The current concept of the metabolic syndrome (MetS) is that of a multicomponent pathophysiological interaction between several parameters, the coincidence of which leads to dramatic escalations in incidences of type 2 diabetes (T2DM) and cardiovascular disease (CVD; a summary term for >12 major syndromes, including, eg, coronary heart disease, vascular diseases, atherosclerosis, stroke, and hypertension).1,2 Overall, CVDs continue being the largest disease burden and cause of mortality worldwide. The international epidemic of MetS,2,3 which has its greatest acceleration rates in threshold and emerging market countries, will even further scale up CVD incidences. It has been found that MetS is more prevalent in individuals with higher personality disposition scores, including anger and hostility traits.4–6 In turn increased high levels of trait hostility the longitudinal risk of myocardial infarction in MetS by a factor of 4 (odds ratio, 4.21).7,8
The interest in personality factors in heart diseases dates back to the early 1950s and is, with more than 6000 published papers, now 1 of the major topics in cardiovascular research. The investigation of anger traits in cardiovascular research had boosted after the neurophysiological demonstration that specifically noradrenergic neurons in the brainstem exert regulatory influence on the heart rate.9 Several anger disposition facets, as well as anger targets, were investigated for CVD risk, but aggressivity as overt anger behavior (ie, behavioral aggression)6,10,11 and hostility as covert trait (ie, cognitive-affective dimension of aggression)4,6,12–16 have received the most focused attention.
A number of influential quantitative meta-analyses17,18 have hitherto summarized the very large numbers of studies triggered by this finding on the relation of hostility-anger and cardiovascular events. Whereas Jorgensen and colleagues18 found significant overall effect sizes for anger and hostility on elevated blood pressure, a later meta-analysis19 concluded that hostility was indeed a significant predictor, yet too practically unimportant for prevention efforts. But further support for the claim of increasing risk due to dispositional anger came from recent longitudinal data: Evaluations of prospective data have supported the notion that factors such as anger and hostility contribute to MetS by means of hypothalamic and sympathetic dysregulation.20 Hazard ratios of prospective studies on hostility contributing to coronary heart disease were 1.19 to 1.24,1 highlighting sex differences, and Cox Relative Indices of Inequality were 1.97 to 2.23 for CVD mortality.21
The purposes of the present study were therefore to (a) replicate previous studies, but in a sample monitored for MetS, and (b) relate single anger/hostility traits to single MetS components, following typical strategies in the study of MetS. The Buss-Durkee Hostility Inventory (BDHI) is considered to reflect overt (behavioral) and covert (cognitive) anger behavior and experience, respectively,22 and this scale was chosen because it is most established internationally. In the light of these recent developments, we considered the following research hypotheses: Because recent studies raised the possibility that opposing personality traits23 could associate with specific risk factors, (1) we hypothesized respective association patterns with different laboratory measurements. More specifically, because a recent study had revealed that dispositional anger and hostility may associate to different immune parameters,24 we expected analogous results in our MetS parameters. As recent epidemiological studies have further suggested a “psychological component”6 in terms of respective latent personality variables involved in MetS, (2) we expected a role of similar latent constructs in our data. Finally, owing to increasing literature on gender differences in personality traits, (3) we expected gender differences from men and women pertaining to differential anger and hostility self-report.
Eligibility Criteria and Recruitment
Patients treated within the countrywide healthcare system of the Ukrainian Ministry of Transport were referred from all parts of Ukraine to the specialized unit at Railway Hospital No. 2 in Kiev. For reasons of preventive control, this national pool of patients were annually reexamined and treated. The treatment cohort had been compiled on endocrinological parameters or family histories indicating specific elevated risk for T2DM. All patients were of white descent, and approximately 80% were of Ukrainian ethnicity (the remainder of Russian, Tatar, and minor ethnicities), well in accordance with published (2001) census figures. We could therefore be confident that our sample represents well the general population structure in Ukraine. The flowchart in Figure 1 summarizes the steps in patient sampling to final sample size, with inclusion and exclusion criteria.
The final sample of 101 volunteering individuals (mean age, 45.6 ± 0.14 SEM years; education level, 4.73 ± 0.11, 4 = junior college level, 67 women) was comparable in size and composition with research samples typically used in biomarker detection studies in MetS and in approximately counterbalancing diabetic and nondiabetic patients by design (eg, Huang et al25). The investigation was conducted in compliance with the Helsinki Declaration (www.wma.net/e/ethicsunit/helsinki.htm). The institutional review board of the National Medical University of the Ukraine had endorsed all procedures. All subjects gave written informed consent to the scientific use of their data and were reimbursed for their participation.
