Introduction

This chapter begins with the results of our literature search and some general description of the included studies. It is then organized by Key Question (KQ) and grouped by intervention (i.e., by type of psychological intervention or by drug class, whichever is relevant). For each KQ, we first give the key points and then proceed with a more detailed synthesis of the literature. Additional details for the studies included in this results chapter are provided in an appendix of evidence tables (Appendix D).

Briefly, we wanted to examine the efficacy and comparative effectiveness of psychological and pharmacological treatments for adults with posttraumatic stress disorder (PTSD). Efficacy of psychological treatments and their comparative effectiveness with each other are addressed in KQ 1. For each type of psychotherapy, we first address efficacy by evaluating studies with inactive comparison groups (e.g., waitlist, usual care). By the term inactive, we mean comparators that do not involve a specific psychotherapeutic intervention that may benefit people with PTSD. Of note, we have stratified our meta-analyses by comparison group, to show how the effect size and confidence interval would differ if we only included studies with a waitlist control, as opposed to including those with both waitlist and usual care controls. We then proceed to address comparative effectiveness of a given psychotherapy by evaluating studies with active comparison groups (i.e., head-to-head studies involving other psychotherapies).

KQ 2 addresses efficacy of pharmacological treatments and their comparative effectiveness with each other. As with KQ 1, we first address efficacy for each type of pharmacotherapy by evaluating studies with placebo controls. We then proceed to address comparative effectiveness by evaluating head-to-head studies (i.e., drug vs. drug).

KQ 3 addresses the direct (head-to-head) evidence on comparative effectiveness of psychological and pharmacological treatments with each other. KQ 4 addresses the direct evidence on comparative effectiveness of combinations of psychological and pharmacological interventions compared with either one alone. KQ 5 addresses whether any of the treatment approaches are more effective than other approaches for victims of particular types of trauma. Finally, KQ 6 synthesizes the evidence on adverse effects associated with treatments for adults with PTSD.

Results of Literature Searches

Results of our searches appear in Figure 2. We included 101 published articles reporting on 92 studies. Of the included studies, all were randomized controlled trials. We assessed the majority as medium risk of bias. We assessed 4 studies as low risk of bias. Additional details describing the included studies are provided in the relevant sections of this results chapter.

Figure 2 is titled “Disposition of articles.” The figure is a flow chart that summarizes the search and selection of articles: There were 21647 citations of randomized trials identified by searching MEDLINE®, IPA, CINAHL®, PsycINFO®, Embase®, the Cochrane Library, Web of Science, and PILOTS. In addition, 67 references were identified by hand searches and scientific information packets (SIPS). After removal of duplicate citations, titles and abstracts of 3048 references were screened for potential inclusion. Of these, 527 were deemed appropriate for full-text review to determine eligibility. After full-text review, 380 were excluded: 210 for no original data or wrong study design; 170 failed to meet at least one PICOTS element; and 46 were deemed poor quality. The 46 poor quality studies were eligible only for sensitivity analyses. One hundred and one articles representing 92 studies are included in this report’s qualitative synthesis and 77 studies are included in quantitative analyses.

Figure 2

Disposition of articles. PICOTS = populations, interventions, comparators, outcomes, timing, settings; SIPS = scientific information packets a Our main quantitative syntheses included 77 studies with low or medium risk of bias. This total does not include (more...)

Table 5 describes the most common outcome measures used in this literature. For further details about these instruments and scales, see Appendix A. Definitive thresholds for clinically significant changes are not well established for many of these measures, although there are some general guideposts. For example, some suggest that a reduction of 15 points on the CAPS constitutes a clinically significant reduction.40 However, this cutoff has not been validated and is somewhat uncertain. For the PTSD Checklist, some have considered a reduction of five or more points to indicate a clinically significant response.41 For the HAM-D and the BDI, a three-point improvement has been considered clinically meaningful.42 For continuous outcomes for which an SMD was calculated (when data from different scales are combined), an effect size of ~0.5 (a “medium” effect size)43 or higher has been considered a threshold for clinically significant benefit.

Table 5. Common outcome measures used in the included trials.

Table 5

Common outcome measures used in the included trials.

Key Question 1. Comparative Effectiveness of Different Psychological Treatments for Posttraumatic Stress Disorder

We organized this section by type of psychological treatment and present the information in the following order: (1) cognitive behavioral therapy (CBT)-cognitive therapy; (2) CBT-coping skills; (3) CBT-exposure; (4) CBT-mixed therapies; (5) eye movement desensitization and reprocessing (EMDR); and (6) other psychotherapies (Seeking Safety, imagery rehearsal therapy, narrative exposure therapy, brief eclectic psychotherapy). Within each section, we focus first on studies with inactive comparison groups (e.g., waitlist, usual care) to determine whether evidence supports the efficacy of each type of intervention. We then address studies with active comparison groups (i.e., head-to-head comparative evidence) or we provide cross-references for where those studies are addressed.

Tables describing characteristics of included studies are presented in a similar order. We first give details on studies that use any inactive comparators (in alphabetical order by last name of the first author)—i.e., those about efficacy—and then the details on any additional studies that only included active comparators.

In the bulleted text below we summarize the main overall key points and then the key points for each type of psychotherapy and report the strength of evidence (SOE) where appropriate. The primary outcomes of interest for determining whether treatments are effective for adults with PTSD are improving PTSD symptoms, inducing remission, and losing PTSD diagnosis; we focus more on these outcomes than on other outcomes in the key points. We also comment on other outcomes of interest, such as prevention or reduction of coexisting medical or psychiatric conditions (especially depression symptoms), quality of life, disability or functional impairment, and return to work or active duty. The findings in these key points are primarily based on meta-analyses of the trials that we rated low or medium risk of bias; those trials are cited in the detailed synthesis and related tables. In the detailed synthesis section for each treatment, we provide section headers for each outcome reported (PTSD symptoms, remission, loss of PTSD diagnosis, prevention or reduction of coexisting medical or psychiatric conditions, quality of life, disability or functional impairment, and return to work or active duty). If an outcome does not appear, no trial reported data on it.

Key Points: Overall—Efficacy

  • The strongest evidence of efficacy for improving PTSD symptoms is for exposure-based therapy (high SOE).
  • Evidence also supports the efficacy of exposure-based therapy for achieving loss of PTSD diagnosis, with a number needed to treat (NNT) of 2 (moderate SOE).
  • Evidence of moderate strength also supports the efficacy of cognitive processing therapy, cognitive therapy, CBT-mixed interventions, EMDR, and narrative exposure therapy for improving PTSD symptoms and/or achieving loss of PTSD diagnosis.
  • For improving PTSD symptoms, the effect sizes were very large for most of the psychological interventions with evidence of efficacy (e.g., 28.9-point reduction in CAPS and Cohen’s d = 1.27 for exposure).
  • For achieving loss of PTSD diagnosis, NNTs were ≤ 4.
  • Evidence was insufficient to determine efficacy for achieving remission for most psychological interventions, as trials typically did not report remission as an outcome.
  • Evidence was insufficient to determine efficacy of relaxation, stress inoculation training, Seeking Safety, or imagery rehearsal therapy.

Table 6 summarizes the efficacy and SOE for psychological treatments for improving PTSD symptoms, inducing remission, and achieving loss of PTSD diagnosis.

Table 6. Summary of efficacy and strength of evidence of psychological treatments for adults with PTSD for improving PTSD symptoms, remission, and loss of PTSD diagnosis.

Table 6

Summary of efficacy and strength of evidence of psychological treatments for adults with PTSD for improving PTSD symptoms, remission, and loss of PTSD diagnosis.

Key Points: Overall—Comparative Effectiveness

  • Most of the direct head-to-head comparative evidence was insufficient to determine whether psychotherapies differ for improving outcomes.
  • With few trials and few total subjects, most of our meta-analyses of head-to-head trials were underpowered to detect anything but medium to large differences.
  • Head-to-head evidence was insufficient to determine whether exposure therapy is more or less effective than cognitive processing therapy, cognitive therapy (CT), stress inoculation training, or EMDR.
  • Exposure therapy was more effective than relaxation for achieving loss of PTSD diagnosis (risk difference [RD], 0.31; 95% confidence interval [CI], 0.04 to 0.58; 2 trials, N=85, moderate SOE) and for improving depression symptoms.
  • For exposure therapy compared with exposure plus cognitive restructuring (CR), evidence supported a conclusion of no significant difference between treatments for achieving loss of diagnosis (RD, −0.01; 95% CI, −0.17 to 0.14; 3 trials, N=146). Although point estimates favored exposure plus CR, evidence was insufficient to determine comparative effectiveness for reduction of PTSD symptoms or depression symptoms, largely because of imprecision.
  • CBT-mixed interventions resulted in greater improvements in PTSD symptoms than relaxation interventions (moderate SOE).
  • For seeking safety compared with active controls (relapse prevention, psychoeducation, and treatment as usual in a VA substance use disorders clinic), evidence supported a conclusion of no significant difference between treatments for PTSD symptom reduction.

Table 7 summarizes the available head-to-head evidence and SOE for improving PTSD symptoms, inducing remission, and achieving loss of PTSD diagnosis.

Table 7. Summary of comparative effectiveness from head-to-head trials and strength of evidence for improving PTSD symptoms, remission, and loss of PTSD diagnosis.

Table 7

Summary of comparative effectiveness from head-to-head trials and strength of evidence for improving PTSD symptoms, remission, and loss of PTSD diagnosis.

Key Points: CBT—Cognitive Therapy

  • Evidence supports the efficacy of cognitive processing therapy for improving PTSD symptoms (WMD, −32.2 compared with waitlist or usual care), achieving loss of PTSD diagnosis, and improving depression symptoms for adults with PTSD (moderate SOE).
  • For achieving loss of diagnosis, 44 percent more subjects treated with cognitive processing therapy than subjects in control groups achieved the outcome. This translates to a NNT of 3.
  • For cognitive processing therapy, evidence was insufficient for remission and for other outcomes (such as anxiety symptoms, quality of life, disability or functioning, and return to work or active duty).
  • Evidence supports the efficacy of other CT interventions (i.e., that were not cognitive processing therapy) for improving PTSD symptoms, achieving loss of PTSD diagnosis, improving depression and anxiety symptoms, and reducing disability for adults with PTSD (moderate SOE).

Key Points: CBT—Coping Skills

  • Evidence was insufficient to determine efficacy of relaxation or stress inoculation training for adults with PTSD. One trial comparing prolonged exposure, stress inoculation training, prolonged exposure plus stress inoculation training, and waitlist suggests that stress inoculation training may be efficacious.49

Key Points: CBT—Exposure

  • Evidence supports the efficacy of exposure therapy for improving PTSD symptoms (standardized mean difference [SMD], −1.27; 95% CI, −1.54 to −1.00; 7 trials, N=387; high SOE), achieving loss of PTSD diagnosis (moderate SOE), and improving depression symptoms for adults with PTSD (high SOE).
  • For achieving loss of PTSD diagnosis, 66 percent more subjects treated with exposure than subjects in waitlist control groups achieved the outcome (RD, 0.66; 95% CI, 0.42 to 0.91; 3 trials, N=197). This translates to a NNT of 2.
  • Evidence was insufficient for other outcomes (remission, anxiety, quality of life, disability or functional impairment, and return to work or active duty).
  • Most efficacy evidence comes from trials of prolonged exposure, which combines imaginal and in vivo exposure.

