The Prospective Contribution of Hostility Characteristics to High Fasting Glucose Levels: The moderating role of marital status

Participants were from the Normative Aging Study, a longitudinal study designed to investigate biomedical and psychosocial changes associated with aging. The study recruited men from the Boston area who were healthy and without a history of chronic illnesses at entry. All participants provided written informed consent. The sampling and design of the study have been reported in detail previously (17).

To be included in this study, participants were required to 1) have completed the Minnesota Multiphasic Personality Inventory (MMPI) in 1986 and 2) have provided a fasting blood sample around 1986 and another one ∼9 years later. Participants were excluded if they had a history of CAD (angina pectoris, ischemic heart disease, and myocardial infarction) or diabetes at baseline.

There were 692 participants at baseline, among whom 207 did not provide blood samples at follow-up and were excluded. The excluded participants did not differ from those with follow-up glucose in BMI, cholesterol, and all psychological variables except that they tended to be older (aged 63.0 ± 8.5 vs. 59.0 ± 6.9 years, t [690] = 6.58; P < 0.01), more likely to be unmarried (30 vs. 20%, χ²[1] = 7.40; P < 0.01), and had slightly higher systolic blood pressure (130 ± 16 vs. 127 ± 15 mmHg, t [690] = 2.48; P < 0.05).

The descriptive statistics of the final sample of 485 men are shown in Table 1. They were predominantly Caucasian between 42 and 76 years of age with a mean of 59.0 ± 7.0 years. The majority were married and had at least a high school education. On average, they consumed 1.5 ± 2.0 alcoholic drinks and smoked 1.1 ± 1.5 cigarettes per day.

Participants represented a fairly healthy population of older men at baseline, with normal average HDL cholesterol, triglyerides, and diastolic blood pressure, while their average BMI, LDL cholesterol, total cholesterol, and systolic blood pressure were somewhat elevated by the current standards. Their average fasting glucose levels were within normal range (5.5 ± 0.5 mmol/l) at baseline, with 64 (13.2%) showing impaired fasting levels (between 6.11 and 6.99 mmol/l). At follow-up, fasting glucose values (5.6 ± 1.0 mmol/l) became slightly higher and more variable. There were 40 (8.2%) incidents of type 2 diabetes (fasting glucose ≥6.9 mmol/l) and 47 new cases of impaired fasting glucose (between 5.6 and 6.9 mmol/l) at follow-up.

After 1986, participants received medical examinations every 3 years. On the night before examination, participants were asked to refrain from eating or drinking after midnight. The examination included an update of medical history, assessment of blood pressure and anthropometric measures, and collection of a fasting blood sample.

Participants completed standard questionnaires to provide demographic background as well as information on cigarette smoking and alcohol consumption. In 1986, active participants received the MMPI form-AX by mail, from which various psychological measures were derived.

Type A behavior, anger, and hostility have long been a research focus in the influence of negative emotionality on physical health. Although hostility or anger has been suggested as the active ingredient in type A behavior, each has shown distinctive predictability of disease status in different studies (e.g., type A behavior in Kawachi et al. [18]). To understand whether these three factors represent either a separate or a common construct in predicting change in glucose levels, we examined their effects separately and in combination.

The 19-item MMPI-2 Type A Scale assesses an individual's sense of time urgency, competitive attitude, impatience, and irritability. Type A individuals are recognized as fast pacing, hard driving, and highly involved at work. They are also impatient, direct, confrontational, argumentative, and short tempered during interpersonal transactions. The Type A Behavior Scale was developed to be conceptually distinct from cynical hostility during the restandardization of MMPI (19). The scale evidenced high 9-day test-retest (r = 0.82) and internal consistency (Cronbach's α = 0.72) reliability (19). Its predictive validity has been established by its independent and prospective association with CAD onset (18).

Hostility was assessed by the Cook-Medley Hostility Scale. It consists of 50 true-or-false items, tapping the disposition characterized by cynical attitudes, misanthropic beliefs, and aggressive responding style. Individuals with high scores tend to be suspicious and distrustful of people, perceive their environment as threatening, and see others as harboring harmful intent. The instrument has established reliability, convergent validity, and discriminant validity (20) and is associated with detrimental health habits, adverse health outcomes, and incidence of coronary disease (21).

