| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Department of Epidemiology, School of Public Health (E.H., M.J.), Biostatistics, School of Public Health (M.A.S.), Psychology (E.H.), Internal Medicine (L.G.), and the Center for Human Growth and Development (N.K.), University of Michigan, Ann Arbor, Michigan.
Address reprint requests to: Ernest Harburg, PhD, 240 East 10th Street, #9B, New York, NY 10003–7702. Email: ernie{at}harburgfoundation.org
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXNOTESREFERENCES |
|---|
METHODS: Anger-coping was measured by responses to hypothetical unfair anger-provoking situations. Cox proportional hazard regressions were used adjusted for seven health risk factors (age, smoking, relative weight, systolic blood pressure (SBP), bronchial problems, FEV1, and cardiovascular (CV) risk).
RESULTS: Men’s suppressed anger interacted significantly with SBP and also with bronchial problems to predict both all-cause and CV mortality. Women showed direct relationships between suppressed anger and early mortality (all-cause, CV, and cancer). Women also showed an interaction of spouse-suppressed anger and SBP for all-cause and CV mortality. Data suggest men who expressed their anger died earlier of cancer (N = 16) deaths.
CONCLUSIONS: Suppressed anger at the time of an unjust attack may become chronic resentment (intermittent rage or hatred) about which little is known and requires research. The design for future research should experimentally measure both suppressed anger-coping responses (after an unfair attack) and morbidity (eg, blood pressure, bronchitis, immune disorder, etc.) to predict prospectively to earlier mortality.
Key Words: anger, • blood pressure, • cancer, • gender, • mortality.
Abbreviations: CV = cardiovascular;; FEV1 = forced expiratory volume in 1 second;; LCES = Life Change Events Study;; SBP = systolic blood pressure;; TCHS = Tecumseh Community Health Study.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXNOTESREFERENCES |
|---|
There is no scientific agreement about the definition of anger; anger involves multidisciplinary knowledge and seems to result in more CV reactivity to imagery and exercise tests than happiness, sadness, or fear (5). Few prospective studies have used measures of "anger" as distinct from "hostility" (6, 7). Heuristic distinctions among "hostility," "anger," and "aggression" by psychologists are proposed by Smith (6). Hostility is defined as "a set of negative attitudes, beliefs, and appraisals concerning others...and connotes a view of others as frequent and likely sources of mistreatment, frustration, and provocation." Anger refers to an "emotion ranging in intensity from irritation to rage, usually in response to perceived mistreatment or provocation...and can be seen as both an emotional state and an enduring personality trait." Aggression refers to "overt behavior...typically defined as attacking, destructive, or hurtful actions."
The focus in this article is on anger-coping responses to an imagined unfair aggressive verbal attack. Our heuristic thesis is that psychophysiological anger responses (including hostility) are automatically induced in unjust attack situations. An "attack" exists when a person appraises (8) an actual loss or a threat of loss of something believed to be owned or possessed (one’s rights, health, status, etc.) through perceived arbitrary (unfair or unjust) action by others (person, group, society) or often by natural events (eg, accidents, death of others). When the attack is sudden and the owned object is strongly valued, then anger will be quick and intense. If these assumptions are valid, then anger-coping responses to an anger-inducing attack can be minimally modeled as either overtly "expressive" or "suppressive" of that anger to the attacker. Alternative measures of anger-coping modes were developed after our baseline measure in 1971 (9–11).
The concept of suppressed anger has a long history (12–14). Our use of the term "suppressed" comes explicitly from Newcomb (15) who conceived that suppression (a semiconscious process) can be observed by what is omitted in structured communication and interaction; we would add "after structured provocation." Other studies on cardiovascular outcomes have used the concept of suppressed anger (conceived and measured in different ways), both in multiyear prospective research (16–18) and in survival after experimental studies (19–20). Besides the studies just cited relating suppressed anger and hypertension, other research has related suppressed anger to rheumatoid arthritis (21), breast cancer (22), and duodenal ulcers (23). These studies allow us to assume that chronic suppressed anger exacerbates a variety of potential or existent pathologic medical conditions and thus eventually leads to early mortality. The precise process of how this exacerbation "interacts" with medical disorders is yet unknown (24).
