- Email: [email protected]
Anger management style and the prediction of treatment outcome among male and female chronic pain patients
BEHAVIOUR RESEARCH AND THERAPY
PERGAMON
Behaviour Research and Therapy 36 (1998) 1051±1062
Anger management style and the prediction of treatment outcome among male and female chronic pain patients John W. Burns a, *, Barbara J. Johnson a, James Devine b, Neil Mahoney b, Ronald Pawl b a
Finch University of Health Sciences/The Chicago Medical School, Department of Psychology, North Chicago, IL, USA b Lake Forest Hospital, Center for Rehabilitation, Lake Forest, IL, USA Received 9 March 1998
Abstract Anger is a prominent emotion experienced by chronic pain patients. Anecdotes suggest that anger predicts poor outcome following multidisciplinary pain programs, but no empirical evidence documents this link. We expected that patient anger expression or suppression would predict poor outcome following a pain program and that gender dierences would emerge. Pre- to posttreatment measures of lifting capacity, walking endurance, depression, pain severity and activity level were collected from 101 chronic pain patients. An `anger expression gender' interaction was found such that anger expression among males was correlated negatively with lifting capacity improvements. `Anger suppression gender' interactions emerged such that anger suppression among males was correlated negatively with improvements in depression and general activities. These eects remained signi®cant after controlling for trait anger. Thus, how anger is managed may exert unique in¯uence on outcomes apart from the eects of mere anger proneness, at least among male pain patients. # 1998 Elsevier Science Ltd. All rights reserved.
1. Introduction Mounting evidence indicates that anger is a salient component of the emotional experience of chronic pain (Wade et al., 1990; Burns et al., 1996a; Kerns et al., 1994). Findings show that dimensions of anger correlate with pain severity and disability among chronic pain patients even when the detrimental eects of depressed mood are controlled (Kerns et al., 1994; Burns et al., 1996a). Clinical anecdotes suggest that anger may also adversely aect outcome * Corresponding author. Tel.: +1-847-578-8751; Fax: +1-847-578-8765.
0005-7967/98/$19.00 # 1998 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 5 - 7 9 6 7 ( 9 8 ) 0 0 0 8 0 - 1
1052
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
following multidisciplinary pain programs (Fernandez and Turk, 1995). Investigators speculate that some patients may not fully bene®t from pain programs because of anger and resentment about the chronicity of their condition, about the quality of healthcare they have received, about employment and compensation, etc. (Fernandez and Turk, 1995). To date, however, no empirical work has been reported which actually documents the putative link between patient anger and poor treatment outcome. In the present study, further analyses of data presented by Burns et al. (1998) were performed to determine whether treatment outcome following a multidisciplinary pain program can be predicted by anger variables. Such programs commonly combine cognitive-behavioral and functional restoration components; treatment modalities which require active, invested participation by patients to achieve success (Turk and Rudy, 1991). Factors which interfere with patients fully engaging in the treatment components or putting forth sucient eort could subvert their potential to show improvement. `Anger management style' is one dimension of anger which is de®ned as the manner in which angry feelings are typically handled. Individuals are said to either verbally express and display anger or to inhibit the expression of anger and suppress angry feelings (Spielberger et al., 1985). Both the inclinations to express or suppress anger have been implicated in the development and/or maintenance of poor physical health (e.g. Siegman et al., 1987; Harburg et al., 1991), including chronic pain conditions (Hatch et al., 1991; Burns et al., 1996a). The manner in which anger is managed could aect pain treatment outcome in a variety of ways. The tendency to express anger could aggravate pain via increased low back muscle tension (Burns, 1997) or the direct expression of vexation and frustration to sta and therapists could interfere with the establishment of sound therapeutic alliances (Burns et al., in press). The inclination to suppress anger may deter patients from fully exploring their frustrations and displeasures with chronic pain during cognitive-behavioral therapy sessions, and this deeply rooted tendency may be inadequately addressed in relatively short (i.e. 4 weeks) multidisciplinary pain programs. Irrespective of mechanism, we expected that high anger expressors or high anger suppressors would fare poorly in multidisciplinary treatment across multiple indexes of outcome. While a plausible argument can be made for the role of anger management style, mere anger proneness may also aect outcome. `Trait anger' is another dimension of anger which describes the tendency to become angry (Siegman, 1993). Although anger expression, suppression and trait anger are conceptually distinct, evidence indicates that measures of these constructs are often correlated. This appears to be the case particularly for scales assessing anger expression and trait anger (e.g. Deenbacher et al., 1996). Given this empirical overlap, signi®cant associations between anger management style and treatment outcome may actually be due to an underlying dimension of anger proneness. That is, treatment outcome may be hampered less by poorly managed anger and more by the tendency to become very angry. To address this concern, we also examined relationships among trait anger and indexes of treatment outcome. If anger expression or suppression exert unique eects on outcome, then we expected these eects to remain signi®cant after controlling for the eects of trait anger. Additionally, we considered the role of gender in moderating links between anger and pain program outcome. In cross-sectional studies, ®ndings suggest that relationships between variables re¯ecting negative aect and adjustment may dier for male and female pain patients.
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
1053
For instance, Haley et al. (1985) found that depression was related positively to pain severity among women but not among men, and Burns et al. (1996a) showed that the interaction of anger management style hostility predicted adjustment dierently in men and women. The few studies which have pursued these promising leads do indeed suggest that variables of negative aect, such as depressed mood (e.g. Burns et al., 1996b), predict outcome dierently in men and women. Given the paucity of research on this topic, however, we oered no explicit hypotheses regarding gender dierences. 2. Method 2.1. Participants Subjects were 101 patients who participated in the multidisciplinary program for the treatment of chronic pain at the Center for Rehabilitation at Lake Forest Hospital, IL. Patients treated in this program suer from benign musculoskeletal pain. Exclusion criteria were: pain due to a malignant condition (e.g. cancer), or to migraine or tension headache, current alcohol or substance abuse problems, a history of psychotic or bipolar disorders and an inability to speak English well enough to complete questionnaires. One hundred twenty-three patients were eligible for this study. Eleven patients dropped out of treatment and 11 had incomplete functional capacity measures due to physical restrictions. The present sample of 101 patients includes 7 patients not included in the sample analyzed in Burns et al. (1988) who had missing data pertinent to the latter study. ANOVAs or w2 tests were used to compare the ®nal sample of 101 patients to the 22 patients who either dropped out of the program or had incomplete data. No signi®cant dierences emerged on pretreatment demographic factors and measures relevant to this study. Demographic information appears in Table 1.
Table 1 Demographic information Statistics Variables Gender (male) Age With at least highschool education Time since injury (months) Underwent one or more surgeries for pain Recipients of worker's compensation Not working Low back pain Anger-out subscale Anger-in subscale Anger content scale
M
S.D.
