HOSTILITY DIMENSIONS AND METABOLIC SYNDROME IN A HEALTHY, MIDLIFE SAMPLE

Abstract

Objective:

Evidence links trait hostility with components of the metabolic syndrome (MetS), a clustering of cardiometabolic risk factors, but which hostility dimensions (e.g., expressive or cognitive hostility) relate to MetS are not well known. Further, there may be age and sex differences in the extent to which hostility dimensions relate to MetS. The present study evaluated associations between dimensions of hostility and the metabolic syndrome and its individual components as well as the moderating effects of sex and age.

Methods:

In a cross-sectional sample of 478 employed adults, a principal components analysis from common trait hostility questionnaires yielded a two-factor solution: Expressive Hostility (anger and aggression) and Cognitive Hostility (cynicism). Each of these two components of hostility were examined as predictors of each of two aggregated MetS outcomes: a dichotomous measure of MetS, based upon the NCEP-ATP III definition, and a continuous measure based upon the average of standardized scores for each component, and they were examined as predictors of individual MetS components as well.

Results:

Expressive Hostility was associated with MetS severity (b = .110, p = .04) and waist circumference (b = 2.75, p = .01). Moderation analyses revealed that elevated Expressive Hostility was associated with elevated waist circumference in women but not men. Cognitive Hostility was not related to any metabolic syndrome component or aggregated outcome, and no moderation was observed.

Conclusions:

Among multiple individual components and two aggregated scores, only trait dispositions to expressed hostile affect and behavior were associated with MetS severity and waist circumference. The effects were small but statistically significant. The association between cognitive hostility and metabolic syndrome measures may not be robust in a large sample of healthy, midlife adults.

Introduction

Trait hostility is a broad psychological characteristic that involves affective states (experience of frequent and intense anger), behavioral tendencies (aggression), and cognitive styles (cynicism). The terms anger, aggression, and cynicism have been used interchangeably to describe trait hostility; however, these related constructs may represent distinct dimensions. Empirically, data reduction techniques conducted across multiple hostility questionnaires have shown a three-factor solution involving affective, behavioral, and cognitive styles [1, 2] as well as a two-factor solution involving expressive and cognitive hostility [1]. These dimensions are important for health because a large body of evidence have shown that individuals high in these dimensions are associated with increased risk for cardiovascular disease (CVD) morbidity and mortality [3, 4]. It is thought that hostile individuals are at increased cardiovascular health risk through exaggerated psychophysiological reactivity, reduced social support, and greater engagement in poor health behaviors [5].

Some studies have shown that trait hostility is associated with increased risk for the metabolic syndrome (MetS), a clustering of traditional cardiovascular risk factors (e.g., insulin resistance, dyslipidemia, central obesity, and elevated blood pressure [6]) that has been shown to predict future CVD [7] and mortality [8]. These findings linking trait hostility and MetS, however, have not been consistent. One explanation for the inconsistency is that the association between trait hostility and MetS may vary by dimension as well as by sex and age. For instance, higher cognitive hostility scores have been associated with MetS in young adult women but not men [9], whereas others have found no sex differences in midlife adults [10, 11]. Aggression has been linked to MetS in young adult males but not females [12]. Baseline trait anger scores predicted increased risk for developing the metabolic syndrome in midlife women [13-15], while others have shown the relationship between anger and MetS only among older adults [16]. These data suggest a range of findings involving trait hostility that may be moderated by hostility dimensions and and demographic factors.

Recent efforts using a multidimensional assessment of hostility have been promising as a means of clarifying the expected association with MetS. In a large sample of patients with stable coronary heart disease, anger expression and cognitive hostility scales were each related to MetS [17]. In a sample of patients treated for metabolic disorder, aggression was related to waist circumference and diastolic blood pressure, and cognitive hostility was related to total cholesterol, triglycerides, and systolic blood pressure [18, 19]. Finally, D'Antono, Moskowitz [20] assessed possible sex and age differences between hostility dimensions and MetS in a small, prospective sample of healthy adults. The authors found that, at baseline, daily measures of quarrelsomeness but not cynical hostility or anger were related to MetS severity (a continuous MetS measure). Though neither age nor sex moderated this relationship at baseline, cynical hostility predicted MetS severity only among mid-life and older adults at the three-year follow-up.