This conceptual framework was used to construct a dichotomous measure of anger-coping responses based partially on prior research (25) and developed by the lead author (26). Briefly, respondents imagine that a hypothetical unfair anger attack by a socially defined power figure has just occurred. Three items inquire as to the most likely coping responses: either show anger or not show anger to the attacker, not feel guilty or feel guilty (later) if their anger was shown directly, and protest or not protest to the attacker. We then scored these anger-coping responses as "expressed" or "suppressed."
Finally, in regard to gender and suppressed anger, one could expect that both the perception of anger-inducing situations and the anger-coping responses would differ between the genders, ie, most women more than most men tend to inhibit their anger (27, 28) and to "somaticize" (29) their depression (resigned sadness) or distress (angry sadness). As one example, in an earlier national survey, women reported feeling "hurt" when treated arbitrarily but no men used this term; this term was then coded as indicating "anger-in" which was related to female rheumatoid arthritis (21). Thus it seemed logical that women’s (rather than men’s) suppressed anger might be more related to morbidity and earlier mortality.
The specific hypotheses of the present analysis guided testing for: 1) an association between expressed/suppressed anger-coping and blood pressure; 2) an association between expressed/suppressed anger-coping and mortality; and 3) an interactive relationship of expressed/suppressed anger and chronic disease conditions (elevated blood pressure or respiratory problems) with mortality. The association between anger-coping indices and mortality would also be explored for each of these hypotheses by gender.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXNOTESREFERENCES |
|---|
Data were collected in the TCHS clinic from 1971 to 1972. Demographic and health-related information was obtained by a standard TCHS questionnaire. Subjects for the LCES also completed the psychosocial questionnaire. Morbidity was diagnosed in the third series of examinations (1967–1969) for cardiovascular risk, bronchial problems, and FEV1.
Mortality status was ascertained from death certificates and medical reports from 1978 to 1979 for virtually every respondent who had ever participated in the Tecumseh Project. After 1979, mortality status was updated by screening daily reports in local newspapers and by follow-up contact with relatives. To estimate the bias in mortality status after 1979, separate survival models were fitted using the mortality status of 1979 and the mortality status of 1988. The point estimates obtained by the two models were similar. For males, the total anger suppression index using mortality 1979 had an adjusted RR = 1.00 (95% CI = 0.70–1.40) and mortality 1988 had an adjusted RR = 1.01 (95% CI = 0.82–1.23). For females, mortality 1979 yielded an adjusted RR = 1.36 (95% CI = 0.83–2.21) and for mortality 1988 an adjusted RR = 1.44 (95% CI = 1.04–1.98). The ascertainment after 1979 does not seem to severely bias the estimates; therefore the mortality status of 1988 is used in this paper. Between 1971 and 1988, 91 (13.1%) respondents of the total LCES (N = 696) had died, 17.3% of the men (N = 56) and 9.4% of the women (N = 35). These death rates were similar to the full project population.
The types of mortality used in this analysis were: 1) all-cause; 2) cardiovascular, which includes six ICD-9 categories: hypertensive disease (401–405), ischemic heart disease (410–414), diseases of the pulmonary circulation (415–417), other forms of heart disease (420–429), cerebrovascular disease (430–438), and diseases of arteries, arterioles, and capillaries (440–448); and 3) cancer, all types.
Anger-coping types were assessed using a format developed by Harburg et al. (26). All subjects responded to two hypothetical anger-provoking situations involving injustices perpetrated by a power figure termed an (unjustified) "attack" throughout the paper (see Appendix A). For each attack role situation (spouse or police) separate anger, guilt, and protest scores were constructed by recoding responses to the six items as follows: items 1 and 4 (show anger), responses 1, 2 were coded as "0" for "show anger" and responses 3 to 5 were coded as "1" for "not show anger." For items 2 and 5 (guilt), responses 1 to 3 were recoded as "1" for "guilt" and response 4 was recoded as "0" for "no guilt." For items 3 and 6 (protest), item responses 1, 2 were coded as "1" for "not protest" and item responses 3, 4 were coded as "0" for "protest." It is assumed that each recoded item is an indicator of either an expressive (score "0") or suppressive (score "1") anger process. Thus, "show anger," "not feel guilt," and "protest" were each scored "0" and "not show anger," "feel guilt," and "not protest" were each scored "1." This method requires recoding to a two-point scale after the respondent chooses from a five-point scale. We expect this recoding to increase test-retest reliability and validity, though this might result in some loss of statistical power and information. The aim is to describe accurately a two-category measure: express or suppress anger-coping. The increase in reliability occurs because respondents can choose to report (eg, either "anger" or "annoyed") which then can be collapsed into a new response (for example, see Appendix A, item 1, where response values 1 and 2 are collapsed into "show anger"). Thus, minor changes in test-retest values will be merged in such new responses. If two of three recoded responses for each attack situation are "suppress," then that person is categorized as coping by suppressing anger; otherwise they are categorized as coping by expressing anger.
