The Impact of Trauma-Focused Psychotherapies on Anger: A Systematic Review and Meta-Analysis

Abstract

Objective:

Anger is one of the most prevalent concerns among individuals with posttraumatic stress disorder (PTSD) and is often a residual symptom following PTSD treatment. The purpose of this systematic review and meta-analysis was to determine how effective trauma-focused PTSD psychotherapies are in reducing anger.

Method:

The study was reported according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. This study conducted a systematic review of studies that reported the effect of trauma-focused treatments on anger outcomes. Additionally, a meta-analysis was conducted with a subset of studies that used randomized controlled trials (RCTs) methodologies to compare trauma-focused PTSD treatments to non-trauma-focused and control conditions.

Results:

The systematic review included 16 studies with a total of 1,846 participants. In 11 of the studies, there was a significant decrease in an anger dimension following treatment. Eight studies with 417 total participants met inclusion criteria for the meta-analysis. The meta-analysis yielded a pooled effect size of PTSD treatment on anger of Hedge’s g = 0.33.

Conclusion:

Overall, trauma-focused treatments for PTSD significantly improve anger, but the magnitude of change is small-to-medium. Additional research is needed to determine how best to maximize anger outcomes following trauma-focused treatment or determine if and when targeted anger treatment is needed.

Difficulty managing anger is a significant issue for individuals with posttraumatic stress disorder (PTSD; Olatunji et al., 2010; Orth & Wieland, 2006), is one of the main issues Veterans have with reintegration (Sayer et al., 2010), and is a priority for treatment (Crawford et al., 2015; Passemato et al., 2018; Rosen et al., 2013). Anger is associated with greater PTSD symptom severity (Koenen et al., 2003; Novaco & Chemtob, 2002) and is a risk factor for violence toward others (Novaco & Chemtob, 2015), criminal legal involvement (Elbogen et al., 2012), and suicidal thoughts and behavior (Dillon et al., 2020). Anger also negatively impacts social functioning, a protective factor against PTSD (Ray & Vanstone, 2009). While anger is often experienced in the aftermath of trauma and can be adaptive (e.g., in response to mistreatment), it may be problematic if it is experienced too frequently or intensely, is difficult to manage, or impacts functioning.

Clinical practice guidelines recommend trauma-focused cognitive behavioral therapies (TFTs) for PTSD (e.g., Department of Veterans Affairs & Department of Defense, 2017; International Society of Traumatic Stress Studies; Bisson et al., 2019), including cognitive processing therapy (CPT; Resick et al., 2017), prolonged exposure (PE; Foa et al., 2019), and eye movement desensitization and reprocessing (EMDR; Shapiro, 2018), as well as other cognitive behavioral interventions that contain a trauma processing component (i.e., trauma-focused cognitive restructuring or exposure). Research demonstrates that TFTs significantly reduce anger, but generally with modest effect sizes (e.g., Cahill et al., 2003; Forbes et al., 2012; Resick et al., 2008; Stapleton et al., 2006). This can be compared to anger-focused interventions, which have yielded large effect sizes on anger (Morland et al., 2010). Anger is also a common residual symptom following PTSD treatment (e.g., Miles et al., 2022). Because individuals often present to PTSD treatment to address problematic anger, better understanding the impacts of TFTs on anger will inform treatment considerations.

To date, there has yet to be a comprehensive integration of findings to determine the overall effect of TFTs on problematic anger. A systematic review and meta-analysis will help determine whether TFTs are sufficient for treating problematic anger or if additional efforts may be required. The purpose of the present study was to determine whether TFTs for PTSD (e.g., CPT, PE, EMDR) reduce problematic anger at posttreatment and follow-up. To do this, we conducted a systematic review of the literature to examine if anger improved following trauma-focused treatments. We also wanted to determine if TFTs were more effective at reducing anger compared to non-trauma focused interventions and controls. Therefore, we conducted a meta-analysis with a subset of studies that utilized a randomized controlled trial (RCT) design comparing TFTs to non-trauma-focused interventions. This study also aimed to identify limitations in the literature.

Methods

Search Strategy

This systematic review was reported using guidelines from the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) statement (Page et al., 2021). A medical librarian composed a sensitive search using a mix of keywords and subject headings to represent the concepts of PTSD, therapy, military personnel, and anger, and another librarian peer-reviewed the search according to a modified Peer Review of Electronic Search Strategies (PRESS) checklist. The focus of the review was the effect of evidence-based TFTs on anger outcomes. MEDLINE via PubMed, Embase via Elsevier, Scopus via Elsevier, PTSDPubs via ProQuest, and APA PsycINFO via EBSCO were searched from inception to April 8, 2021, with an updated search run in September 2022. All results were compiled in EndNote and imported into Covidence systematic review software (Veritas Health Innovation) for screening. (See Supplemental materials for full search criteria). The protocol for this review was not pre-registered.

