Latinos remains relatively unexplored.

Anger has been defined as “an unpleasant emotion ranging in intensity from irritation or annoyance to fury or rage” (Smith et al., 2004), and can be conceptualized from a trait or state perspective. Hostility is a trait-like cognitive characteristic described as “an expectation that others are likely sources of wrong-doing, a relational view of opposition toward others” (Smith et al., 2004). Consequently, anger is viewed as being emotional in nature, whereas hostility, and in particular, cynical hostility, the dimension that involves mistrust of others and their perceived selfish motives, is considered to have a greater cognitive or attitudinal component (Sirois and Burg, 2003).

Reviews of the relationship between anger and CVD have found that anger is predictive of CVD morbidity in males and females (Sirois and Burg, 2003; Smith et al., 2004; Steptoe and Brydon, 2009), with one prospective 7- year study reporting a 2.66 times increased risk of coronary events in men reporting the greatest versus the lowest levels of anger (Kubzansky and Kawachi, 2000). Similarly, hostility has been shown to have a consistent association (r=.18) with coronary heart disease (CHD) outcomes in males and females (Miller et al., 1996a), with one study reporting twice the risk of a recurrent myocardial infarction (MI) in women with CHD in the highest versus the lowest quartile of hostility (Chaput et al., 2002), and has also been associated with greater CVD mortality (Matthews et al., 2004; Smith et al., 2004). Cynical hostility, specifically, has been linked to CVD risk markers including visceral adipose tissue, carotid intimal-medial thickness, and the metabolic syndrome, as well as CHD incidence and mortality (Lewis et al., 2009; Nelson et al., 2004; Raikkonen et al., 1999; Tindle et al., 2009).

It is unclear, however, whether these associations are similar across gender groups, as the majority of research has not disaggregated effects for men versus women (Kubzansky and Kawachi, 2000; Miller et al., 1996b). A recent meta-analysis (Chida and Steptoe, 2009) that evaluated anger and hostility in aggregate found that hostility-related elevated risk was 22% in healthy males but nonsignificant in healthy females at 7%; however, only 5 of the 21 studies reviewed contained information on females specifically, and the authors did not report results for research in diseased samples as too few studies evaluated gender differences. There is some evidence that differential relationships with CVD exist by gender, especially for anger and its dimensions (Culic et al., 2005; Davidson and Mostofsky, 2010; Harburg et al., 2003; Matthews et al., 1998; Player et al., 2007; Siegman et al., 2000; Williams et al., 2007).

Moreover, the majority of research has focused on non-Latino white samples (Miller et al., 1996a). A few studies examining relationships between anger, hostility and CVD outcomes have reported stronger effects among non-Latino blacks when compared with non-Latino whites (Magai et al., 2003; Markovitz et al., 1997; Tindle et al., 2009; Williams et al., 2007). To our knowledge only one study has evaluated anger and hostility as CVD risk factors among Latinos, and found that the dimension of “anger-out” was related to left ventricular mass, but only in females (Gleiberman et al., 2008). Consequently, additional work evaluating how these psychosocial factors and their dimensions relate to CVD risk in Latino women is greatly needed.

One mechanism by which anger and hostility may affect CVD risk is through the promulgation of systemic inflammation. It is believed that chronic anger and hostility lead to overstimulation of sympathetic-adrenal-medullary (SAM) and hypothalamicpituitary-adrenal (HPA) axis systems, triggering the frequent release of catecholamines and cortisol. Catecholamines are putative catalysts of systemic inflammatory processes (Kubzansky and Kawachi, 2000; Sirois and Burg, 2003; Smith and Ruiz, 2002), and cortisol, while acutely anti-inflammatory, is thought to result in increased inflammation with chronic secretion (Black, 2006). Continuous cortisol exposure is purported to lead to adrenal organ exhaustion and glucocorticoid receptor resistance, thereby causing disruption of the glucocorticoid anti-inflammatory feedback system, and increased inflammation; in addition, chronic cortisol secretion is believed to lead to increased susceptibility to secondary inflammations through down-regulation of the immune response (Hansel et al., 2010). Although causality is debated (e.g., Danesh and Pepys, 2009; Rattazzi et al., 2003), inflammation is thought to be involved in the creation of atherosclerotic plaques through promotion of smooth muscle proliferation at the site of endothelial dysfunction, and initiation of macrophage and lymphocyte accumulation, leading to plaque growth, instability and rupture (Lind, 2003; Ross, 1999). Intercellular adhesion molecule-1 (ICAM-1), which can be assessed in circulation in its soluble form (sICAM-1), promotes recruitment of inflammatory cells to the endothelium by facilitating firm leukocyte adhesion to the endothelial wall, leading to the initiation of pro-inflammatory signaling cascades (Keaney et al., 2004; Lawson and Wolf, 2009). Higher sICAM-1 has been linked to increased risk of CVD (Hajilooi et al., 2003; Jin et al., 2009; Luo et al., 2004; Shai et al., 2006; Tanne et al., 2002).

