A comparison of group anger management treatments for combat veterans with PTSD: Results from a quasi-experimental trial

Abstract

Objective:

Difficulty controlling anger is a significant concern among combat veterans with posttraumatic stress disorder (PTSD), yet few controlled studies have examined the efficacy of anger treatments for this population. This study examined the effects of a group cognitive behavioral therapy (CBT) intervention compared to a group present centered therapy (PCT) control condition in male and female combat veterans with PTSD.

Method:

Thirty-six combat veterans with PTSD and anger difficulties began group treatment (CBT, n = 19; PCT, n = 17). Separate multilevel models of self-rated anger, PTSD symptoms, and disability were conducted using data from baseline, each of 12 treatment sessions, post-treatment, and 3- and 6-month follow-up time points.

Results:

Significant decreases in anger and PTSD symptoms were observed over time, but no significant differences between CBT and PCT were observed on these outcomes. A significant interaction of therapy by time favoring the PCT condition was observed on disability scores. Gender differences were observed in drop-out rates (i.e., 100% of female participants dropped out of CBT).

Conclusions:

Findings suggest that both CBT and PCT group therapy may be effective in reducing anger in combat veterans with PTSD. Results also highlight potential gender differences in response to group anger treatment.

Introduction

U.S. Iraq and Afghanistan combat veterans have reported that controlling anger and aggressive urges are primary readjustment concerns (Sayer et al., 2010; Sippel et al., 2016), and problems with anger and aggression have been found to be particularly pronounced in combat veterans with posttraumatic stress disorder (PTSD; Andrews & Brewin, 2012; Elbogen et al., 2010; Orth & Wieland, 2006). Anger in combat veterans with PTSD is associated with diminished quality of life and functional impairments across domains (Thomas et al., 2010), including ineffective parenting, unsafe driving, employment difficulties, and social alienation (Rodriguez, Holowka, & Marx, 2012). In particular, PTSD and elevated anger has been identified as a risk factor for interpersonal violence among veterans (Miles, Menefee, Wanner, Teten Tharp, & Kent, 2015; Novaco & Chemtob, 2015; Taft et al., 2015), and recent longitudinal research found that anger-related cognitions partially mediated the association between PTSD and aggression in Iraq and Afghanistan-era veterans (Van Voorhees et al., 2016).

Anger-related re-adjustment problems can persist for decades (Biddle, Elliott, Creamer, Forbes, & Devilly, 2002; Koenen, Stellman, Stellman, & Sommer, 2003; Novaco & Chemtob, 2002): A study conducted almost 30 years after the end of hostilities in Vietnam found that Vietnam veterans, their spouses, and their clinicians identified problems with anger as the highest priority among several potential psychiatric concerns, including anxiety, depression, and alcohol problems (Biddle et al., 2002). Moreover, anger has been found to increase the likelihood of drop-out from treatment for PTSD (Rizvi, Vogt, & Resick, 2009) and to diminish the effectiveness of therapy among those who do engage (Foa, Riggs, Massie, & Yarczower, 1995; Forbes, Creamer, Hawthorne, Allen, & McHugh, 2003; Forbes et al., 2008). The increasing awareness among researchers of the impact of anger on the lives of veterans with PTSD is reflected in clinical practice, such that an estimated 35–65% of Department of Veterans Affairs (VA) PTSD specialists report providing anger management to their patients (Rosen et al., 2004). Yet few studies have examined the efficacy of treatments for anger problems in veterans with PTSD, leaving clinicians to address this pressing concern with limited empirical guidance.

For decades, the only study of anger treatment for veterans with PTSD was a small trial comparing 12 individually-delivered sessions of cognitive behavioral therapy (CBT; n=8) to treatment as usual (n=7) in male Vietnam veterans with combat-related PTSD (Chemtob, Novaco, Hamada, & Gross, 1997). In this study there was a high (46%) drop-out rate. Improvements were found at post-treatment on the total score and the anger control subscale of the Speilberger Anger Expression Inventory (STAXI), but not on any of the other subscales or on the Novaco Anger Scale. Gains on the STAXI anger control subscale were maintained 18-month follow-up.