Abstract Review

During the abstract review phase, articles were included if they: (1) were peer-reviewed, (2) included empirical, quantitative data, (3) were in English, (4) comprised adults ages 18 or older, (5) described a standalone TFT (i.e., no other treatments incorporated) or one arm of the study met these criteria. Exclusion criteria at the abstract review phase included: (1) combined PTSD treatment and pharmacotherapy or pharmacotherapy alone, (2) case studies or case series designs, or (3) treatments targeting a subset of PTSD symptoms (e.g., insomnia treatment, anger management).

An initial 4,808 articles were identified and imported to Covidence. After duplicate articles (n = 1,416) were removed, 3,392 abstracts were eligible for review. Several authors (SYW, KHD, LM, TP, TW, AP, KK) reviewed 10 abstracts to calibrate on agreement for the inclusion/exclusion criteria. Four reviewers (TP, TW, AP, KK) screened 100 abstracts as an initial review to increase calibration. The larger group discussed discrepancies. Five reviewers (TP, TW, AP, KK, GP) then screened the remaining abstracts. Any conflicts were resolved by SYW and KHD. One of these authors reviewed the abstract independently to determine if the article met the inclusion for full-text review. A total of 724 articles were eligible for full-text review.

Full-Text Review

The inclusion/exclusion criteria were reassessed in addition to the following: (1) the study contained an assessment of problematic anger at pre- and posttreatment, (2) participants met PTSD diagnostic criteria based on clinician-administered interview or probable PTSD diagnostic cutoff on a validated self-report measure, and (3) participants received TFT recommended by the Department of Veterans Affairs/Department of Defense (Department of Veterans Affairs & Department of Defense, 2017) or International Society of Traumatic Stress Studies (Bisson et al., 2019). These guidelines were selected because Department of Veterans Affairs, Department of Defense, and ISTSS have a specific focus on the treatment of PTSD. Any study that reported changes in anger following PTSD treatments at pre to post-treatment or pre to follow-up were included, such as open trials, pilot trials, observational clinic studies, and RCTs. Full texts were reviewed by SYW, KHD, CR, STL, TP, and GP. Initially each reviewer reviewed 10 articles and then the reviewers met to jointly resolve any discrepancies to calibrate. Raters then rated articles independently. Any conflicts after this phase were resolved by SYW and KHD. These authors reviewed the articles independently to review the criteria. If there was any uncertainty about if an article should be included, SYW and KHD discussed and made a joint decision. A total of 16 studies were included. If multiple papers were published from the same dataset, only one was included, which was the original article rather than a secondary analysis. If a single article included sub-studies with distinct samples, each was considered separately and reported separately in this paper (i.e., Glassman et al., 2020).

Quality Assessment

Two authors (SYW and KHD) conducted independent bias ratings to assess the quality of studies, using the Cochrane Risk of Bias tool (Higgins et al., 2011). Ratings assessed the following elements: (1) random sequence generation, (2) allocation concealment, (3) masking of participants and personnel, (4) masking of outcome assessments, (5) attrition/incomplete outcomes, and (6) selective outcome. Quality assessments were conducted for handling missing data, sample size, and sample representativeness. SYW and KHD reconciled any discrepancies together by consensus and came to a consensus after discussion. For studies where a consensus could not be reached, a third rater (STL) reviewed the article and made a final decision. Each domain was rated using the following scale: 1 (reported in paper/low bias), .5 (unclear), or 0 (absent from paper/high bias). Similar to other reviews (e.g., Viswanathan, et al., 2012), if a domain of bias was not mentioned, it was assumed to have not been considered and rated as high bias.

Data Extraction

Two authors (SYW and KHD) independently extracted the following data from each study: funding source, geographic region, participant characteristics, study design and intervention (e.g., diagnostic assessment, treatment type, delivery modality, comparison group, time points), anger assessment (e.g., anger measure, type of assessment, time points), effect sizes, and study methods. In the absence of effect sizes, the available data were used to calculate Cohen’s d effect sizes.

Synthesis of Results for Systematic Review

Results were reviewed across anger dimensions/subscales and then for specific anger dimensions/subscales. Studies were categorized by treatment type: CPT, PE, EMDR, cognitive behavioral conjoint therapy (CBCT; Monson & Fredman, 2012), cognitive behavioral therapy (CBT), and imaginal exposure (IE). Studies were categorized based on patient population: civilians, military populations (e.g., active duty, Veteran, Reserve, National Guard), or both. Effect sizes were narratively described using the following convention: small (<.50), medium (.50–.79), or large (.80 or higher) Cohen’s d effect sizes. The number of reported effect sizes exceeds the number of studies because some studies reported multiple anger effect sizes from one or multiple conditions. Where sufficient data points existed, forest plots consisting of mean within-group effect size estimates and their associated 95% confidence intervals (CIs) were included to visually summarize and quantify the precision of effect size estimates across the large number of anger dimensions evaluated.