All patients had been consensus-diagnosed with at least 1 of the primary diagnoses obesity, hyperinsulinemia, dyslipidemia, and elevated arterial blood pressure, according to International Classification of Diseases 10th revision criteria by at least 2 cardiological and/or endocrinological specialists not involved in the study. MetS criteria conforming to International Diabetes Federation (www.idf.org/publications) consensus were coded from the patient files. The diagnosis of MetS was established as the presence of 3 or more of the following features: waist circumference greater than 80 cm in women and greater than 94 cm in men; fasting serum triglyceride (TG) level 1.7 mmol/L or greater; serum high-density lipoprotein level less than 1.29 mmol/L in women and less than 1.03 mmol/L in men; systolic blood pressure (SBP) 130/85 mm Hg or higher or treatment; fasting plasma glucose level 5.6 mmol/L or higher; or T2DM.
The official Russian-language version of the BDHI26 (www.psyjournals.ru/files/30547/psyedu_ru_2010_3_Kalyagin_Enikolopov.pdf) was used to assess different facets of anger and hostility traits. The 66-item BDHI is the most extensively used self-report measurement of aggressivity and hostility worldwide, with a well-replicable 2-factor structure.22 Its psychometric properties are well established, with overall internal consistencies ranging between 0.96 and 0.70, and reportedly high retest stability.27 The BDHI consists of the theoretically derived subscales Assault, Verbal Aggressivity, Indirect Aggressivity, Negativism, Irritation, Suspicion, Resentment, and Guilt. The factor analytically derived Aggressivity Index (AI) integrates subscales Assault, Verbal Aggressivity, and Indirect Aggressivity, whereas the factor analytically derived Hostility Index (HI) is composed of subscales Irritation, Suspicion, and Resentment.
Biological Laboratory Data and Statistical Analysis
Laboratory analyses of all biological specimens were performed in-house with enzymatic methods using commercially available reagents. The laboratory of Railway Hospital No. 2 was subject to quality control standards imposed by the Ukrainian Ministry of Public Health and was certified accordingly. STATA/MP 12.1 (64-bit version; StataCorp LP, College Station, Texas) was used for the statistical methods hierarchical regression analysis (HRA), multivariable logistic regression (MVLR), and structural equation modeling (SEM). IBM SPSS 20 for Intel Macintosh computers (IBM Corp, Armonk, New York) was used to compute correlations, analyses of variance (ANOVA), and principal components analysis (PCA).
Statistical Methods and Models
Using the recommended standard textbook approach,28 we examined the data in the steps correlation, PCA, HRA, and SEM. The SEM module in STATA, which is based on the Bentler-Weeks regression approach29,30 to decomposing variance-covariance matrices, uses the maximum likelihood statistics in fitting expected and observed information matrices with the Newton-Raphson stepping algorithm. Our overall SEM approach was the specification of a measurement-based causal model, using the latent constructs indicated by the 2-factor solution for aggressivity and hostility in the preceding PCAs. Error terms were included to estimate measurement error for both observed and latent variables. The final SEM model was fitted with constraints for latent variables after adjustment for different clustering in the sexes in the robust ordinary least square regression paths.
Scale Reliabilities for the Hostility and Aggressivity Measures
The internal consistencies for the subscales of the BDHI, as ascertained by Cronbach’s coefficient α, were satisfactory to moderate (Assault, α = .853; Verbal Aggressivity, α = .876; Indirect Aggressivity, α = .840; Negativism, α = .531; Irritation, α = .737; Suspicion α = .827; Resentment, α = .891; Guilt, α = .745). Because of the more robust scale reliabilities for the 2 index measures (AI, α = .871; HI, α = .906), only these were used for further statistical analyses. Our subscale reliabilities are in excellent agreement with published meta-analyses of the BDHI. Note that the relative low scale reliability for Negativism is a typically observed value across studies,27 as are all of the other scale reliability coefficients.