Key Points: CBT—Mixed

  • Evidence25,46,47,4969 supports the efficacy of CBT-mixed treatments for improving PTSD symptoms (mean change from baseline in CAPS: WMD, −31.1; 8 trials, N=476; mean change from baseline in any PTSD symptom measure: SMD, −1.09; 14 trials, N=825, moderate SOE).
  • Evidence also supports the efficacy of CBT-mixed interventions for achieving loss of PTSD diagnosis (moderate SOE), remission (moderate SOE), reduction of depression symptoms (moderate SOE), reduction of disability or functional impairment (low SOE), and anxiety symptoms (low SOE).
  • For achieving loss of diagnosis, 26 percent more subjects treated with CBT-mixed therapies than subjects in inactive control groups achieved the outcome (RD, 0.26; 6 trials, N=290). This translates to a NNT of 4.

Key Points: Eye Movement Desensitization and Reprocessing (EMDR)

  • Evidence supports the efficacy of EMDR for reduction of PTSD symptoms, but SOE is low because of some inconsistency and imprecision.
  • Evidence supports the efficacy of EMDR for achieving loss of PTSD diagnosis and improving depression symptoms (moderate SOE for both); 64 percent more subjects treated with EMDR experienced this outcome than did subjects in waitlist control groups. This translates to a NNT of 2.
  • Evidence was insufficient to determine the efficacy of EMDR for other outcomes (remission, anxiety, quality of life, disability or functioning, and return to work or active duty).

Key Points: Other Psychological Therapies

  • Evidence supports the efficacy of narrative exposure therapy for improving PTSD symptoms (Posttraumatic Diagnostic Scale [PDS], mean change from baseline: WMD, −10.2; 95% CI, −13.1 to −7.4; 3 trials, N=227, moderate SOE) and for achieving loss of PTSD diagnosis (RD, 0.15; 95% CI, 0.01 to 0.30; 3 trials, N=227, low SOE).
  • Some evidence (3 trials, N=96) supports the efficacy of brief eclectic psychotherapy for improving PTSD symptoms, achieving loss of diagnosis, reducing depression and anxiety symptoms, and returning to work (all low SOE).
  • Evidence was insufficient to determine the efficacy of Seeking Safety or imagery rehearsal therapy.

Detailed Synthesis: CBT—Cognitive Therapy

Characteristics of Trials

Table 8 summarizes the characteristics of the nine cognitive therapy (CT) trials meeting our inclusion criteria. Five trials included a comparison with a waitlist condition (two of which also included an active comparison arm).52,7073 Two trials included a comparison with usual care or treatment as usual.74,75 Two trials included only comparisons with active interventions.46,76 Further details describing the included trials are provided in Appendix D.

Table 8. Characteristics of included cognitive therapy trials.

Table 8

Characteristics of included cognitive therapy trials.

Three trials compared cognitive processing therapy with a waitlist control.7072 Of these, one trial enrolled male (n=54) and female (n=6) military veterans;70 one enrolled women with histories of childhood sexual abuse (n=71);71 and one enrolled subjects with histories of adult sexual assault (n=121).72 All three trials were conducted in the United States. The subjects in the trial enrolling military veterans had a higher average age (54 years) than those in the other two trials (~32 years). Subjects were allowed to participate in two of the trials if they had been on a stable medication regimen for 2 or 3 months.70,71 Subjects were excluded from the trial enrolling those with histories of adult sexual assault if they were in an abusive relationship or were being stalked. The primary outcomes for the trials were the Clinician Administered PTSD Scale (CAPS), PTSD Checklist (PCL), Modified PTSD Symptom Scale (MPSS), and PTSD Symptom Scale (PSS).

One trial compared cognitive processing therapy with usual treatment at veterans’ community-based counseling services.74 The trial randomized 59 people with military-related PTSD living in three states in Australia.

Two trials from the same research group in the United Kingdom compared cognitive therapy (CT) treatments with waitlist controls. The first enrolled survivors of motor vehicle accidents, and compared CT with a waitlist condition of repeated symptom assessments and with a self-help booklet, “Understanding Your Reactions to Trauma” (SHB group), which the authors reported was based on cognitive behavioral principles for treating patients with PTSD.73 This study was designed as an “early intervention” and included only subjects who started therapy within 6 months of their MVA. Subjects were excluded from the study if they had been unconscious for more than 15 minutes after the accident or had no memory of it. The second trial from the same research group compared CT with waitlist.52 The trial enrolled 28 consecutive referrals from General Practitioners and Community Mental Health Teams. Subjects were required to have PTSD resulting from trauma that occurred at least 6 months before study entry.

Another trial of CT randomized 108 people with PTSD from various traumatic events to 12 to 16 sessions of CT or usual care.75 In addition to PTSD, all subjects also had diagnoses of either major mood disorder (85%) or schizophrenia or schizoaffective disorder (15%). The therapy intervention was a program involving CT that had previously been designed and pilot tested for PTSD in people with severe mental illness.75

Of the two trials that included only comparisons with active interventions, one four-arm study compared prolonged exposure alone, CR alone, prolonged exposure and CR together, and a relaxation group;46 the other compared CT with imaginal exposure (IE).76,80 Both enrolled heterogeneous samples of men and women in the United Kingdom who had experienced a variety of traumatic events (physical assault, witnessing a trauma, road accident, nonroad accident, sexual assault, being held hostage, bombing, combat, and “miscellaneous”;46 crime, accident, and other events76,80).

Results for Cognitive Therapy Compared With Inactive Comparators

Under each outcome header below, we first present our data synthesis for studies of cognitive processing therapy. Then we present results for the other CT studies with inactive comparator groups.52,73,75

PTSD Symptom Reduction

All included trials reported measures of PTSD symptom reduction. Of the four trials comparing cognitive processing therapy with controls, all found that subjects in the active treatment arm had a greater reduction in symptoms of PTSD than those in control groups.7074

Our meta-analysis of CAPS scores (Figure 3) found a much greater reduction in PTSD symptoms for subjects treated with cognitive processing therapy than for those in control groups (WMD, −32.2; corresponding Cohen’s d = −1.40; 95% CI, −1.95 to −0.85; Appendix F). The meta-analysis had considerable statistical heterogeneity (I2=86.5%), but the direction of effects was consistent. The differences were only in the magnitude of benefit; all trials found moderate or large magnitudes of benefit. The pooled effect size was slightly larger when only including the three studies with a waitlist comparator (WMD, −35.9) than when also including the one study with a usual care comparator.

Figure 3 is titled “Mean change from baseline in CAPS for cognitive processing therapy compared with controls, by type of comparator.” The figure displays a forest plot reporting the weighted mean difference in CAPS scores stratified by cognitive processing therapy versus waitlist and cognitive processing therapy versus usual care. Data for Figure 3 are presented in Appendix F. This figure is described further in the “PTSD Symptom Reduction” section as follows: “Our meta-analysis of CAPS scores found a much greater reduction in PTSD symptoms for subjects treated with cognitive processing therapy than for those in control groups (weighted mean difference −32.2, corresponding Cohen’s d = −1.40; 95% CI, −1.95 to −0.85. The meta-analysis had considerable statistical heterogeneity (I2=86.5%), but the direction of effects was consistent. The differences were only in the magnitude of benefit; all trials found moderate or large magnitudes of benefit. The pooled effect size was slightly larger when only including the three studies with a waitlist comparator (WMD −35.9) than when also including the one study with a usual care comparator. “

Figure 3

Mean change from baseline in CAPS for cognitive processing therapy compared with controls, by type of comparator. Note: Timing of outcome assessment: 17 weeks (Chard, 2005), 10 weeks (Monson, 2006), 6 weeks (Resick, 2002), 12 weeks (Forbes, 2012).

For two of the three trials comparing cognitive processing therapy with waitlist control, the authors reported that changes were maintained at a 3-month posttreatment followup.71,72 In one trial, subjects continued to improve from posttreatment to the 3-month followup (p=0.02); no significant difference on CAPS scores was observed between the 3-month and 1-year follow-up points.71 In the other trial, both of the active interventions exhibited a strong decrease in CAPS scores from baseline to posttreatment (p<0.0001) with some increase from posttreatment to the 3-month assessment (p<0.005), and no change between 3 and 9 months.72 A later publication from the trial reported that decreases in symptoms were maintained throughout a long-term followup of 5 to 10 years after participation in the study.78

The study comparing cognitive processing therapy with usual care reported similar, but slightly lower CAPS scores, at 3-month posttreatment followup compared with posttreatment assessments for both study groups.74

Each of the four trials involving cognitive processing therapy also reported one other measure of PTSD symptom reduction. The trials used several different measures (PCL,70,74 MPSS,71 and PSS72)—see Appendix D for details.

Overall, we concluded that evidence of moderate strength supports the efficacy of cognitive processing therapy for reduction of PTSD symptoms based on consistent and direct evidence from four trials. Even though findings were not precise, the differences in magnitudes of benefit suggest a moderate or large benefit.

All three studies comparing other CT interventions (i.e., that were not cognitive processing therapy) with inactive control groups reported greater improvement for those treated with CT than those in control groups.52,73,75 The trial involving CT, self-help booklet, or repeated assessments as an early intervention measured PTSD symptoms with the PDS (Posttraumatic Diagnostic Scale) and CAPS; data were reported separately for “CAPS assessor frequency” and “CAPS assessor intensity.” The CT group showed better outcomes on all PTSD symptom measures at posttreatment followup and 3- and 9-month followup (p<0.001).73 The trial using CT for those with severe mental illness showed that CT was more effective than treatment as usual in decreasing total-CAPS score (p=0.005).

Our meta-analysis of PTSD symptom measures found a greater reduction in PTSD symptoms for subjects treated with CT than for those in waitlist, self-help booklet, and usual care control groups (Cohen’s d=−1.22; 95% CI, −1.91 to −0.53, using the CAPS intensity scores from Ehlers et al. 2003 and Ehlers et al. 2005 and using the total CAPS from Mueser et al. 2008, Appendix F). The meta-analysis had considerable statistical heterogeneity (I2=79.6%), but the direction of effects was consistent. When only compared with waitlist controls, the effect size was larger (Cohen’s d=−1.54; 95% CI, −2.17 to −0.92; Appendix F).

Overall, we concluded that evidence of moderate strength supports the efficacy of CT for reduction of PTSD symptoms based on consistent and direct evidence. Even though findings were not precise, the differences in magnitudes of benefit suggest a moderate or large benefit.

Loss of PTSD Diagnosis

All trials in this section reported data on posttreatment diagnostic status. The four trials that compared people receiving cognitive processing therapy with controls reported a reduction in the number of subjects meeting the criteria for PTSD at the end of treatment and at later follow-up assessments in both the cognitive processing therapy and control groups, with fewer subjects meeting diagnostic criteria in the intervention arm than in the control arm.

Our meta-analysis for achieving loss of PTSD diagnosis (Figure 4) found that 52 percent more subjects treated with cognitive processing therapy achieved loss of PTSD diagnosis than subjects in waitlist groups (RD, 0.52). This translates to a NNT of 2. When also including the study with a usual care comparator, the effect size decreased to 44 percent (Figure 4).

Figure 4 is titled “Loss of PTSD diagnosis for cognitive processing therapy compared with controls.” The figure displays a forest plot reporting the risk difference of loss of PTSD diagnosis stratified by cognitive processing therapy versus waitlist and cognitive processing therapy versus usual care. Data for Figure 4 are presented in Appendix F. This figure is described further in the “Loss of PTSD diagnosis” section as follows: “Our meta-analysis for achieving loss of PTSD diagnosis found that 52 percent more subjects treated with cognitive processing therapy achieved loss of PTSD diagnosis than subjects in waitlist groups (RD, 0.52). This translates to a NNT of 1.9. When also including the study with a usual care comparator, the effect size decreased to 44 percent.”