Meta Analysis Data Analysis

A subset of studies (n = 8) utilizing a RCT design to compare the effects of TFTs to non-trauma-focused treatments or control groups on problematic anger were included in the meta-analysis. Standardized mean differences between pre-and posttreatment anger outcomes formed the indices of effect size. Standardized mean differences were estimated (n = 15) using Hedge’s g to stabilize variance and estimate CIs (Hedges, 1981). Because higher scores on anger measures indicated more severe symptoms, Hedge’s g estimates were multiplied by −1 so that a positive Hedge’s g reflected TFT improved anger symptoms compared with the non-trauma-focused treatment. A multi-level random effects meta-analysis was performed using the metafor package (Viechtbauer, 2010) in R version 4.2.2 (R Core Team, 2022), with study entered as a random effect. Random effects meta-analyses allow for the exploration of heterogeneity introduced by different methods or samples and for the inclusion of multiple effect estimates from a single sample (Hasselblad & Hedges, 1995; Hedges & Vevea, 1998; Vichtbauer, 2010). A sample-weighted overall average effect size of TFT on anger was calculated. Between-study heterogeneity was assessed using the Q test and I². Funnel plots were inspected to examine potential publication bias (Egger et al., 1997).

Results

Characteristics of Studies

The 16 studies included were conducted in four countries: USA (n = 12), Canada (n =1), USA and Canada (n = 1), Norway (n =1), and Australia (n = 1). The studies involved six TFTs: CBT (n = 1), CBCT (n = 3), CPT (n = 6; with a trauma account [CPT+A], without a trauma account [CPT], or a written account only), PE (n = 6), EMDR (n = 1), and IE (n = 1). Of the 16 studies, 10 were RCTs. Eight RCTs compared a TFT to a non-trauma-focused treatment, and two compared TFTs to one another. Additionally, 12 of the studies included follow-up assessments at: 1 month (n = 1), 3 months (n = 8), 4 months (n = 2), 6 months (n = 5), and 12 months (n = 2). See Supplemental Table 1 for study characteristics.

Participants

A total of 1,846 individuals participated in the studies, and 1,565 received a TFT. Participants’ ages ranged from 21–80 years (M = 43.95), 39.90% identified as male, and 64.43% identified as White. Studies were comprised of civilians (n = 9), veterans and/or active-duty military (n = 5), or both (n = 2). The veteran and/or active-duty military samples were 83.64% male.

Anger Assessments

Across studies, problematic anger was assessed using self-report measures. The four assessment measures used were the State-Trait Anger Expression Inventory (STAXI; n = 15; STAXI-2; n = 1), Hostility subscale of the Symptom Checklist 90 (SCL-90; n = 1),¹ Anger/Irritability subscale of the Trauma Symptom Inventory (TSI; n = 1), and the Dimensions of Anger Reaction scale (DAR; n = 1). There was variation in which STAXI subscales were used; seven studies used the Anger Expression subscale (three reported anger expression dimensions, such as Anger-In, Anger-Out, and Anger-Control), six used the State subscale, and seven used the Trait subscale. State anger is defined as the extent to which an individual is currently experiencing angry feelings. Trait anger is the individual predisposition to experience anger in various situations (i.e., the tendency to experience elevations in state anger). Anger Expression is a measure of how anger tends to be expressed when it is felt. Anger-In is the tendency to hold in or suppress angry feelings. Anger-Out is the tendency to express angry feelings verbally or physically. Anger-Control is the tendency to control or manage angry feelings rather than expressing them. Higher scores on all dimensions/subscales, except for Anger-Control, reflect greater problematic anger. Lower scores on Anger-Control reflect greater problematic anger.

Risk of Bias

As can been seen in Table 1, studies included in the present review demonstrated a mixed risk of bias. Ten out of the 16 studies (62.50%) were a RCT and were evaluated using the Cochrane Risk of Bias tool (see Table 1). All the studies were evaluated for handling of missing data, adequate sample size, and sample representativeness. Nine of the 16 studies (56.25%) were determined to have low risk of bias in how missing data were addressed by using intent-to-treat analyses. Six (37.50%) of the remaining studies did not provide sufficient information to determine the risk of bias, and one study (6.25%) was rated as having a high risk of bias due to only looking at completer analyses in a small sample (Ford et al., 2018). The risk of bias was mixed about whether samples were adequately sized. Specifically, four studies were determined to be at low risk (25.00%), five studies were at high risk (31.25%), and for the remaining seven (43.75%) it was unclear whether the sample size affected risk of bias. Finally, the risk of bias related to the representativeness of the sample within studies was evaluated, with seven of the studies being at high risk of bias (46.67%) due to homogenous samples. Five studies (31.25%) were determined to be at low risk of bias of representativeness, and the remaining four (25%) did not provide sufficient detail to make an informed judgment regarding the representativeness of the sample.