Cumulative anger-guilt-protest indices of suppressed anger were also developed separately for each attack role situation (spouse or police) and for the total index across both situations. These measures were constructed by summing the recoded responses described above. Thus, for each role situation (spouse or police) those persons with a high score (2–3 of 3 items) on an anger-guilt-protest index are more likely to not show their anger, feel guilty, or not protest an unjustified attack. We label these indices as "suppressed anger" and designate the specific role situation as suppressed anger spouse index or suppressed anger police index. The range was 0 to 3 for spouse and police suppressed anger indices. The score of the six items derived from both role situations is labeled as the suppressed anger total index (Appendix B). All suppressed anger indices were used as continuous variables in regression models because no significant differences between using a continuous or ordinal scale were found. The strength of this anger-coping measure relies mostly on its external validity; the measure has been applied by a number of different researchers across a variety of populations (26, 31–35).
Health risk factors were ascertained during the medical test series interviews and medical tests (1971–1972) and included age in years, cigarette smoking, relative weight, blood pressure, and education in years. Blood pressure was measured with a mercury sphygmomanometer. All readings were taken with the subject seated and using his/her right arm. Both systolic pressure and diastolic (fifth phase) pressure were recorded. Three indicators of physical status and morbidity obtained at the third series of examinations from 1967 to 1969 were also included as possible confounding variables. They are: 1) a diagnosis of suspect or probable CHD, defined as a probable history of myocardial infarction, angina, or electrocardiographic evidence of myocardial infarction (Minnesota codes 1–1 or 1–2); 2) bronchial problems (suspect or possible), chronic bronchitis, or persistent cough or phlegm; and 3) FEV1 values adjusted for gender, age, and height using the FEV1 values of the nonsmoking respondents without respiratory disease or symptoms and reversed so that higher scores indicate lower FEV1.
Statistical Methods
The Cox proportional hazard model was used to model the survival curves from 1971 until 1988. The survival time for each individual was calculated as the difference between the date of death and the starting date for the cases who died, and the difference between the date of last contact and the starting date for those who were still alive in 1988, ie, those were censored at the end of the study. The health-related factors (health risk and medical) measured in this study were investigated using Cox modeling to assess the impact of each factor separately on survival for all three types of mortality. Then Cox models including these health-related factors and each anger-coping measure singly and as interaction terms were derived. Finally, as systolic and diastolic measures were highly correlated (for males, r = 0.72; for females, r = 0.78), the inclusion of both in multiple adjustment equations could result in multicollinearity; therefore, in this paper, only regression results for systolic pressure are presented. The education variable was also removed from the model because it was highly insignificant and did not contribute significantly to the log likelihood function.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXNOTESREFERENCES |
|---|
|
.05); the police index showed RR = 1.54 (95% CI = 1.18–2.02, p .01). Again, no effects of the spouse anger index were observed. For cancer deaths, there were no significant effects of any of the anger indices on mortality. In summary, these data show evidence to support our second hypothesis that there is an association of anger-coping and mortality.