Since that time only three other randomized controlled trials (RCTs) have evaluated individual or group interventions targeting anger or aggression in veterans. Shea and colleagues (2013) adapted Chemtob and colleagues’ (1997) CBT intervention and compared it to a supportive intervention (SI) in 23 male Iraq and Afghanistan War veterans with anger problems and at least one other hyperarousal symptom. Results indicated a dropout rate of 33% from the CBT and 27% from SI. Among those who completed treatment, findings suggested greater improvement in CBT vs. SI from pre- to post-treatment on the State-Trait Anger Expression Inventory-2 (STAXI-2) Anger Expression Index, as well as on each of the Index’s component subscales (Expression-In, Expression-Out, Anger Control-In, Anger Control-Out) (Shea, Lambert, & Reddy, 2013). Taft and colleagues (2016) developed the 12-week cognitive behavioral Strength at Home (SAH) group therapy program to specifically address partner aggression in male veterans with PTSD. An RCT comparing SAH to enhanced treatment as usual (ETAU) in 135 male veterans with a history of partner violence found reductions in partner violence in both arms at post treatment and 3-month follow up. Veterans in the SAH demonstrated significantly lower incidence of partner aggression at post-treatment compared to ETAU, though this difference did not remain at 3-month follow-up (Taft, Macdonald, Creech, Monson, & Murphy, 2016). It should be noted that this study evaluated aggression as the primary outcome, and did not specifically measure anger. Finally, Ford and colleagues (2018) compared 10 sessions of individual therapy using a manualized present-centered emotion regulation therapy (TARGET) to prolonged exposure (PE) in 31 male service members who had been deployed to Iraq or Afghanistan. Results indicated lower dropout in TARGET than in PE; similar improvements in emotion regulation and hostility in both conditions; and no significant improvement in anger as measured by the STAXI trait anger subscale (Ford, Grasso, Greene, Slivinsky, & DeViva, 2018). Of the four RCTs evaluating interventions for anger or aggression in veterans, only the SAH intervention was delivered in a group format. Other uncontrolled trials have found that group CBT for anger may reduce anger and aggression from pre-to post-treatment in male veterans, including measures of state anger (Gerlock, 1994; Marshall et al., 2010; Morland et al., 2010), trait anger (Gerlock, 1994; Marshall et al., 2010), anger expression (Morland et al., 2010), and physical aggression (Marshall et al., 2010).

Taken together, prior studies have suggested that cognitive-behavioral interventions may be effective in reducing anger in veterans. However, only three published studies to date have compared CBT to an active control (Chemtob, Novaco, Hamada, & Gross, 1997; Shea, Lambert, & Reddy, 2013; Taft et al., 2016), and one of these evaluated intimate partner aggression and not anger as the primary outcome (Taft et al., 2016). One controlled study found that present centered emotion regulation therapy reduced hostility but not anger in veterans post-deployment (Ford et al, 2018). Finally, several uncontrolled studies found CBT to be effective in reducing anger in veterans (Gerlock, 1994; Marshall et al., 2010; Morland et al., 2010). Of note, although research suggests that women veterans struggle with anger and aggression as much as male veterans (Kirby et al., 2012; Sullivan & Elbogen, 2013; Worthen et al., 2014, 2015), no study of individual or group therapy for anger or aggression in veterans to date has included women.

The current pilot study was designed to determine the feasibility of a larger RCT, and to characterize the effects of two group interventions for anger and aggression in male and female combat veterans with PTSD. A 12-session, manualized group CBT intervention designed to address information processing biases and physiological hyperarousal associated with anger in combat veterans was compared to a 12-session, manualized group present-centered therapy (PCT) intervention. We hypothesized that veterans in the CBT condition would show greater reductions in self-reported anger and aggression than veterans in the PCT condition. Secondary outcomes included self-report measures of PTSD symptoms and overall disability.

Method

This study was a pilot, two-arm parallel controlled trial comparing group cognitive behavioral therapy (CBT) to group present centered therapy (PCT). The study was approved by the [blinded for review] Institutional Review Board, and has been registered with ClinicalTrials.gov [registration number blinded for review]. All patients gave written informed consent before participation. Data collection occurred between May 1, 2014 and July 5, 2017. Participants were compensated for their participation. Target enrollment was 6 groups (4 all-male groups, 2 all-female groups) of 5 to 6 veterans per group (N=36), with 3 groups of CBT and 3 groups of PCT. Optimal randomization of participants to groups would have required enrollment of 12 eligible participants prior to group assignment (Schnurr, Friedman, Lavori, & Hsieh, 2001), which would have resulted in significant delays between screening and the initiation of treatment. However, requiring participants to wait for several months between study enrollment and treatment for anger and aggression may be ethically questionable, and may also increase the likelihood that a high percentage would drop out of the study prior to the first session to seek treatment elsewhere. As a result, we began each group immediately upon enrollment of 6 participants, alternating the initiation of CBT and PCT groups.