Previous studies of relationships between anger and/or hostility and inflammation (Coccaro, 2006; Graham et al., 2006; Marsland et al., 2008; Miller et al., 2003a; Stewart et al., 2008; Suarez et al., 2004) have focused on other inflammatory markers (e.g., interleukins, Tumor Necrosis Factor Alpha, C-Reactive Protein [CRP]), and not ICAMs, which represent a separate component of the inflammatory pathway and may be early indicators of endothelial damage (Blann et al., 2002; Kop, 1999). To our knowledge, only one study to date has considered anger in relation to sICAM-1 reactivity, and found no association; however, trait anger was not measured, and state anger was assumed to be provoked by a laboratory task (Kop et al., 2008).

The purpose of the current study was to investigate whether greater levels of cynical hostility and trait anger are associated with higher levels of sICAM-1 in Mexican-American women, an understudied population at high risk for CVD. As well as considering overall levels of trait anger, the study examined anger temperament and anger reaction subdimensions to determine whether their relationships with sICAM-1 differed. Given the conflicting nature of results thus far regarding the relationship of anger and cynical hostility to inflammation and the dearth of research in Latinos, we do not advance any a priori hypotheses regarding expected direction of effects.

Methods

Participants and Recruitment

Participants were recruited using simple random sampling as part of a larger study on psychosocial factors and CVD risk in middle-aged Mexican-American women living near the San Diego, CA - Mexico border. Participants were identified from a list obtained through a commercial database that used public records to gather information such as names, addresses, and general demographics for a particular region. Potential participants were randomly selected from households with Latino surnames that had a female occupant in the targeted age range. They were subsequently contacted through targeted direct mailings and phone solicitation, with the goal of obtaining a sample representative across levels of socioeconomic status. Eligible participants were female, between the ages of 40–65, Mexican-American (selfidentified), residing in selected recruitment areas, and free of cardiovascular diseases, diabetes, kidney disease, chronic inflammatory conditions, or other major illnesses, pregnancy, or medications with autonomic nervous system effects. Six-hundred and fifty-six women were screened, 363 (55.3%) were deemed eligible, and 321 (88.4%) participated in some or all portions of the study. Of these, 17 did not complete the physical exam or blood draw and 10 had incomplete data on covariates, resulting in a final sample of N=294 for the current study.

Procedures

After enrollment, participants were scheduled for two home visits by study staff. Participants were given instructions over the phone to avoid the use of anti-inflammatory medications for at least 48 hours, strenuous exercise and alcohol consumption for at least 24 hours, and caffeine and tobacco consumption for at least 30 minutes prior to the physical exam (i.e., second visit). Participants were rescheduled if they reported any acute illness. During the first visit, participants were consented and given a battery of questionnaires to complete (in Spanish or English) assessing sociodemographic characteristics, behavioral risk factors, health history, and psychosocial variables. They were also given instructions for a 12-hour, overnight fast. During the second visit, a licensed phlebotomist obtained a fasting blood draw, and a research assistant performed physical measurements (blood pressure, height, weight, waist circumference). Research staff was trained in all study procedures. The SDSU and UCSD institutional review boards approved all study procedures.