Figure 4

Loss of PTSD diagnosis for cognitive processing therapy compared with controls, by type of comparator. Note: Timing of outcome assessment: 17 weeks (Chard, 2005), 12 sessions (Monson, 2006), 6 weeks (Resick, 2002), 12 weeks (Forbes, 2012).

All three trials comparing cognitive processing therapy with waitlist reported posttreatment follow-up assessments indicating that, over time, the changes seen in loss of PTSD diagnosis were maintained. One trial reported that 30 percent of subjects treated with cognitive processing therapy and 3 percent of waitlist subjects did not meet criteria for PTSD diagnosis 1 month posttreatment (p=0.01).70

Two of the cognitive processing therapy trials reported posttreatment followups of 3 months or longer. In one trial, 93 percent of subjects treated with cognitive processing therapy (and 36 percent of those in the minimal attention group) no longer met criteria for PTSD posttreatment; later values for the intervention group were 97 percent at 3 months posttreatment and 94 percent at 1-year followup.71 Another trial reported that 58 percent and 55 percent of subjects treated with cognitive processing therapy no longer met criteria for PTSD at 3 and 9 months after treatment, respectively (immediately posttreatment, 53% no longer met criteria for PTSD).72 A later publication from the trial reported that 77.8 percent no longer met criteria for PTSD at long-term followup of 5 to 10 years after participation in the study.73,78 From the above findings and our meta-analysis, evidence of moderate strength supports the efficacy of cognitive processing therapy for achieving loss of PTSD diagnosis. This grade is based on consistent, direct, and fairly precise evidence from four trials.

All three studies comparing other CT interventions (i.e., that were not cognitive processing therapy) with inactive control groups reported data on loss of PTSD diagnosis.52,73,75 The study comparing CT, a self-help booklet, and repeated assessments reported that 78.6 percent and 89.3 percent of subjects treated with CT no longer met criteria for PTSD at 3 and 9 months after treatment, respectively.73 The study comparing CT for people with severe mental illness with treatment as usual reported that 63.3 percent and 72.7 percent of subjects treated with CT no longer met criteria for PTSD at 3 and 6 months after treatment, respectively.75

Our meta-analysis for achieving loss of PTSD diagnosis found that 51 percent (95% CI, 24% to 78%) more subjects treated with CT achieved loss of PTSD diagnosis than subjects in waitlist, self-help booklet, and usual care control groups by 3 months after treatment (Appendix F). This translates to a NNT of 2. The meta-analysis had considerable statistical heterogeneity (I2=84.7%), but the direction of effects was consistent. When only compared with waitlist controls, the effect size was larger (risk difference 0.66; 95% CI, 0.50 to 0.82; I2=0%, Appendix F).

Overall, we concluded that evidence of moderate strength supports the efficacy of CT for achieving loss of PTSD diagnosis based on consistent and direct evidence.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

All cognitive therapy trials assessed the impact on coexisting psychiatric conditions—anxiety, depression, or both. No trial reported the reduction or prevention of a comorbid medical condition as one of their outcomes of interest. All trials assessed the impact of cognitive therapy on symptoms of depression as measured by the Beck Depression Inventory (BDI) or BDI-II. Of the four studies comparing people receiving cognitive processing therapy with those in a control group, all found that subjects in the active treatment arm had a greater reduction in symptoms of depression than those in the control arm.7074

Our meta-analysis of cognitive processing therapy trials reporting BDI or BDI-II scores (Figure 5) found greater improvement for subjects treated with cognitive processing therapy than for those in the waitlist groups (WMD, −11.9; 95% CI, −18.9 to −4.9). When including the study with a usual care comparison71 the magnitude of benefit decreased slightly (WMD, −10.7; 95% CI, −16.5 to −4.9). The statistical heterogeneity in the analysis was considerable. Regardless of the reason, all four trials found substantial benefits for reducing depression symptoms in adults with PTSD.

Figure 5 is titled “Mean change from baseline in depression (measured by the Beck Depression Inventory) for cognitive processing therapy compared with control, by type of comparator.” The figure displays a forest plot reporting weighted mean difference in depression scores (measured by the Beck Depression Inventory) stratified by cognitive processing therapy versus waitlist and cognitive processing therapy versus usual care. Data for Figure 5 are presented in Appendix F. This figure is described further in the “Prevention or reduction of comorbid medical or psychiatric conditions” section as follows: “Our meta-analysis of cognitive processing therapy trials reporting BDI or BDI-II scores found greater improvement for subjects treated with cognitive processing therapy than for those in the waitlist groups (WMD, −11.9). When including the study with a usual care comparison the magnitude of benefit decreased slightly (WMD, −10.7). The statistical heterogeneity in the analysis was considerable. “

Figure 5

Mean change from baseline in depression (measured by the Beck Depression Inventory) for cognitive processing therapy compared with control, by type of comparator. Note: Timing of outcome assessment: 17 weeks (Chard, 2005), 12 sessions (Monson, 2006), (more...)

These changes were maintained from the posttreatment assessment at 3 months7274 and 9 months72 in two trials. In another trial, the pre- to posttreatment effect size was 1.00; this figure declined to 0.49 at the 3-month follow-up interval.70 The authors attributed this trend to improving depression scores in the waitlist group, not to worsening of depression in the cognitive processing therapy group.

From the above findings and our meta-analysis, we concluded that evidence of moderate strength supports the efficacy of cognitive processing therapy for reducing depression symptoms. This determination is based on consistent, direct, and precise evidence from four trials.

Two trials of cognitive processing therapy assessed anxiety as an outcome using73 the State-Trait Anxiety Inventory (STAI).70,74 One found cognitive processing therapy to be no more effective in reducing symptoms of anxiety than waitlist;70 the other found greater improvement in anxiety for subjects treated with cognitive processing therapy than those receiving usual treatment from intake to posttreatment (p=0.018).73,74 We concluded that evidence is insufficient to determine the efficacy of cognitive processing therapy for reducing anxiety symptoms, based on lack of consistency and imprecise findings of two trials.

All three studies comparing other CT interventions (i.e., that were not cognitive processing therapy) with inactive control groups assessed both depression and anxiety symptoms.52,73,75 In the study comparing CT to self-help booklet and repeated assessment, greater improvement in anxiety (Beck Anxiety Inventory [BAI]) and depression (BDI) were seen among those treated with CT compared with either the self-help booklet or repeated assessments at both 3 and 9 months (p<0.001 for both assessments).73 The study comparing CT for the mentally ill with treatment as usual was effective for reducing depression (BDI-II), anxiety (BAI), and overall psychiatric symptoms (BPRS).75

Our meta-analysis of depression symptom measures found a greater reduction in depression symptoms for subjects treated with CT than for those in waitlist, self-help booklet, and usual care control groups (Cohen’s d = −0.91; 95% CI, −1.20 to −0.62, Appendix F). When only compared with waitlist controls, the effect size was larger (Cohen’s d = −1.06; 95% CI, −1.52 to −0.60, Appendix F).

Our meta-analysis of anxiety symptom measures found a greater reduction in anxiety symptoms for subjects treated with CT than for those in waitlist, self-help booklet, and usual care control groups (Cohen’s d = −0.93; 95% CI, −1.36 to −0.50, Appendix F). When only compared with waitlist controls, the effect size was larger (Cohen’s d = −1.20; 95% CI, −1.67 to −0.73, Appendix F).

Overall, we concluded that evidence of moderate strength supports the efficacy of CT for reducing depression and anxiety symptoms based on consistent and direct evidence.

Quality of Life

One trial of cognitive processing therapy assessed quality of life using the Abbreviated Dyadic Adjustment Scale (ADAS) and the short form of the World Health Organization Quality of Life Scale (WHOQOL).74 The trial reported significant time by condition interactions for social quality of life measures, but not for physical quality of life measures. With data from a single trial (N=59), unknown consistency, and imprecision, evidence was insufficient to determine the efficacy of cognitive processing therapy for improving quality of life.

The trial comparing CT for people with severe mental illness with treatment as usual reported outcomes using the SF-12. The CT group had slightly better quality-of-life outcomes than the usual care group for the SF-12 Physical Component (p=0.002), but not for the Mental Component (p=0.13). With data from a single trial, unknown consistency, and imprecision, evidence was insufficient to determine the efficacy of this particular CT treatment for improving quality of life.

Disability or Functional Impairment

None of the trials that assessed cognitive processing therapy reported outcomes for this category.

Two studies comparing other CT interventions (i.e., that were not cognitive processing therapy) with inactive control groups assessed disability using the Sheehan Disability Scale.52,73 The trial evaluating CT, a self-help booklet, and repeated assessments measured disability or functional impairment with the Sheehan Disability Scale at posttreatment and at 3- and 9-month follow-up assessments.73 At 3 and 9 months, those in the CT group had greater reduction in disability scores than those in the repeat assessments group (p<0.001). The trial comparing CT with a waitlist control also reported greater reduction in disability scores for those in the CT group at 3 months (p<0.0005).52

Our meta-analysis of disability measures found a greater improvement for subjects treated with CT than for those in waitlist and self-help booklet control groups (Cohen’s d = −1.13; 95% CI, −1.76 to −0.51, Appendix F). When only compared with waitlist controls, the effect size was larger (Cohen’s d = −1.41; 95% CI, −2.41 to −0.41, Appendix F).

Overall, we concluded that evidence of moderate strength supports the efficacy of CT for reducing disability based on consistent and direct evidence.

Results for Cognitive Therapy Compared With Active Comparators

Three trials compared CT with exposure therapy.46,72,76 Assessment of these studies appears in the CBT-Exposure section below.

One trial compared CR (N=13) with a relaxation group (N=21) and a combination of prolonged exposure and CR (N=24);46 these results appear in the CBT-Coping Skills section (below). The authors did not report data on the comparative effectiveness of CR and the combination of prolonged exposure and CR. Briefly, because of unknown consistency, imprecision, and data from a single trial (with 13 CR subjects), we conclude that evidence is insufficient about the comparative effectiveness of CR relative to either relaxation or the prolonged exposure-CR combination for reducing PTSD symptoms.

Detailed Synthesis: CBT—Coping Skills

Characteristics of Trials

Table 9 summarizes the characteristics of the four trials meeting our inclusion criteria.4446,49 Further details describing the included studies are provided in Appendix D.

Table 9. Characteristics of included coping skills trials.

Table 9

Characteristics of included coping skills trials.

The trials in this section had a “coping skills” arm(s)—either relaxation training or stress inoculation training. Stress inoculation training is a cognitive behavioral intervention for PTSD in which the basic goal is to help subjects gain confidence in their ability to cope with anxiety and fear stemming from trauma-related reminders. In stress inoculation training, the therapist helps patients increase their awareness of trauma-related cues for fear and anxiety. In addition, clients learn a variety of coping skills that are useful in managing anxiety, such as muscle relaxing and deep breathing.