No study has looked at metabolic syndrome and its individual components using empirically-derived dimensions of hostility while considering the moderating effects of sex and age in a healthy, community sample. In the current study, we explore (a) dimensions of hostility associated with individual metabolic syndrome components as well as aggregated metabolic syndrome measures as outcomes, and (b) possible moderating effects of sex and age on these associations in a large sample of healthy, midlife adults.

Methods

Participants

Participants were drawn from the Adult Health and Behavior Project – Phase 2 (AHAB-II). Eligibility criteria included age between 30 and 54 years and working at least 25 hours per week outside the home. Exclusion criteria for this sample have been previously described [21]. Briefly, individuals were excluded if they had a history of CVD, Stage 2 hypertension (systolic/diastolic BP ≥160 mmHg/ ≥100 mmHg), or were prescribed insulin or antihypertensive, lipid-lowering, or prescription weight-loss medications. Informed consent was obtained from all individual participants included in the study. The overall sample consisted of 494 participants.

Measures

Hostility.

Participants were instructed to complete three common self-report trait hostility questionnaires: the 29-item Buss Perry Aggression Questionnaire (BPAQ), the 50-item Cook-Medley Hostility Scale (CMHS), and the 44-item Spielberger State-Trait Anger Expression Inventory (STAXI). The BPAQ consists of four empirically designed subscales (Physical Aggression, Verbal Aggression, Anger, and Hostility). The CMHS is thought to represent the cognitive dimension of hostility [22]. Four relevant trait subscales of the STAXI include Trait Anger, Anger-in, Anger-out, and Anger Control. The STAXI Trait Anger subscale is thought to represent the frequency of anger, whereas the other three STAXI subscales are thought to represent the style in which one expresses anger [23].

We performed a principal components analysis (PCA) with a goal of deriving underlying dimensions of hostility from nine hostility subscales: BPAQ (four subscales), CMHS (one score) and STAXI (four scores). Using an eigenvalue criterion of 1 and an oblique varimax rotation (see Table 1), we extracted a two-factor solution: Expressive Hostility (anger and aggression) and Cognitive Hostility (cynicism). Unit-weighted scores were derived for each of these two factors by averaging z-scores across subscales within each of these two groupings. STAXI Trait Anger was excluded from the calculation of these factor scale scores because of loading on both factors. Expressive Hostility and Cognitive Hostility were moderately correlated (r = 0.47, p < .0001).

Table 1.

Hostility Factor Scores (Oblique Varimax Rotated Factor Pattern)

Measure	Expressive	Cognitive
Spielberger Anger Out subscale	0.82	0.27
Buss-Perry Anger subscale	0.80	0.49
Spielberger Anger Control subscale	−0.75	−0.29
Buss-Perry Verbal Aggression subscale	0.72	0.23
Buss-Perry Physical Aggression subscale	0.64	0.49
Buss-Perry Hostility subscale	0.40	0.83
Cook-Medley Hostility scale	0.40	0.82
Spielberger Anger In subscale	−0.21	0.71
Spielberger Trait Anger subscale	0.64	0.67

Note. N=478. Boldface designates the factor loadings (above .50) associated with component measures that were selected for inclusion in each of the two derived hostility factor scores. For the Expressive Hostility factor, the eigenvalue and variance explained consisted of 4.33 and 3.59, respectively. For the Cognitive Hostility factor, the eigenvalue and variance explained consisted of 1.43 and 3.01, respectively.

Metabolic Syndrome.

A trained staff member collected participants’ height, weight, waist circumference, and venous blood sample in the morning following a 12-hour, overnight fast. Resting blood pressure was measured as the average of two seated BP measurements following a 10-minute resting period. Determination of serum high-density lipoprotein (HDL) cholesterol, triglycerides, and glucose was performed by the Heinz Nutrition Laboratory, University of Pittsburgh Graduate School of Public Health.