Anger and Cynical hostility

Anger was assessed using the Spielberger Trait Anger Scale, a 10-item self-report measure of trait anger (Speilberger et al., 1983), comprised of two subscales, anger temperament and anger reaction. Anger temperament is the predisposition toward quick, unprovoked or minimally provoked anger. Anger reaction refers to anger aroused in response to frustration, criticism, or unfair treatment. For each item, participants marked a response ranging from (0) strongly disagree to (4) strongly agree. Cynical hostility was measured using an abbreviated 6-item version of the 13-item Cook-Medley cynicism subscale, a scale that measures negative beliefs and attitudes toward others (Barefoot et al., 1989). This abbreviated version has been used in other studies on mechanisms of CVD risk (Janicki-Deverts et al., 2010; Schulz et al., 2008). For each item, participants indicated whether they felt the item to be true of them (1) or false (0), resulting in a score from 0 to 6. Cronbach’s alpha reliability coefficients were acceptable for the overall trait anger (.82), anger reaction (.67), and anger temperament (.81), as was the Kuder-Richardson 20 coefficient for the cynical hostility scale (.72). Previously validated Spanish versions of the cynical hostility (Butcher et al., 2007) and anger measures were utilized (Mind Garden, n.d.).

Soluble Inter-cellular Adhesion Molecule-1, Lipids, Leukocytes and HbA1c

Circulating levels of sICAM-1, lipids, and leukocytes were obtained from a venous blood draw using either EDTA as a preservative or serum. Samples for sICAM-1 were immediately put on ice and centrifuged and the plasma stored at −70 degree C until assay. Levels of sICAM-1 were determined at the University of California San Diego (UCSD) Clinical Biomarker Laboratory in non-freeze thawed plasma samples by commercial Enzyme-Linked Immunosorbent Assay (MSD, Gaithersburg, Maryland). Inter- and intra-assay coefficients of variation were <10%.

Serum low-density lipoprotein cholesterol, high-density lipoprotein cholesterol (HDL-c,) and triglycerides were measured by modified enzymatic methods (intra and inter assay CV <3%) on AU5400 random access analyzer, following standards set by the Lipid Standardization Program of the Centers for Disease Control and Prevention (Warnick, 2000). Hemoglobin A1c (HbA1c), a measure of average blood glucose concentration, was assayed using a Biorad Diomat high-pressure liquid chromatography analyzer. Complete blood count, including leukocytes, and differential counts were measured by flow Cytometry method (precision < 6% red blood cell count and varied in analytes) on a Beckman Coulter LH750 Hematology Analyzer, with improved cell enumeration accuracy, equipped with AccuCount technology.

Covariates

Analyses controlled for demographic, behavioral and biological factors associated with anger, hostility, or inflammation in prior research that could represent potential confounding influences or biobehavioral pathways from anger/hostility to inflammation, and that have been controlled for in studies of anger/hostility and inflammation (Fortmann et al., 2004; Graham et al., 2006; Marsland et al., 2008; Stewart et al., 2008; Suarez et al., 2004)¹. Demographic covariates included age, nativity (U.S. or Mexico-born), and socioeconomic status (SES). SES was assessed by education (highest level of education achieved, from no education to a doctoral or professional degree, recoded into six categories for analysis) and yearly income (measured on an ordinal scale, ranging from less than $6000/year to more than $96000/year, recoded into six categories for analysis). Income was imputed using the expectation-maximization algorithm method (Dempster et al., 1977) for 5 participants with missing data using age and education as predictors. Behavioral covariates consisted of current smoking and alcohol use (both coded yes/no). Exercise frequency and intensity was assessed using the Leisure Time Exercise Questionnaire (Godin and Shephard, 1985), which asks respondents to state the number of times during a typical week they participated in strenuous or moderate exercise for at least 15 minutes, and provides an estimate of total Metabolic Equivalent of Task Units (METs) per week (an estimate of the intensity and energy expenditure of physical activity that is comparable across differing body sizes). For the Spanish language version of the scales, translation was conducted following a standard protocol which involved forward/back translation by trained bilingual staff, review for semantic equivalence by an expert committee, pilot-testing of items in bilingual and monolingual focus groups, and evaluation of equivalence of reliability and validity of scales by language version.