Table 1.

Summary of Assessment of Risk of Bias in included Studies

Study	Sequence Generation	Allocation	Blinding of Participants	Blinding of Outcome	Incomplete data	Selective Reporting	Missing Data Handled	Adequate Sample Size	Sample Representativeness
Cahill et al. 2003	High	High	High	Low	High	High	Unclear	High	High
Feske, 2008	High	High	Unclear	Low	Low	Low	Unclear	High	High
Foa et al., 1995	––	––	––	––	––	––	Unclear	High	High
Forbes et al., 2012	Low	Unclear	Low	Low	Low	Low	Low	Unclear	High
Ford et al., 2018	Low	High	High	Low	Low	Low	High	High	High
Fredman et al., 2020	––	––	––	––	––	––	Low	Unclear	Unclear
Galovski et al., 2013	––	––	––	––	––	––	Low	Unclear	Low
Glassman et al., 2020a,c	––	––	––	––	––	––	Low	Low	Low
Glassman et al. 2020b,c	––	––	––	––	––	––	Low	Low	Unclear
Langkaas et al., 2017	Low	Low	Unclear	Low	Low	Low	Low	Unclear	Unclear
McDonagh et al., 2005	High	High	Unclear	Low	Low	Low	Unclear	Unclear	High
Monson et al., 2011	––	––	––	––	––	––	Unclear	High	Unclear
Monson et al., 2012	Low	Low	Unclear	Low	Low	Low	Low	Unclear	Low
Resick et al., 2008	High	Unclear	Unclear	Low	Low	Unclear	Low	Low	High
Schnurr et al., 2022	Low	Unclear	Unclear	Low	Low	Low	Low	Low	Low
Stapleton et al., 2006	Unclear	Unclear	Low	Low	Unclear	Low	Unclear	Unclear	Low

Note. Due to some studies not being randomized controlled trials, Cochrane criteria were not applied and assessed. The last three columns represent additional areas of quality assessed by authors that were used for all studies.

^aThis refers to Study 1 of Glassman et al. 2020.

^bThis refers to Study 2 of Glassman et al. 2020.

^cGlassman et al. 2020 Study 1 and Study 2 both utilized data from parent studies of RCTs comparing in-person care to telehealth but the two modalities were collapsed in Study 1 and Study 2 of Glassman et al. 2020 due to the in-person and telehealth modalities being non-inferior in the parent studies. Therefore, Glassman et al. 2020 Study 1 and Study 2 were not reported as RCTs in this paper.

Pre- to Posttreatment Within Group Anger Reductions and Effect Sizes

Fifteen studies reported the statistical significance of change in an anger dimension/subscale from pre- to posttreatment following TFT. Of these, 11 (73.33%) reported a decrease in an anger dimension/subscale from pre- to posttreatment following TFT, and four did not (26.67%). Across studies, 36 effect sizes were reported, ranging from 0.06–0.97. The top row of Figure 1 shows the mean pre to post effect size collapsed across all measures was g = 0.47 (95% CI [0.39, 0.55]), representing a small but trending towards medium change in anger symptoms. See Table 2 for outcomes.

Figure 1.

Forest Plot for Within Group Mean Standardized Effect Size Estimates (Cohen’s d) with 95% Confidence Intervals for Anger Dimensions by Time Period (Pre- to Posttreatment and Pre to Follow-Up) for All Trauma-Focused Therapy Studies

Note. All combined = all effect size estimates collapsed across measures.

Table 2.