Two of the four trials compared coping skills interventions with inactive comparators.44,49 One compared prolonged exposure, stress inoculation training, combined prolonged exposure and stress inoculation training, and a waitlist group49 and the other compared relaxation, EMDR, and treatment as usual.44 Both trials were conducted in the United States; one enrolled women who were victims of sexual or nonsexual assault49 and the other enrolled male combat veterans.44 Duration of treatment ranged from 6 to 9 weeks and both studies included posttreatment follow-up assessments at 3 months, although one study also conducted assessments at 6 and 12 months.49 The primary outcome measure for one study was the PSS-I;49 for the other it was the CAPS.44

All four included trials made comparisons with active psychotherapy interventions, such as prolonged exposure or EMDR. Three were conducted in the United States44,46,49 and one in Canada.45 Sample sizes ranged from 35 to 96. Duration of treatment ranged from 6 to 16 weeks. All four trials included posttreatment follow-up assessments at 3 months; three conducted follow-up assessments as far out as 12 months.46 One study enrolled male combat veterans;44 one enrolled victims of sexual and nonsexual assault;49 the other two enrolled heterogeneous groups of subjects with a variety of index trauma types (e.g., physical assault, road accidents, nonroad accident, witnessing a trauma or homicide, sexual assault, being held hostage, bombing, combat). Mean age for subjects in three trials was mid- to late 30s; one sample included slightly older males (age 45 to 52).44 In two trials, 75 percent or more of subjects were female.45,49 The primary outcome for three trials was the CAPS; one study used the PSS-I.49

We rated five coping skills trials otherwise meeting criteria for this section as high risk of bias (Table 10). Two of the five trials compared coping skills interventions with inactive comparators.44,49 One compared prolonged exposure, stress inoculation training, combined prolonged exposure and stress inoculation training, and a waitlist group49 and the other compared relaxation, EMDR, and treatment as usual.44 Both trials were conducted in the United States; one enrolled women who were victims of sexual or nonsexual assault and the other enrolled male combat veterans.44 Duration of treatment ranged from 6 to 9 weeks and both studies included posttreatment follow-up assessments at 3 months, although one study also conducted assessments at 6 and 12 months.49 The primary outcome measure for one study was the PSS-I;49 for the other it was the CAPS.44 We excluded them from our main data synthesis and used them only for sensitivity analyses.

Table 10. Characteristics of coping skills trials excluded from main analyses because of high risk of bias.

Table 10

Characteristics of coping skills trials excluded from main analyses because of high risk of bias.

Results for Coping Skills Compared With Inactive Comparators

PTSD Symptom Reduction

Both trials that compared a coping skills intervention with inactive comparators reported measures of symptom reduction (Table 11).44,49 The trial that compared prolonged exposure, stress inoculation training, combined prolonged exposure and stress inoculation training, and waitlist found greater improvement in PTSD symptoms for subjects treated with stress inoculation training than for those in the waitlist group.49

Table 11. Results at end of treatment for PTSD symptoms for coping skills interventions compared with inactive controls.

Table 11

Results at end of treatment for PTSD symptoms for coping skills interventions compared with inactive controls.

The trial that compared relaxation, EMDR, and treatment as usual found no statistically significant difference between relaxation and treatment as usual using the Impact of Event Scale (IES)-total (Table 11).44 Using the Mississippi scale, both groups had a similar small decrease in symptoms.

Neither study reported follow-up data after the posttreatment assessment for the inactive comparator group—only the active intervention groups were assessed.

For stress inoculation training, with data from a single trial (N=41 subjects in the stress inoculation training and waitlist arms combined), unknown consistency, and imprecision, evidence was insufficient to determine its efficacy. However, the single trial of stress inoculation training suggests that it may be efficacious, but further research is needed to confirm or refute the findings. For relaxation, the trial provides insufficient evidence to determine the efficacy of relaxation—evidence was inconsistent and imprecise.

Loss of PTSD Diagnosis

Both trials reported loss of diagnosis. In one trial, 42 percent of the subjects in the stress inoculation training group and 0 percent in the waitlist group lost their PTSD diagnosis (p<0.001).49

In the other trial, 2 of 9 patients in the relaxation group who completed treatment (out of 13 patients randomized to relaxation) no longer met criteria for PTSD diagnosis. The study did not report data for the treatment as usual group.

For stress inoculation training, with data from a single trial (N=41 subjects in the stress inoculation training and waitlist arms combined), unknown consistency, and imprecision, evidence was insufficient to determine its efficacy. However, the single trial of stress inoculation training suggests that it may be efficacious, but further research is needed to confirm or refute the findings.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Both trials reported on coexisting anxiety and depression symptoms (Table 12).44,49 The trial that included stress inoculation training and waitlist arms found that subjects treated with stress inoculation training had greater reduction in their symptoms of depression than those in the waitlist group; reduction in anxiety symptoms was not statistically significantly different between groups.49

Table 12. Results at end of treatment for depression and anxiety symptoms for coping skills interventions compared with inactive controls.

Table 12

Results at end of treatment for depression and anxiety symptoms for coping skills interventions compared with inactive controls.

The trial comparing relaxation and treatment as usual found a reduction in both depression and anxiety symptoms in the relaxation group; however, the authors reported no statistically significant between-group difference on measures of anxiety and did not provide data on between-group differences for depression.44

For stress inoculation training, with data from a single trial (N=41 subjects in the stress inoculation training and waitlist arms combined), unknown consistency, and imprecision, evidence was insufficient to determine its efficacy. The single trial of stress inoculation training suggests that it may be efficacious, but further research is needed to confirm or refute the findings.

Neither trial reported data on the prevention or reduction of a coexisting medical condition.

Results for Coping Skills Compared With Active Comparators

Of the four included trials comparing a coping skills therapy with an active comparator, three included comparisons with exposure-based interventions;45,46,49 two included comparisons with EMDR;44, 45 two included comparisons with CBT-mixed therapies;46,49 and one included a comparison with CR.46 For assessment of the comparisons with exposure-based therapies, see the CBT Exposure section (below). For assessment of the comparisons with CBT-mixed therapies, see the CBT-Mixed section (below). For assessment of the comparisons with EMDR, see the EMDR section (below).

One trial comparing a relaxation intervention with CR randomly assigned subjects (N=81) to prolonged exposure, CR, prolonged exposure plus CR, or relaxation.46 In summary, direct evidence was insufficient to determine the comparative effectiveness of CR and relaxation. Consistency of the evidence is unknown (limited to this single trial) and results were imprecise, with 34 total subjects in the CR and relaxation groups. Of note, indirect evidence (described in other sections of this report) from comparisons with inactive controls (e.g., waitlist) was insufficient to determine the efficacy of relaxation. In addition, the head-to-head trial described here reported outcomes for the relaxation group that were consistently less favorable than those for the other three groups.

PTSD Symptom Reduction

The trial defined the percentage of patients whose PTSD symptoms improved using the CAPS and IES based on a criterion of 2 standard deviations or more improvement since week 0. Using the IES, the authors reported that 50 percent of the subjects in the CR group and 20 percent of the subjects in the relaxation group improved (p=0.04).46

The trial also reported data on end-state function, determined by a 50 percent drop in PTSD Symptoms Scale, a BDI score of 7 or less, and a STAI score of 35 or more at week 11. A higher percentage of subjects were improved in the prolonged exposure, CR, and prolonged exposure plus CR arms than in the relaxation arm, but the differences were not statistically significant (53% vs. 32% vs. 32% vs. 15%, p=NS).

Loss of PTSD Diagnosis

At week 11, more subjects in the CR group than in the relaxation group no longer met criteria for PTSD; the difference was not statistically significant (65% vs. 55%, p=NS).

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Subjects treated with relaxation did consistently less well than comparators on the BDI; mean change scores for exposure therapy were 13 (95% CI, 8 to 18); for CR, 17 (95% CI, 11 to 22); prolonged exposure plus CR, 18 (95% CI, 13 to 23); and for relaxation, 7 (95% CI, 3 to 11).46

The trial did not report on anxiety symptoms or medical conditions.

Detailed Synthesis: CBT—Exposure

Characteristics of Trials

Table 13 summarizes the characteristics of the 15 trials meeting our inclusion criteria. Further details are provided in Appendix D. Of the 15 included trials, 11 compared exposure therapy (imaginal, in vivo, or prolonged exposure [which includes both components]) with waitlist,25,46,49,72,87,88 usual care,89 treatment as usual,90 present-centered therapy,91,92 or supportive counseling.63 Among these studies, many also included active comparators, including EMDR,45,87 relaxation,45 CR or CT,46,72,76 prolonged exposure plus CR,25,46 stress inoculation training,49 and prolonged exposure plus stress inoculation training.49 Two of the 10 prolonged exposure studies had only active comparators—1 compared prolonged exposure with EMDR and relaxation;45 the other compared prolonged exposure, prolonged exposure plus CR, imaginal exposure, and in vivo exposure.66 One additional study compared virtual reality with imaginal exposure and waitlist among combat veterans in Portugal.93 Finally, 1 study compared a version of prolonged exposure conducted in a group setting with present-centered therapy.92

Table 13. Characteristics of included CBT-exposure trials.

Table 13

Characteristics of included CBT-exposure trials.

These trials generally enrolled subjects with severe or extreme PTSD symptoms. The majority of the trials assessing exposure therapy were conducted in the United States; 1 each was conducted in Japan,89 Canada,45 Israel,90 the United Kingdom,46 and Australia.63 Sample sizes ranged from 24 to 284. Each trial included posttreatment followups after 3, 6, 9, or 12 months. Seven of the trials enrolled a heterogeneous group of subjects with a variety of index trauma types (e.g., accident, disaster, physical assault, sexual assault, witnessing death or serious injury); 4 trials enrolled a majority of subjects with sexual assault-related PTSD;25,49,72,87 2 enrolled subjects with combat-related PTSD;90,92,93 1 enrolled subjects with combat- or terror-related PTSD;90 and 1 enrolled natural disaster victims.88 Mean age ranged from 27 to 63. Eight trials enrolled two-thirds or more female subjects. The primary outcome for the majority of trials was some version of the CAPS (CAPS, CAPS-2, or CAPS-Sx); 3 trials identified the PSS-I as the primary outcome measure.25,49,90

Regarding the type of exposure therapy evaluated by the included trials, the majority evaluated prolonged exposure. One trial compared a modified version of prolonged exposure conducted in a group format to an inactive control condition for combat veterans (group exposure vs. PCT).92 Four examined imaginal exposure.63,66,76,93 Of these 4, 1 trial (N=10) compared imaginal exposure with virtual reality exposure and waitlist;93 1 (N=68) compared it with imaginal exposure plus cognitive restructuring and supportive control;63 1 (N=72) compared it with cognitive therapy;76 and 1 compared it with prolonged exposure, prolonged exposure plus cognitive restructuring, and in vivo exposure.66 Of these 4 trials, 2 were conducted in Australia, 1 in Portugal, and 1 in England. One trial assessed in vivo exposure compared with waitlist among natural disaster victims in Turkey.88

Twelve trials otherwise meeting criteria for this section were rated high risk of bias (Table 14); we excluded them from our main data synthesis but used them in sensitivity analyses.

Table 14. Characteristics of CBT-exposure trials excluded from main analyses because of high risk of bias.

Table 14

Characteristics of CBT-exposure trials excluded from main analyses because of high risk of bias.

Results for Exposure Therapy Compared With Inactive Controls

PTSD Symptom Reduction

Eight of the 11 trials comparing various exposure therapies with an inactive comparator reported measures of PTSD symptom change. All 8 reported greater improvement in PTSD symptoms in the exposure group than in the control group.25,49,72,8791

Our meta-analysis including all trials with sufficient data (available outcome measures were CAPS and PSS-I) (Figure 6) that compared exposure therapy with waitlist or usual care found a greater reduction in PTSD symptoms for subjects treated with exposure than for those in control groups; the effect size was very large (SMD −1.27).89,91

Figure 6 is titled “Mean change from baseline to end of treatment in PTSD symptoms (any measure) for exposure therapy compared with control, by type of comparator.” The figure displays a forest plot reporting the standardized mean difference in PTSD symptoms stratified by exposure therapy versus waitlist and exposure therapy versus usual care. Data for Figure 6 are presented in Appendix F. This figure is described further in the “PTSD symptom reduction” section as follows: “Our meta-analysis including all trials with sufficient data (available outcome measures were CAPS and PSS-I) that compared exposure therapy with waitlist or usual care found a greater reduction in PTSD symptoms for subjects treated with exposure than for those in control groups; the effect size was very large (SMD −1.27).