Biological covariates included body mass index (BMI), which was calculated using the standard algorithm [i.e., weight (kilograms)/height (meters)²]. Systolic and diastolic blood pressure (SBP and DBP) readings were taken while the participant was in a seated position, with arm elevated to heart height, following 30 minutes of rest, using an automatic sphygmomanometer shown to be valid and reliable (i.e., Omron HEM 705-CP, Omron Corporation, Kyoto Japan). For parsimony, mean arterial pressure (MAP; 2/3*DBP + SBP/3) was used to indicate average blood pressure. Lipids, which have been associated with both hostility/anger and sICAM-1 (Bunde and Suls, 2006; Dujovne and Houston, 1991; Hsu et al., 2009; Shai et al., 2006; Suarez et al., 1998; Thorand et al., 2006), and HbA1c were also included as covariates. To account for the effects of acute infection on sICAM-1 levels, leukocyte count was included as a control variable, as in previous research (e.g., Alley et al., 2006).

Statistical Analyses

All data were analyzed using Predictive Analytics SoftWare, Version 17.0 for Macintosh (PASW, Chicago, Illinois, USA). Model assumptions about linearity, normality, independence, and homoscedasticity of errors were assessed graphically and analytically and were adequately satisfied. Ordinary Least Squares (OLS) hierarchical multiple linear regression was used to test the relationship between each predictor and sICAM-1. For each predictor (i.e., anger, anger temperament, anger reaction, cynical hostility), a model adjusting only for demographics (i.e., age, nativity, SES) was first estimated (Model 1). Biological variables (i.e., BMI, MAP, lipids, HbA1c, leukocyte count) that represented possible mediators or confounds of the relationship between the predictors and sICAM-1 were then added to Model 1 (Model 2). In addition to covariates included in Models 1 and 2, the final model (Model 3) included health behaviors that could fall in the causal pathway (i.e., smoking, alcohol use, exercise) between anger and CVD, to test the direct association of predictors with sICAM-1 independent of these intermediate mechanisms. These models were estimated for each predictor separately, and a final set of models was estimated entering all predictors simultaneously with the exception of the full anger scale, which was redundant with the anger subscales.

All non-categorical variables were centered about the sample mean to increase the interpretability of coefficients. Due to significant skew, exercise was categorized into quintiles for the purpose of analysis. Although the anger and cynical hostility scales were positively skewed, results using log-transformed versions were not appreciably different than those without transformation; consequently, for ease of interpretability and brevity, only analyses using non-transformed predictors are presented below.

Results

Table 1 displays means and standard deviations or frequencies and N’s, for all demographic, behavioral, and biological covariates as well as predictor and outcome variables. Correlations between anger scales, cynical hostility, sICAM-1 and all covariates are also listed. Means of anger (M= 15.63, SD= 4.54), anger temperament (M= 6.17, SD= 2.32), anger reaction (M= 6.76, SD= 2.30) and cynical hostility (M= 2.29, SD= 1.80) were below the mid-point of each scale. The means for the trait anger and anger reaction scales were also less than equivalent normative sample means for normal (i.e., non-psychiatric) women aged 30 years and older from a variety of occupational backgrounds ranging from professional and managerial to clerical as well as undergraduate and graduate students (17.01, 8.46, respectively; Speilberger, 1999), whereas the anger temperament mean was somewhat higher (5.71). Average age of the current sample was 49.66 years, and the majority had at least some college (53.39%). The mean value of sICAM-1 (274.97) was comparable to levels reported in non-Latino whites but higher than those reported for non-Latino blacks in other female-only (Miller et al., 2003b) and mixed gender samples (Bielinski et al., 2008; Wong et al., 2006). This trend is consistent with prior research that has shown Latinos to have sICAM-1 levels comparable to those of non-Latino whites, but higher than those of non-Latino blacks (Albert, 2007; Albert et al., 2007).

Table 1.