Study Findings

Citation	Type of Trauma Focused Tx	Significant anger improvement pre-post treatment?	If so, maintained at follow-up?	TF vs. TF RCT	TF vs. non-TF RCT	Within Group Effect Sizes	Between Group Effect Sizes
Galovski et al., 2013	CPT+A	Anger-in ↓ Anger-out ↓ Anger-control ↑ Anger ↓	Yes	No	No	Anger-in: .73 (post), .68 (fu)  Anger-out: .49 (post), .42 (fu)  Anger-control: −.31 (post), −.33 (fu)  Anger: .93 (post), .94 (fu)	N/A
Glassman et al., 2020a+	CPT	Trait anger ↓	N/A	Yes	No	.26	N/A
Glassman et al., 2020b+	CPT+A	Trait anger ↓	N/A	Yes	No	.39	N/A
Forbes et al., 2012	CPT+A	Anger ↓	Yes	No	Yes (CPT+A vs. TAU)	.49 (post), 1.00 (fu)	−.17 (post)a, −.32 (fu)
Resick et al., 2008	CPT	Anger-in ↓ Anger-out ns	Not reported	Yes (CPT vs. CPT+A vs. WA)	No	CPT  Anger-in: .63  Anger-out: .34 CPT+A  Anger-in: .56  Anger-out: .23 WA  Anger-in: .97  Anger-out: .21	CPT+A vs. WA  Anger-in: −.48 (post), .13 (fu)  Anger-out: −.18 (post), −.12 (fu) CPT+A vs. CPT  Anger-in: .14 (post), .15 (fu)  Anger-out: .15 (post), .15 (fu)  WA vs. CPT  Anger-in: .55 (post), .02 (fu)  Anger-out: .33 (post), .26 (fu)
Schnurr et al., 2022	CPT+A & PE	State anger ↓ Trait anger ↓ Anger expression ↓	Not reported	Yes (CPT+A vs. PE)	No	PE:  State anger: .23  Trait anger: .44  Anger expression: .43 CPT+A:  State anger: .25  Trait anger: .38  Anger expression: .36	PE vs. CPT+A  State anger: .08  Trait anger: .05  Anger expression: .11
Cahill et al., 2003	PE	State anger ↓	Yes	No	Yes (PE vs. PE/SIT vs. SIT vs. WL)	PE: .74 (post), .73 (fu)  PE/SIT: .35 (post), .34 (fu)  SIT: 1.36 (post), .88 (fu)  WL: .03 (post)	PE vs. SIT: .06 (post)a, .48 (fu)  PE vs. PE/SIT: .66 (post), .60 (fu)  PE vs. WL: 1.16a (post)
Feske, 2008	PE	Anger expression ↓	Yes	No	Yes (PE vs. TAU)	PE: .90 (post), .84 (fu)  TAU: .02 (post), .03 (fu)	0.98a (post), .96 (fu)
Ford et al., 2018	PE	Trait anger n/r	Not reported	No	Yes (PE vs. TARGET)	PE: .38 TARGET: .26	.10a
Langkaas et al., 2017	PE	Anger-in ↓ Anger-out ns Anger-control ns Anger ↓	Not reported	No	Yes (PE vs. IR)	PE  Anger-in: .64 (post); .58 (12-mo fu)  Anger-out: .29 (post); .09 (12-mo fu)  Anger-control: .06 (post); −.13 (12-mo fu)  Anger: .72 (post); .71 (12-mo fu) IR  Anger-in: .36 (post); .38 (12-mo fu)  Anger-out: .47 (post); .29 (12-mo fu)  Anger-control: −.50 (post); −.37 (12-mo fu)  Anger: .55 (post); .04 (12-mo fu)	Anger-in: .50a(post), .44 (fu) Anger-out: .26a (post), .17 (fu) Anger-control: .22a (post), .20 (fu) Hostility/anger: .30a (post), .69 (fu)
Stapleton et al., 2006	PE & EMDR	Trait anger ns	N/A	Yes (PE vs. EMDR vs. Relaxation)	Yes (PE vs. EMDR vs. Relaxation)	PE  Trait anger: .23 (post), .39 (fu) EMDR  Trait anger: .06 (post), .27 (fu) Relaxation  Trait anger: .23 (post), .34 (fu)	PE vs. EMDR  Trait anger: .29 (post), .32 (fu) EMDR vs. Relaxation  Trait anger: .64a(post), −.84 (fu) PE vs. Relaxation  Trait anger: .32a(post), −.42 (fu)
Fredman et al., 2020	CBCT	State anger ↓	Yes	No	No	.45 (1-mo fu), .60 (3-mo fu)	N/A
Monson et al., 2011	CBCT	State anger ns Trait anger ns Anger expression ns	N/A	No	No	State anger: .80 Trait anger: .42 Anger expression: .31	N/A
Monson et al., 2012	CBCT	Trait anger ↓ Anger expression ↓	Yes	No	Yes (CBCT vs. WL)	CBCT  Trait anger: .84  Anger expression: .68 WL  Trait anger: −.06  Anger expression: .13	Trait anger: .56a Anger expression: .67a
Foa et al., 1995	IE	State anger ns	N/A	No	No	.68	N/A
McDonagh et al., 2005	CBT	State anger ns	N/A	No	Yes (CBT vs. PCT vs. WL)	CBT: .18 (post), .53 (3-mo fu), .61 (6-mo fu)  PCT: .12 (post), −.11 (3-mo fu), −.30 (6-mo fu)  WL: −.13 (post)	CBT vs. PCT  −.08a (post), .11 (3-mo fu), .58 (6-mo fu) CBT vs. WL  −.17a (post) PCT vs. WL  .16 (post)

Note. CPT = cognitive processing therapy (cognitive therapy only); CPT+A = cognitive processing therapy with written trauma account; PE = prolonged exposure; EMDR = eye movement desensitization and reprocessing; CBCT = cognitive-behavioral couple therapy; IE = imaginal exposure; CBT = cognitive behavioral therapy; ↓ = significant reduction; ↑ = significant increase; ns = not significant; nr = not reported; WA= written account only; TAU = treatment as usual; PE/SIT = combined prolonged exposure and stress inoculation therapy; SIT = stress inoculation therapy; WL = waitlist; TARGET = Trauma Affect Regulation: Guide for Education and Therapy; IR = imagery rescripting; PCT = present-centered therapy.