40.5
9.3
23.6
30.1
15.9 17.3 16.8
4.4 4.5 7.3
%
n
61.0
61
87.0
88
55.4 70.0 71.3 62.0
56 71 72 63
1054
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
2.2. Measures 2.2.1. Independent variables Anger management style was measured with the Anger Expression Inventory (Spielberger et al., 1985). This inventory has subscales which assess the tendency to express anger (anger-out subscale; AOS) and to suppress anger (anger-in subscale; AIS). Spielberger et al. (1985) reported adequate internal consistency coecients for these subscales and Faber and Burns (1996) showed that the AOS correlated with verbal expressions of angry/hostile aect during harassment. Trait anger was assessed with the 16-item, anger content scale (ACS) of the MMPI-2. Adequate psychometrics for this scale have been reported (Butcher et al., 1989). Means and S.D.s for the AOS, AIS and ACS appear in Table 1. 2.2.2. Dependent variables 2.2.2.1. Self-report. The Beck Depression Inventory (BDI; Beck et al., 1961) was used to assess current depressive symptoms and has well-established psychometric properties (Beck et al., 1988). The BDI was given at pre- and posttreatment. The Multidimensional Pain Inventory (MPI; Kerns et al., 1985) was also administered at pre- and posttreatment. Pain severity was measured with the pain severity subscale (PSS) of the MPI. The degree to which patients were able to perform general activities was assessed with the 18 items of Section 3 of the MPI; hereafter referred to as the general activities subscale (GAS). See Table 2 for means and S.D.s. 2.2.2.2. Functional capacity. The Progressive Isoinertial Lifting Evaluation (PILE; Mayer et al., 1988; Curtis et al., 1994) was used to assess lifting capacity. The test consisted of lifting weights in a plastic box which was covered so that patients could not see the number of weights. In this study, women began the PILE with a 2.5 pound weight, while men began with a 5 pound weight. Patients lifted the box from ¯oor to waist height, then lifted the box from waist height to place it on a shelf at shoulder height and then returned the box to the ¯oor. This procedure was completed 10 within 1 min before weight was added. Weight was added in increments of 5 pounds for women and men. The ¯oor to waist to shoulder procedure was then repeated. Table 2 Pre- and posttreatment values for dependent variables (n = 101)
BDI PSS GAS Treadmill time (mins) Treadmill grade Treadmill speed (mph) PILE (lbs)
Pretreatment
Posttreatment
F-test
16.2 4.4 1.9 11.9 0.8 2.1 9.9
11.5 3.9 2.4 19.4 2.8 3.0 18.1
F(1,100) = 35.1a F(1,100) = 23.7a F(1,100) = 23.7a F(1,100) = 132.8a F(1,100) = 87.8a F(1,100) = 130.2a F(1,100) = 170.8a
(9.2) (0.9) (0.9) (3.7) (1.7) (0.7) (5.6)
(8.8) (1.3) (1.0) (6.1) (2.3) (0.7) (10.1)
BDI is the Beck Depression Inventory, PSS the pain severity subscale, GAS the general activity subscale and PILE the Progressive Isoinertial Lifting Evaluation. a p < 0.0001.
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
1055
The PILE stopped when the patient reported fatigue or pain. The result of a PILE was recorded as the maximum weight lifted for 10 repetitions within 1 min. We assessed test±retest reliability of PILE results between the ®rst and third days of the program. The correlation was r = 0.95 ( p < 0.0001), which suggests that the PILE oered a reliable assessment of lifting capacity. Pre- and posttreatment PILE values were de®ned, respectively, as the mean of values recorded on the ®rst and third days of the ®rst week and on the third and ®fth days of the fourth week in the program. See Table 2 for means and S.D.s. A treadmill test was used to tap `walking endurance'. The treadmill test consisted of patients walking on the treadmill at a speed of at least 2 miles hÿ1. They were instructed to walk until fatigued or in pain, while adjusting speed and grade. Complete data on walking endurance was available in the form of `time on treadmill', treadmill speed and ramp grade. As shown in Table 2, means of the 3 treadmill variables increased from pre- to posttreatment. Since `time on treadmill', speed and grade all represent important dimensions re¯ecting the amount of energy expended, we z-scored and summed the 3 variables to form a single composite of treadmill exertion, hereafter referred to as the `treadmill composite'. We assessed test±retest reliability of treadmill composite results between the ®rst and third days in the program. The correlation was r = 0.81 ( p < 0.0001), which suggests that the treadmill composite is a reliable index of walking endurance. Pre- and posttreatment treadmill composite values were de®ned, respectively, as the mean of values recorded on the ®rst and third days of the ®rst week and on the third and ®fth days of the fourth week in the program. 2.3. Procedure Subjects completed questionnaire measures about 1 week prior to starting the program. Subjects participated in an intensive multidisciplinary program (4 weeks, 5 days/week) which was intended primarily to improve psychological and physical functioning. The program included treatment by a physician, physical therapy, occupational therapy, individual and group cognitive-behavioral therapy provided by clinical psychologists, biofeedback and education about chronic pain. Patients participated in at least 2 h of individual and 2 h of group cognitive-behavioral therapy per week and engaged in physical therapies about 5 h per day. On the day of discharge (®fth day of fourth week), subjects again completed questionnaires. 2.4. Statistical analyses To determine whether depression and pain severity decreased and general activities and functional capacity increased from pre- to posttreatment, within-subjects ANOVAs were performed. Next, regressions were performed in which pretreatment values were regressed on posttreatment values to derive residualized change scores. To determine whether certain covariates (see below), anger expression, anger suppression or trait anger predicted these change scores, zero-order correlations were computed. Pretreatment BDI scores were evaluated as a potential covariate so that the eects of depression could be controlled in regressions, which would allow for the assessment of the unique contributions of expressed anger, suppressed anger and trait anger to the prediction of pre- to posttreatment change. Hypotheses that anger management style and gender may interact to predict pre- to posttreatment changes
1056
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
were evaluated with hierarchical regressions. For example, to test whether the `AOS gender' interaction predicted PSS residualized change scores, relevant covariates were entered on the ®rst step, main eect terms on the second step (AOS and a dummy-coded gender variable) and the 2-way interaction term on the third step.
3. Results 3.1. Pre- to posttreatment changes for dependent variables Within-subjects ANOVAs with two levels (pretreatment; posttreatment) were performed for each dependent variable (see Table 3). As shown in Table 2, BDI and PSS scores decreased signi®cantly while GAS, PILE and treadmill time, speed and grade increased signi®cantly from pre- to posttreatment. These ®ndings suggest that the multidisciplinary pain program engendered improvement in depressed mood, pain severity, disability and functional capacity. 3.2. Zero-order correlations Zero-order correlations were generated to evaluate potential covariates and to examine the correlations among all variables (see Table 3). Residualized change scores for the dependent variables were computed by regressing pretreatment values on posttreatment values. Regarding covariates, neither patient age, time since injury, nor pretreatment BDI scores emerged as signi®cant correlates of any change score. Thus, none of these potential covariates were used in subsequent regressions. Noteworthy is that only one signi®cant correlation emerged for the prediction of pre- to posttreatment changes by AOS, AIS or ACS scale scores: namely, that AOS scores were correlated negatively with PILE increases. Thus, high anger expressors improved less on lifting capacity than low anger expressors. Our hypothesis that high anger expressors or high anger suppressors would generally fare more poorly than low expressors or suppressors was not supported. 3.3. Tests of anger management style gender eects Hierarchical multiple regressions were used to test the interaction models. Gender was dummy-coded (women = 0; men = 1) and interaction terms were computed by multiplying relevant variables. The eects of two models, `AOS gender' and `AIS gender', on changes in BDI, PSS, GAS, PILE and treadmill composite values were evaluated. Residualized change scores (see above) were used as dependent variables. The hierarchical procedure involved entering the main eect terms in the ®rst step (e.g. AOS scores, gender) and the 2-way interaction term in the second step (e.g. AOS gender). 3.3.1. Anger expression gender The AOS gender term was a signi®cant predictor of PILE changes (see Table 4). Simple eects tests were conducted by performing regressions separately for men and women. For men, AOS scores were associated negatively with PILE increases (r = ÿ 0.32; p < 0.01),
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
1057
Table 3 Zero-order correlations Variable
(1)
(1) BDI change ± (2) PSS change (3) GAS change (4) Treadmill change (5) PILE change (6) AOS (7) AIS (8) ACS (9) Time since injury (10) Age (11) Pretreatment BDI
(2)
(3)
(4)
0.41c ±
ÿ0.43c ÿ0.04 0.43c ÿ0.10 ± 0.13 ±
(5)
(6)
(7)
0.02 0.01 0.13 0.18 ±
ÿ0.01 0.16 ÿ0.10 0.10 ÿ0.11 ÿ0.17 0.12 0.08 ÿ0.20a ÿ0.03 ± 0.36c ±
(8)
(9)
(10)
(11)
ÿ0.01 ÿ0.05 0.09 0.06 ÿ0.04 0.57c 0.22a ±
ÿ0.15 ÿ0.08 0.12 0.04 ÿ0.01 0.02 ÿ0.05 ÿ0.04 ±
ÿ0.03 0.17 ÿ0.05 ÿ0.05 ÿ0.05 ÿ0.20a ÿ0.06 ÿ0.14 0.03 ±
0.00 ÿ0.02 ÿ0.17 ÿ0.05 ÿ0.17 0.43c 0.52c 0.32b 0.04 ÿ0.11 ±
BDI change, PSS change, GAS change, Treadmill change, PILE change are residualized change scores for the Beck depression inventory, pain severity subscale, general activity subscale, Treadmill composite and progressive isoinertial lifting evaluation, AOS is the anger-out subscale, AIS the anger-in subscale, Ho the Cook-Medley hostility scale and Pretreatment BDI the pretreatment Beck depression inventory. a p < 0.05. bp < 0.001. cp < 0.0001.