Intercorrelations of Biological Variables
Waist circumference (mean ± SD, range, 93.703 ± 11.110 cm, 73–160 cm) showed a negative association with high-density lipoprotein cholesterol (HDL; 1.049 ± 0.119 mmol/L, 0.66–1.26 mmol/L) (r = −0.514, P < .0001) and correlated positively with blood pressure minimum (diastolic blood pressure [DBP], 89.554 ± 5.834 mm Hg, 70–105 mm Hg) (r = 0.304, P < .002). High-density lipoprotein cholesterol exhibited an inverse correlation with DBP (r = −0.203, P < .042). There were significant positive correlations of TG (3.204 ± 0.855 mmol/L, 1.6–4.5 mmol/L) with microalbuminuria (33.66% ± 0.474%) (r = 0.308, P < .002), blood pressure maximum (SBP, 146.188 ± 13.155 mm Hg, 120–180 mm Hg) (r = 0.334, P < .001), and T2DM status (57.425% ± 0.496%) (r = 0.341, P < .0001). Microalbuminuria was positively associated with T2DM status (Cramer’s V = 0.359, P < .0001). Blood pressure minima and maxima were correlated (r = 0.564, P < .001). Total cholesterol (6.160 ± 1.175 mmol/L, 3.2–12.1 mmol/L) was positively associated with body mass index (BMI) (32.040 ± 5.495 kg/m2, 21.6–46.1 kg/m2) (r = 0.259, P < .009), TG (r = 0.483, P < .0001), low-density lipoprotein (5.838 ± 2.512 mmol/L, 2.6–14.5 mmol/L) (r = 0.627, P < .0001), SBP (r = 0.325, P < .001), and DBP (r = 0.0.564, P < .0001). Descriptive values of all measurements specified to sex differences have been previously published elsewhere.31
Confounding of Self-report With Sociodemographic Variables
The AI (54.135 ± 17.396, 16–99.7) and HI (58.316 ± 19.316, 0–89.5) were examined for possible confounding with sociodemographic variables, age, sex, education level, family size, and number of children. Both compound indices were unrelated to any of these, but significant gender differences in aggressivity (r = −0.306, P < .05) emerged in terms of lesser self-reported overt anger in women.
The PCA, ANOVA, HRA, and MVLR models were tested in preparation of SEMs, as suggested by standard approaches. The results of these analyses are reported in the supplemental digital content linked to this article (Supplemental Digital Content 1, http://links.lww.com/JCN/A10, Supplemental Digital Content 2, http://links.lww.com/JCN/A12, and Supplemental Digital Content 3, http://links.lww.com/JCN/A13). The profile plots of the ANOVAs for sex differences in disease severity of MetS are depicted in Figure, Supplemental Digital Content 2, http://links.lww.com/JCN/A12. In the resulting 2-factor solution, latent component 1 was dominated by AI, and latent component 2, by HI. Testing the factor loadings systematically by the HRA and MVLR models revealed 2 distinct regressive clusters grouped around AI and HI (results reported in Document, Supplemental Digital Content 1, http://links.lww.com/JCN/A10).
Structural Equation Modeling
The aim of the SEM was the setup and testing of a parsimonious causal model with less than 10 paths. Steps in the SEM included the setup of measurement-based regression path models in each of the aggressivity and the hostility chains and the inclusion of 2 latent variables (both being both exogenous and endogenous), as suggested by the 2-factor solution of the PCA (ie, a 2-factor causal model). Given the significant moderate intercorrelation (r = 0.399, P < .001) of the 2 factors, these were connected by adding a respective covariance path. The final comprehensive SEM (model 1) is depicted in Figure 2, with Table 1 reporting path coefficients and their significances and Table 2 reporting equation-level goodness-of-fit indices. When setting the latent variables to 1, this constrained model with adjustment of standard errors to differential clustering in the 2 sexes yielded an overall unique variance explanation of 72.3% in robust ordinary least square path estimation. The sex-adjusted model 1 was distinctly superior to a constrained model without introduction of sex differences (model 2) and a relaxed model (model 3; see Document, Supplemental Digital Content 1, http://links.lww.com/JCN/A10). Please note that none of the traditional SEM goodness-of-fit criteria are reported by STATA in robust estimation. In conclusion, we can state that, of our 3 computations, the constrained and sex-adjusted model 1 yielded superior fit, as indicated by the Bentler-Raykov squared multiple correlation coefficients and the high magnitudes in the prediction-correlations hovering around 0.9 (Table 2).