Anger was measured with the 16-item MMPI-2 Anger Content Scale that assesses excessive anger expression and inability to control outbursts (19). Individuals with higher scores are impulsive, testy, easily annoyed, and likely to engage in physical and verbal confrontations. When provoked, they are likely to become verbally or physically aggressive and engage in swearing and fighting. The scale has high 9-day test-retest reliability (r = 0.85) and internal consistency (α = 0.76).

Considering the high correlations among hostility, anger, and type A behavior (r = 0.64–0.73), we conducted a principal components analysis and extracted a single component explaining 79% of the total variance. A factor score was derived to represent an overall index for hostility, anger, and type A tendencies.

Depressive symptom severity was assessed with the 33-item MMPI-2 Depression Content Scale (19). It assesses affective, cognitive, and behavioral symptoms of depression, including dysphoric mood, lack of interest, low motivation, feelings of guilt and worthlessness, and thoughts of suicide. The scale evidenced high 9-day test-retest reliability (r = 0.87) and internal consistency (α = 0.85) (19).

Blood pressure was measured to the nearest 2 mmHg with a standard mercury sphygmomanometer. Systolic and diastolic blood pressure was obtained from both arms in a sitting position. An average reading from both arms was calculated for analysis.

Height was measured to the nearest 0.1 inch, and weight was measured to the nearest 0.5 lb, with the participant standing in bare feet and in undershorts. BMI was calculated from dividing weight in kilograms by the square of height in meters.

Fasting blood samples were obtained at 8:00 a.m. Values of glucose and cholesterol were obtained using standardized procedures (18). Serum glucose was measured in duplicate on an autoanalyzer using the hexokinase method. Serum cholesterol was analyzed with enzymatic method (SCALVO Diagnostics, Wayne, NJ). After precipitation of LDL cholesterol and VLDL fractions, HDL cholesterol was measured in supernatant with the Abbott Biochromatic Analyzer 100 (Abbott Laboratories, South Pasadena, CA). LDL cholesterol was estimated with the Friedewald method.

Before analysis, variables were inspected for normality. Fasting glucose values were transformed with a natural log function because of nonnormality. All psychological measures were transformed to Z scores to facilitate interpretation.

A number of sociodemographic and biomedical variables may confound the relationships between psychological variables and glucose levels. We included the following covariates in analyses as standard control variables: age, education, marital status, BMI, total cholesterol, and systolic blood pressure.

Hierarchical multiple regression analysis was conducted to examine the associations between psychological variables (hostility, anger, type A behavior, and depressive symptom severity) and fasting glucose levels. Each psychological variable was tested separately for its contribution to either concurrent or follow-up glucose levels. All models were controlled for standard sociodemographic and biomedical variables. In the analyses of follow-up glucose levels, baseline glucose levels were also controlled.

Additional analyses were conducted to investigate whether age, education, and marital status moderated the relationship between a psychological factor and fasting glucose values. Moderation of personality by demographic characteristics could either enhance or buffer their effects on glucose. Procedures described by Cohen et al. (22) were used to test interactions in regression models. In short, to test a particular interaction (e.g., anger × marital status), we computed an appropriate interaction term by multiplying the corresponding psychological factor (e.g., anger) and sociodemographic variable (e.g., marital status) and then entered it in the regression model with all control variables and independent variables involved in the interaction. Continuous variables were centered at the mean before the interactions were calculated.

The correlations between psychological variables and fasting glucose levels at baseline and follow-up are presented in online appendix Table 1A (available at http://dx.doi.org/10.2337/dc07-1945).

None of the psychological variables were found to be associated with concurrent fasting glucose levels at the baseline in regression models with standard covariates (β = 0.017–0.050, all P > 0.05).

After controlling for baseline glucose values, age, education, marital status, BMI, total cholesterol, and systolic blood pressure, Cook-Medley hostility (β = 0.105, P = 0.021; model R² = 0.158, F [9,484] = 9.884, P < 0.001), anger (β = 0.091, P = 0.036; model R² = 0.164, F [9,484] = 10.368, P < 0.001), type A behavior (β = 0.152, P = 0.002; model R² = 0.169, F [9,484] = 10.696, P < 0.001), and overall hostility factor (β = 0.124, P = 0.005; model R² = 0.170, F [9,484] = 10.837, P < 0.001) each were found to predict fasting glucose at the follow-up. In contrast, depressive symptom severity (β = 0.004, P = 0.929; model R² = 0.135, F [9,484] = 8.209, P < 0.001) was not a significant predictor of follow-up fasting glucose. The R² and change in R² in each step of the regression models are presented in online appendix Table 2A. In addition to psychological variables, being unmarried (β = 0.114–0.121, P = 0.005–0.008), BMI (β = 0.124–0.147, P = 0.001–0.009), and baseline glucose (β = 0.292–0.294, P = 0.000) were also significant predictors of higher fasting glucose at follow-up in various models.