Two measures of aggregated metabolic syndrome used in this investigation were calculated from the individual metabolic components: systolic blood pressure (SBP), diastolic blood pressure (DBP), glucose, triglycerides, waist circumference, and high-density lipoprotein (HDL) cholesterol. First, a dichotomous measure (diMetS) was defined according to the National Cholesterol Education Program – Adult Treatment Panel III guidelines (NCEP – ATP III) [24, 25]. The NCEP – ATP III defines the metabolic syndrome as the presence of three or more of the following features: fasting plasma glucose ≥ 100 mg/dL; serum triglycerides ≥ 150 mg/dL; serum HDL cholesterol < 40 mg/dL in men and < 50 mg/dL in women; systolic blood pressure ≥ 130 mmHg or diastolic blood pressure ≥ 85 mmHg; waist circumference ≥ 102 cm in men and ≥ 88 cm in women. A dichotomous metabolic syndrome (diMetS) variable was derived where 0 = two or less symptoms and 1 = 3 or more symptoms.

Second, we standardized each metabolic syndrome components, with HDL cholesterol and waist circumference standardized separately by gender. HDL cholesterol was reverse scored so that higher values denoted lower HDL. The SBP and DBP z-scores were averaged so that BP was represented by one z-score. The five z-scores produced for each component were averaged across participant to create a continuous metabolic syndrome score (zMetS).

Analysis Plan

Analyses were derived in separate models using the statistical software SAS (9.4). Generalized linear models were used for the main effects of the hostility components on the individual MetS components and zMetS. Logistic regression was used for analyses involving diMetS. The moderating effects between hostility dimensions and sex or age were tested in separate models while retaining the main effect; significant interactions were followed up with subgroup analyses or simple slope analyses, respectively. All models included demographic covariates: sex, age, race, and education.

Of the total sample size of 494 subjects, one subject was removed due to missing metabolic data, another subject due to an outlying zMetS score that was more than three standard deviations above the mean, and 14 subjects were removed due to missing hostility questionnaire data. The final analytic sample consisted of 478 participants. Sample characteristics are provided in Table 2.

Table 2.

Sample Characteristics

Demographics	Mean or N	SD or %
Sex (N, % female)	255	53.3
Age, years	42.7	7.35
Race, (N, % White/Caucasian)	390	81.6
Bachelor's degree or higher, (N, %)	342	71.5
Metabolic Syndrome Measures
Systolic Blood Pressure, mmHg	115.2	10.7
Diastolic Blood Pressure, mmHg	72.4	7.64
Glucose, mg/dl	98.2	10.4
Triglycerides, mg/dl	108.9	66.9
HDL, mg/dl	55.9	15.1
Waist Circumference, cm	90.4	14.1
Continuous MetS (zMS)	−0.008	0.62
Dichotomous Metabolic Syndrome Criteria
Dichotomous MetS (diMS; N, %)	73	15.3
Blood Pressure (N, %), ≥130/≥85mmHg	53	11.1
Glucose (N, %), ≥100 mg/dL	196	41.0
Triglycerides (N, %), ≥150 mg/dL	90	18.8
HDL Criteria (N, %)
Men (< 40 mg/dL)	56	25.1
Women (< 50 mg/dL)	52	20.4
Waist Circumference Criteria (N, %)
Men (≥ 102 cm)	54	24.2
Women (≥ 88 cm)	102	40.0

Note. N = 478. Raw values are reported for continuous measures. MetS = Metabolic Syndrome. HDL = High-Density Lipoprotein. Dichotomous MetS = 3 or more of the individual MetS measures.

Results

Of the MetS measures, Expressive Hostility was significantly associated with zMetS (b = 0.110, 95% CI [0.004, 0.21], p = .04) and waist circumference (b = 2.75, 95% CI [0.57, 4.95], p = .01; see Table 3a). Cognitive Hostility was unrelated to the individual components and aggregated measures (see Table 3b). A significant interaction between sex and Expressive Hostility emerged for waist circumference (b = 5.98, 95% CI [1.74, 10.22], p = .0006). Subgroup analyses reveal a significant association between Expressive Hostility and waist circumference for women (b = 5.33, 95% [1.97, 8.69], p = .002) but not for men (b = 0.17; 95% CI [−2.57, 2.90], p = .90). As indicated in Table 3, no significant interactions involving age emerged between either hostility component or any MetS measure.