^aThis denotes this effect size was used in the meta-analysis.

⁺Glassman et al. 2020 utilized data from two parent RCTs that compared CPT delivered via telehealth or in-person. However, the Glassman et al. 2020 paper collapsed these two conditions together because telehealth was non-inferior to in-person.

Pre to Follow-Up Within Group Anger Reductions and Effect Sizes

Twelve studies reported changes in anger from pre to the follow-up after TFT. Six (50%) studies reported the maintenance of treatment gains from posttreatment. Effect sizes ranged from 0.09–1.00. As depicted in the second row of Figure 1, the mean effect size was 0.55 (95% CI [0.42, 0.68]), which represents a medium change in anger across time. The 95% CIs for these two effect size estimates cover much of the same outcome space, with a larger interval for the follow-up mean estimate, which was partially influenced by fewer individual estimates than was reported in the pre to post data.

Pre- to Posttreatment Within Group Effect Sizes by Anger Dimensions/Subscales

For each of the seven anger dimensions/subscales, mean effect size estimates from pre to posttreatment are summarized in Figure 1. With the exception of Anger-Out (d = 0.31), which only included small effect size estimates, the other pre to post effect size estimates included medium effects in their CIs. Also, of the four anger dimensions (i.e., Anger-In, State Anger, Trait Anger, and Anger Expression) with five or more individual estimates each, three (75.00%) of the means were small towards medium sized (i.e., 0.48, 0.54, and 0.71) while the other one (Trait Anger; 25.00%) was small (0.38). Anger-In and general problematic anger changed the most from pre- to posttreatment (ds = 0.71). Large CIs for many estimates also indicate possible heterogeneity in effects.

Pre to Follow-up Within Group Effect Sizes by Anger Dimensions/Subscales.

For each anger dimension/subscale reviewed, mean effect size estimates from pre to follow-up are summarized in Figure 1. Of note, the mean estimate across measures contains imprecision as there were three or fewer effect sizes for each measure. Given the single effect size, a 95% CI for the mean effect size estimate was not calculated. Despite variability within estimates, which clouds distinctions, the mean estimates tended to be somewhat smaller than the pre- to posttreatment estimates.

Within Group Findings by Treatment.

CPT.

Six studies examined CPT (i.e., CPT or CPT+A), and all showed decreases in an anger dimension from pre- to posttreatment. Sixteen within group effect sizes for anger dimensions from pre- to posttreatment were reported, ranging from 0.21–0.97, and the mean effect size estimates across measures was 0.47 (95% CI [0.35, 0.59]). See the first row of Figure 2. Thus, across CPT studies, the average reduction of anger symptoms was small but approached a medium effect size.

Figure 2.

Forest Plot for Within Group Mean Standardized Effect Size Estimates (Cohen’s d) with 95% Confidence Intervals for Anger Dimensions by Time Period (Pre- to Posttreatment and Pre to Follow-Up) for Cognitive Processing Therapy Studies

Note. All combined = all effect size estimates collapsed across measures.

The second row of Figure 2 summarizes the pre to follow-up effect size estimates for CPT, with a mean of 0.67 (95% CI [0.26, 0.94]). The imprecision in the mean effect size estimates can reflect several factors, including the small number of individual effect size estimates (n = 5) and the pattern of three small effect size estimates for the three STAXI subscales and the two large effect size estimates for the other anger measures.

PE.

Six studies examined PE, four (67.67%) reported a decrease in an anger dimension from pre- to posttreatment. The 11 within group effect sizes range from 0.06–0.90. The first row of Figure 3 displays the mean within group pre to post effect size estimate of 0.46 (95% CI [0.31, 0.61]), indicating an average small but trending towards medium reduction in anger symptoms.

Figure 3.

Forest Plot for Within Group Mean Standardized Effect Size Estimates (Hedge’s g) with 95% Confidence Intervals for Anger Dimensions by Time Period (Pre- to Posttreatment and Pre to Follow-Up) for Prolonged Exposure Studies

Note. All combined = all effect size estimates collapsed across measures.

The second row in Figure 3 summarizes the effect sizes for pre to follow-up PE studies with a mean effect size estimate of 0.50 (95% CI [0.27, 0.72]). Of note, there was a lot of variability in the small number of effect sizes gathered for pre to follow-up time periods.

Other Treatments.