whereas among women, this association was nonsigni®cant (r = 0.00; n.s.). For men, an additional regression analysis showed that when ACS scores were entered ®rst and AOS scores second, the latter still accounted for signi®cant variance in PILE changes (R2 increment: 0.09; p < 0.02). The AOS gender term did not signi®cantly predict other outcome variables. Results imply that male patients who are high anger expressors show smaller improvements in lifting capacity than males who are low anger expressors and this eect does not appear to be due to eects of trait anger. 3.3.2. Anger suppression gender The AIS gender term was a signi®cant predictor of GAS and BDI changes (see Table 5). Simple eects tests were conducted by performing regressions separately for men and women. Table 4 Summary of hierarchical regression analyses: AOS gender predicting pre- to posttreatment PILE changes R2
R2 change for step
Signi®cance of R2 change for step
DV is residualized change scores for PILE values Step 1 Gender 0.34 0.20 0.17 AOS ÿ0.05 0.02 ÿ0.22a
0.074
0.074
p < 0.02
Step 2 Gender AOS
0.109
0.035
p < 0.05
Variable
B
ÿ0.09
SE B
0.04
b
ÿ0.74a
PILE is the progressive isoinertial lifting evaluation and AOS the anger-out subscale. a p < 0.05.
1058
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
Table 5 Summary of hierarchical regression analyses: AIS gender predicting pre- to post-treatment GAS and BDI changes Variable
B
SE B
b
DV = Residualized changes scores for GAS scores Step 1 Gender ÿ0.10 0.20 ÿ0.05 AIS ÿ0.04 0.02 ÿ0.16 Step 2 Gender AIS ÿ0.08 0.05 ÿ0.821 DV = Residualized changes scores for BDI scores Step 1 Gender 0.28 0.20 0.14 AIS 0.03 0.02 0.15 Step 2 Gender AIS 0.09 0.04 0.901
R change for step
Signi®cance of R2 change for step
0.031
0.031
ns
0.067
0.037
p < 0.05
0.044
0.044
ns
0.087
0.043
p < 0.04
R
2
2
Note. GAS = General Activity Subscale; AIS = Anger-in Subscale; BDI = Beck Depression Inventory. 1 p < 0.05.
For men, AIS scores correlated negatively with GAS increases (r = ÿ 0.32; p < 0.01) and negatively with BDI decreases (r = ÿ 0.33; p < 0.01). For women, these links were nonsigni®cant (r's = 0.08, 0.11; respectively). For men, additional regression analyses showed that when ACS scores were entered ®rst and AIS scores second, the latter still accounted for signi®cant variance in GAS and BDI changes (R2 increments: 0.10, p < 0.01; 0.09, p < 0.02; respectively). The AIS gender term did not signi®cantly predict other outcome variables. Results imply that male patients who are high anger suppressors exhibit smaller improvements in general activity and depressed mood than low anger suppressors. Moreover, these eects do not appear to be due to eects of trait anger. 4. Discussion Anger has been linked to poor adjustment among chronic pain patients and clinical anecdotes hint that it may detrimentally aect response to multidisciplinary pain programs. Our goal was to determine whether patient anger management style predicted pre- to posttreatment improvements across multiple indexes and did so after taking into account levels of trait anger. We found that anger expression was related signi®cantly to only one outcome variable as a main eect and anger suppression was related to none. However, we found support for the notion that men and women dier with regard to how anger management style aects pre- to posttreatment changes. Among men, our ®ndings suggest that the tendency to express anger may predict poor outcomes in functional capacity, whereas anger suppression may adversely aect improvements in general activity and depressed mood. Among women, anger management style did not signi®cantly predict changes in any outcome measure. Finally, anger expression and suppression remained signi®cant predictors after controlling for trait
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
1059
anger, which suggests that how anger is managed exerts a unique in¯uence on outcome apart from the eects of mere anger proneness. We found that changes in lifting capacity and walking endurance were virtually uncorrelated with changes in pain severity, general activity and depression. These results suggest that improvements in functional capacities and improvements in pain, disability and mood may represent largely independent domains of pain treatment outcome. Thus, our ®ndings suggest that eects of anger expression on outcome are con®ned to only one area: namely, functional capacity. In functional restoration settings, therapy often entails patients experiencing increased pain during stretching, weight lifting and walking. Physical and occupational therapists must encourage patients to exert themselves to improve functional capacities despite transient increases in discomfort. Patients may become irritated, exasperated and angry as a result. On one level, therapists may be less willing to encourage or even approach patients who lash out with verbal or behavioral expressions of anger and so these patients may not bene®t suciently from interactions with therapists. On another level, frequent expressions of anger by patients toward therapists could foster an atmosphere of con¯ict between patient and therapist, thereby undermining the quality of the therapeutic relationship. To the extent that the therapeutic relationship is an active ingredient of diverse therapies ranging from psychotherapy (Horvath, 1994) to pharmacotherapy (Krupnick et al., 1996), poor or damaged relationships in a pain program may adversely impact patients' chances of reaping adequate treatment gains. Our null ®ndings for links between anger suppression and lifting capacity may be interpreted as consistent with the notion that patients who tend to express anger may help create a milieu of contention and antagonism with their therapists. Suppressed anger simply may not aect patient±therapist exchanges and the therapeutic relationship so directly and adversely. Alternatively, it may be that mere anger could have damaging eects on performance during functional restoration such that general patient irritability or resentment toward a paininducing treatment regimen could contribute to low motivation and eort. Trait anger, however, was not related signi®cantly to changes in lifting capacity, nor did its statistical control attenuate the association between anger expression and such changes. Thus, the expression of anger, not merely a proneness to become angry, may uniquely in¯uence changes in some functional capacities. Anger expression, however, signi®cantly predicted lifting capacity only among men. This gender dierence could be attributed to men having a greater likelihood of expressing anger than women, but the mean AOS score for men (M = 16.0, S.D. = 4.32) was indistinguishable from the mean score for women [M = 15.68, S.D. = 4.58; F(1,99) < 1], which is consistent with past ®ndings (e.g. Thomas and Williams, 1991). It may the case instead that the expression of anger by male patients to physical and occupational therapists, who were exclusively female at our study site, may appear more aggressive and truculent than similarly frequent expressions by female patients. Findings of Deenbacher et al. (1996) suggest not only that males have higher physical and verbal assaultive expression than women, but that males experience greater negative consequences of doing so than women. Thus, the belligerence of male anger expressions may have a profound negative impact on the therapeutic relationship. Gender dierences in how anger is conveyed by patients and interpreted by therapists will require further research.