Descriptive statistics (means, standard deviations or frequencies, percentiles) for all variables, and bivariate correlations between anger, anger subscales, cynical hostility, sICAM-1, and covariates

Variable	M(SD)/N(%)	Anger r	Tempa r	Reactb r	Hostilityc r	sICAM-1 r
Age (years)	49.66 (6.53)	−0.07	−0.08	−0.02	0.01	0.20**
Educationd,e		−0.07	−0.13*	−0.02	−0.18**	−0.18**
≤8th grade	47 (15.99)
Some high school	54 (18.37)
GED/high school diploma	36 (12.24)
Some college	93 (31.63)
4 year degree	46 (15.64)
Graduate degree	18 (6.12)
Yearly Incomed,e		−0.10	−0.10	−0.04	−0.09	−0.25**
<$15000	30 (10.20)
$15000–$25000	40 (13.60)
$25000–$34999	50 (17.01)
$35000–$49999	55 (18.71)
$50000–$74999	56 (19.05)
≥$75000	63 (21.43)
Nativityd
% Born in U.S.	76 (25.85)	−0.13*	−0.12*	−0.14*	−0.09	0.02
Body Mass Index	29.03 (6.39)	0.01	0.04	−0.02	0.01	0.21**
Mean Arterial	87.85 (9.82)	−0.02	−0.02	−0.00	0.02	0.22**
Pressure (mmHg)
HDL-c (mg/dL)	56.85 (15.30)	0.02	0.04	−0.01	0.05	−0.07
LDL-c (mg/dL)	121.33 (32.12)	0.02	−0.01	0.05	0.03	−0.00
Triglycerides (mg/dL)	130.45 (61.84)	−0.06	−0.08	−0.03	−0.01	0.16**
Hemoglobin A1c (% of hemoglobin)	5.73 (0.66)	−0.03	−0.02	−0.03	0.09	0.25**
Leukocyte Count (×103/L)	5.94 (1.57)	−0.06	−0.11	0.00	0.06	0.04
Current Smokere	25 (8.50)	−0.05	−0.05	−0.03	−0.03	0.09
Current Drinkere	169 (57.48)	0.09	0.05	0.07	−0.10	−0.12*
Exercise (METs/week)	19.70 (20.57)	−0.03	−0.03	−0.03	−0.08	−0.05
Anger	15.63 (4.54)	-	-	-	-	-
Temperament	6.17 (2.32)	0.86**	-	-	-	-
Reaction	6.76 (2.30)	0.84**	0.49**	-	-	-
Cynical hostility	2.29 (1.80)	0.26**	0.25**	0.21**	-	-
sICAM-1 (ng/mL)	274.97 (93.84)	0.04	−0.01	0.07	0.14*	-

^*p<.05,

^**p<.01

^aAnger Temperament

^bAnger Reaction

^cCynical Hostility

^dN (%)

^eSpearman rank-order correlation

Of all covariates, only nativity and education were significantly related to any of the anger scales. Individuals who were born in the U.S. had lower trait anger (r=−0.13), anger temperament (r=−0.12), and anger reaction (r=−0.14) scores than those who were born in Mexico (all ps < .05). Higher education was related to lower anger temperament (r=−0.13), and was the only covariate significantly (inversely) associated with cynical hostility (r=−0.18). SICAM-1 was significantly related to all covariates with the exception of nativity, lipids, smoking, exercise, and leukocyte count. Relationships were in the expected direction with greater age, less education and income, and greater values on CVD risk factors (i.e., BMI, MAP, triglycerides, HbA1c) being associated with higher sICAM-1 levels. One exception was alcohol consumption, wherein lower sICAM-1 values were seen in drinkers (r=−0.12).

Anger, Cynical hostility and sICAM-1

Results of the primary analyses are presented in Table 2. As a set, covariates accounted for 19.2% of the variance in sICAM-1 (p<.001).

Table 2.

Results of multiple linear regression analyses regressing sICAM-1 separately on anger, anger subscales, and cynical hostility, and simultaneously on anger subscales and cynical hostility, adjusting for demographic (Model 1), biological (Model 2), and behavioral (Model 3) covariates