The remaining studies examined EMDR, CBCT, IE, or Trauma-Focused CBT. Because there were so few effect sizes reported for these treatments, no forest plots were created for EMDR, CBCT, IE, or Trauma-Focused CBT. One study examined EMDR and trait anger, which did not decrease. The effect sizes from pre- to posttreatment (0.06) and pre to follow-up (0.27) for EMDR were small. Three studies examined CBCT. Two of the three studies that examined CBCT showed reductions in an anger dimension from pre- to posttreatment. Six within group effect sizes were included across all studies, three (50.00%) were small, one (16.67%) was medium, and two (33.33%) were large. One study reported within group effect size (i.e., state anger) from pre to follow-up for CBCT and found a medium effect (0.60). Other anger dimensions were not measured from pre to follow-up in the CBCT studies. One study examined IE and state anger, which did not decrease from pre- to posttreatment. One study examined Trauma-Focused CBT and did not find state anger to decrease from pre- to posttreatment, and the effect size was small (0.18). From pretreatment to follow-up, two state anger effect sizes were reported, both of which were medium (0.53–0.61).

Findings by Population

Military Samples.

Five studies reported 10 within group effect sizes from pre- to posttreatment ranging from 0.23–0.49. The first row of Figure 4 depicts the small mean pre- to posttreatment effect size estimate of 0.38 (95% CI [0.33, 0.43]), and the second row shows the large mean pre to follow-up effect size estimate of 0.80 (95% CI [0.41, 1.19]). The wide CI, however, indicates that the estimate is imprecise.

Figure 4.

Forest Plot for Within Group Mean Standardized Effect Size Estimates (Hedge’s g) with 95% Confidence Intervals for Anger Dimensions by Time Period (Pre- to Posttreatment and Pre to Follow-Up) for Military, Civilian and Mixed Military and Civilian Studies

Note. All combined = all effect size estimates collapsed across measures.

Civilian Samples.

Nine studies reported 23 within group effect sizes from pre- to posttreatment, ranging from 0.06–0.97. Estimates suggest medium sized reductions in anger at posttreatment, with the third row of Figure 4 depicting the effect size estimate of 0.50 (95% CI [0.38, 0.61]), and the fourth row of Figure 4 depicting a pre to follow-up effect size estimate of 0.52 (95% CI [0.38, 0.65]). However, there were no clear differences in outcomes between civilian and military populations.

Combined Civilian and Military Samples.

Two studies reported three within group effect sizes from pre- to posttreatment ranging from 0.26–0.84 but did not report combined group effect sizes. Due to the small number of studies, a mean effect size was not calculated.

Meta-Analysis Findings

Based on 15 effect sizes from eight studies and a total of 417 participants (see Figure 5), a random-effects meta-analysis yielded a small-to-medium overall average Hedge’s g (0.33, 95% CI [0.12, 0.53], p < .001). This suggests that compared with non-trauma-focused treatments, TFTs reduce anger, with effect sizes ranging from −0.08 (95% CI [−0.64, 0.47]) to 1.16 (95% CI [0.40, 1.92]). Effect estimates from all studies except one (McDonagh et al., 2005) were positive, indicating larger effect sizes for reducing anger in TFTs compared to non-trauma-focused treatments. Funnel plot inspection did not detect a significant publication bias (See Supplementary Figure 1). Heterogeneity between studies accounted for 21.08% of the variation in effect sizes (Q = 15.69, p = .33; I²= 21.08), while less than 1% of the variation in effect sizes was due to within-study heterogeneity.

Figure 5.

Forest Plot Comparing Trauma-Focused Treatments to Non-Trauma-Focused Treatments for Reducing Anger.

Note: Estimates displayed as squares to the right of the dotted line indicate larger between-group estimates observed for trauma-focused treatments compared to non-trauma-focused treatments. Diagonal polygon represents summary sample-weighted Hedge’s g estimate. TFT = trauma-focused treatment, comparison = non-trauma-focused treatment, WL = waitlist, TAU = treatment as usual, SIT = Stress Inoculation Therapy; TARGET = Trauma Affect Regulation: Guide for Education and Therapy, IR = imagery rescripting, STAXI = State-Trait Anger Expression Inventory, SCL-90 = Symptom Checklist-90.

Discussion

A systematic review and a meta-analysis were conducted to determine how well TFTs reduce problematic anger. The majority of studies indicated that TFTs reduced anger from pre- to posttreatment. While anger was reduced following treatment, almost two-thirds of these studies suggested that the magnitude of change was small or small trending towards medium based on standardized effect size estimates. Of studies that examined follow-up assessments, all showed that reductions in problematic anger were maintained from 1 to up to 12 months after treatment with a medium effect. Results from the meta-analysis also showed a small-to-medium difference between TFT compared to non-trauma focused treatments. Taken together, these findings indicate that TFTs can reduce anger, but there is room for growth regarding the magnitude of improvement.