1060
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
Suppressed anger was related signi®cantly to poor improvements in indexes of the other outcome domain: namely, general activity and depression. These factors are typically addressed by psychologists in multidisciplinary programs with cognitive-behavioral therapy. Patients who suppress anger, however, may not bene®t fully from such techniques. Anger suppressors may be reluctant to express their frustrations and distress regarding chronic pain in therapy, and may not give voice to, and fully explore the cognitions which support these negative emotions (Beutler et al., 1986; Corbishley et al., 1990). Patients who suppress anger may therefore ®nd it dicult to reappraise their pain and disability in ways which invoke less emotional distress and to modify maladaptive coping strategies. To the extent that general activity and depression are improved through changes in pain-related cognition and coping (e.g. Jensen et al., 1994; Burns et al., 1988), the tendency to suppress anger may undermine this therapeutic process. Again, however, it may be that merely being angry could thwart therapeutic gains during cognitivebehavioral therapy. Patients' general irritability about their pain condition or resentment toward a treatment regimen they do not understand could contribute to low motivation and eort. Nonetheless, as with anger expression, trait anger was not related signi®cantly to changes in general activity and depression, nor did its statistical control reduce the links between anger suppression and such changes. Thus, the suppression of anger, not merely anger proneness, may uniquely in¯uence changes in general activities and depression. The eects for anger suppression were, like those for anger expression, con®ned to men. Also like results for anger expression, this gender dierence could not be ascribed to men scoring higher on the AIS than women; that is, to male patients as a whole having a greater tendency to suppress anger than female patients. AIS scores for men (M = 17.50, S.D. = 4.73) did not dier from those of women [M = 16.83, S.D. = 4.31; F(1,99) < 1], which is consistent with past ®ndings (e.g. Thomas and Williams, 1991). Currently, it is not clear why anger suppression would aect depression and activity outcomes only among men. It may be that the tendency to suppress anger is part of dierent constellations of characteristics in men and women. Since anger suppression among women is consistent with gender role stereotypes and may thus be construed as `normal' for them, anger suppression may be part of a relatively unpathological pro®le. Among men, however, the tendency to suppress anger may be a component of a more pathological pro®le. That is, a pro®le described by a general inability to articulate emotions, explore cognitions and reappraise pain, which could undermine the process of short-term cognitive-behavioral treatment (Beutler et al., 1986). Our results suggest that both anger expression and suppression may represent legitimate targets for intervention in multidisciplinary pain programs, particularly among men. One weakness with our study, however, is that the mechanisms by which patient anger actually aects outcome were not delineated (e.g. poor therapeutic relationships). Before current treatment practices for chronic pain are amended, results need to replicated and mechanisms need to be understood. Still, our results provide preliminary evidence that deliberate attempts to alter patients' anger management style may contribute to overall treatment ecacy. Acknowledgements This investigation was supported in part by BRSG S07 RR05366-28 awarded to the ®rst author by Biomedical Research Grant Program, Division of Research Resources, National
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
1061
Institutes of Health. The authors wish to thank Kathleen Kiselica, M.S., Administrative Director, and the sta of the Center for Rehabilitation of Lake Forest Hospital, IL, for facilitating the conduct of this study. The authors are especially indebted to Lisa Wheeler, K.T., and Mery Slowminski, P.T. for collecting the physical capacity data and for providing technical assistance.
References Beck, A. T., Steer, R. A., & Garbin, M. G. (1988). Psychometric properties of the beck depression inventory: twenty-®ve years of evaluation. Clinical Psychology Review, 8, 77±100. Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4, 561±571. Beutler, L. E., Engle, D., Oro'-Beutler, M. E., & Daldrup, R. (1986). Inability to express intense aect: a common link between depression and pain?. Journal of Consulting and Clinical Psychology, 53, 354±364. Burns, J. W. (1997). Anger management style and hostility: predicting symptom-speci®c physiological reactivity among chronic low back pain patients. Journal of Behavioral Medicine, 20, 505±522. Burns, J. W., Higdon, L., Mullen, J., Lansky, D., & Wei, J. M. (in press). Eects of patient hostility, anger expression and depression on the working alliance in a work hardening program. Annals of Behavioral Medicine. Burns, J. W., Johnson, B. J., Mahoney, N., Devine, J., & Pawl, R. (1996a). Anger management style, hostility and spouse responses: gender dierences in predictors of adjustment among chronic pain patients. Pain, 64, 445±453. Burns, J. W., Johnson, B. J., Mahoney, N., Devine, J., & Pawl, R. (1996b). Gender dierences in predictors of response to multidisciplinary treatment of chronic pain. Pain Research and Management, 1, 149±154. Burns, J. W., Johnson, B. J., Mahoney, N., Devine, J., & Pawl, R. (1988). Cognitive and physical capacity process variables predict long-term outcome following treatment of chronic pain. Journal of Consulting and Clinical Psychology, 66, 434±439. Butcher, J. N., Graham, J. R., Williams, C. L., & Ben-Porath, Y. S. (1989). Development and use of MMPI-2 content scales. Minneapolis: University of Minnesota Press. Corbishley, M. A., Hendrickson, R., Beutler, L. E., & Engle, D. (1990). Behavior, aect and cognition among psychogenic pain patients in group expressive psychotherapy. Journal of Pain and Symptom Management, 5, 241±248. Curtis, L., Mayer, T. G., & Gatchel, R. J. (1994). Physical progress and residual impairment quanti®cation after functional restoration. Part III. Isokinetic and isoinertial lifting capacity. Spine, 19, 401±405. Deenbacher, J. L., Oetting, E. R., Lynch, R. S., & Morris, C. (1996). The expression of anger and its consequences. Behaviour Research and Therapy, 34, 575±590. Faber, S. D., & Burns, J. W. (1996). Anger management style, degree of expressed anger and gender in¯uence cardiovascular recovery from interpersonal harassment. Journal of Behavioral Medicine, 19, 31±53. Fernandez, E., & Turk, D. C. (1995). The scope and signi®cance of anger in the experience of chronic pain. Pain, 61, 165±175. Haley, W. E., Turner, J. A., & Romano, J. M. (1985). Depression in chronic pain patients: relation to pain, activity and sex dierences. Pain, 23, 337±343. Harburg, E. H., Gleiberman, L., Russell, M., & Cooper, L. (1991). Anger coping styles and blood pressure in black and white males. Psychosomatic Medicine, 53, 153±164. Hatch, J. P., Schoenfeld, L. S., Boutros, N. N., Seleshi, E., Moore, P. J., & Cyr-Provost, M. (1991). Anger and hostility in tension-type headache. Headache, 31, 302±304. Horvath, A. O. (1994). Research on the alliance. In A. O. Horvath, & L. S. Greenberg (Eds.), The working alliance: theory, research and practice. New York: Wiley. Jensen, M. P., Turner, J. A., & Romano, J. M. (1994). Correlates of improvement in multidisciplinary treatment of chronic pain. Journal of Consulting and Clinical Psychology, 62, 172±179. Kerns, R. D., Rosenberg, R., & Jacob, M. C. (1994). Anger expression and chronic pain. Journal of Behavioral Medicine, 17, 57±67. Kerns, R. D., Turk, D. C., & Rudy, T. E. (1985). The West Haven-Yale multidimensional pain inventory (WHYMPI). Pain, 23, 345± 356. Krupnick, J. L., Sotsky, S. M., Simmens, S., Moyer, J., Elkin, I., Watkins, J., & Pilkonis, P. A. (1996). The role of the therapeutic alliance in psychotherapy and pharmacotherapy outcome: ®ndings in the National Institute of Health Treatment of Depression Collaborative Research Program. Journal of Consulting and Clinical Psychology, 64, 532±539. Mayer, T. G., Barnes, D., Kishino, N., Nichols, G., Gatchel, R., Mayer, H., & Mooney, V. (1988). Progressive isoinertial lifting evaluation. I. A standardized protocol and normative database. Spine, 13, 993±997.
1062
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
Siegman, A. W. (1993). Cardiovascular consequences of expressing, experiencing and repressing anger. Journal of Behavioral Medicine, 16, 539±569. Siegman, A. W. , Dembroski, T. M. , & Ringel, N. (1987). Components of hostility and the severity of coronary artery disease. Psychosomatic Medicine, 49, 127±135. Spielberger, C. D., Johnson, E. H., Russell, S. F., Crane, R. J., Jacobs, G. A., & Worden, T. J. (1985). The experience and expression of anger: construction and validation of an anger expression scale. In M. A. Chesney & R. H. Rosenman (Eds.), Anger and hostility in cardiovascular and behavioral disorders. Washington, DC: Hemisphere Publishing. Thomas, S. P., & Williams, R. L. (1991). Perceived stress, trait anger, modes of anger expression and health status of college men and women. Nursing Research, 40, 303±307. Turk, D. C., & Rudy, T. E. (1991). Neglected topics in the treatment of chronic pain patients Ð relapse, noncompliance and adherence enhancement. Pain, 44, 5±28. Wade, J. B., Price, D. D., Hamer, R. M., Schwartz, S. M., & Hart, R. P. (1990). An emotional component analysis of chronic pain. Pain, 40, 303±310.