Participants

Eligible participants were veterans 18 years of age or older who met criteria for current posttraumatic stress disorder based on the Clinician Administered PTSD Scale for the DSM-5 (CAPS; Weathers et al., 2013), had served in a combat zone, and were able to speak and write fluent conversational English. Participants must also have reported problems with irritability, anger, or aggression within the past month as defined by a CAPS score ≥ 2 (moderate/threshold) on item 15 (E1): “Irritable behavior and angry outbursts (with little or no provocation) typically expressed as verbal or physical aggression toward people or objects.” Veterans were ineligible for participation if they expected to be unstable on their medication regimen during the active treatment phase of the study; currently met criteria for bipolar I disorder or a primary psychotic disorder as determined by the Structured Clinical Interview for the DSM-IV (SCID; (First, Gibbon, Spitzer, & Williams, 1996); were receiving or planned to initiate other anger management psychotherapy or empirically-supported therapy for PTSD during the treatment component of the study; met criteria for substance dependence (other than nicotine) of sufficient severity to prevent full engagement in the treatment protocol; or were determined to have moderate or severe impairment related to traumatic brain injury as measured by the Brief Traumatic Brain Injury Screen (Schwab et al., 2007) and medical record review or consultation with their provider. Participants were recruited through referral by clinicians at the [blinded for review] VA Medical Center (VAMC) and the local Vet Center, as well as through flyers and brochures posted in the VA medical center and the local Vet Center and distributed at community events for veterans. IRB-approved letters were also sent to potentially eligible veterans identified using the VAMC’s Fileman system or the Corporate Data Warehouse, and the [blinded for review] Recruitment Database. We oversampled women for a target recruitment of 33% women, to include one group of women and two groups of men in each treatment arm.

Treatment

Both treatments were administered in 12 weekly sessions using a group format. Previous research has suggested that women and men veterans may differ in their expression of anger and aggression (Sullivan & Elbogen, 2013); that much of the anger expressed by men in anger management groups is directed at women (Gerlock, 1994); and that inclusion of women with sexual trauma histories in trauma-related therapy groups with men is not recommended (Chard, 2010). Therefore, groups were comprised of all-male and all-female cohorts. Each group session lasted 120 minutes. Participants were asked to miss no more than 3 of the 12 group sessions. Participants who missed more than 3 group sessions were given the opportunity for one “make-up” individual session with the group facilitator; participants who missed more than 3 sessions were dropped from the group but were invited to complete follow-up measures.

Cognitive behavioral therapy (CBT).

This 12-week manualized group treatment was developed specifically to target problems with PTSD-related anger and aggression in combat veterans. The first session oriented participants to the structure and philosophy of the program, provided an historical overview of PTSD, and introduced the concept of the “survival mode” of functioning (Chemtob, Novaco, Hamada, Gross, & Smith, 1997). The remaining 11 sessions followed a standard format: 1) practice relaxation training (15–20 minutes); 2) review homework, introduce new material, and engage in group activities focused on implementing new skills and behaviors (70–80 minutes); and 3) review problems or concerns of group members. A summary of each session, as well as information on how sessions were tailored specifically for combat veterans with PTSD, is available in the supplementary material.

Present-centered therapy (PCT).

Participants assigned to the control condition received 12 weeks of group PCT. PCT was selected as the control condition because it has been used as an active control condition in a number of psychotherapy trials for PTSD (Foa, McLean, Zang, & et al., 2018) and anger (Shea et al., 2013) in veterans. It was initially developed to provide an active, manualized treatment comparison condition for psychotherapy trials by controlling for nonspecific factors such as contact with a trained therapist, rationale for treatment, and instillation of expectancy for therapeutic gains. The therapeutic approach was drawn from Yalom’s group therapy model, which utilizes interpersonal process, supportive techniques, identification of response options, encouragement of adaptive reactions, and focus on the “here-and-now” (Yalom, 1975). A recent meta-analysis suggested that PCT may be efficacious in the treatment of PTSD (Frost, Laska, & Wampold, 2014), and a survey of practice patterns within the VA suggests that similar present-focused approaches are routinely employed by VA mental health providers (Rosen et al., 2004). The PCT manual used in this trial was adapted from a manual used in previous psychotherapy trials for PTSD and anger (Shea et al., 2013). The use of cognitive behavioral strategies such as Socratic questioning or discussion of traumatic material was proscribed in this arm. Instead, therapists modeled and encouraged supportive therapeutic interactions by facilitating problem-solving and application of formerly used or new coping methods; using and modelling active listening; encouraging expression of feelings; pointing out themes; providing praise; and encouraging contact with supportive others.

Therapist Training and Treatment Fidelity.