Figure 6

Mean change from baseline to end of treatment in PTSD symptoms (any measure) for exposure therapy compared with control, by type of comparator. Note: Timing of outcome assessment: 8 weeks (Basoglu, 2007), 12 weeks (Foa, 2005), 9 weeks (Foa, 1999), 6 weeks (more...)

Sensitivity analyses that added trials comparing exposure therapy with present-centered therapy, those rated as high risk of bias, or both had little impact on the results; effect sizes were still large, ranging from −1.19 to −1.09 (Appendix F and Figure 7).

Figure 7 is titled “Sensitivity analysis for mean change from baseline to end of treatment in PTSD symptoms (any measure) for exposure therapy compared with control, by type of comparator: including present centered therapy comparators and high risk of bias studies.” The figure displays a forest plot reporting the standardized mean difference in PTSD symptoms stratified by exposure therapy versus waitlist, exposure therapy versus usual care, exposure therapy versus present centered therapy high risk of bias studies. Data for Figure 7 are presented in Appendix F. This figure is described further in the “PTSD symptom reduction” section as follows: Sensitivity analyses that added trials comparing exposure therapy with present centered therapy, those rated as high risk of bias, or both had little impact on the results; effect sizes were still large, ranging from −24 to −27.9.”

Figure 7

Sensitivity analysis for mean change from baseline to end of treatment in PTSD symptoms (any measure) for exposure therapy compared with control, by type of comparator: including present-centered therapy comparators and high risk of bias studies. Note: (more...)

Our meta-analysis of the trials reporting CAPS scores found a 28.9-point greater reduction for subjects treated with exposure than for those in control groups (WMD, −28.9; 95% CI, −35.5 to −22.3; 4 trials, N=212, Appendix F).

Sensitivity analyses that added trials comparing exposure therapy with present-centered therapy, those rated as high risk of bias, or both had little impact on the results; effect sizes were still large, ranging from −24 to −27.9 (Appendix F).

In general, the effects for reduction of PTSD symptoms were maintained at longer-term followup of 3, 6, 9, or 12 months.

Overall, we concluded that the SOE is high to support the efficacy of exposure therapy for reduction of PTSD symptoms. This conclusion is based on consistent, direct, and precise evidence from trials that used common comparators and found large effect sizes.

Loss of PTSD Diagnosis

Five of the trials comparing people receiving exposure therapy with those in inactive control groups reported on achieving loss of diagnosis. In each one, a substantial percentage of participants treated with exposure therapy lost their PTSD diagnosis (range, 41% to 95%); this was a significantly higher percentage than among controls.

Our meta-analysis for achieving loss of diagnosis found that 66 percent more subjects treated with exposure therapy achieved loss of PTSD diagnosis than in waitlist control groups over 4 to 9 weeks (RD, 0.66; 95% CI, 0.42 to 0.91; 3 trials, N=197, Appendix F). This translates to a NNT of 2. Our sensitivity analysis adding trials that compared exposure therapy with present-centered therapy or with supportive control (there were not sufficient data to conduct sensitivity analyses by adding high risk of bias trials) resulted in a reduced effect size (RD 0.46, Appendix F).

Prevention or Reduction of Comorbid Conditions

Eight trials reported on changes in depression symptoms.25,49,72,8791 All reported a significantly greater decrease in depression symptoms for exposure intervention patients than for controls. Results of our meta-analysis indicated a greater reduction in BDI scores for subjects treated with exposure than for those in waitlist or usual care groups (WMD, −8.2; 95% CI, −10.3 to −6.1; I2=0%, N=363, Appendix F). Together, these trials provide high SOE of the efficacy of exposure therapy for decreasing symptoms of depression in adults with PTSD.

No trial reported on anxiety symptoms or coexisting medical conditions.

Quality of Life

No studies of exposure therapy meeting inclusion criteria and with a waitlist or usual care control reported quality of life outcomes. One trial comparing prolonged exposure with present-centered therapy included a measure of quality of life.91 The study reported that groups did not differ across time (Cohen’s d = 0.09, NS). Evidence was insufficient (because of unknown consistency and imprecision) to determine the efficacy of exposure therapy for improving quality-of-life outcomes.

Disability or Functional Impairment

One trial comparing in vivo exposure with waitlist included a measure of work and social adjustment.88 It found that in vivo exposure led to greater improvement in functional impairment than the waitlist control at 4 weeks (Cohen’s d = 0.8) and 8 weeks (Cohen’s d = 0.6).

One trial comparing prolonged exposure, prolonged exposure plus CR, and waitlist (N=190) included the Social Adjustment Scale. The trial reported numerically greater improvements for the two intervention groups than for the waitlist group, but the differences were not statistically significant (see Appendix D for details).

Evidence was insufficient (because of inconsistency and imprecision) to determine the efficacy of exposure therapy for improving disability or functional impairment.

Results for Exposure Therapy Compared With Active Comparators: Exposure Therapy Versus Cognitive Therapy

Three trials compared exposure therapy and either CR, CT, or cognitive processing therapy.46,72,76 Of these, one compared prolonged exposure with CR,46 one compared imaginal exposure with CT,76 and one compared prolonged exposure with cognitive processing therapy.72 The results from these head-to-head trials did not find either treatment to be statistically significantly better than the other. Our meta-analyses (below) for some outcomes found point estimates favoring exposure therapies (loss of PTSD diagnosis) and for other outcomes favoring cognitive therapies (reduction of PTSD symptoms and depression symptoms). We concluded that the evidence was largely insufficient to determine the comparative effectiveness of therapies for each individual outcome. Nevertheless, considering all of the outcomes across these studies suggests that if a difference in effectiveness exists between treatments, it is small.

PTSD Symptom Reduction

All three trials found that both exposure therapies and cognitive therapies led to substantial decreases in PTSD symptoms from baseline to posttreatment, with no statistically significant difference between the interventions.

Results of our meta-analyses (Figure 8) found no statistically significant difference between exposure therapy and CT (WMD, 4.8; 95% CI, −4.5 to 14.2) or between exposure therapy and cognitive processing therapy (WMD, 3.97; 95% CI, −5.95 to 13.9). Mainly because of imprecision of these findings, we concluded that these trials provide insufficient head-to-head data to determine whether exposure therapy is better or worse than cognitive therapy for reducing PTSD symptoms.

Figure 8 is titled “Mean change from baseline in CAPS for exposure therapy compared with cognitive therapy.” The figure displays a forest plot reporting weighted mean difference in CAPS stratified by exposure therapy versus cognitive therapy and exposure therapy versus cognitive processing therapy.” Data for Figure 8 are presented in Appendix F. This figure is described further in the “PTSD Symptom Reduction” section as follows: “Results of our meta-analyses found no statistically significant difference between exposure therapy and cognitive restructuring (WMD, 4.8; 95% CI, −4.5, 14.2) or between exposure therapy and cognitive processing therapy (WMD, 3.97; 95% CI, −5.95, 13.9).”

Figure 8

Mean change from baseline in CAPS for exposure therapy compared with cognitive therapy. Note: Timing of outcome assessment: mean 16 weeks (Marks, 1998), following 16 sessions (Tarrier, 1999), 6 weeks (Resick, 2002).

Loss of PTSD Diagnosis

All three trials reported data on achieving loss of PTSD diagnosis.46,72,76 Loss of PTSD diagnosis for exposure therapy-treated subjects was equal to or greater than that for CT-treated subjects in all three trials (range 53% to 75%).

Our meta-analysis (Figure 9) found no statistically significant difference between exposure therapy and CT (RD, 0.13; 95% CI, −0.06 to 0.32) or between exposure therapy and cognitive processing therapy (RD 0.0).

Figure 9 is titled “Loss of PTSD diagnosis for exposure therapy compared with cognitive therapy.” The figure displays a forest plot reporting risk difference in lost of PTSD diagnosis stratified by exposure therapy versus cognitive therapy and exposure therapy versus cognitive processing therapy. Data for Figure 9 are presented in Appendix F. This figure is described further in the “Lost of PTSD Diagnosis” section as follows: “Results of our meta-analysis (Figure 9) indicate a trend favoring exposure over CR, but the difference between groups was not statistically significant (RD, 0.13; 95% CI, −0.06 to 0.32). There was no difference between exposure therapy and cognitive processing therapy (RD 0.0).”

Figure 9

Loss of PTSD diagnosis for exposure therapy compared with cognitive therapy.

Prevention or Reduction of Comorbid Conditions

All three trials used the BDI to measure change in depression symptom scores. Although point estimates favored CT and cognitive processing therapy, no study found a statistically significant difference between the interventions.

Our meta-analysis (Figure 10) found no statistically significant difference between interventions. The point estimates favored CT (WMD 2.75) and cognitive processing therapy (WMD 2.94). We concluded, however, that evidence is insufficient (mainly because of imprecision) to determine whether either treatment is more effective for reducing depressive symptoms.

Figure 10 is titled “Mean change in Beck Depression Inventory for exposure therapy compared with cognitive therapy. “ The figure displays a forest plot reporting weighted mean difference in depression (as measured by the Beck Depression Inventory) stratified by exposure therapy versus cognitive therapy and exposure therapy versus cognitive processing therapy. Data for Figure 10 are presented in Appendix F. This figure is described further in the “Prevention or Reduction of Comorbid Conditions” section as follows: “Our meta-analysis (Figure 10) found no statistically significant difference between interventions. The point estimates favored CR (WMD 2.75) and cognitive processing therapy (WMD 2.94).”

Figure 10

Mean change in Beck Depression Inventory for exposure therapy compared with cognitive therapy. Note: Timing of outcome assessment: mean 16 weeks (Marks, 1998), following 16 sessions (Tarrier, 1999), 6 weeks (Resick, 2002).

One trial comparing imaginal exposure with CT used the Beck Anxiety Inventory as a measure of anxiety symptoms.76 It found no significant difference between groups posttreatment or at 12-month followup.

No trial reported on reduction or prevention of a comorbid medical condition as one of their outcomes of interest.

Return to Work or Active Duty

One trial of CT and imaginal exposure (N=72) reported the impact of interventions on one of these outcomes.76 The percentage of patients working was significantly better at 6-month followup (40%) than before treatment (15%); differences between treatment groups were not statistically significant (CT, 37%; imaginal exposure, 44%).

Results for Exposure Therapy Compared With Active Comparators: Exposure Therapy Versus Coping Skills Therapies

Three trials compared exposure therapy with a coping skills therapy.45,46,49 One compared prolonged exposure with stress inoculation training and two compared prolonged exposure with relaxation.

PTSD Symptom Reduction

Figure 11 shows results from the trial comparing prolonged exposure with stress inoculation training. Results did not show a statistically significant difference between treatments. Sensitivity analysis including trials rated as high risk of bias found no difference between treatments (SMD 0.04, 95% CI, −0.46 to 0.54, 2 trials, N=75, Appendix F).

Figure 11 is titled “Mean change from baseline in PTSD symptoms for exposure therapy compared with stress inoculation training.” The figure displays a forest plot reporting the standardized mean difference in PTSD symptoms as measured by the PSS-I, exposure therapy compared with stress inoculation training. Data for Figure 11 are presented in Appendix F. This figure is described further in the “PTSD Symptom Reduction” section as follows: “Figure 11 show results from the trial comparing prolonged exposure with stress inoculation training. Results trended in favor of exposure therapy, but did not reach statistical significance.” There was a standardized mean difference of −0.14 (95% CI −0.69 to 0.41).

Figure 11

Mean change from baseline in PTSD symptoms for exposure therapy compared with stress inoculation training. Note: Timing of outcome assessment: 9 weeks

Our meta-analysis of the two studies comparing exposure therapy with relaxation found a summary effect favoring exposure, but the difference was not statistically significant (WMD −9.7, 95% CI, −22.3, 2.9, 2 trials, N=85, Appendix F).