PERGAMON
Behaviour Research and Therapy 36 (1998) 1051±1062
Anger management style and the prediction of treatment outcome among male and female chronic pain patients John W. Burns a, *, Barbara J. Johnson a, James Devine b, Neil Mahoney b, Ronald Pawl b a
Finch University of Health Sciences/The Chicago Medical School, Department of Psychology, North Chicago, IL, USA b Lake Forest Hospital, Center for Rehabilitation, Lake Forest, IL, USA Received 9 March 1998
Abstract Anger is a prominent emotion experienced by chronic pain patients. Anecdotes suggest that anger predicts poor outcome following multidisciplinary pain programs, but no empirical evidence documents this link. We expected that patient anger expression or suppression would predict poor outcome following a pain program and that gender dierences would emerge. Pre- to posttreatment measures of lifting capacity, walking endurance, depression, pain severity and activity level were collected from 101 chronic pain patients. An `anger expression gender' interaction was found such that anger expression among males was correlated negatively with lifting capacity improvements. `Anger suppression gender' interactions emerged such that anger suppression among males was correlated negatively with improvements in depression and general activities. These eects remained signi®cant after controlling for trait anger. Thus, how anger is managed may exert unique in¯uence on outcomes apart from the eects of mere anger proneness, at least among male pain patients. # 1998 Elsevier Science Ltd. All rights reserved.
1. Introduction Mounting evidence indicates that anger is a salient component of the emotional experience of chronic pain (Wade et al., 1990; Burns et al., 1996a; Kerns et al., 1994). Findings show that dimensions of anger correlate with pain severity and disability among chronic pain patients even when the detrimental eects of depressed mood are controlled (Kerns et al., 1994; Burns et al., 1996a). Clinical anecdotes suggest that anger may also adversely aect outcome * Corresponding author. Tel.: +1-847-578-8751; Fax: +1-847-578-8765.
0005-7967/98/$19.00 # 1998 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 5 - 7 9 6 7 ( 9 8 ) 0 0 0 8 0 - 1
1052
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
following multidisciplinary pain programs (Fernandez and Turk, 1995). Investigators speculate that some patients may not fully bene®t from pain programs because of anger and resentment about the chronicity of their condition, about the quality of healthcare they have received, about employment and compensation, etc. (Fernandez and Turk, 1995). To date, however, no empirical work has been reported which actually documents the putative link between patient anger and poor treatment outcome. In the present study, further analyses of data presented by Burns et al. (1998) were performed to determine whether treatment outcome following a multidisciplinary pain program can be predicted by anger variables. Such programs commonly combine cognitive-behavioral and functional restoration components; treatment modalities which require active, invested participation by patients to achieve success (Turk and Rudy, 1991). Factors which interfere with patients fully engaging in the treatment components or putting forth sucient eort could subvert their potential to show improvement. `Anger management style' is one dimension of anger which is de®ned as the manner in which angry feelings are typically handled. Individuals are said to either verbally express and display anger or to inhibit the expression of anger and suppress angry feelings (Spielberger et al., 1985). Both the inclinations to express or suppress anger have been implicated in the development and/or maintenance of poor physical health (e.g. Siegman et al., 1987; Harburg et al., 1991), including chronic pain conditions (Hatch et al., 1991; Burns et al., 1996a). The manner in which anger is managed could aect pain treatment outcome in a variety of ways. The tendency to express anger could aggravate pain via increased low back muscle tension (Burns, 1997) or the direct expression of vexation and frustration to sta and therapists could interfere with the establishment of sound therapeutic alliances (Burns et al., in press). The inclination to suppress anger may deter patients from fully exploring their frustrations and displeasures with chronic pain during cognitive-behavioral therapy sessions, and this deeply rooted tendency may be inadequately addressed in relatively short (i.e. 4 weeks) multidisciplinary pain programs. Irrespective of mechanism, we expected that high anger expressors or high anger suppressors would fare poorly in multidisciplinary treatment across multiple indexes of outcome. While a plausible argument can be made for the role of anger management style, mere anger proneness may also aect outcome. `Trait anger' is another dimension of anger which describes the tendency to become angry (Siegman, 1993). Although anger expression, suppression and trait anger are conceptually distinct, evidence indicates that measures of these constructs are often correlated. This appears to be the case particularly for scales assessing anger expression and trait anger (e.g. Deenbacher et al., 1996). Given this empirical overlap, signi®cant associations between anger management style and treatment outcome may actually be due to an underlying dimension of anger proneness. That is, treatment outcome may be hampered less by poorly managed anger and more by the tendency to become very angry. To address this concern, we also examined relationships among trait anger and indexes of treatment outcome. If anger expression or suppression exert unique eects on outcome, then we expected these eects to remain signi®cant after controlling for the eects of trait anger. Additionally, we considered the role of gender in moderating links between anger and pain program outcome. In cross-sectional studies, ®ndings suggest that relationships between variables re¯ecting negative aect and adjustment may dier for male and female pain patients.
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
1053
For instance, Haley et al. (1985) found that depression was related positively to pain severity among women but not among men, and Burns et al. (1996a) showed that the interaction of anger management style hostility predicted adjustment dierently in men and women. The few studies which have pursued these promising leads do indeed suggest that variables of negative aect, such as depressed mood (e.g. Burns et al., 1996b), predict outcome dierently in men and women. Given the paucity of research on this topic, however, we oered no explicit hypotheses regarding gender dierences. 2. Method 2.1. Participants Subjects were 101 patients who participated in the multidisciplinary program for the treatment of chronic pain at the Center for Rehabilitation at Lake Forest Hospital, IL. Patients treated in this program suer from benign musculoskeletal pain. Exclusion criteria were: pain due to a malignant condition (e.g. cancer), or to migraine or tension headache, current alcohol or substance abuse problems, a history of psychotic or bipolar disorders and an inability to speak English well enough to complete questionnaires. One hundred twenty-three patients were eligible for this study. Eleven patients dropped out of treatment and 11 had incomplete functional capacity measures due to physical restrictions. The present sample of 101 patients includes 7 patients not included in the sample analyzed in Burns et al. (1988) who had missing data pertinent to the latter study. ANOVAs or w2 tests were used to compare the ®nal sample of 101 patients to the 22 patients who either dropped out of the program or had incomplete data. No signi®cant dierences emerged on pretreatment demographic factors and measures relevant to this study. Demographic information appears in Table 1.
Table 1 Demographic information Statistics Variables Gender (male) Age With at least highschool education Time since injury (months) Underwent one or more surgeries for pain Recipients of worker's compensation Not working Low back pain Anger-out subscale Anger-in subscale Anger content scale
M
S.D.
