اثربخشی درمان خشم برای مشکلات خاص خشم: بررسی فراتحلیلی

This meta-analytic review analyzed the effects of anger treatment on various aspects of anger with 65% of studies not previously reviewed. To improve on past reviews, this review included only noninstitutionalized adults with demonstrable anger as determined by standardized measures. The studies were compiled from a computer search of published and unpublished anger treatment studies conducted between January 1980 and August 2002. The search resulted in 23 studies containing one or more treatment groups and a control group, with effect sizes derived for each anger problem within each treatment category. The meta-analysis resulted in medium to large effect sizes across therapies. Further analyses of effect sizes within treatment groups by the kind of anger reported support the implementation of cognitive therapies for driving anger, anger suppression, and trait anger. In contrast, relaxation is recommended in cases of state anger. Other implications for treatment and future research directions are discussed, including a special need for research with treatment-seeking individuals and clinical populations.

Anger is often a precipitant of family, work, health, and legal problems. Anger is a negative emotional state that varies in intensity and duration and usually is associated with emotional arousal and a perception of being wronged by another (Kassinove & Sukhodolsky, 1995). The adverse consequences of anger have been described since antiquity. From great philosophers and religious leaders like Seneca, Descartes, and Gandhi came various admonitions to keep one's anger in check (Tavris, 1989). However, following the influences of Freud, the power of the instinct ascended to prominence, and the need to control anger began to be questioned. In fact, the 1960s and 1970s saw the development of a therapy industry designed to aid individuals in venting their angry feelings. Those therapies were called into question after many psychologists showed that expression of anger generally increases anger (e.g., Bushman, Baumeister, & Phillips, 2001). In the last 20 years, we have seen a pendulum swing with increasing numbers of anger management programs. However, before turning to an evaluation of treatments for anger, evidence regarding the negative consequences of anger in the family and health arenas is reviewed. Anger has been linked to various forms of aggression including spouse abuse, child abuse, road rage, and murder. Dobash and Dobash (1984) found that arguments preceded physical aggression in couples 67% of the time. Cascardi, Vivian, and Meyer (1991) found that 100% of husbands and 67% of wives who engaged in acts of physical aggression, such as pushing and slapping, reported that the acts occurred in the context of a verbal argument. Although not all individuals are angry when they argue, when arguments escalate to the point of physical aggression, it is assumed that anger is usually involved. In fact, Boyle and Vivian (1996) found that men in physically aggressive relationships had significantly higher anger scores than a community control group. Regarding parental anger, mothers use physical discipline most in response to child behaviors that make them angry (Peterson, Ewigman, & Vandiver, 1994). Further, parental anger is significantly associated with child abuse risk Kolko, 1996 and Rodriguez & Green, 1997. High anger drivers, compared with low anger drivers, report more automobile accidents, more aggressive driving, and more intense and frequent angry experiences Deffenbacher et al., 2000 and Deffenbacher et al., 2003. In addition, according to the U.S. Department of Justice (2000), 29% of murders were preceded by an argument or disagreement. In summary, anger is associated with a wide variety of negative behaviors that often have negative psychosocial and interpersonal consequences. While anger often leads to negative consequences, it can also encourage positive behaviors and cognitions such as an increase in motivation and goal-setting behaviors. Averill (1983) found that a vast majority of anger-provoking episodes did not result in aggressive acts. More specifically, only 10% of the episodes resulted in a physically aggressive act. Further, anger can result in beneficial interpersonal interactions. For example, about one-third of anger episodes were reported by subjects to have positive outcomes leading to behavioral compliance by others (Tafrate, Kassinove, & Dundin, 2002). In addition, hostile/angry/contemptful interchanges in marriage may lead to immediate compliance; however, in the long-term, they are also predictive of divorce (Gottman, 1994). Anger is not only linked to negative psychological consequences but also increases one's vulnerability to illnesses, compromises the immune system, increases pain, and increases the risk of death from cardiovascular disease (Suinn, 2001). The reasons for such consequences, as suggested by Suinn (2001), may be that anger leads to poorer health behaviors and/or is associated with psychosocial characteristics related to health vulnerability (e.g., frequency of high conflicts). The suppression of anger has also been shown to result in negative consequences for the individual. Anger suppression correlates positively with pain assessment and intensity, pain behaviors, and interference with daily functioning and negatively with pain tolerance Gelkopf, 1997 and Kerns et al., 1994. Furthermore, anger suppression is a stronger predictor of pain intensity than pain history, anger intensity, and depression (Kerns et al., 1994). Suppression of anger is associated with greater ambulatory systolic