Table 3a. Main Effect and Moderating Effects of Expressive Hostility and Metabolic Outcome
	Expressive Hostility			Sex*Expressive Hostility			Age*Expressive Hostility
Variables	b	p	95% CI	b	p	95% CI	b	p	95% CI
Standardized MetS	0.110	0.04	[0.004, 0.21]	0.18	0.07	[−0.02, 0.38]	0.008	0.24	[−0.005, 0.02]
Dichotomous MetS	0.420	0.08	[−0.06, 0.89]	0.37	0.29	[−0.33, 1.05]	0.008	0.24	[−0.002, 0.02]
Waist Circumference	2.75	0.01	[0.57, 4.95]	5.98	0.006	[1.74, 10.22]	0.28	0.06	[−0.01, 0.58]
Diastolic Blood Pressure	1.03	0.12	[−0.26, 2.32]	2.29	0.07	[−0.21, 4.80]	0.07	0.44	[−0.11, 0.24]
Systolic Blood Pressure	1.31	0.14	[−0.42, 3.04]	1.84	0.28	[−1.53, 5.21]	0.22	0.07	[−0.01, 0.45]
Glucose	0.44	0.62	[−1.29, 2.18]	3.13	0.07	[−0.24, 6.51]	−0.009	0.94	[−0.24, 0.23]
Triglycerides	6.88	0.23	[−4.34, 18.10]	0.281	0.80	[−19.10, 24.72]	0.89	0.91	[−1.44, 1.62]
HDL Cholesterol	−0.84	0.46	[−3.09, 1.41]	0.23	0.92	[−4.18, 4.63]	−0.09	0.57	[−0.39, 0.22]

Table 3b. Main Effect and Moderating Effects of Cognitive Hostility and Metabolic Outcome
	Cognitive Hostility			Sex*Cognitive Hostility			Age*Cognitive Hostility
Variables	b	p	95% CI	b	p	95% CI	b	p	95% CI
Standardized MetS	0.009	0.32	[−0.009, 0.03]	0.009	0.61	[−0.02, 0.04]	0.0007	0.52	[−0.001, 0.79]
Dichotomous MetS	−0.01	0.76	[0.10, 0.07]	0.02	0.79	[−0.11, 0.14]	−0.0004	0.69	[−0.002, 0.001]
Waist Circumference	0.24	0.23	[−0.16, 0.64]	0.55	0.16	[−0.22, 1.31]	0.03	0.26	[−0.02, 0.08]
Diastolic Blood Pressure	−0.08	0.52	[−0.31, 0.16]	0.37	0.10	[−0.07, 0.82]	0.01	0.32	[−0.01, 0.04]
Systolic Blood Pressure	−0.06	0.73	[−0.37, 0.26]	0.34	0.27	[−0.26, 0.94]	0.03	0.15	[−0.01, 0.07]
Glucose	0.11	0.49	[−0.21, 0.43]	0.16	0.60	[−0.44, 0.77]	−0.02	0.23	[−0.06, 0.01]
Triglycerides	0.91	0.38	[−1.14, 2.97]	−0.46	0.82	[−4.40, 3.47]	−0.05	0.70	[−0.31, 0.21]
HDL Cholesterol	−0.14	0.51	[−0.55, 0.28]	0.36	0.37	[−0.42, 1.15]	0.03	0.18	[−0.09, 0.02]

Note. Boldface denotes statistically significant values (p < .05). MetS = Metabolic Syndrome. CI = Confidence Interval. The models involving the interaction terms simultaneously controlled for the main effect of the respective hostility dimension.

Exploratory Analyses

Given its prevalence in the literature, the STAXI Trait Anger subscale was analyzed as a predictor using similar models as the factor analytic hostility dimensions. Consistent with Expressive Hostility, Trait Anger had a significant main effect on zMetS (b = 0.02; 95% CI [0.002, 0.30], p = .03) and waist circumference (b = 0.39; 95% CI [−0.08, 0.68], p = .01). Of the moderator analyses, significant sex differences again emerged for zMetS (b = 0.03; 95% CI [0.002, 0.06], p = .03) and waist circumference (b = 0.66, 95% CI [0.06, 1.25], p = .03). Subgroup analyses revealed these associations to be only for women (zMetS: p = .003; waist circumference: p = .003) but not for men (zMetS: p = .81; waist circumference: p = .50).