Study therapists for both conditions were doctoral-level licensed clinical psychologists or Master’s-level rehabilitation therapists. All therapists had extensive experience administering manualized, empirically-supported psychotherapy for PTSD. Specifically, one Ph.D. level psychologist was present in each group, and all Ph.D. level therapists had successfully completed VA’s training program for Cognitive Processing Therapy (CPT) and been added to the VA’s roster of approved CPT therapists. One of the Master’s level therapists was also on this roster. Training on the CBT protocol was provided by [blinded for review], the primary author of the treatment protocol. Sessions were videotaped, and a random selection of 20% of the group sessions in each treatment arm were rated by a licensed clinical psychologist using fidelity checklists adapted from the Yale Adherence and Competence Scale (Carroll et al., 2000).

Measures

Diagnostic assessments.

The Clinician-Administered PTSD Scale (CAPS) for the Fifth edition of the DSM-5 (Weathers et al., 2013) was administered at baseline to diagnose PTSD. The CAPS is a semi-structured interview to assess PTSD. The excellent reliability and validity of the CAPS has established it as the “gold standard” for PTSD assessment (Blake et al., 1995; Weathers, Keane, & Davidson, 2001). Selected modules of the Structured Clinical Interview for DSM-IV (SCID) (First et al., 1996) were administered to assess for bipolar 1 disorder, primary psychotic disorders, and substance use disorders. The Brief Traumatic Brain Injury Screen (BTBIS; Schwab et al., 2007) is a three-item self-report questionnaire that was administered to screen for the possibility of a moderate or severe TBI that could interfere with the ability to engage meaningfully in cognitive-behavioral treatment. As necessary based on the results of the BTBIS, medical records were examined or providers consulted to evaluate for the presence of moderate to severe TBI, or referrals were made for further evaluation.

Outcome measures.

Three scales were administered at baseline, during each of the 12 psychotherapy sessions, at post-treatment, and at 3- and 6-month follow-ups: the Dimensions of Anger Reactions Scale (DAR) (Forbes et al., 2004); the PTSD Checklist for the DSM-5 (PCL-5; Weathers et al., 2013); and the World Health Organization Disability Assessment Schedule, Version 2.0 (WHODAS). The DAR (Forbes et al., 2004) is a 7-item scale measuring the frequency, duration, and behavioral response to anger, and anger-related functional impairment on social relationships, health, and work. Participants rated each item on a scale of 0 (not at all) to 8 (exactly so), for a total scale ranging from 0 to 56. The DAR has been found to have high internal reliability (α = .92), and good concurrent validity (r ranging from .69-.77); and to correlate highly with measures of functional impairment, in a large sample of treatment-seeking soldiers who served in Iraq or Afghanistan (Novaco, Swanson, Gonzalez, Gahm, & Reger, 2012). The PTSD Checklist for the DSM-5 (PCL-5; Weathers et al., 2013) is a 20-item measure of PTSD symptoms on which participants first report an autobiographical narrative of a trauma, and then rate how much they were bothered over the past week by each DSM-5 PTSD symptom on a scale of 0 (not at all) – 4 (extremely). Total scores could range from 0 to 80. Previous versions of the PCL have been found to have good reliability and validity (Blanchard, Jones-Alexander, Buckley, & Forneris, 1996), and preliminary findings suggest adequate reliability and validity for this version as well (i.e. internal consistency α = .94; test-retest reliability r = .82; and convergent validity r ranging from .74 to .85). (Blevins, Weathers, Davis, Witte, & Domino, 2015). The 12-item version of the WHODAS (Üstün & World Health Organization, 2010) was administered to assess broad functioning in the areas of understanding and communication, self-care; mobility (getting around), interpersonal relationships (getting along with others), work and household roles (life activities), and community and civic roles (participation). Participants rate each item on a scale of 0 (none/no difficulty) to 4 (extreme or cannot do), for a total score ranging from 0 to 36. The WHODAS has been found to be reliable and valid in the assessment of functioning across a range of chronic physical and psychological conditions, and to be sensitive to change (Andrews, Kemp, Sunderland, Von Korff, & Ustun, 2009).