The main findings of the present study consist of the confirmation that CVD-risk biomarker variables in this MetS sample (a) chain into 2 distinct clusters and that (b) 1 cluster associates with self-reported overt anger, whereas the other associates with self-reported covert hostility. Beginning with correlation analyses, and based on PCAs, SEM modeling revealed that 2 latent constructs are involved in regression paths: 1 for aggressivity plus obesity-related biological measures and another for hostility plus cholesterol-related biological measures. Although aggressivity and hostility do moderately covary, the 2 regressive path models remain otherwise distinct. Regarding the a priori assumptions, hypothesis 1, which states that different biomarkers associate to different traits, was supported, with the exception of BMI, which was linearly related to both trait index variables. Hypothesis 2, which postulates a role for latent constructs, could be supported. Regarding sex differences (hypothesis 3), respective ANOVAs partly yielded such evidence. In addition, model fit in SEM was notably improved when the model was adjusted for sex differences.
A European multicenter genome-wide association study on coronary heart disease and lipid genes reported separable genomic bases for lipid transport and for cholesterol metabolism.32 Related to the genomic bases, distinct risk profiles for lipid metabolism (including waist girth and BMI) and for cholesterol metabolism (associating total cholesterol, TG, low-density lipoprotein, and HDL) are described. Also in our data set, the application of PCA, HRA, and SEM models supports the notion of differential association patterns, identical with the risk profiles identified in that cardiological multicenter study.
Consistent with our results, as well, trait anger has previously been found to be associated with BMI.33 In our data, the latent aggressivity component is also linearly related to waist. Waist girth, in turn, strongly correlates with BMI but is an independent indicator of CVD risk, as it reflects intra-abdominal fat deposition.2 The latent aggressivity component is also characterized by DBP, which is in line with previous findings. It is a consistently replicated result34 that anger arousal raises specifically DBP—more than any other emotion. Trait hostility was previously related to TGs and negatively with HDL,35 a pattern also observed in our data. Associated with the hostility component in the presented SEM was SBP. In longitudinal research, SBP strongly predicted CVD events, but these events were not found related to overt anger in that study.36 Dyslipidemia and hypercholesterolemia being associated with or being a longitudinal consequence of trait hostility is also a well-replicated finding.35,37 Contrary to Williams and colleagues,38 but in line with many other studies, we found sex differences more pronounced in anger and related biomarker variables and less pronounced in hostility. Thus, we can conclude in an overall summary that our findings synthesize well previous singular clinical results into a more comprehensive picture.
In conclusion, it can be stated that the present investigation extends existing knowledge by integrating and unifying several previous findings into a more comprehensive theoretical picture of anger traits and CVD risk markers. We were able to demonstrate that aggressivity associates with lipid transport/obesity biomarkers, whereas hostility associates with cholesterol biomarkers. This more differential picture could contribute to specification of more personalized prevention and care.
- The results of the present analyses suggest differential susceptibility for overt and covert anger types for specific risk profiles in cardiovascular biomarkers.
- Aligning to recent genomic analyses, lipid transport and cholesterol metabolism biomarkers are separable risk systems, requiring specified preventive approaches.
- The finding of this article will enable nurses and specialists to better adjust personalized approaches to psychosocial and internal care and prevention.