As STAXI Trait Anger was cross-loaded on each Expressive Hostility and Cognitive Hostility dimensions, we sought to understand if the above pattern of findings could be attributed to Expressive or Cognitive Hostility. We re-analyzed the data while controlling for Expressive Hostility and Cognitive Hostility in separate models. The associations between STAXI Trait Anger and zMetS and waist circumference persisted after controlling for Cognitive Hostility but not after controlling for Expressive Hostility, suggesting that the STAXI Trait Anger associations are attributable to the Expressive Hostility dimension. Altogether, these data suggest that STAXI Trait Anger subscale findings were generally consistent with the PCA-derived Expressive Hostility findings.

Discussion

The current investigation found that among subscales from common hostility measures, a dimension of hostility representing expressions of hostile affect and behavior was associated only with MeTs severity and waist circumference. Moderation analyses suggest that the association between expressive hostility and waist circumference was found only for women. Unlike the expressive dimension of hostility, the cognitive dimension of hostility was unrelated to MetS or its individual components. Though a three-factor solution was not observed, a two-factor solution involving cognitive and expressive hostility shown here is consistent with prior work [1].

These findings are consistent with other work showing anger to be related to waist circumference in midlife adults, particularly women. Cross-sectional and longitudinal studies have shown indices of central obesity, such as waist circumference [26], waist-hip ratio [27], or visceral adipose tissue [28], to be related to higher STAXI Trait Anger scores in midlife women. In addition to replicating the association between elevated waist circumference in women with elevated Trait Anger scores, which is thought to best represent the frequency of anger, our findings suggest that women high in Expressive Hostility, which may better represent the style in which one expresses anger as well as aggression, are also at risk for elevated waist circumference. As individuals with excess abdominal fat are at substantially higher risk for insulin resistance, hyperglycemia, and development of Type 2 diabetes and CVD [29, 30], our findings may highlight another avenue by which women high in anger and aggressive tendencies may be at increased adverse health risk later in life.

The null findings involving the cognitive dimensions of hostility is somewhat inconsistent with the prior literature showing CMHS, a common hostility questionnaire representative of the cognitive dimension, associated with metabolic syndrome [10, 11, 20] and its components [e.g., 31]. Unlike prior work, our study analyzed affective and behavioral components as a construct distinct from the cognitive aspects of hostility. Given the modest correlation between the two hostility components, the extent to which previous findings regarding the CMHS may be attributed to Expressive Hostility effects cannot be determined. Another explanation for the null findings here may be that the exclusionary criteria (e.g., Stage 2 hypertension, lipid-lowering medications) resulted in less frequent or severe MetS cases. The relative good health of the sample may have curbed the ability to more consistently detect relationships between hostility dimensions and measures of MetS.

The current investigation is not without limitations. In addition to the relative health of this sample, the effect sizes shown here and elsewhere in the literature are relatively small [see 31]. Further, the association the association between Expressive Hostility and MetS severity shown here was at the boundary of significance and thus should be interpreted with caution. Taken together, the relationship between hostility dimensions and MetS may not be robust. Additionally, despite our efforts to reduce the number of models through the factor analytic approach, the large number of tested models may have increased the risk for familywise error.

Despite the aforementioned limitations, this study may make notable contributions to the literature. Multidimensional hostility assessments may elucidate which hostility dimensions are most important in prediction of clinical markers or outcomes. Even in a healthy, midlife community sample, trait dispositions to express anger and aggression may be associated with elevated waist circumference, a key correlate of metabolic dysregulation [29]. These findings may identify a pathway by which relatively healthy individuals high in hostility are prone to cardiovascular disease. Our findings support the possibility that any observed associations with metabolic syndrome may be most robust for measures of expressive hostility and with measures of waist circumference, especially for women.

In addition to replicating these findings involving sex differences between expressive hostility and waist circumference, future work involving trait hostility should continue to employ multidimensional models of hostility. Such work may help us understand which specific dimensions of hostility may pose the greatest risk, not only for metabolic syndrome, but for a number of other health pathways linked with clinical heart disease, such as inflammatory, atherogenic and hemostatic mechanisms, as well as health behaviors. In addition, future research can extend these cross-sectional findings by testing the prospective association between hostility dimensions and MetS measures. Such longitudinal data may afford the opportunity to understand which hostility dimensions may influence the progression of MetS over time.