|
Figure 1 reinforces the findings of Table 2 using survival curves to describe the actual death experience along a time dimension; curves are dichotomized by expressed/suppressed anger indices for presentation and adjusted for the seven health risk factors. (The statistical significance of these relations are indicated in Table 2.) These curves show that the tendency for earlier deaths by those suppressing anger for all-cause, CV, and cancer mortality occurred throughout the 17-year period, significant only for females. Figure 1, however, also shows the gender differences in the opposite direction for cancer mortality – specifically, for men, those who reported suppression of anger to unjustified verbal attack by spouse and police show more longevity than men who expressed their anger. For men only (Table 2), the police index is significant (RR = 0.52, 95% CI = 0.28–0.97, p < .05); but this result is different in direction from females whose suppressed anger risk ratio for police is RR = 1.50 (p < .10).
|
|
140 mm Hg compared with <140 mm Hg predicts to higher mortality rates as expected. However, data in Figure 2 for males show that the combination of suppressed anger and elevated SBP predicts to significantly higher all-cause mortality than for those with elevated SBP who express their anger. Note the seemingly paradoxical results for men: those who suppress their anger (total index) with low SBP (<140 mm Hg) have almost half (8.6%) the all-cause mortality than the average male mortality rate (17.3%) and their total index shows almost five times fewer deaths than men who also suppress their anger and who have high SBP (140 mm Hg). This effect appears for all three suppressed anger indices and is perhaps due to the contrary direction of anger for cancer mortality. For females, for all-cause and CV mortality, similar bar charts (not shown) supported the significant interaction effects of SBP and the spouse anger index seen in Table 2.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXNOTESREFERENCES |
|---|
Our findings also show as hypothesized that suppressed anger interacted with SBP to predict early mortality for both men and women (Table 3). Conclusions from a small sample (16 male, 16 female college students) suggest that hemodynamic mechanisms after stressors differed consistently between genders (36). The highest reactivity occurred during an experimental role play of a socially unjust situation (but not an aggressive attack). Females showed greater increases in heart rate, cardiac output, and decreased total peripheral resistance compared with men who showed consistent increase in total peripheral resistance to all five stressor modes. Women were more "heart stricken" by these stressor events than men. A study of reactivity as well as recovery from anger provocative stimuli and anger release also showed differential gender-related anger-coping CV responses (37). Are these gender-specific physiologic responses reiterated when the unjust stressor event is rearoused through memory? Are these responses increased with repetitions of the actual unfair attacks as in daily marital or occupational life? If the emotion of anger/fear occurring in these daily events is chronically suppressed, do morbid physical conditions emerge? Or does chronic suppression "interact" with an incipient morbid disequilibria to generate frank morbid disorder?
In another psychophysiological study, 58 college women (38) were asked to "recall a time when they had become very angry" in a probing interview coded later into four anger-coping responses: suppression, cognition, assertion, aggression; heart rate and blood pressure were monitored at baseline and after the interview. Results showed higher SBP reactivity for those women who coped using a suppressed mode and the lowest SBP reactivity with an assertive mode. The highest SBP reactivity was for the combination of Type A women who also suppressed their anger; the lowest SBP was for those Type B women who were assertive. Whether reactivity in experimental results is useful, however, to predict morbidity or mortality requires a combined risk analysis in prospective research (39, 40). In the present prospective data for both men and women, the statistical interaction of a suppressive mode with elevated SBP was significantly related to earlier deaths (Table 3). This result is suggestive of finding early CHD mortality in a 27-year follow-up of men where a triple combination of Type A, high hostility, and elevated SBP showed the highest rates of earlier deaths in survival curves across six health risk groups (41).
It is reasonable to assume that a suppressor response to attack rather than an expressive response would likely lead to an internal process of resentment (10). Our theory explicitly assumes that chronic perception of arbitrary attacks by others results in intermittently but continuously aroused anger/hostility. Each single attack (police) or several over time (spouse) where one’s anger is kept suppressed may be retained in memory, revived in imagery or fantasy, and reexperienced as the original event (5). This private recall of unfair dispossession (of self-esteem, social status, property) can keep rearousing the anger, including hostile attitudes. This private, chronic, iterative rearousal of anger/hostility due to the perception of chronic social attack is the sociopsychobiological toxic process we term chronic resentment. Colloquially, this is termed "hatred" or "rage," which most Americans do not want or are unable to report, and is a form of extreme anger that is only rarely the focus of sociopsychosomatic research.