To illustrate the effect of hostility factor on glucose change, we calculated the mean increase in fasting glucose of participants in each quartile of the overall hostility factor. They were 0.02, 0.03, 0.23, and 0.36 mmol/l from the lowest to highest quartile, respectively, indicating greater increases in groups with higher overall hostility (F [3,481] = 3.68, P = 0.012). We also compared those who developed new incidents of type 2 diabetes (n = 40) and impaired fasting glucose (n = 47) with those who remained healthy at follow-up. The former evidenced significantly higher anger (t [483] = −1.99, P = 0.047), type A behavior (t [483] = −3.02, P = 0.003), Cook-Medley hostility (t [483] = −2.20, P = 0.028), and overall hostility (t [483] = −2.71, P = 0.007) scores and were more likely to be unmarried (34.5 vs. 17.3%, χ² = 12.92, P < 0.001) at baseline. These two groups, however, did not differ in depressive symptom severity (t [483] = −0.441, P = 0.660). After controlling for baseline glucose and BMI, the group differences remained significant for overall hostility (P = 0.035) and type A behavior (P = 0.011), while the significance was attenuated for anger (P = 0.098) and Cook-Medley hostility (P = 0.120).

Marital status was found to moderate the effects of Cook-Medley hostility, anger, type A behavior, and the overall hostility factor on follow-up glucose levels. These psychological variables were significantly associated with higher fasting glucose at follow-up among unmarried men (β = 0.348, 0.444, 0.439, and 0.453, respectively; all P < 0.001) but not in those who were married (β = 0.078, 0.060, 0.106, 0.093, respectively; all P > 0.05). Neither age nor education was found to moderate the effects of psychological variables on the follow-up glucose values. Finally, further controlling for cigarette smoking and alcohol consumption per day among participants with valid smoking and drinking data (n = 446) did not alter the results.

This study demonstrated that anger, hostility, and type A characteristics, but not depressive symptom severity, independently and significantly predicted higher fasting glucose levels among older men over 9 years after controlling for baseline glucose, age, education, BMI, cholesterol, and blood pressure. This relationship was moderated by marital status such that the hostility-related characteristics predicted higher follow-up fasting glucose only for men who were not married. Several strengths of the study bolstered the validity of the findings, including a longitudinal design, stringent statistical controls, and a relatively large sample size. Results suggest that hostility-related psychological qualities may confer higher prospective risk for impaired glucose metabolism.

Several previous studies showed that hostility, anger, and type A behavior were cross-sectionally correlated with higher fasting glucose (7–11). Although hostility-related characteristics were not correlated with concurrent fasting glucose in this study, they, in combination with marital status, predicted higher fasting glucose ∼9 years later. Several reasons may explain this discrepancy. First, as prior studies compared sex differences, they demonstrated that the link between personality risk factors and indexes of glucose metabolism was seen among women but not men (10,11,13). The lack of cross-sectional findings in our sample of older men appeared to be consistent with the literature. Second, this study applied stringent statistical control for covariates, which may have attenuated the relationship between psychological factors and glucose levels. Third, past studies were mostly based on smaller samples that might have produced less reliable observations. Our findings suggest that although the effects of hostility, anger, and type A behavior may not be immediate, their influences on glucose metabolism can be seen cumulatively over time and shape the course of illness onset in a gradual and progressive manner.

Instead of investigating only one psychological factor as in most prior studies, we examined hostility, anger, and type A behavior either separately or in combination. The type A scale in the study was developed to be differentiated from cynical hostility or anger while focusing on behavioral signs of time urgency, impatience, and confrontational attitudes. Results indicated that their effects are similar and overlapping, suggesting that they reflect a common underlying psychological construct affecting glucose metabolism.