The Conflict Tactics Scale, CTS (Straus, 1979) was administered at baseline, post-treatment, and 3- and 6-month follow up to assess psychological and physical aggression. The CTS includes 18 items that measure both psychological aggression (i.e. insulting, swearing at, or threatening someone; destroying objects) and physical aggression (i.e. throwing something at someone, pushing, grabbing, shoving, slapping, kicking, biting, hitting, beating up, threatening with a gun or knife, or using a gun or knife on someone). Participants indicate how often they have engaged in each behavior using the following scale: 0 (never); 1 (once); 2 (twice); 3 (3–5 times); 4 (6–10 times); 5 (11–20 times); 6 (more than 20 times). Scores on psychological aggression could range from 0 to 36; scores on physical aggression could range from 0 to 42. Consistent with previous studies on veterans and violence (Beckham, Feldman, Kirby, Hertzberg, & Moore, 1997), the original CTS instructions were modified by expanding the target of the aggressive behavior from “partner” to “anyone”; and reducing the time frame of the response from one year to one month to detect potential changes associated with treatment. Coefficient alpha for the CTS has been found to range from .62 to .88, with good evidence of construct validity.

Data Analysis

Data were screened during data entry for completeness and accuracy, and when possible participants were contacted about inconsistent or improbable response options to ensure data quality and participant safety. This process resulted in either corroboration of responses (e.g., verification that a veteran had pulled a knife on an individual over the past month when he felt threatened in his neighborhood), or elimination of a specific measure for a single time point from analysis (e.g., a veteran indicated that he must have misunderstood the questionnaire when called to verify self-report on the CTS that he had engaged in all forms of physical aggression more than 20 times over the previous month). In one case inspection of data prior to analysis resulted in the determination that a participant in the male CBT arm had likely provided invalid responses to most or all measures. For example, on one measure he circled a single response option and then drew an arrow down the page and through the next page indicating the same response option for the remainder of the measure. This veteran could not be contacted to corroborate his responses, so his data were eliminated from analysis.

Participants who attended at least one group therapy session were included in the primary analyses. An intent-to-treat, last-observation-carried-forward approach was used, based on the conservative assumption that participants who dropped out of treatment would remain at the same level on all outcome measures. Inspection of residuals by means of Q-Q plots indicated adequate model fit for time modeled as a linear effect for the DAR, PCL-5, and WHODAS outcomes. Time was modeled as a categorical variable for the CTS psychological and physical aggression outcomes, which were measured at fewer time points. Separate multilevel models of the DAR, PCL-5, and WHODAS were examined using scores from baseline, each of the 12 sessions, post-treatment, and 3- and 6-month follow-up. Each model examined the effects of gender, therapy type (CBT vs. PCT), time, and therapy type by time on the outcome variable.

Separate multilevel models of CTS psychological aggression and CTS physical aggression were examined using data from baseline, post-treatment, and 3- and 6-month follow-up. In these models, the independent variables were gender, therapy type (CBT vs. PCT), categorical time, and therapy type by time. For all outcomes, we calculated the effect size (Cohen’s d) of change from baseline to post-treatment using estimates calculated within the context of the multilevel models.

Two of the three previous controlled trials of therapy for anger or aggression in veterans presented findings based on analyses only of participants who completed the full course of therapy (Chemtob, Novaco, Hamada, & Gross, 1997; Shea et al., 2013). To make our findings consistent with reporting in these studies, we re-ran multilevel models using only individuals who had completed at least 9 of the 12 sessions (n = 23).

Results

Participants

Thirty-six participants began treatment, with 19 assigned to the CBT arm, and 17 assigned to the PCT arm. Slow initial recruitment and a drop-out resulted in 3 participants in one of the first male CBT groups; therefore, we recruited for an additional male CBT group for a total of 3 male CBT groups of 3 to 6 members each; 2 male PCT groups of 5 to 6 members each; 1 female CBT group of 5 members; and 1 female PCT group of 6 members. As described in the Data Analysis section above, one participant’s data was eliminated from analysis due to invalid responding. This resulted in a total sample size of 35, with 18 in the CBT arm and 17 in the PCT arm. There were no statistically significant differences in demographic or baseline clinical characteristics between treatment groups (Table 1).

Table 1.

Sample characteristics by treatment condition

	CBT (N=18) M (SD) N (%)	PCT (N=17) M (SD) N (%)	t or χ2	p-value
Age	51.6 (11.5)	52.6 (12.3)	0.24	0.810
Years of Education	13.1 (3.5)	15.4 (2.9)	1.98	0.058
Male	13 (72%)	11 (65%)	0.23	0.632
Race
Black/African-American	12 (68%)	9 (53%)
White/Caucasian	6 (33%)	6 (35%)	2.40	0.301
Biracial/Other	0 (0%)	2 (12%)
Married	3 (17%)	8 (47%)	3.75	0.053
Employed (Full- or Part-Time)	2 (11%)	3 (18%)	0.31	0.581
Baseline Symptoms and Functioning
DAR	40.2 (12.6)	41.2 (10.4)	0.24	0.814
PCL5	58.8 (13.2)	56.0 (12.1)	0.65	0.519
WHODAS 2.0	19.3 (8.3)	24.6 (7.3)	2.01	0.053
CTS psychological aggression	8.9 (7.7)	8.3 (5.1)	0.26	0.794
CTS physical aggression	0.5 (1.0)	1.23 (4.1)	0.71	0.484

DAR = Dimensions of Anger Reactions scale; PCL5 = PTSD Checklist for DSM-5; WHODAS 2.0 = World Health Organization Disability Assessment Schedule; CTS = Conflict Tactics Scale.