blood pressure in women but not in men (Helmers, Baker, O'Kelly, & Tobe, 2000). High systolic blood pressure, also called systolic hypertension, has recently been found to be a better blood pressure indicator than diastolic blood pressure for the risk of heart disease and stroke, which are leading causes of death, for middle aged and older individuals (National Institutes of Health, 2000). While the anger construct shares properties with hostility and aggression, the terms are not synonymous, the constructs anger, hostility, and aggression overlap (Spielberger, Reheiser, & Sydeman, 1995), and the distinction among the three concepts can be summarized by referring to anger as the emotion, hostility as the attitude, and aggression as the behavior. Anger is described as an emotional state that can underlie both aggression and hostility. Anger is not considered a behavior or a personality trait (Edmondson & Conger, 1996). Hostility refers to a pervasive aggressive attitude that directs an individual toward aggressive behaviors (Spielberger, Jacobs, Russell, & Crane, 1983), while aggression is defined as an observable behavior with an intention to do harm. Indeed, anger and hostility are moderately correlated in clinic and nonclinic samples, but the constructs share less than 50% of their variance (Boyle & Vivian, 1996). In addition, Spielberger et al. (1983) found moderately high correlations between anger and hostility. To further analyze the relations between the two variables, college students were given self-report measures of anger, hostility, anxiety, and curiosity. A four-factor model resulted in separate factors for anger and hostility, supporting the notion that the two constructs are distinct but related. In brief, in several different samples, anger, hostility, and aggression are correlated but represent quite different concepts. Currently, no criteria for a diagnosis of an anger disorder per se are outlined in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV; American Psychiatric Association, 1994). Anger serves only as a contributor and not a defining characteristic for any mental disorder in the DSM-IV (i.e., oppositional defiant disorder, post-traumatic stress disorder, and borderline personality disorder; pp. 94, 428, and 654, respectively). However, a cluster of symptoms has successfully differentiated high anger from low anger populations. As described by Deffenbacher (1996), compared with low anger students, high anger students have more intense anger across various situations, tend to suppress anger or express it negatively and with more serious consequences, cope poorly with common stressors, report lower self-esteem, and are at greater risk to abuse drugs and alcohol. High anger individuals have also been shown to exhibit biases in their appraisals of situations, blaming others for bad events, and judging that the behavior was intentional (Hazebroek, Howells, & Day, 2001). In addition, a community sample of high anger adults reported more frequent, intense, and long-lasting anger episodes than their low anger counterparts (Tafrate et al., 2002). Further, high anger individuals “were almost twice as likely to engage in some type of negative verbal response, three times more likely to act physically aggressive, and three times more likely to use substances” (p. 1581). Anger clearly has several important correlates and is predictive of individual and psychosocial problems. Several meta-analytic reviews of the treatment of anger Beck & Fernandez, 1998, Edmondson & Conger, 1996 and Tafrate, 1995 have been published, and the general findings for each are briefly discussed in chronological order. The initial review by Tafrate (1995) indicated effect sizes ranging from 0.82 to 1.16 (i.e., moderate to strong effect sizes) for cognitive, relaxation-based, skills training, and multicomponent therapies. In the second meta-analysis, Edmondson and Conger (1996) reviewed the relations between assessment methods and treatment outcomes for anger problems. This review indicated effect sizes ranging from 0.64 to 0.80 for the treatment of anger and anger-related issues using cognitive, cognitive-relaxation, social skills, and relaxation therapies. The review by Beck and Fernandez (1998) included only treatments that contained both cognitive and behavioral components (e.g., cognitive restructuring plus relaxation). The reviewers found a weighted mean effect size of 0.70 for the treatment of anger in children and adults. In summary, these meta-analytic reviews provide us with very important findings regarding the utility of anger treatments.1 They show similar effect sizes that are in the medium to large range, and the results suggest that effective interventions are available for the treatment of adult anger problems. However, these reviews have several limitations that are discussed below. First, the reviews did not limit their scope to samples that met specific criteria for anger. Thus, it is not known if the treated individuals had clinically significant levels of anger. Secondly, the reviews by Beck and Fernandez (1998) and Tafrate (1995) included studies with highly diverse populations such as children, inmates, inpatients, and child-abusing parents. These groups are so different that they might warrant special consideration, and separate analyses of treatment effects with varied populations now seem in order. Third, Edmondson and Conger (1996) and Tafrate limited their reviews to published studies, thereby excluding dissertations and unpublished presentations. Glass, McGaw, and Smith (1981) argue that using only published studies, on average, inflates effect sizes approximately