40.5
9.3
23.6
30.1
15.9 17.3 16.8
4.4 4.5 7.3
%
n
61.0
61
87.0
88
55.4 70.0 71.3 62.0
56 71 72 63
1054
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
2.2. Measures 2.2.1. Independent variables Anger management style was measured with the Anger Expression Inventory (Spielberger et al., 1985). This inventory has subscales which assess the tendency to express anger (anger-out subscale; AOS) and to suppress anger (anger-in subscale; AIS). Spielberger et al. (1985) reported adequate internal consistency coecients for these subscales and Faber and Burns (1996) showed that the AOS correlated with verbal expressions of angry/hostile aect during harassment. Trait anger was assessed with the 16-item, anger content scale (ACS) of the MMPI-2. Adequate psychometrics for this scale have been reported (Butcher et al., 1989). Means and S.D.s for the AOS, AIS and ACS appear in Table 1. 2.2.2. Dependent variables 2.2.2.1. Self-report. The Beck Depression Inventory (BDI; Beck et al., 1961) was used to assess current depressive symptoms and has well-established psychometric properties (Beck et al., 1988). The BDI was given at pre- and posttreatment. The Multidimensional Pain Inventory (MPI; Kerns et al., 1985) was also administered at pre- and posttreatment. Pain severity was measured with the pain severity subscale (PSS) of the MPI. The degree to which patients were able to perform general activities was assessed with the 18 items of Section 3 of the MPI; hereafter referred to as the general activities subscale (GAS). See Table 2 for means and S.D.s. 2.2.2.2. Functional capacity. The Progressive Isoinertial Lifting Evaluation (PILE; Mayer et al., 1988; Curtis et al., 1994) was used to assess lifting capacity. The test consisted of lifting weights in a plastic box which was covered so that patients could not see the number of weights. In this study, women began the PILE with a 2.5 pound weight, while men began with a 5 pound weight. Patients lifted the box from ¯oor to waist height, then lifted the box from waist height to place it on a shelf at shoulder height and then returned the box to the ¯oor. This procedure was completed 10 within 1 min before weight was added. Weight was added in increments of 5 pounds for women and men. The ¯oor to waist to shoulder procedure was then repeated. Table 2 Pre- and posttreatment values for dependent variables (n = 101)
BDI PSS GAS Treadmill time (mins) Treadmill grade Treadmill speed (mph) PILE (lbs)
Pretreatment
Posttreatment
F-test
16.2 4.4 1.9 11.9 0.8 2.1 9.9
11.5 3.9 2.4 19.4 2.8 3.0 18.1
F(1,100) = 35.1a F(1,100) = 23.7a F(1,100) = 23.7a F(1,100) = 132.8a F(1,100) = 87.8a F(1,100) = 130.2a F(1,100) = 170.8a
(9.2) (0.9) (0.9) (3.7) (1.7) (0.7) (5.6)
(8.8) (1.3) (1.0) (6.1) (2.3) (0.7) (10.1)
BDI is the Beck Depression Inventory, PSS the pain severity subscale, GAS the general activity subscale and PILE the Progressive Isoinertial Lifting Evaluation. a p < 0.0001.
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
1055
The PILE stopped when the patient reported fatigue or pain. The result of a PILE was recorded as the maximum weight lifted for 10 repetitions within 1 min. We assessed test±retest reliability of PILE results between the ®rst and third days of the program. The correlation was r = 0.95 ( p < 0.0001), which suggests that the PILE oered a reliable assessment of lifting capacity. Pre- and posttreatment PILE values were de®ned, respectively, as the mean of values recorded on the ®rst and third days of the ®rst week and on the third and ®fth days of the fourth week in the program. See Table 2 for means and S.D.s. A treadmill test was used to tap `walking endurance'. The treadmill test consisted of patients walking on the treadmill at a speed of at least 2 miles hÿ1. They were instructed to walk until fatigued or in pain, while adjusting speed and grade. Complete data on walking endurance was available in the form of `time on treadmill', treadmill speed and ramp grade. As shown in Table 2, means of the 3 treadmill variables increased from pre- to posttreatment. Since `time on treadmill', speed and grade all represent important dimensions re¯ecting the amount of energy expended, we z-scored and summed the 3 variables to form a single composite of treadmill exertion, hereafter referred to as the `treadmill composite'. We assessed test±retest reliability of treadmill composite results between the ®rst and third days in the program. The correlation was r = 0.81 ( p < 0.0001), which suggests that the treadmill composite is a reliable index of walking endurance. Pre- and posttreatment treadmill composite values were de®ned, respectively, as the mean of values recorded on the ®rst and third days of the ®rst week and on the third and ®fth days of the fourth week in the program. 2.3. Procedure Subjects completed questionnaire measures about 1 week prior to starting the program. Subjects participated in an intensive multidisciplinary program (4 weeks, 5 days/week) which was intended primarily to improve psychological and physical functioning. The program included treatment by a physician, physical therapy, occupational therapy, individual and group cognitive-behavioral therapy provided by clinical psychologists, biofeedback and education about chronic pain. Patients participated in at least 2 h of individual and 2 h of group cognitive-behavioral therapy per week and engaged in physical therapies about 5 h per day. On the day of discharge (®fth day of fourth week), subjects again completed questionnaires. 2.4. Statistical analyses To determine whether depression and pain severity decreased and general activities and functional capacity increased from pre- to posttreatment, within-subjects ANOVAs were performed. Next, regressions were performed in which pretreatment values were regressed on posttreatment values to derive residualized change scores. To determine whether certain covariates (see below), anger expression, anger suppression or trait anger predicted these change scores, zero-order correlations were computed. Pretreatment BDI scores were evaluated as a potential covariate so that the eects of depression could be controlled in regressions, which would allow for the assessment of the unique contributions of expressed anger, suppressed anger and trait anger to the prediction of pre- to posttreatment change. Hypotheses that anger management style and gender may interact to predict pre- to posttreatment changes
1056
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
were evaluated with hierarchical regressions. For example, to test whether the `AOS gender' interaction predicted PSS residualized change scores, relevant covariates were entered on the ®rst step, main eect terms on the second step (AOS and a dummy-coded gender variable) and the 2-way interaction term on the third step.
3. Results 3.1. Pre- to posttreatment changes for dependent variables Within-subjects ANOVAs with two levels (pretreatment; posttreatment) were performed for each dependent variable (see Table 3). As shown in Table 2, BDI and PSS scores decreased signi®cantly while GAS, PILE and treadmill time, speed and grade increased signi®cantly from pre- to posttreatment. These ®ndings suggest that the multidisciplinary pain program engendered improvement in depressed mood, pain severity, disability and functional capacity. 3.2. Zero-order correlations Zero-order correlations were generated to evaluate potential covariates and to examine the correlations among all variables (see Table 3). Residualized change scores for the dependent variables were computed by regressing pretreatment values on posttreatment values. Regarding covariates, neither patient age, time since injury, nor pretreatment BDI scores emerged as signi®cant correlates of any change score. Thus, none of these potential covariates were used in subsequent regressions. Noteworthy is that only one signi®cant correlation emerged for the prediction of pre- to posttreatment changes by AOS, AIS or ACS scale scores: namely, that AOS scores were correlated negatively with PILE increases. Thus, high anger expressors improved less on lifting capacity than low anger expressors. Our hypothesis that high anger expressors or high anger suppressors would generally fare more poorly than low expressors or suppressors was not supported. 3.3. Tests of anger management style gender eects Hierarchical multiple regressions were used to test the interaction models. Gender was dummy-coded (women = 0; men = 1) and interaction terms were computed by multiplying relevant variables. The eects of two models, `AOS gender' and `AIS gender', on changes in BDI, PSS, GAS, PILE and treadmill composite values were evaluated. Residualized change scores (see above) were used as dependent variables. The hierarchical procedure involved entering the main eect terms in the ®rst step (e.g. AOS scores, gender) and the 2-way interaction term in the second step (e.g. AOS gender). 3.3.1. Anger expression gender The AOS gender term was a signi®cant predictor of PILE changes (see Table 4). Simple eects tests were conducted by performing regressions separately for men and women. For men, AOS scores were associated negatively with PILE increases (r = ÿ 0.32; p < 0.01),
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
1057
Table 3 Zero-order correlations Variable
(1)
(1) BDI change ± (2) PSS change (3) GAS change (4) Treadmill change (5) PILE change (6) AOS (7) AIS (8) ACS (9) Time since injury (10) Age (11) Pretreatment BDI
(2)
(3)
(4)
0.41c ±
ÿ0.43c ÿ0.04 0.43c ÿ0.10 ± 0.13 ±
(5)
(6)
(7)
0.02 0.01 0.13 0.18 ±
ÿ0.01 0.16 ÿ0.10 0.10 ÿ0.11 ÿ0.17 0.12 0.08 ÿ0.20a ÿ0.03 ± 0.36c ±
(8)
(9)
(10)
(11)
ÿ0.01 ÿ0.05 0.09 0.06 ÿ0.04 0.57c 0.22a ±
ÿ0.15 ÿ0.08 0.12 0.04 ÿ0.01 0.02 ÿ0.05 ÿ0.04 ±
ÿ0.03 0.17 ÿ0.05 ÿ0.05 ÿ0.05 ÿ0.20a ÿ0.06 ÿ0.14 0.03 ±
0.00 ÿ0.02 ÿ0.17 ÿ0.05 ÿ0.17 0.43c 0.52c 0.32b 0.04 ÿ0.11 ±
BDI change, PSS change, GAS change, Treadmill change, PILE change are residualized change scores for the Beck depression inventory, pain severity subscale, general activity subscale, Treadmill composite and progressive isoinertial lifting evaluation, AOS is the anger-out subscale, AIS the anger-in subscale, Ho the Cook-Medley hostility scale and Pretreatment BDI the pretreatment Beck depression inventory. a p < 0.05. bp < 0.001. cp < 0.0001.