Treatment Retention

Of the 18 participants assigned to the CBT arm, 10 (55.6%) were categorized as completers (attended 9 or more sessions); of the 17 assigned to the PCT arm, 13 (76.5%) were completers. The difference in completion rates between the two arms was not statistically significant, χ² (1) = 1.70, p = 0.19. Completion rates differed significantly by gender, χ² (1) = 6.13, p = 0.01. Across both treatment arms, 4 of 11 women (36.4%) were completers, while 19 of 24 men (79.2%) were completers. Two men (14%) and 5 women (100%) dropped out of the CBT arm voluntarily. One woman (17%) and one man (9%) dropped out of the PCT arm voluntarily, and one man and one woman were withdrawn from the PCT arm by the PI after being admitted for inpatient substance abuse treatment and medical hospitalization, respectively.

Efficacy of CBT vs. PCT

Intent-to-treat analyses.

Results of multilevel models for the DAR, PCL-5, and WHODAS are presented in Table 2. Results showed no main effect for therapy type. Significant main effects of time were observed for anger disposition (DAR), PTSD scores (PCL-5), and disability scores (WHODAS), suggesting that, for the reference condition (PCT), participants improved approximately 0.31 points on the DAR per time point, an estimated 0.19 points on the PCL-5 per time point, and an estimated 0.10 points on the WHODAS per time point. Significant main effects were observed for gender on the DAR and the PCL-5: Women scored an estimated 10.1 points higher on the DAR and an estimated 15.9 points higher on the PCL-5. No significant interaction effects were observed for therapy by time for the DAR or the PCL-5. A significant interaction effect of therapy by time was observed for the WHODAS, such that participants demonstrated about a 0.3-point difference per session favoring PCT. Figure 1 presents modeled contrasts of scores by therapy type from baseline to post-treatment, baseline to 3-month follow-up, and baseline to 6-month follow-up for the DAR, PCL-5, and WHODAS.

Table 2.

Multilevel Model of Therapy Type (CBT vs. PCT) For Measures Given at Each Session.

	DAR		PCL5		WHODAS 2.0
	Intent-to-Treat Analysis	Completer Analysis	Intent-to-Treat Analysis	Completer Analysis	Intent-to-Treat Analysis	Completer Analysis
Intercept	36.8 (3.1)***	35.0 (4.0)***	48.1 (3.9)***	45.9 (4.8)***	20.5 (2.2)***	19.7 (2.8)***
Gendera	10.1 (4.0)*	13.0 (7.0)	15.9 (5.1)**	19.2 (8.4)*	5.76 (2.9)	6.39 (5.0)
Therapy Typeb	−1.38 (3.8)	0.62 (5.5)	2.52 (4.8)	3.58 (6.6)	−3.32 (2.8)	−3.68 (3.9)
Time	−0.31 (0.07)***	−0.51 (0.11)***	−0.19 (0.09)*	−0.25 (0.12)*	−0.10 (0.04)*	−0.20 (0.6)**
Time x Therapy Typeb	0.11 (0.10)	0.28 (0.07)	0.16 (0.12)	0.07 (0.18)	0.28 (0.06)***	0.36 (0.09)***

Model coefficients and standard errors (in parentheses). DAR = Dimensions of Anger Reactions scale;

PCL5 = PTSD Checklist for DSM-5; WHODAS 2.0 = World Health Organization Disability Assessment Schedule.

^aMale is used as the reference category;

^bPCT is used as the reference condition.

^*p < .05

^**p < .01

^***p < .001

Figure 1a–c. Change in scores on anger, PTSD, and disability: Baseline to post-treatment, 3 months, and 6 months.

Note: Bar graphs present modeled contrast scores from baseline to post-treatment, 3 months, and 6 months for measures of anger (a), PTSD symptoms (b), and disability (c). DAR = Dimensions of Anger Reactions scale; PCL5 = PTSD Checklist for DSM-5; WHODAS = World Health Organization Disability Assessment Schedule; CTS = Conflict Tactics Scale. For all measures, decrease in scores reflects improvement.