Findings should be interpreted in the context that the studies examined were primarily PTSD-focused studies, and individuals were often recruited based on PTSD symptoms rather than anger. Therefore, some participants may not have had clinically elevated levels of anger. Baseline levels of anger were not routinely reported, but low baseline levels of anger could minimize the ability to detect a change across time. Further, it may be important to compare TFTs to anger-focused interventions among those with elevated anger to determine if more targeted anger treatment is needed for maximum benefit.

Due to a limited number of studies and sample size, we were unable to compare specific TFTs (e.g., CPT vs. PE) via meta-analysis. We were, however, able to narratively describe findings across these two evidence-based treatments. From pre- to posttreatment, mean effect size estimates appeared to converge towards small-to-medium effects on anger for both CPT and PE. The next most studied treatment (i.e., CBCT) showed reductions in anger for two-thirds of studies, with mean effect sizes ranging from small to large depending on the anger construct examined. There was not enough evidence for the studies on IE and CBT. Finally, despite the wide use of EMDR in practice, there was only one study of EMDR, which did not show a reduction in anger. Overall, this review suggests more studies are needed to compare the effects of different TFTs on anger. Because the majority of studies were comprised of civilians only, we could not compare outcomes between civilian and military populations. Future studies should consider other demographic differences, such as gender, race, and trauma type.

Variability in anger assessments across studies made it difficult to draw conclusions. While we were unable to compare types of anger outcomes via meta-analysis, our narrative review found that effect sizes across the different types of anger ranged from small to large, and there was a large variability for similar anger dimensions across studies. One of the challenges in comparing effect sizes for anger outcomes is that the assessment timeframe can vary from the present (i.e., state anger) to across time (i.e., trait anger), which complicates the ability to draw conclusions about the durability to treatment effects.

While this systematic review and meta-analysis provides useful information about the impact of TFTs on anger, there are several limitations to note. A primary limitation of the current evidence base was inconsistency in anger assessments and reported subscales. Additionally, assessments used in these studies are face valid, so anger may be underreported due to social desirability. Anger outcomes were not routinely reported for PTSD treatment responders compared to non-responders. While the findings from the meta-analysis are informative, we were only able to synthesize a subset of the studies with an RCT design comparing TFTs to control, and the meta-analysis only examined pre- to posttreatment outcomes due to variability in follow-up. Due to a smaller sample size of studies in the meta-analysis, we also were unable to examine moderators of effect sizes (e.g., comparison of TFTs). There is a need for more studies with follow-up data to better determine the impacts of TFTs on anger long-term, particularly for specific treatments or anger domains. Regarding the quality of the studies that were included in the review, there were some risks of bias detected. First, only 60% of the studies included were RCTs. Further, only 53% of the studies demonstrated evidence of handling missing data using intent-to-treat, 20% were judged to have adequate sample size, and only 27% were judged to have a representative sample. In the future, researchers may benefit from attending to these limitations to enhance the scientific rigor of future studies.

In considering the modest effects of TFTs on anger, it may be that clinicians focus more on other emotions. For example, clinicians delivering CPT may have a greater focus on guilt and shame because the protocol encourages clinicians to prioritize resolving erroneous beliefs about the cause of a traumatic event, like unrealistic self-blame. Similarly, clinicians delivering PE may focus more on patients’ experience of fear and anxiety, given the protocol’s basis in the exposure model. However, both models have the flexibility to increase the focus on anger. For example, anger can be addressed in the post-exposure processing of PE or in any of the worksheets in CPT. Anger is a complex emotion that can be directed inward or outward and may be considered adaptive or maladaptive depending on the circumstance. Clinicians may need additional guidance during training on how to address anger within these and other models.

In addition to current PTSD treatment models, it is also possible that some patients may benefit or even require adjunctive anger interventions, such as anger management, which a prior study found a large effect size for among individuals with PTSD (Morland et al., 2010). Future research is needed to determine whether anger management adds value beyond TFTs, and, if so, the ideal timing to deliver this intervention (e.g., concurrent with TFT, after TFT to address residual anger). Additionally, future studies could examine gender differences to determine if TFTs are effective in reducing anger across genders. It is also critical to ensure that future studies expand the diversity of their samples to examine factors that play a critical role in psychological phenomena, such as race (Roberts & Rizzo, 2020). Regardless, our findings suggest additional efforts may be needed to maximize the effectiveness of treatments to help patients address their anger.

Clinical Impact Statement.

Anger is a symptom of PTSD that often drives people to seek treatment. Our findings suggest that clinicians can use trauma-focused PTSD treatments to improve anger and that improvements following these treatments are greater than effects following non-trauma focused PTSD treatments. However, given the small difference between these trauma-focused treatments and non-trauma focused PTSD treatments, clinicians should assess for anger during and after PTSD treatment to determine if it is sufficient, or if anger management treatment is needed.