1. Chida Y, Steptoe A. The association of anger and hostility with future coronary heart disease. J Am Coll Cardiol
2. Despres JP, Lemieux I, Bergeron J, et al. Abdominal obesity and the metabolic syndrome
. Arterioscler Thromb Vasc Biol
3. Capoulade R, Clavel MA, Dumesnil JG, et al. Impact of metabolic syndrome
on progression of aortic stenosis. J Am Coll Cardiol
4. Gremigni P. Cynical hostility and the metabolic syndrome
. Monaldi Arch Chest Dis
5. Cohen BE, Panguluri P, Na B, Whooley MA. Psychological risk factors and the metabolic syndrome
. Psychiatry Res
6. Sutin AR, Costa PT Jr, Uda M, Ferrucci L, Schlessinger D, Terracciano A. Personality and metabolic syndrome
. Age (Dordr)
7. Kubzansky LD, Sparrow D, Jackson B, Cohen S, Weiss ST, Wright RJ. Angry breathing. Thorax
8. Todaro JF, Con A, Niaura R, Spiro A 3rd, Ward KD, Roytberg A. Combined effect of the metabolic syndrome
and hostility on the incidence of myocardial infarction. Am J Cardiol
9. Reis DJ, Miura M, Weinbren M, Gunne LM. Brain catecholamines: relation to defense reaction evoked by acute brainstem transection. Science
10. Zhang J, Muldoon MF, McKeown RE, Cuffe SP. Association of serum cholesterol and history of school suspension. Am J Epidemiol
11. Haukkala A, Konttinen H, Laatikainen T, Kawachi I, Uutela A. Hostility, anger control, and anger expression as predictors of cardiovascular disease. Psychosom Med
12. Anderson DE, Metter EJ, Hougaku H, Najjar SS. Suppressed anger is associated with increased carotid arterial stiffness. Am J Hypertens
13. Brydon L, Strike PC, Bhattacharyya MR, et al. Hostility and physiological responses to laboratory stress in acute coronary syndrome patients. J Psychosom Res
14. Grothe KB, Bodenlos JS, Whitehead D, Olivier J, Brantley PJ. The psychosocial vulnerability model of hostility as a predictor of coronary heart disease. J Clin Psychol Med Settings
15. Guerrero C, Palmero F. Impact of defensive hostility in cardiovascular disease. Behav Med
16. Meesters CM, Smulders J. Hostility and myocardial infarction. J Psychosom Res
17. Suls J, Wan CK. The relationship between trait hostility and cardiovascular reactivity. Psychophysiology
18. Jorgensen RS, Johnson BT, Kolodziej ME, Schreer GE. Elevated blood pressure and personality. Psychol Bull
19. Myrtek M. Meta-analyses of prospective studies on coronary heart disease, type A personality, and hostility. Int J Cardiol
20. Goldbacher EM, Matthews KA. Are psychological characteristics related to risk of the metabolic syndrome
? Ann Behav Med
21. Lemogne C, Nabi H, Zins M, et al. Hostility may explain the association between depressive mood and mortality. Psychother Psychosom
22. Bushman BJ, Cooper HM, Lemke KM. Meta-analysis for factor analysis: an illustration using the Buss-Durkee Hostility Scale
. Pers Soc Psychol Bull
23. Davidson KW, Mostofsky E. Anger expression and risk of coronary heart disease. Am Heart J
24. Shivpuri S, Gallo LC, Mills PJ, Matthews KA, Elder JP, Talavera GA. Trait anger, cynical hostility and inflammation. Brain Behav Immun
25. Huang CF, Cheng ML, Fan CM, Hong CY, Shiao MS. Nicotinuric acid: a potential marker of metabolic syndrome
. Diabetes Care
26. Buss AH, Durkee A. An inventory for assessing different kinds of hostility. J Consult Psychol
27. Vassar M, Hale W. Reliability reporting across studies using the Buss Durkee Hostility Inventory. J Interpers Violence
28. Weiber R, Mühlhaus D. Strukturgleichungsmodellierung
. Heidelberg, Germany: Springer-Verlag; 2010.
29. Bentler PM, Weeks DG. Linear structural equations with latent variables. Psychometrika
30. Bentler PM, Weeks DG. Interrelations among models for the analysis of moment structures. Multivariate Behav Res
31. Lemche AV, Chaban OS, Lemche E. Alexithymia as a risk factor for type 2 diabetes mellitus in the metabolic syndrome
. Psychiatry Res
32. Aulchenko YS, Ripatti S, Lindqvist I, et al. Loci influencing lipid levels and coronary heart disease in 16 European population cohorts. Nat Genet
33. Golden SH, Williams JE, Ford DE, et al. Anger temperament is modestly associated with the risk of type 2 diabetes mellitus. Psychoneuroendocrinology
34. Gambaro S, Rabin AI. Diastolic blood pressure responses following direct and displaced aggression after anger arousal. J Pers Soc Psychol
35. Niaura R, Banks SM, Ward KD, et al. Hostility and the metabolic syndrome
. Psychosom Med
36. Shaffer JA, Wasson LT, Davidson KW, Schwartz JE, Kirkland S, Shimbo D. Blood pressure reactivity to an anger provocation interview does not predict incident cardiovascular events. Int J Hypertens
37. Dujovne VF, Houston BK. Hostility-related variables and plasma lipid levels. J Behav Med
38. Williams ED, Steptoe A, Chambers JC, Kooner JS. Ethnic and gender differences in the relationship between hostility and metabolic and autonomic risk. Psychosom Med