Because there were no significant therapy by time interaction effects for the DAR and the PCL-5, we collapsed across therapy type when calculating baseline to post-treatment effect size (d) estimates for these outcomes. Baseline to post-treatment effect sizes were 0.86 (p < 0.001) for the DAR and 0.29 (p < 0.08) for the PCL-5. Given the significant interaction effect of therapy by time for the WHODAS, effect size estimates were calculated separately for each arm. For PCT, baseline to post-treatment effect size was 0.56; for CBT, baseline to post-treatment effect size was 0.48. Note that for CBT the direction of the change was positive, suggesting that scores on the WHODAS got worse over time. On the CTS, no significant main effects or interaction effects were observed on any variable.

Completer analyses.

The analyses of the subsample of treatment completers (n=26) produced a similar pattern of results to the intent-to-treat analyses. In the multilevel models including completers only, the main effects of time remained significant but were slightly larger. In the reference condition (PCT), participants improved an estimated 0.51 points on the DAR per time point; an estimated 0.25 points on the PCL-5 per time point; and an estimated 0.20 points on the WHODAS per time point. A main effect of gender was found for the PCL-5, where women scored an estimated 19.2 points higher than men. No significant interaction effect was observed for therapy by time for the DAR or the PCL-5. A significant interaction effect of therapy by time was observed for the WHODAS, such that participants demonstrated about a 0.36-point difference per session favoring PCT. There were no significant interaction effects in the completer subsample on the CTS except for an interaction on the physical aggression scale at 6 months suggesting that veterans in the CBT (vs.PCT) group showed approximately 1.4 points less improvement (p<0.05).

Discussion

This study examined the efficacy of group CBT for anger and aggression in male and female combat veterans with PTSD using PCT as an active control condition. To our knowledge, this is the first study to include female veterans in an individual or group therapy intervention trial for anger or aggression in a veteran population.

Overall, while significant reductions in anger, PTSD symptoms, and disability were observed over time, results do not support the hypothesis that CBT is more effective than PCT in reducing these outcomes in combat veterans with PTSD. Specifically, multilevel models using an intent-to-treat approach indicated that participants in the PCT condition reported significant decreases in anger, but the absence of a significant interaction effect with time suggests that there were no differences between CBT and PCT on this outcome. Collapsed across both CBT and PCT therapy arms, effect sizes for anger reductions over time were large. PTSD symptoms and overall disability scores also improved significantly over time in the PCT active control condition. No interaction effect of therapy type by time was observed for PTSD symptoms, and the significance of the effect of time was reduced to a trend when scores were collapsed across therapy arm.

Surprisingly, findings related to changes in self-rated disability were in the opposite direction of the hypothesis, such that interaction effects of therapy by time favored the PCT condition. Effect size estimates suggest a significant, medium sized increase in disability in the CBT condition, and a significant medium size decrease in disability in the PCT condition. Although this finding is difficult to interpret, it is possible that the interaction may be the result of (non-significant) higher baseline scores in the PCT group vs. the CBT group, and a regression towards the group mean over time in both groups.

Finally, intent-to-treat analyses revealed no changes over time and no interaction effects for psychological or physical aggression as measured by the Conflict Tactics Scale (CTS). Analyses using only participants who completed at least 9 out of 12 sessions of therapy showed similar results, except that a significant difference was observed in changes in physical aggression scores on the CTS at 6 months favoring PCT. It is important to keep in mind, however, that baseline rates of physical aggression in this sample were low, and changes in aggression at each time point were small. As such, the observed difference on changes in physical aggression at 6 months should be interpreted with caution. Several previous intervention studies targeting psychological aggression in couples and families have also reported low baseline rates of physical aggression in their samples (Hayes et al., 2015; Taft, Creech, et al., 2016; Taft et al., 2014). It may be that though PTSD increases risk for physical aggression in veterans (Beckham et al., 1997; Blakey, Love, Lindquist, Beckham, & Elbogen, 2018; Macmanus et al., 2013; MacManus et al., 2015), base rates of physically aggressive acts are sufficiently rare among combat veterans with PTSD that they are difficult to capture in a measure evaluating behavior within the previous month.