We concluded that the data are insufficient to determine the comparative effectiveness of exposure relative to stress inoculation training or relaxation for reducing PTSD symptoms, mainly because of imprecision. The analyses were underpowered to detect a small or medium difference in effect size.

Loss of PTSD Diagnosis

All three trials reported data on achieving loss of PTSD diagnosis.45,46,49 In each study, a greater proportion of subjects treated with exposure lost their PTSD diagnosis at posttreatment (87%, 60%, and 75%, respectively) than subjects receiving coping skills interventions (40%, 42%, and 55%, respectively).

The trial comparing prolonged exposure with stress inoculation training found no statistically significant difference between the two therapies (RD, 0.18 favoring exposure therapy; 95% CI, −0.09 to 0.45, N=51, Appendix F). Sensitivity analysis including trials rated as high risk of bias (which added 1 trial to the analysis) found that 26 percent more patients treated with exposure lost their PTSD diagnosis than patients treated with stress inoculation training (RD, 0.26; 95% CI, −0.04 to 0.48, 2 trials, N=75, Appendix F).

Our meta-analysis of the trials comparing exposure with relaxation (Figure 12) found that 31 percent more patients treated with exposure lost their PTSD diagnosis than patients treated with relaxation.

Figure 12 is titled “Loss of PTSD diagnosis for exposure therapy compared with relaxation. “ The figure displays a forest plot reporting the risk difference in loss of PTSD diagnosis, exposure therapy compared with relaxation. Data for Figure 12 are presented in Appendix F. This figure is described further in the “Loss of PTSD Diagnosis” section as follows: “Our meta-analysis of the three trials comparing exposure with relaxation (Figure 12) found that 31 percent more patients treated with exposure lost their PTSD diagnosis than patients treated with relaxation.”

Figure 12

Loss of PTSD diagnosis for exposure therapy compared with relaxation.

We concluded that the data are insufficient to determine the comparative effectiveness of exposure relative to stress inoculation training for achieving loss of PTSD diagnosis, because of unknown consistency and imprecision. The analyses were underpowered to detect a small or medium difference in effect size.

Taken together, consistent, direct, precise findings indicate that exposure therapy is more effective for achieving loss of PTSD diagnosis than relaxation (moderate SOE).

Prevention or Reduction of Comorbid Conditions

All three studies reported BDI-related measures of depression symptoms.45,46,49 The trial comparing exposure with stress inoculation training found no difference between treatments (WMD, −0.15; 95% CI, −5.8 to 5.5, Appendix F).

Our meta-analysis comparing exposure therapy with relaxation found that subjects treated with exposure therapy had greater reduction in depression symptoms than those treated with relaxation (WMD, −5.5; 95% CI, −10.2 to −0.79; 2 trials, N=85, Appendix F).

Because of inconsistency and imprecision, the evidence was insufficient to determine whether exposure therapy is more effective than stress inoculation training for reducing depression symptoms.

Consistent, direct, precise findings indicate that exposure therapy is more effective for improving depression symptoms than relaxation (moderate SOE).

Results for Exposure Therapy Compared With Active Comparators: Exposure Therapy Compared With Eye Movement Desensitization and Reprocessing

Two trials (total N=91) compared prolonged exposure with EMDR.45,87

PTSD Symptom Reduction

Both trials found that prolonged exposure and EMDR led to significant decreases in CAPS scores from baseline to end of treatment, but found no statistically significant difference between interventions. In one trial, prolonged exposure led to greater reductions in re-experiencing and avoidance symptoms of PTSD among completers.45 The results of these two trials provide insufficient data on the comparative effectiveness of prolonged exposure over EMDR for reducing PTSD symptoms, mainly because of imprecision.

Loss of PTSD Diagnosis

In both trials, more participants in the prolonged exposure group than in the EMDR group achieved loss of PTSD diagnosis (~88% vs. ~60%, p>0.05;45 95% vs. 75%, p=0.10887).

Our meta-analysis of these two trials (Figure 13) did not find a statistically significance difference between treatments.

Figure 13 is titled “Percentage of subjects achieving loss of diagnosis for exposure compared with EMDR. ” The figure displays a forest plot reporting the risk difference in loss of PTSD diagnosis, exposure therapy compared with EMDR. Data for Figure 13 are presented in Appendix F. This figure is described further in the “Loss of PTSD Diagnosis” section as follows: “Our meta-analysis of these two trials (Figure 13) found a trend toward a greater percentage of prolonged exposure subjects than EMDR subjects achieving loss of PTSD diagnosis; this did not reach statistical significance.”

Figure 13

Percentage of subjects achieving loss of diagnosis for exposure compared with EMDR. Note: Timing of outcome assessment: 8 weeks (Taylor, 2003), 4.5 weeks (Rothbaum, 2005).

Prevention or Reduction of Comorbid Conditions

Both trials used the BDI to assess change in depression symptom scores. In both trials, prolonged exposure and EMDR led to significant decreases in these symptoms, but the intervention groups did not differ on this measure.45,87

Results for Exposure Therapy Compared With Active Comparators: Exposure Therapy Versus Exposure Plus Cognitive Restructuring

Four trials compared exposure therapy with exposure plus CR.25,46,63,66

PTSD Symptom Reduction

Two trials found no difference between subjects treated with exposure and those treated with exposure plus CR on measures of PTSD symptom reduction.25,46 Another trial found no difference at the end of treatment but an advantage for exposure plus CR at posttreatment followup.63 Finally, one trial found that exposure plus CR led to significantly greater decreases in PTSD symptoms at the end of treatment.66

Our meta-analysis of PTSD symptom reduction found no statistically significant difference between therapies (SMD, 0.25; 95% CI, −0.29 to 0.80; 3 trials, N=259, Appendix F). On this basis, we concluded that evidence is insufficient to determine the comparative effectiveness of exposure with exposure plus CR for reducing PTSD symptoms, as the evidence was both inconsistent and imprecise.

Loss of PTSD Diagnosis

Three of these four trials reported data on achieving loss of PTSD diagnosis.46,63,66 Only one found greater benefit for exposure plus CR over exposure alone (69% lost diagnosis vs. 37%).66

Results of our meta-analysis (Figure 14) indicate that the two interventions did not differ for achieving loss of PTSD diagnosis (RD, −0.01; 95% CI, −0.17 to 0.14; 3 trials, N=146). Taken together, consistent evidence (I2=0%) from three trials supports a conclusion of no significant difference for achieving loss of PTSD diagnosis for exposure therapy alone compared with exposure plus CR (moderate SOE).

Figure 14 is titled “Loss of PTSD diagnosis for exposure compared with exposure plus cognitive restructuring. “ The figure displays a forest plot reporting the risk difference in loss of PTSD diagnosis, exposure therapy compared with exposure plus cognitive restructuring. Data for Figure 14 are presented in Appendix F. This figure is described further in the “Loss of PTSD Diagnosis” section as follows: “Results of our meta-analysis indicate that the two interventions did not differ for achieving loss of PTSD diagnosis (risk difference, −0.01; 95% CI, −0.17 to 0.14; 3 trials, N=146). Taken together, consistent evidence (I2=0%) from three trials supports a conclusion of no significant difference for achieving loss of PTSD diagnosis for exposure therapy alone compared with exposure plus CR.”

Figure 14

Loss of PTSD diagnosis for exposure compared with exposure plus cognitive restructuring. Note: Timing of outcome assessment: 8 weeks (Bryant, 2008 and Bryant, 2003), mean 16 weeks (Marks, 1998).

Prevention or Reduction of Comorbid Conditions

All four trials used the BDI as a measure of change in depression symptoms. Each found no statistically significant difference between interventions from baseline to the end of treatment.

Our meta-analysis found no statistically significant difference between groups for change in BDI score (WMD, 2.78; 95% CI, −1.68 to 7.25; 4 trials, N=299, Appendix F). Overall we concluded that evidence is insufficient to determine comparative effectiveness of exposure and exposure plus CR, largely because of inconsistent results and imprecision.

Results for Exposure Therapy Compared With Active Comparators: Prolonged Exposure Versus Imaginal Exposure Versus In Vivo Exposure

One trial (N=58) compared prolonged exposure, imaginal exposure alone, and in vivo exposure alone.66 All three types of exposure therapy led to substantial decreases in PTSD symptoms, but the authors found no significant differences between the three groups. In addition, the proportions of subjects who no longer met criteria for PTSD after treatment did not differ significantly (41% vs. 37% vs. 35%); the groups also did not differ with respect to reduction in BDI scores. We concluded that evidence is insufficient to determine the comparative effectiveness of these three types of exposure based on this single trial.

Detailed Synthesis: CBT—Mixed Interventions

Characteristics of Trials

Table 15 summarizes the characteristics of the 23 trials meeting our inclusion criteria. Further details about these trials appear in Appendix D. The trials in this section are somewhat heterogeneous in several ways: how authors define and describe “cognitive behavioral therapy,” duration of the intervention, and mode of delivery. Elements of the CBT arm of the studies considered here include: psychoeducation, self-monitoring, stress management, relaxation training, skills training, exposure (imaginal, or in vivo, or both), cognitive restructuring, guided imagery, mindfulness training, breathing retraining, crisis/safety planning, and relapse prevention. The studies varied as to how many sessions (if any) were dedicated to these elements and whether homework was assigned as part of the intervention.

Table 15. Characteristics of included CBT-mixed intervention trials.

Table 15

Characteristics of included CBT-mixed intervention trials.

Eighteen of these 23 trials included an inactive comparator, such as a waitlist (13 trials), usual care (2 trials), or supportive control (3 trials).25,4951,5365,69 Ten of the 24 trials made comparisons with active interventions (i.e., other psychotherapies).25,46,47,49,58,60,63,6668 Of these 10 trials, 5 included an exposure-based intervention as the comparison;25,46,49,63,66 1 used “structured writing therapy”;60 1 used a present-centered therapy;58 2 used relaxation;46,47 and 2 used another CBT-mixed intervention.67,68

Of the 18 trials with inactive comparators, 11 were conducted in the United States; 1 was conducted in Switzerland,57 1 in Canada,53 1 in the Netherlands,60 2 in Australia,59,63 1 in Germany,62 and 1 in France.64 Sample sizes ranged from 23 to 190. Duration of treatment ranged from 4 to 16 weeks. All trials also included posttreatment follow-up assessments after 1, 3, 6, 9, or 12 months, although the follow-up interval for 1 was unclear.60 The majority of trials enrolled a heterogeneous group of subjects with a variety of index trauma types (e.g., childhood abuse [physical, sexual, or mental], physical assault, road accidents, nonroad accident, sexual assault, being held hostage, bombing, combat, witnessing genocide, nonsexual assault, and motor vehicle accidents). Mean age ranged from 30 to 50 years. Most trials enrolled a large majority of female subjects. The primary outcome measure for 9 of these trials was some version of the CAPS (CAPS, CAPS-2, or CAPS-Sx);50,51,53, 2001,54,56,58,61,63,69 4 trials used a form of the PSS (PSS-I or PSS-SR);25,49,55,65 1 trial used the PDS;57 2 trials used the PCL;59,64 1 trial used the Harvard Trauma Questionnaire (HTQ);62 and 1 the IES.60

Of the 10 trials with active comparators, 6 were conducted in the United States; 1 was conducted in the United Kingdom;46 1 in the Netherlands;60 and 2 in Australia.63,66 Sample sizes ranged from 24 to 190. Duration of treatment ranged from 8 to 16 weeks. All trials also included posttreatment follow-up assessments. The majority of trials enrolled a heterogeneous group of subjects with a variety of index trauma types. Mean age ranged from 33.2 to 51.4. Most trials enrolled a large majority of female subjects. The primary outcome for 6 trials was some version of the CAPS (CAPS, CAPS-2, or CAPS-Sx); 2 used the PSS-I,25,49 1 the PCL,47 and 1 the IES.60

Ten trials otherwise meeting criteria for this section were rated high risk of bias (Table 16). We excluded them from our main data synthesis but used them in sensitivity analyses.