whereas among women, this association was nonsigni®cant (r = 0.00; n.s.). For men, an additional regression analysis showed that when ACS scores were entered ®rst and AOS scores second, the latter still accounted for signi®cant variance in PILE changes (R2 increment: 0.09; p < 0.02). The AOS gender term did not signi®cantly predict other outcome variables. Results imply that male patients who are high anger expressors show smaller improvements in lifting capacity than males who are low anger expressors and this eect does not appear to be due to eects of trait anger. 3.3.2. Anger suppression gender The AIS gender term was a signi®cant predictor of GAS and BDI changes (see Table 5). Simple eects tests were conducted by performing regressions separately for men and women. Table 4 Summary of hierarchical regression analyses: AOS gender predicting pre- to posttreatment PILE changes R2
R2 change for step
Signi®cance of R2 change for step
DV is residualized change scores for PILE values Step 1 Gender 0.34 0.20 0.17 AOS ÿ0.05 0.02 ÿ0.22a
0.074
0.074
p < 0.02
Step 2 Gender AOS
0.109
0.035
p < 0.05
Variable
B
ÿ0.09
SE B
0.04
b
ÿ0.74a
PILE is the progressive isoinertial lifting evaluation and AOS the anger-out subscale. a p < 0.05.
1058
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
Table 5 Summary of hierarchical regression analyses: AIS gender predicting pre- to post-treatment GAS and BDI changes Variable
B
SE B
b
DV = Residualized changes scores for GAS scores Step 1 Gender ÿ0.10 0.20 ÿ0.05 AIS ÿ0.04 0.02 ÿ0.16 Step 2 Gender AIS ÿ0.08 0.05 ÿ0.821 DV = Residualized changes scores for BDI scores Step 1 Gender 0.28 0.20 0.14 AIS 0.03 0.02 0.15 Step 2 Gender AIS 0.09 0.04 0.901
R change for step
Signi®cance of R2 change for step
0.031
0.031
ns
0.067
0.037
p < 0.05
0.044
0.044
ns
0.087
0.043
p < 0.04
R
2
2
Note. GAS = General Activity Subscale; AIS = Anger-in Subscale; BDI = Beck Depression Inventory. 1 p < 0.05.
For men, AIS scores correlated negatively with GAS increases (r = ÿ 0.32; p < 0.01) and negatively with BDI decreases (r = ÿ 0.33; p < 0.01). For women, these links were nonsigni®cant (r's = 0.08, 0.11; respectively). For men, additional regression analyses showed that when ACS scores were entered ®rst and AIS scores second, the latter still accounted for signi®cant variance in GAS and BDI changes (R2 increments: 0.10, p < 0.01; 0.09, p < 0.02; respectively). The AIS gender term did not signi®cantly predict other outcome variables. Results imply that male patients who are high anger suppressors exhibit smaller improvements in general activity and depressed mood than low anger suppressors. Moreover, these eects do not appear to be due to eects of trait anger. 4. Discussion Anger has been linked to poor adjustment among chronic pain patients and clinical anecdotes hint that it may detrimentally aect response to multidisciplinary pain programs. Our goal was to determine whether patient anger management style predicted pre- to posttreatment improvements across multiple indexes and did so after taking into account levels of trait anger. We found that anger expression was related signi®cantly to only one outcome variable as a main eect and anger suppression was related to none. However, we found support for the notion that men and women dier with regard to how anger management style aects pre- to posttreatment changes. Among men, our ®ndings suggest that the tendency to express anger may predict poor outcomes in functional capacity, whereas anger suppression may adversely aect improvements in general activity and depressed mood. Among women, anger management style did not signi®cantly predict changes in any outcome measure. Finally, anger expression and suppression remained signi®cant predictors after controlling for trait
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
1059
anger, which suggests that how anger is managed exerts a unique in¯uence on outcome apart from the eects of mere anger proneness. We found that changes in lifting capacity and walking endurance were virtually uncorrelated with changes in pain severity, general activity and depression. These results suggest that improvements in functional capacities and improvements in pain, disability and mood may represent largely independent domains of pain treatment outcome. Thus, our ®ndings suggest that eects of anger expression on outcome are con®ned to only one area: namely, functional capacity. In functional restoration settings, therapy often entails patients experiencing increased pain during stretching, weight lifting and walking. Physical and occupational therapists must encourage patients to exert themselves to improve functional capacities despite transient increases in discomfort. Patients may become irritated, exasperated and angry as a result. On one level, therapists may be less willing to encourage or even approach patients who lash out with verbal or behavioral expressions of anger and so these patients may not bene®t suciently from interactions with therapists. On another level, frequent expressions of anger by patients toward therapists could foster an atmosphere of con¯ict between patient and therapist, thereby undermining the quality of the therapeutic relationship. To the extent that the therapeutic relationship is an active ingredient of diverse therapies ranging from psychotherapy (Horvath, 1994) to pharmacotherapy (Krupnick et al., 1996), poor or damaged relationships in a pain program may adversely impact patients' chances of reaping adequate treatment gains. Our null ®ndings for links between anger suppression and lifting capacity may be interpreted as consistent with the notion that patients who tend to express anger may help create a milieu of contention and antagonism with their therapists. Suppressed anger simply may not aect patient±therapist exchanges and the therapeutic relationship so directly and adversely. Alternatively, it may be that mere anger could have damaging eects on performance during functional restoration such that general patient irritability or resentment toward a paininducing treatment regimen could contribute to low motivation and eort. Trait anger, however, was not related signi®cantly to changes in lifting capacity, nor did its statistical control attenuate the association between anger expression and such changes. Thus, the expression of anger, not merely a proneness to become angry, may uniquely in¯uence changes in some functional capacities. Anger expression, however, signi®cantly predicted lifting capacity only among men. This gender dierence could be attributed to men having a greater likelihood of expressing anger than women, but the mean AOS score for men (M = 16.0, S.D. = 4.32) was indistinguishable from the mean score for women [M = 15.68, S.D. = 4.58; F(1,99) < 1], which is consistent with past ®ndings (e.g. Thomas and Williams, 1991). It may the case instead that the expression of anger by male patients to physical and occupational therapists, who were exclusively female at our study site, may appear more aggressive and truculent than similarly frequent expressions by female patients. Findings of Deenbacher et al. (1996) suggest not only that males have higher physical and verbal assaultive expression than women, but that males experience greater negative consequences of doing so than women. Thus, the belligerence of male anger expressions may have a profound negative impact on the therapeutic relationship. Gender dierences in how anger is conveyed by patients and interpreted by therapists will require further research.