Several mechanisms may explain the adverse effects of hostility-related characteristics on glucose metabolism. Individuals with high hostility, anger, and type A characteristics may perceive their environment as threatening and experience heightened stress and hassles. Elevated stress may in turn induce maladaptive psychophysiological responses that adversely affect hypothalamic–pituitary–adrenal axis activation and sympathetic arousal. Prolonged hypothalamic–pituitary–adrenal activation is accompanied by elevated adrenocorticotropic hormone, cortisol, and catecholamines that may disturb glucose metabolism and increase insulin resistance (11). This hypothesis has been supported by studies in which hostile individuals showed heightened sympathetic arousal, cardiovascular reactivity, and neuroendocrine response when facing laboratory stressors (23). Furthermore, a maladaptive stress response may increase chronic inflammation, which is implicated in impaired glucose metabolism (24). An increased inflammatory response, therefore, may be another promising mechanism linking hostility characteristics and impaired glucose metabolism.

It is noteworthy that being married had a main effect such that it was generally associated with lower fasting glucose at follow-up. Marriage also interacted with hostility factors to buffer their detrimental effects on subsequent increase in fasting glucose. The salutary health benefits of marriage have been well documented. Marriage is a primary source of support for men (14) and may help ameliorate the impact of stress and protect against excessive neurohormonal arousal (25). Unmarried men may also be less likely to monitor their health and more likely to engage in unhealthy habits such as overeating and smoking (14), thus rendering them at higher risk for illness. We speculate that married men benefited from their spouses, who helped them maintain a healthier lifestyle and curtail detrimental habits.

In addition, marriage appeared to be particularly beneficial for individuals with high hostility characteristics by counteracting their negative impact on fasting glucose. Although hostile individuals were more likely to experience higher distress and engage in unhealthy behaviors, their spouses may counteract these damaging effects by providing instrumental and emotional support, alleviating distress, and curtailing high-risk behaviors (26).

In contrast to past studies, we did not find that depressive symptom severity was related to concurrent or prospective fasting glucose levels. First, our study sample consisted of healthy men, whereas the majority of past studies examined individuals with diabetes (27). Depressive symptoms may bear a stronger relationship with poor glycemic control among those with diabetes than in a healthy population. Moreover, previous studies primarily observed a cross-sectional relationship, while none associated depression with change in glucose levels over time (28). Another reason may lie in the depression instrument used in the study, which predominantly assesses affective and cognitive, but not somatic, symptoms of depression. Intriguingly, some studies do suggest that depression is a risk factor for diabetes incidence (1). In light of our findings, it would be interesting to examine whether depression remains a significant predictor of diabetes onset even after controlling for hostility, anger, and type A behavior.

There are a few limitations that may also point to directions for future research. First, the sample consisted of relatively healthy and predominantly older Caucasian men without major illness, which may limit the generalizability of findings to other populations. Considering the differences in risk of type 2 diabetes across specific populations and observations of group-specific relationship between psychological factors and glucose levels (11), future research is needed to examine whether these results can be replicated among diverse samples, such as younger or older individuals, ethnic minorities, women, or those with chronic disease. Second, the assessment of marital status was limited. It was measured at baseline and a crude indicator of various aspects of marital quality. Future studies should document change in marital status over time and the different facets of marital quality in order to discern their exact influences on health outcomes. Third, diagnosis of depression was not assessed. Results, therefore, may not be extrapolated to those who are clinically depressed. Finally, we speculate that a number of stress-induced neuronendocrine and inflammatory responses may mediate the link between psychological factors and glucose metabolism. Future research should investigate these elements in glycemic control.

In summary, this study demonstrated that hostility-related psychological characteristics are associated with increased fasting blood glucose levels among unmarried older men over ∼9 years. It suggests that glucose metabolism may be a legitimate mediating mechanism explaining the link between hostility and diabetes-related illnesses. It also underlines the importance of identifying a psychosocial profile of older men at higher risk for impaired glucose metabolism.

This work was supported by a grant from the American Heart Association and an award from the National Alliance for Research on Schizophrenia and Depression: The Mental Health Research Association. The Veterans Affairs Normative Aging Study is supported by the Cooperative Studies Program/Epidemiology Research and Information Center, U.S. Department of Veterans Affairs, and is a research component of the Massachusetts Veterans Epidemiology Research and Information Center.