Table 16. Characteristics of CBT-mixed intervention trials excluded from main analyses because of high risk of bias.

Table 16

Characteristics of CBT-mixed intervention trials excluded from main analyses because of high risk of bias.

Results for CBT-Mixed Interventions Compared With Inactive Comparators

PTSD Symptom Reduction

Of 18 trials with inactive comparators, 8 reported the CAPS and gave sufficient data to include in meta-analyses. Among these 8 trials, 6 reported reductions in CAPS scores that were statistically significant.

Our meta-analysis (Figure 15) found greater reduction in PTSD symptoms for CBT-mixed interventions than for inactive controls (WMD, −31.1). Statistical heterogeneity was substantial (I2=87%). Much of the heterogeneity may be explained by the diversity of both interventions (as explained above, these interventions used various CBT components). Five trials found a similarly large improvement in CAPS for CBT-mixed intervention groups compared with waitlist controls—about a 30-point greater reduction.50,51,53,54,69 One trial with a waitlist control found even greater benefits (about a 68 point reduction).56 Two of the 8 trials found little to no benefit.58,61,75 One of these compared CBT-mixed interventions with usual care (in which the control patients were often receiving some form of treatment) rather than with waitlist; this likely biased results toward the null.61,75

Figure 15 is titled “Mean change from baseline in CAPS for CBT-Mixed interventions compared with control, by comparator. “ The figure displays a forest plot reporting the weighted mean difference in CAPS stratified by CBT-mixed versus usual care and CBT-mixed versus waitlist. Data for Figure 15 are presented in Appendix F. This figure is described further in the “PTSD Symptom Reduction” section as follows: “Our meta-analysis found greater reduction in PTSD symptoms for CBT-mixed interventions than for inactive controls (weighted mean difference, −31.1). Statistical heterogeneity was substantial (I2=87%).”

Figure 15

Mean change from baseline in CAPS for CBT-mixed interventions compared with control, by comparator. Note: Timing of outcome assessment: 7 weeks (Johnson, 2011), 8 to 12 weeks (Blanchard, 2003), 12 weeks (Cloitre, 2002), 4 weeks (Fecteau, 1999), 12 weeks (more...)

Sensitivity analyses including high risk of bias studies or adding studies with supportive counseling control groups and sensitivity analyses removing each individual study one at a time did not result in any significant differences in findings (Appendix F).

For posttreatment followup at 3 to 6 months, just 2 of the 8 trials reported sufficient CAPS data to permit meta-analysis (Appendix F).50,61 Of these, 1 found significant differences between a CBT-mixed intervention and waitlist (WMD, −22; 95% CI, −36.4 to −7.6).50 One found no significant difference between a CBT-mixed intervention and usual care (WMD, 1.41; 95% CI, −9.8 to 12.6).61 Thus, drawing any strong conclusions about whether reduction of symptoms is maintained at long-term followup is difficult. A third trial reported 3- and 6-month follow-up data, reporting no significant differences between groups, but the control group had all received the intervention by that time.69

We conducted additional meta-analyses to calculate an effect size (Cohen’s d) for change in PTSD symptoms using additional outcome measures reported across all trials with waitlist (CAPS, PSS-I, IES, PCL, PDS). Our meta-analysis found greater reduction in PTSD symptoms for CBT-mixed interventions compared with waitlist (13 trials) and usual care (1 trial) controls, with a very large effect size (SMD, −1.09; 95% CI, −1.4 to −0.78; 14 trials, N=825, Appendix F). Similar to the meta-analysis in Figure 15, statistical heterogeneity was substantial (I2=75.3%). However, also like that analysis, the differences in findings were in the magnitude (not the direction) of the effect; all point estimates favored CBT-mixed interventions, and the vast majority of individual trials reached statistical significance. When the 2 trials with sufficient data with supportive counseling comparators were also included, the effect size decreased slightly (SMD −0.98; 95% CI, −1.28 to −0.68, Appendix F). Sensitivity analyses including high risk of bias studies were similar (Appendix F).

For posttreatment followup at 3 to 6 months, just 4 of the trials reported sufficient data about PTSD symptom measures to permit meta-analysis. Thus, determining with confidence how much of the reduction in symptoms is maintained at long-term followup is difficult, partly because of potential for reporting bias (with the other trials not reporting sufficient data). Of the 4, 3 found statistically significant differences between CBT-mixed interventions and waitlist50,55 or supportive counseling,65 and 1 found no difference between a CBT-mixed intervention and usual care; meta-analysis of the 4 trials found that improvements were maintained, but with a smaller effect size, although still in the medium to large range (SMD, −1.02; 95% CI, −1.43 to −0.61 for the 2 trials with waitlist control; −0.65; 95% CI, −1.21 to −0.09 when including all 4 trials; Appendix F).

Overall, we concluded that evidence of moderate strength supports the efficacy of CBT-mixed interventions for reducing PTSD symptoms. Although magnitude of the effect was somewhat inconsistent, trials were consistent in the direction of effect; our meta-analyses provided fairly precise estimates with moderate to large effect sizes.

Remission (No Longer Having Symptoms)

Two trials comparing CBT-mixed interventions with an inactive comparator reported data on remission of PTSD.59,62 Both trials reported that greater percentages of subjects in CBT-mixed groups than controls achieved remission (61% vs. 21%, p=NR using the PCL;59 82.4% vs. 0%, p<0.001 using the HTQ62). Evidence of moderate strength supports the efficacy of CBT-mixed interventions for achieving remission.

Loss of PTSD Diagnosis

Six trials reported sufficient data on achieving loss of PTSD diagnosis to permit meta-analysis.54,55,58,6365 Our meta-analysis (Figure 16) found that 26 percent more CBT-mixed intervention subjects than waitlist or supportive counseling control subjects achieved loss of PTSD diagnosis (29% when just pooling the three trials with waitlist controls). This translates to a NNT of 4 (and was also 4 when only considering the waitlist controls). We concluded that evidence of moderate strength supports the efficacy of CBT-mixed interventions for achieving loss of PTSD diagnosis.

Figure 16 is titled “Loss of PTSD diagnosis for CBT-mixed interventions compared with control, by type of comparator. “ The figure displays a forest plot reporting risk difference in loss of PTSD diagnosis stratified by CBT-mixed versus waitlist and CBT-mixed versus supportive counseling. Data for Figure 16 are presented in Appendix F. This figure is described further in the “ Loss of PTSD Diagnosis” section as follows: “Our meta-analysis found that 26 percent more CBT-mixed intervention subjects than waitlist or supportive counseling subjects achieved loss of PTSD diagnosis.”

Figure 16

Loss of PTSD diagnosis for CBT-mixed interventions compared with control, by type of comparator.

Two of the trials also reported 3- to 6-month followup data. These findings suggested that the improvements from the CBT-mixed interventions were sustained over time. Our meta-analysis of these trials found a similar result (RD, 0.24; 95% CI, 0.04 to 0.43; Appendix F).

Overall, we concluded that evidence of moderate strength supports the efficacy of CBT-mixed interventions for achieving loss of PTSD diagnosis. Although the magnitude of the effect was somewhat inconsistent across trials, the direction of effect was consistent; results of our meta-analyses provided a fairly precise estimate of the effect.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Twelve of the 18 trials that compared CBT-mixed interventions with an inactive control reported data on depression symptoms using the BDI. All but one of these reported point estimates favoring subjects treated with CBT-mixed interventions; the vast majority reported these findings to be statistically significant. Meta-analysis of these trials found greater improvement in depression symptoms for subjects treated with CBT-mixed interventions than for those in control groups (BDI, mean change from baseline; WMD, −10.4; 95% CI, −14.4 to −6.4; 10 trials, N=662; Appendix F). Sensitivity analysis including two trials with supportive counseling controls found similar results (WMD −9.23; 95% CI, −13.0 to −5.5, Appendix F).

Four of the trials reported sufficient 3- to 6-month postintervention follow-up data for meta-analysis. Meta-analysis of the four trials found that improvements were maintained, but with a slightly smaller effect size (WMD, −5.1; 95% CI, −8.1 to −2.1; Appendix F).

Overall, evidence of moderate strength supports the efficacy of CBT-mixed interventions for improvement in depression symptoms for adults with PTSD. Although the magnitude of the effects across trials was somewhat inconsistent, trials were consistent in the direction of effect, and results of our meta-analyses provided a fairly precise estimate of the effect.

A number of trials also reported reduction in anxiety symptoms; a variety of different measures were used (see Appendix D for details). The most commonly reported measure was the STAI, reported with sufficient data for meta-analysis by 4 of the trials that compared CBT-mixed interventions with an inactive condition. Meta-analysis of these 4 trials found greater improvement in anxiety symptoms for subjects treated with CBT-mixed interventions than for those in control groups (STAI, mean change from baseline; WMD, −11.2; 95% CI, −20 to −2.4; 4 trials, N=172; Appendix F). Based on data from medium risk-of-bias trials, some inconsistency in findings, and imprecision, we determined that the SOE supporting the efficacy of CBT-mixed interventions for improvement in anxiety symptoms for adults with PTSD is low.

Quality of Life

Three trials reported data on quality of life.58,62,64 All three used different measures of quality of life. Two trials found no differences between groups; one reported some differences between groups. Taken together, this evidence is insufficient to determine the efficacy of CBT-mixed interventions for improving quality of life.

One trial (N=60) found no significant difference in change from baseline on the Marks’ Quality of Life Scale (−6.7 vs. −9.6, p=0.26);64 another found no difference in change from baseline on the Quality of Life Inventory (QOLI: 3.4 vs. 0.4, p=0.63).58 One trial (N=70) enrolling Bosnian refugees reported positive effect sizes for both the mental and the physical component summary scales of the SF-36 for CBT subjects compared with usual care subjects (Cohen’s d = 2.1 vs. −0.1, p<0.001, and 1.4 vs. 0.2, p<0.001, respectively).62,75

Disability or Functional Impairment

Five trials reported data on disability or functional impairment50,51,55,59,64 using a variety of measures (Table 17).

Table 17. Results at end of treatment for disability or functional impairment outcomes for CBT-mixed interventions compared with inactive controls.

Table 17

Results at end of treatment for disability or functional impairment outcomes for CBT-mixed interventions compared with inactive controls.

Four of the five trials compared CBT-mixed interventions with waitlist controls; one compared a CBT-mixed intervention with standard care. All four trials with waitlist controls found greater improvements in disability or functional outcomes for subjects who received CBT-mixed interventions—all but one reached statistical significance,59 p=0.07). The trial that compared CBT-mixed with standard care found similar changes in both groups.64 Taken together, results suggest CBT-mixed interventions are efficacious for reducing disability and functional impairment; SOE was low because of some inconsistency and imprecision (low SOE).