1060
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
Suppressed anger was related signi®cantly to poor improvements in indexes of the other outcome domain: namely, general activity and depression. These factors are typically addressed by psychologists in multidisciplinary programs with cognitive-behavioral therapy. Patients who suppress anger, however, may not bene®t fully from such techniques. Anger suppressors may be reluctant to express their frustrations and distress regarding chronic pain in therapy, and may not give voice to, and fully explore the cognitions which support these negative emotions (Beutler et al., 1986; Corbishley et al., 1990). Patients who suppress anger may therefore ®nd it dicult to reappraise their pain and disability in ways which invoke less emotional distress and to modify maladaptive coping strategies. To the extent that general activity and depression are improved through changes in pain-related cognition and coping (e.g. Jensen et al., 1994; Burns et al., 1988), the tendency to suppress anger may undermine this therapeutic process. Again, however, it may be that merely being angry could thwart therapeutic gains during cognitivebehavioral therapy. Patients' general irritability about their pain condition or resentment toward a treatment regimen they do not understand could contribute to low motivation and eort. Nonetheless, as with anger expression, trait anger was not related signi®cantly to changes in general activity and depression, nor did its statistical control reduce the links between anger suppression and such changes. Thus, the suppression of anger, not merely anger proneness, may uniquely in¯uence changes in general activities and depression. The eects for anger suppression were, like those for anger expression, con®ned to men. Also like results for anger expression, this gender dierence could not be ascribed to men scoring higher on the AIS than women; that is, to male patients as a whole having a greater tendency to suppress anger than female patients. AIS scores for men (M = 17.50, S.D. = 4.73) did not dier from those of women [M = 16.83, S.D. = 4.31; F(1,99) < 1], which is consistent with past ®ndings (e.g. Thomas and Williams, 1991). Currently, it is not clear why anger suppression would aect depression and activity outcomes only among men. It may be that the tendency to suppress anger is part of dierent constellations of characteristics in men and women. Since anger suppression among women is consistent with gender role stereotypes and may thus be construed as `normal' for them, anger suppression may be part of a relatively unpathological pro®le. Among men, however, the tendency to suppress anger may be a component of a more pathological pro®le. That is, a pro®le described by a general inability to articulate emotions, explore cognitions and reappraise pain, which could undermine the process of short-term cognitive-behavioral treatment (Beutler et al., 1986). Our results suggest that both anger expression and suppression may represent legitimate targets for intervention in multidisciplinary pain programs, particularly among men. One weakness with our study, however, is that the mechanisms by which patient anger actually aects outcome were not delineated (e.g. poor therapeutic relationships). Before current treatment practices for chronic pain are amended, results need to replicated and mechanisms need to be understood. Still, our results provide preliminary evidence that deliberate attempts to alter patients' anger management style may contribute to overall treatment ecacy. Acknowledgements This investigation was supported in part by BRSG S07 RR05366-28 awarded to the ®rst author by Biomedical Research Grant Program, Division of Research Resources, National
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
1061
Institutes of Health. The authors wish to thank Kathleen Kiselica, M.S., Administrative Director, and the sta of the Center for Rehabilitation of Lake Forest Hospital, IL, for facilitating the conduct of this study. The authors are especially indebted to Lisa Wheeler, K.T., and Mery Slowminski, P.T. for collecting the physical capacity data and for providing technical assistance.
References Beck, A. T., Steer, R. A., & Garbin, M. G. (1988). Psychometric properties of the beck depression inventory: twenty-®ve years of evaluation. Clinical Psychology Review, 8, 77±100. Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4, 561±571. Beutler, L. E., Engle, D., Oro'-Beutler, M. E., & Daldrup, R. (1986). Inability to express intense aect: a common link between depression and pain?. Journal of Consulting and Clinical Psychology, 53, 354±364. Burns, J. W. (1997). Anger management style and hostility: predicting symptom-speci®c physiological reactivity among chronic low back pain patients. Journal of Behavioral Medicine, 20, 505±522. Burns, J. W., Higdon, L., Mullen, J., Lansky, D., & Wei, J. M. (in press). Eects of patient hostility, anger expression and depression on the working alliance in a work hardening program. Annals of Behavioral Medicine. Burns, J. W., Johnson, B. J., Mahoney, N., Devine, J., & Pawl, R. (1996a). Anger management style, hostility and spouse responses: gender dierences in predictors of adjustment among chronic pain patients. Pain, 64, 445±453. Burns, J. W., Johnson, B. J., Mahoney, N., Devine, J., & Pawl, R. (1996b). Gender dierences in predictors of response to multidisciplinary treatment of chronic pain. Pain Research and Management, 1, 149±154. Burns, J. W., Johnson, B. J., Mahoney, N., Devine, J., & Pawl, R. (1988). Cognitive and physical capacity process variables predict long-term outcome following treatment of chronic pain. Journal of Consulting and Clinical Psychology, 66, 434±439. Butcher, J. N., Graham, J. R., Williams, C. L., & Ben-Porath, Y. S. (1989). Development and use of MMPI-2 content scales. Minneapolis: University of Minnesota Press. Corbishley, M. A., Hendrickson, R., Beutler, L. E., & Engle, D. (1990). Behavior, aect and cognition among psychogenic pain patients in group expressive psychotherapy. Journal of Pain and Symptom Management, 5, 241±248. Curtis, L., Mayer, T. G., & Gatchel, R. J. (1994). Physical progress and residual impairment quanti®cation after functional restoration. Part III. Isokinetic and isoinertial lifting capacity. Spine, 19, 401±405. Deenbacher, J. L., Oetting, E. R., Lynch, R. S., & Morris, C. (1996). The expression of anger and its consequences. Behaviour Research and Therapy, 34, 575±590. Faber, S. D., & Burns, J. W. (1996). Anger management style, degree of expressed anger and gender in¯uence cardiovascular recovery from interpersonal harassment. Journal of Behavioral Medicine, 19, 31±53. Fernandez, E., & Turk, D. C. (1995). The scope and signi®cance of anger in the experience of chronic pain. Pain, 61, 165±175. Haley, W. E., Turner, J. A., & Romano, J. M. (1985). Depression in chronic pain patients: relation to pain, activity and sex dierences. Pain, 23, 337±343. Harburg, E. H., Gleiberman, L., Russell, M., & Cooper, L. (1991). Anger coping styles and blood pressure in black and white males. Psychosomatic Medicine, 53, 153±164. Hatch, J. P., Schoenfeld, L. S., Boutros, N. N., Seleshi, E., Moore, P. J., & Cyr-Provost, M. (1991). Anger and hostility in tension-type headache. Headache, 31, 302±304. Horvath, A. O. (1994). Research on the alliance. In A. O. Horvath, & L. S. Greenberg (Eds.), The working alliance: theory, research and practice. New York: Wiley. Jensen, M. P., Turner, J. A., & Romano, J. M. (1994). Correlates of improvement in multidisciplinary treatment of chronic pain. Journal of Consulting and Clinical Psychology, 62, 172±179. Kerns, R. D., Rosenberg, R., & Jacob, M. C. (1994). Anger expression and chronic pain. Journal of Behavioral Medicine, 17, 57±67. Kerns, R. D., Turk, D. C., & Rudy, T. E. (1985). The West Haven-Yale multidimensional pain inventory (WHYMPI). Pain, 23, 345± 356. Krupnick, J. L., Sotsky, S. M., Simmens, S., Moyer, J., Elkin, I., Watkins, J., & Pilkonis, P. A. (1996). The role of the therapeutic alliance in psychotherapy and pharmacotherapy outcome: ®ndings in the National Institute of Health Treatment of Depression Collaborative Research Program. Journal of Consulting and Clinical Psychology, 64, 532±539. Mayer, T. G., Barnes, D., Kishino, N., Nichols, G., Gatchel, R., Mayer, H., & Mooney, V. (1988). Progressive isoinertial lifting evaluation. I. A standardized protocol and normative database. Spine, 13, 993±997.
1062
J.W. Burns et al. / Behaviour Research and Therapy 36 (1998) 1051±1062
Siegman, A. W. (1993). Cardiovascular consequences of expressing, experiencing and repressing anger. Journal of Behavioral Medicine, 16, 539±569. Siegman, A. W. , Dembroski, T. M. , & Ringel, N. (1987). Components of hostility and the severity of coronary artery disease. Psychosomatic Medicine, 49, 127±135. Spielberger, C. D., Johnson, E. H., Russell, S. F., Crane, R. J., Jacobs, G. A., & Worden, T. J. (1985). The experience and expression of anger: construction and validation of an anger expression scale. In M. A. Chesney & R. H. Rosenman (Eds.), Anger and hostility in cardiovascular and behavioral disorders. Washington, DC: Hemisphere Publishing. Thomas, S. P., & Williams, R. L. (1991). Perceived stress, trait anger, modes of anger expression and health status of college men and women. Nursing Research, 40, 303±307. Turk, D. C., & Rudy, T. E. (1991). Neglected topics in the treatment of chronic pain patients Ð relapse, noncompliance and adherence enhancement. Pain, 44, 5±28. Wade, J. B., Price, D. D., Hamer, R. M., Schwartz, S. M., & Hart, R. P. (1990). An emotional component analysis of chronic pain. Pain, 40, 303±310.