	Model 1		Model 2		Model 3
Covariates	B (S.E.)		B (S.E.)		B (S.E.)
Demographic
Age	2.86** (0.80)	0.20	2.66** (0.86)	0.18	2.65** (0.86)	0.18
Education	−7.09§ (3.99)	−0.11	−5.56 (4.00)	−0.09	−5.48 (4.05)	−0.09
Income	−12.83** (3.65)	−0.22	−10.21** (3.65)	−0.18	−9.48* (3.72)	−0.16
Nativitya	28.95* (12.53)	0.13	24.49* (12.40)	0.11	23.81§ (12.46)	0.11
Biological
Body Mass Index			1.29 (0.92)	0.09	1.31 (0.92)	0.09
Mean Arterial Pressure			0.89 (0.58)	0.09	0.93 (0.59)	0.10
HDL-c			0.10 (0.39)	0.02	0.08 (0.40)	0.01
LDL-c			−0.34* (0.17)	−0.12	−0.33§ (0.17)	−0.11
Triglycerides			0.10 (0.10)	0.07	0.10 (0.10)	0.06
Hemoglobin A1c			19.60* (8.28)	0.14	19.31* (8.32)	0.14
Lymphocyte Count			0.48 (3.38)	0.01	−0.31 (3.43)	−0.01
Behavioral
Current Smokerb					28.25 (18.52)	0.08
Current Drinkerb					−6.91 (10.82)	−0.04
Exercise					0.90 (3.68)	0.01
Predictors
Angerc	0.92 (1.16)	0.04	1.24 (1.14)	0.06	1.42 (1.15)	0.07
Anger Temperamentc	−0.41 (2.28)	−0.01	−0.01 (2.26)	0.00	0.26 (2.27)	0.01
Anger Reactionc	3.33 (2.26)	0.08	3.94§ (2.21)	0.10	4.20§ (2.22)	0.10
Cynical Hostilityc	5.94* (2.91)	0.11	5.72* (2.87)	0.11	5.89* (2.89)	0.11
Anger	−3.39 (2.61)	−0.08	−3.28 (2.59)	−0.08	−3.07	−0.08
Temperamentd	4.05 (2.59)	0.10	4.65§ (2.54)	0.11	(2.60)	0.12
Anger Reactiond	5.85§ (3.00)	0.11	5.43§ (2.96)	0.10	4.77§ (2.54)	0.11
Cynical Hostilityd					5.50§ (2.98)

^§p< .10

^*p< .05,

^**p<.01

^aMexico Born (0), U.S. Born (1),

^bNo (0), Yes (1),

^cPredictors entered in separate models,

^dPredictors entered simultaneously

Neither overall trait anger nor anger temperament was significantly associated with sICAM-1 in any model (all ps >.05). Anger reaction was not related to sICAM-1 in Model 1 (model adjusted for demographics only), but the association approached statistical significance in Model 2 (model adjusted for demographic and biological factors) and Model 3 (fully adjusted model). Those who endorsed greater levels of anger reaction tended to have higher sICAM-1 levels (β =4.77, p=.06, in Model 3). Cynical hostility was significantly related to sICAM-1 in all models (all ps <.05), with greater cynical hostility associated with higher sICAM-1 (β = 5.89, p=.04, in Model 3). The significant linear trend for cynical hostility is depicted in Figure 1. When both anger dimensions and cynical hostility were entered simultaneously, the associations for cynical hostility were attenuated in all models, and became marginally significant (β =5.50, p=.07, in Model 3), indicating that the anger and cynical hostility constructs overlap to some extent in their relationship with sICAM-1. No other associations notably differed from models that considered each variable individually. In aggregate, the cynical hostility, anger reaction, and anger temperament scores accounted for 2.20% of the variance explained in sICAM-1 (p=.05), after accounting for all covariates.

Figure 1.

Mean adjusted values of sICAM-1 by cynical hostility quintile level (quintiles displayed for purposes of illustration; full continuum presented in analyses).

Discussion

Mexican-Americans, and Mexican-American women in particular, have been identified in some studies as having risks for CVD mortality higher or equivalent to their non-Latino white counterparts (Hunt et al., 2003; Pandey et al., 2001). The current study is, to our knowledge, the first to examine the association between overall trait anger, anger dimensions and cynical hostility with sICAM-1, in this subpopulation. As a set, anger and cynical hostility accounted for just over 2% of the variance in sICAM-1. Although not large, this effect is comparable to those found in other studies of anger/hostility and markers of inflammation in non-Latino samples (e.g., Miller et al., 2003a; Stewart et al., 2008).