one-third of a S.D. in favor of treatment effects. Fourth, Beck and Fernandez and Edmonson and Conger included measures of assertiveness, hostility, and/or aggression as measures of the anger construct. As such, the effect sizes derived for these measures are not clearly indicators of change in anger alone. As outlined previously, the anger construct is considered distinct from the concepts of hostility and aggression and therefore merits separate analyses. Fifth, the studies included in the meta-analyses Edmondson & Conger, 1996 and Tafrate, 1995 varied in treatment characteristics that may have affected treatment outcome, such as session type (group vs. individual) and session length. These characteristics were not evaluated for their differential effects. In addition, the effect sizes obtained by Tafrate for relaxation-based therapies ranged from 0.36 to 3.47. Effect sizes were not tested for homogeneity, and it is unclear as to whether moderators may be responsible for the variance. The two largest effect sizes (2.02 and 3.47) were derived from samples of female nursing students given 15 individual treatment sessions. These studies involving individual treatment of nursing students differed from the remaining studies, which were primarily college students receiving six to eight group sessions. The overall average effect size for relaxation-based therapies of 1.16 decreases to 0.71 when the above two studies are removed. Finally, Beck and Fernandez did not limit studies to those that contained control groups and used statistics derived from pretest and post-test measures for several effect size computations. Repeated-measures designs may produce larger effect sizes and therefore positively skew the results (Rosenthal, 1994). It is unclear as to what effect the statistical procedures used for this meta-analysis had on the final reported average effect size. To improve on previous meta-analyses, the present review included only studies in which the subjects displayed clinically significant levels of anger as evidenced by scores on standardized anger measures prior to treatment. Standardized measures were used to obtain information regarding clinically relevant cutoff scores and to minimize the variation between studies due to differences in reliability and validity. In addition, effect sizes were derived only from anger measures, not from hostility or aggression scales. Results for both published and unpublished studies were included. Furthermore, new studies on driving anger were included that have not yet been evaluated in a meta-analytic review. In this review, moderator analyses were performed on obtained effect sizes to determine whether there were differential effects as a result of session type and length, sample type, publication status, author, or type of control condition (i.e., minimal treatment vs. no treatment). Lastly, the present article provides effect sizes by type of anger problem and treatment category. This analysis of effect sizes allows for recommendations of different treatment approaches for different anger problems. Of the final sample of 23 articles, 6 were included in the Beck and Fernandez (1998) meta-analysis, 8 were included in the Edmondson and Conger (1996) meta-analysis, and 8 were included in the Tafrate (1995) meta-analysis. This meta-analysis contains 15 articles (65%) that are not represented in previous reviews. The present meta-analytic review focuses on adult outpatients. It does not include child, incarcerated, or institutionalized populations. We chose to focus on adult outpatient populations for several reasons. Clinicians generally identify themselves as adult or child clinicians; in turn, they seek to interpret treatment outcome results for the population they routinely treat. Further, discussion of the advantages and disadvantages of anger treatments are made much easier when one considers one general population (i.e., adult outpatients; as did Edmondson & Conger, 1996) rather than highly varied populations. Finally, the field has grown to an extent that an analysis of different populations is now possible because of the number of studies with adult outpatients. The goal of this article was to review the empirical literature on anger treatment studies from January 1980 to August 2002. We identified treatment effects differentiated by anger subtype presentations to test matching the importance of specific treatments for specific anger problems. In this article, empirical outcome studies were reviewed using meta-analytic procedures. Meta-analysis standardizes the treatment findings allowing for the integration and interpretation of treatment results across studies. This method offers an advantage over the commonly used “vote-counting method” that simply contrasts the number of studies showing significant results to the number of studies that do not. Such a method is insensitive to differences in the power to detect treatment effects as a result of varying sample sizes. Small samples may produce treatment effects that are not rendered significant due to their limited power (Gillett, 2001). Meta-analysis provides sensitivity to sample size differences by weighting effect sizes by their corresponding sample size while increasing the power to detect treatment effects that would be overlooked in a traditional review. Meta-analytic reviews also provide an opportunity to assess the magnitude of treatment effects of the body of literature as a whole as well as of a subpopulation of studies. This analysis enables reviewers to evaluate differences across studies and develop hypotheses regarding moderators and mediators to treatment effectiveness.