Chronic resentment can impede communication and interaction by omission of material required for a social resolution of problems and thereby serves to exacerbate the internal intermittent anger responses. To solve this chronic resentful condition involves a perceived restoration of status (eg, apology, reappraisal, restitution, etc.) (42). Being a member of a status group that faces chronic attack due to gender, ethnicity, religion, etc. (43) can induce, in an unknown proportion of persons, the basis for chronic resentment, perhaps pressured by the daily stress of low income (44), low "life chances," and chronic relational discord. Chronic resentment is complex (45) and can induce chronic disequilibrium in bodily systems (24) (eg, circulatory, bronchial, immune) (46–48) generating or influencing chronic morbidity (49) leading to early death. We suggest that the "interaction" of suppressed anger and high blood pressure levels, as in the present research, may be partially generated by chronic resentment and an absence of a reflective mode of anger-coping (try to resolve the problem).
Our results for cancer deaths must be viewed with caution because of the low number of deaths (male = 16, female = 14). These findings, however, show a marked gender disparity: For women, a suppressed anger index was related to cancer death; for men, survival curves and risk ratios show that expressed anger was related to earlier cancer death. An earlier study on men and lung cancer, however, contradicts this direction, ie, suppressed anger was related to increased lung cancer (50). An earlier study of cancer in women showed that the diagnosis of breast cancer was significantly related to a priori measures of both "extreme suppression" of anger and also to a smaller portion with "extreme expression" (51). This result would be termed "resentful" anger (10).
The findings in this study are subject to a number of limitations that urge caution in assessment. The research was carried out in a small, semirural, ethnically homogeneous (Anglo-Saxon) community with little socioeconomic diversity or conflict; generalizing to larger, ethnically diverse, socioeconomically disparate populations would be problematic. Furthermore, whereas the study had complete ascertainment of death certificates up until 1979, after this date to 1988 the deaths were obtained through newspaper obituaries; although death confirmation evidence from relatives was rigorously gathered within this small community, selection effects could not be avoided. Patently, our mortality data are sparse. Finally, the measure of anger-coping used in this study is not part of standard psychological inventory; it has a different measurement strategy and is derived from theory (see Methods).
This study is preliminary. The findings, however, provide further evidence of the relationships between anger and early mortality as suggested by others (1–3). Future research should involve experimental anger-coping-induced reactivity with recovery data and also measure precursors to morbidity or morbid conditions and follow-up with early CV mortality. A measure of sociopsychophysiologic chronic resentment should be developed. This effort first requires concepts and measures of anger-coping response which describe both psychologic and physiologic/chemical changes simultaneously after perceived chronic or traumatic, unfair, aggressive attack.
|
APPENDIX |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXNOTESREFERENCES |
|---|
Appendix B
A total score of the six items from the spouse and police responses (see Methods) was formulated as shown.
|
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXNOTESREFERENCES |
|---|
b The attacker could be a "doctor," "nurse," "boss," "father," or "mother" or any alter-role power figure acting arbitrarily or unfairly in an angry/insulting attack.
c "Guilty or sorry" should be changed to "guilty or afraid."
Received for publication February 24, 2002.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  M. D. Taylor, M. C. Whiteman, G. R. Fowkes, A. J. Lee, M. Allerhand, and I. J. Deary Five Factor Model Personality Traits and All-Cause Mortality in the Edinburgh Artery Study Cohort Psychosom Med, July 1, 2009; 71(6): 631 - 641. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. A. Sanz Martinez, B. H. Schneider, Y. Santana Gonzales, and M. d. P. Soteras de Toro Modalities of anger expression and the psychosocial adjustment of early adolescents in eastern Cuba International Journal of Behavioral Development, May 1, 2008; 32(3): 207 - 217. [Abstract] [PDF]
|
|
|
|
 |
|
 N. S. Consedine, C. Magai, and D. Horton Ethnic Variation in the Impact of Emotion and Emotion Regulation on Health: A Replication and Extension J Gerontol B Psychol Sci Soc Sci, July 1, 2005; 60(4): P165 - P173. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Denollet DS14: Standard Assessment of Negative Affectivity, Social Inhibition, and Type D Personality Psychosom Med, January 1, 2005; 67(1): 89 - 97. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