The finding that PCT performed equally well with CBT in reducing symptoms associated with PTSD including anger is not unprecedented (Ford et al., 2018). While PCT was originally designed to be an active control condition in psychotherapy trials, a recent meta-analysis found that PCT performed as well as evidenced-based treatment for PTSD in 3 of 5 trials reviewed, and that it was significantly better than a no-treatment control condition with a large effect size (Frost et al., 2014). A more recent RCT found that PCT did not differ significantly from Prolonged Exposure (PE) therapy delivered over an 8 week period (Foa et al., 2018), and another study found that a present-centered intervention performed as well as PE in reducing hostility in Iraq/Afghanistan War veterans, with a lower dropout rate (Ford et al., 2018). Taken together, our findings add to the growing literature suggesting that PCT should be evaluated as a potentially efficacious therapy for trauma and trauma-related dysfunction.

Although dropout rates for male veterans in our study were comparable to other trials of anger treatment for veterans with PTSD, significantly more women than men dropped out of treatment. This difference in dropout by gender was entirely explained by the 100% dropout of female participants from the CBT group. Further, despite the exclusion of female veterans from previous anger and intervention trials, our findings indicated that women veterans scored significantly higher scores on the DAR, PCL-5, and WHODAS than did male veterans.

While the high dropout rate among women assigned to the CBT arm could be viewed as a limitation, these findings could also be interpreted to suggest that treatment retention may be an especially important consideration for women veterans who are seeking treatment for trauma-related anger. It may be relevant to consider that the CBT treatment presented here was based on Chemtob and colleagues’ (1997) “survival mode” model of anger in veterans, which has been called into question as a relevant framework for understand anger in women veterans (Worthen et al, 2015). The survival mode model assumes that hostile attribution bias and excessive combat trauma-related physiological arousal leads to the experience and expression of anger in benign or ambiguous situations where it is not warranted or appropriate (Chemtob, Novaco, Hamada, Gross, & Smith, 1997). However, this assumption may not necessarily hold for women veterans who have experienced a military culture where power dynamics may not allow for the expression of anger, even in situations where it is appropriate. Additionally, there is substantial evidence that women in the military are not uncommonly subjected to unfair or even abusive gender-based treatment in situations where they have limited recourse (Street, Vogt, & Dutra, 2009). Research in civilians has found that powerlessness and injustice play a role in anger expression particularly in women (Kring, 2000; Strachan & Dutton, 1992; Thomas & Gonzalez-Prendes, 2009; Thomas, 2005). It is also important to note that all but one of the women in the study were Black, while the composition of men was more racially balanced. The “Angry Black Woman” is a pernicious stereotype that may serve to further delegitimize or silence anger in Black women, even in cases where it is clearly warranted (Ashley, 2014). Taken together, this suggests that rather than challenging women veterans’ anger as excessive and inappropriate, it may be important to first validate that their anger as warranted and even adaptive in some situations, and to help them develop skills to assert themselves in ways that are safe and constructive. It is likely that the PCT group’s central treatment method was better equipped to provide that necessary validation, which would explain the differential dropout rates between groups for women veterans in the study.

This study has important limitations. First, the study is not a randomized trial. The lack of randomization to treatment arm prevents us from eliminating alternative explanations for our findings, including the possibility that the source of referral or recruitment differed systematically at different time points. We alternated assignment to intervention to attempt to mitigate this concern, but the baseline group differences between PCT and CBT on education level, marital status, and WHODAS scores, while not statistically significant, suggest that this process may not have been completely successful at ensuring equal groups prior to treatment. Second, though to our knowledge this is the largest controlled study of anger management therapy for veterans with PTSD to date, and the only sample to include women, the total sample size remains small and therefore limits both power and the degree to which findings are generalizable. Relatedly, while the high drop-out rate of women from this study is an important finding, it may also be viewed as an important limitation. Finally, despite the consistent association observed between PTSD and aggression in the literature, the baseline levels of physical aggression were low in this sample. Future studies should measure physical aggression over time period of at least 3 or 6 months to increase the chances of observing changes in this critical but relatively rare behavior.

Despite these limitations, our findings suggest that group therapy can be effective in reducing anger in combat veterans with PTSD, and that the theoretical orientation of the group (i.e. cognitive behavioral, present-centered) may not be as important to recovery as the group therapy process itself (Ahn & Wampold, 2001; Hayes & Hofmann, 2017). Our findings did not suggest that the putative active ingredients in CBT were critical in facilitating change, and in fact we found evidence that the techniques specific to CBT may have decreased treatment engagement in women veterans. Though women have been excluded from psychotherapy trials for anger in veterans, results presented here suggest that it may not be appropriate to assume that anger interventions designed for male veterans will necessarily be effective for women veterans. Identifying mechanisms of change in anger interventions for veterans (Kazdin, 2007), and considering how they may operate differently in racially and gender-diverse populations, will be critically important to the VA’s mission of serving all of those who served.