Interestingly, although overall trait anger was not associated with sICAM-1 in the current sample, cynical hostility was. Associations between overall hostility and inflammation appear to vary according to which marker or combination of markers is examined, with various studies reporting positive, negative, and nonsignificant relationships between hostility and IL-6, CRP, and TNF-a (Brummett et al., 2010; Graham et al., 2006; Kiecolt-Glaser et al., 2005; Mommersteeg et al., 2008; Suarez et al., 2004; Suarez et al., 2002). With the exception of the current study there was no prior research to our knowledge demonstrating associations between any dimension of hostility and sICAM-1. Additional work examining other inflammatory markers representing various stages in the inflammatory process (e.g., serum amyloid A, interferon-gamma) would further bolster evidence of an overall link between hostility and inflammation.

Relative to hostility, trait anger has received less attention in the literature as it relates to inflammation. A few studies have found relationships between anger and CRP (Hapuarachchi et al., 2003; Suarez, 2004). It is possible that differences in anger associations with inflammation across studies may be due to unrecognized moderating effects, as interactions have been identified with factors as variant as hypertension status (Williams et al., 2001), multivitamin supplement use (Suarez, 2003), and level of insulin resistance (Suarez et al., 2006). Findings in the literature are equivocal, and may indicate the need for further identification of potential moderating factors.

Whereas overall trait anger did not relate to inflammation in the current study, there was a trend for higher levels of anger reaction, but not anger temperament, to relate to elevated sICAM-1 levels. Anger reaction trends were not present in demographic-only controlled models, but became evident in models including biobehavioral covariates, which is likely the result of a reduction in error variance in sICAM-1 with the addition of these controls. Previous work on dimensions of anger or hostility in relation to inflammation has produced varying results based on dimension type. For example, anger-out but not anger-in was positively related to CRP levels in an Australian sample, and anger control (tendency to control expressions of anger) was associated with lower levels of CRP (Hapuarachchi et al., 2003). A study of the relationship between the cognitive, affective, and behavioral components of hostility and IL-6 and CRP found that only the behavioral component was significantly related to inflammatory markers (Marsland et al., 2008); however only the cognitive component was associated with TNF-α and interferon gamma in another study (Janicki-Deverts et al., 2010). Thus, it is apparent that differences across dimensions exist, although no clear pattern has emerged. With further research, identification of “high-risk dimensions” could help inform interventional efforts toward anger reduction for patients with known heightened CVD vulnerability.

The possible role of cultural influences in the context of the current findings warrants additional comment. A Latino cultural belief, especially prevalent in rural populations, is that the frequent expression of strong emotions like anger can predispose an individual to diseases such as diabetes (Coronado et al., 2004), or psychological distress, such as the culturally bound syndrome “ataque de nervios” (nerve attack; Arce and Torres-Matrullo, 1982; de Snyder et al., 2000). Thus, although they may be angry, women in the study may have been less likely to endorse these feelings, as they may view this as being culturally non-sanctioned or harmful, resulting in possible confounding of the relationship between trait anger, anger temperament and sICAM-1. However, the measurement of specific cultural beliefs and attitudes and their possible moderating effects should be further explored.

The current study possesses several strengths, including the examination of a novel marker of inflammation, the exploration of associations by subdimensions of anger, and an ethnic minority sample in which the relationship between anger, hostility and inflammation has been understudied. Nonetheless, several limitations should also be noted. The study is cross-sectional in nature, and causality cannot be determined. It is possible that inflammation can promote anger, as studies have reported increased anger, aggression and hostility in patients given cytokine therapy to treat medical conditions (Zalcman and Siegel, 2006). In addition, the measures of anger and cynical hostility were self-report, and may be affected by social desirability or recall bias; the employment of multi-method measures (e.g., significant-other report, relative-report, child-report) will be useful for future research. Finally, the current sample consisted only of Mexican American women, from selected communities in a border region, and thus, results may not generalize to males or other Latino subgroups.

The current study provides additional evidence that negative emotions and cognitions, particularly, specific aspects of anger and hostility, relate to physiological pathways that have the potential to influence health over the long term. Several questions remain to be answered regarding specific physiological pathways, differences that may exist by other anger subdimensions, specific populations studied (e.g., healthy vs. post-coronary event, young vs. elderly, white vs. ethnic minority, etc.), and inflammatory biomarker type and function. Answers to these questions can eventually inform and tailor interventions to promote CVD risk reduction.