Results for CBT-Mixed Interventions Compared With Active Comparators

Of the 10 trials comparing a CBT-mixed intervention with an active comparator, 5 compared it with an exposure-based intervention.25,46,49,63,66 Assessment of head-to-head comparisons with exposure-based interventions is covered in the CBT Exposure section (above). Several of the other trials made comparisons with interventions for which we did not aim to assess comparative effectiveness57,58,60,67 (e.g., comparisons with other CBT-mixed interventions57,67 or “structured writing therapy”).60 In this section, we address the 2 trials comparing CBT-mixed interventions and relaxation interventions.46,47

PTSD Symptom Reduction

Both trials reported that CBT-mixed interventions were more effective than relaxation in reducing symptoms of PTSD. One reported improvement from baseline in CAPS scores of 38 (95% CI, 26 to 50) for the CBT group and 14 (95% CI, 4 to 25) for relaxation.46 The other trial used the PCL as the outcome measure and found a large effect size favoring subjects treated with CBT (between-group effect size: d = 1.6). These between-group treatment differences were maintained at followup (p<0.05). From these two trials, we concluded that CBT-mixed interventions are more effective than relaxation for improving PTSD symptoms (moderate SOE).

Disability or Functional Impairment

One trial reported data on disability or functional impairment using the GHQ Global Improvement measure.46 A greater percentage of subjects in the CBT arm than in the relaxation arm improved functioning, but the difference was not statistically significant (70% to 80% vs. 50% to 55%, respectively, p=NS). Evidence from this single trial was insufficient to determine whether CBT-mixed interventions are more effective than relaxation for improving disability or functional impairment because of unknown consistency (single study) and imprecision.

Detailed Synthesis: Eye Movement Desensitization and Reprocessing (EMDR)

Characteristics of Trials

Table 18 summarizes the characteristics of the seven trials meeting our inclusion criteria. Further details describing the included trials are provided in Appendix D. Five trials had an inactive comparator, such as waitlist,87,111,112 usual care44 or placebo.113 Four had active comparisons with either prolonged exposure,45,87 brief eclectic psychotherapy,48 or relaxation.44,45

Table 18. Characteristics of included EMDR trials.

Table 18

Characteristics of included EMDR trials.

Four of the five trials with inactive comparators were conducted in the United States; one was conducted in Sweden.111 Sample sizes ranged from 21 to 88. Duration of treatment ranged from 4 to 8 weeks. All but one of the studies111 included posttreatment followups after 3, 6, or 9 months. Two of the trials enrolled a heterogeneous group of subjects with a variety of index trauma types (e.g., sexual assault, physical assault, witnessing traumatic events, accidents, and combat); one trial enrolled a majority of subjects with combat-related PTSD;44 one enrolled Swedish public transportation workers who witnessed train accidents or were physically assaulted;111 and two enrolled female victims of sexual assault.87,112 Mean age was roughly similar across trials, ranging from 34 to 49 years. Three trials enrolled 75 percent or more female subjects.87,112,113 The primary outcome for the majority of trials was some version of the CAPS (CAPS, CAPS-2, or CAPS-Sx); two trials identified other primary outcomes, including the PSS-I,112 or IES.111

Among the trials described above, two also included an active comparator arm of either prolonged exposure87 or relaxation.44 One other trial compared EMDR with either prolonged exposure or relaxation in a sample (N=60) of individuals with PTSD from mixed trauma types.45 Treatment duration was 8 weeks with a follow-up assessment at 3 months. Seventy-five percent of the sample was female.

Eight trials otherwise meeting criteria for this section were rated high risk of bias (Table 19), and thus were not included in our main data synthesis, and were only included in sensitivity analyses.

Table 19. Characteristics of EMDR trials excluded from main analyses because of high risk of bias.

Table 19

Characteristics of EMDR trials excluded from main analyses because of high risk of bias.

Results for EMDR Compared With Inactive Comparators

PTSD Symptom Reduction

All seven trials measured PTSD symptom change. Of the trials comparing EMDR with either a waitlist, usual care, or a placebo, all found a greater reduction in PTSD symptom score for EMDR than for comparators.44,87,111113 Not all differences reached statistical significance within individual studies, and point estimates varied widely across trials.

Our meta-analysis (Figure 17) found greater reduction in PTSD symptoms for EMDR than for controls (SMD, −1.08). Treatment gains were maintained for studies reporting followup at 3, 6, or 9 months.

Figure 17 is titled “Mean change from baseline in PTSD symptoms for EMDR compared with control, by type of comparator. “ The figure displays a forest plot reporting the standardized mean difference in PTSD symptoms stratified by EMDR compared with waitlist and EMDR compared with usual care. Data for Figure 17 are presented in Appendix F. This figure is described further in the “ Loss of PTSD Diagnosis” section as follows: Our meta-analysis found greater reduction in PTSD symptoms for EMDR than for controls (SMD, −1.08).

Figure 17

Mean change from baseline in PTSD symptoms for EMDR compared with control, by type of comparator. Note: Timing of outcome assessment: 2 months (Hogberg, 2007), 4.5 weeks (Rothbaum, 2005), 4 weeks (Rothbaum, 1997), 6 weeks (Carlson, 1998).

The effect size we report here is Cohen’s d—a small effect size is 0.2, medium is 0.5, and large is 0.8. Thus, the pooled effect size was very large. However, statistical heterogeneity was substantial (I2=70%) and the confidence interval ranged from almost a small effect size to a very large one.

Our sensitivity analysis including the placebo-controlled trial113 resulted in a slightly lower effect size (SMD, −0.92; 95% CI, −1.55 to −0.29, Appendix F). Our sensitivity analysis also including trials with high risk of bias found a slightly larger benefit of EMDR. The confidence interval ranged from a medium to very large effect size (SMD, −1.13; 95% CI, −1.62 to −0.64; eight trials, N=361; Appendix F).

Overall, we concluded that the evidence supports the efficacy of EMDR over inactive controls for reduction of PTSD symptoms. However, the SOE is low because of lack of consistency and imprecision.

Loss of PTSD Diagnosis

Of the studies that compared EMDR with waitlist, all three reported sufficient data to permit meta-analysis. All three found a greater reduction in the number of subjects meeting criteria for PTSD at posttreatment and at follow-up assessments in the EMDR groups than in control groups.87,111,112

Our meta-analysis (Figure 18) found 64 percent more subjects treated with EMDR than in waitlist control groups achieved loss of PTSD diagnosis. This translates to a NNT of 2. Sensitivity analyses removing each study one at a time, adding the placebo-controlled trial,113 and adding high risk of bias trials produced similar results (RDs ranged from 0.46 to 0.68, Appendix F).

Figure 18 is titled “Loss of PTSD diagnosis for EMDR compared with controls (all were waitlist controls). “ The figure displays a forest plot reporting risk difference in loss of PTSD diagnosis, EMDR compared with controls. Data for Figure 18 are presented in Appendix F. This figure is described further in the “ Loss of PTSD Diagnosis” section as follows: “Our meta-analysis found 64 percent more subjects treated with EMDR than in waitlist control groups achieved loss of PTSD diagnosis.”

Figure 18

Loss of PTSD diagnosis for EMDR compared with control (all were waitlist controls).

We concluded evidence of moderate strength supports the efficacy of EMDR for achieving loss of PTSD diagnosis. This conclusion is based on direct, fairly precise evidence from randomized controlled trials.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

All five studies comparing EMDR with inactive control conditions included a measure of depression symptoms (BDI or HAM-D). Each found greater decreases in symptom scores for EMDR than for inactive controls.

Our meta-analysis (Figure 19) found EMDR had a greater reduction in depression scores than did inactive controls, with a large effect size (SMD, −1.13). Our sensitivity analysis including high-risk-of-bias trials and the placebo-controlled trial found a smaller, but still large, effect size (SMD, −0.87; 95% CI, −1.34 to −0.39, Appendix F).

Figure 19 is titled “Mean change from baseline in depression symptoms for EMDR compared with control, by type of comparator. “ The figure displays a forest plot reporting the standardized mean difference in depression stratified by EMDR compared with waitlist and EMDR compared with usual care. Data for Figure 19 are presented in Appendix F. This figure is described further in the “ Loss of PTSD Diagnosis” section as follows: “Our meta-analysis found EMDR had a greater reduction in depression scores than did inactive controls, with a large effect size (SMD, −1.13).”

Figure 19

Mean change from baseline in depression symptoms for EMDR compared with control, by type of comparator. Note: Timing of outcome assessment: 2 months (Hogberg, 2007), 4.5 weeks (Rothbaum, 2005), 4 weeks (Rothbaum, 1997), 6 weeks (Carlson, 1998).

Overall, we concluded that consistent, direct, and precise evidence supports the efficacy of EMDR over inactive controls for reducing depression symptoms (moderate SOE).

Three trials used STAI to assess anxiety symptoms. Our meta-analysis found that EMDR improved anxiety symptoms more than inactive controls, although results did not reach statistical significance (WMD, −11.1; 95%CI, −23.1 to 0.90; three trials, N=93; Appendix F). Overall findings were inconsistent and imprecise, however, leading us to conclude that evidence is insufficient to determine the efficacy of EMDR over inactive controls for this outcome.

Results for EMDR Compared With Active Comparators: Relaxation

Of the trials comparing EMDR with an active comparator, two compared EMDR and exposure therapy45,87 as assessed in the CBT Exposure section (above); one trial compared EMDR with brief eclectic psychotherapy48 as assessed in the brief eclectic psychotherapy section below. Two trials compared EMDR and relaxation.44,45

PTSD Symptom Reduction

One trial found no statistically significant difference in PTSD symptom reduction between subjects treated with EMDR (N=22) and those treated with relaxation (N=19) using the CAPS45; one found that EMDR (N=10) led to greater PTSD symptom reduction than relaxation (N=13) on the Mississippi Scale for Combat Related PTSD, but not on the IES.44

Pooled analyses of these two trials favored EMDR but found no statistically significant difference (SMD, −0.57; 95% CI, −1.44 to 0.29 using the Mississippi Scale for Combat Related PTSD from the study reporting two measures; SMD, −0.3; 95% CI, −0.8 to 0.2 using the IES; Appendix F). We concluded that evidence is insufficient to determine the comparative effectiveness of EMDR and relaxation for reducing PTSD symptoms; evidence was inconsistent and imprecise.

Loss of PTSD Diagnosis

Two trials comparing EMDR with relaxation both reported achieving loss of PTSD diagnosis at some assessments.44,45 One reported loss of diagnosis at the end of treatment—finding 60 percent of subjects treated with EMDR and 40 percent of subjects treated with relaxation no longer met criteria for PTSD diagnosis.45

Both studies reported loss of diagnosis at 3 months after treatment. Our meta-analysis of 3-month follow-up data (using intention-to-treat data, assuming those lost to followup still had a PTSD diagnosis) found a greater percentage of subjects treated with EMDR than with relaxation no longer having a PTSD diagnosis, but the difference was not statistically significant (RD, 0.34; 95% CI, −0.04 to 0.72; Appendix F). Overall, because of lack of consistency and imprecision, evidence is insufficient to draw conclusions about the comparative effectiveness of EMDR and relaxation for achieving loss of PTSD diagnosis.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Both trials used the BDI to measure depression symptoms; one also reported on anxiety symptoms using the STAI.44 Neither trial found a statistically significant difference between groups for reducing depression symptoms. One trial reported a large between-group effect size (>0.90 using BDI) that was not statistically significant.44 The other trial did not report data for the depression symptoms measure.45

The study reporting anxiety symptoms (N=23) found that relaxation was less effective than EMDR (Cohen’s d=1.15, p<0.01) for reducing symptoms of anxiety at the end of treatment.44

Because of limited evidence from two trials, lack of consistency, and imprecision, head-to-head evidence was insufficient to determine whether EMDR is more effective than relaxation for improving depression or anxiety symptoms.

Detailed Synthesis: Other Psychological Interventions

Characteristics of Trials

Table 20 summarizes the characteristics of 14 trials meeting our inclusion criteria. Further details describing the included studies are provided in Appendix D.

Table 20. Characteristics of included studies of other psychological interventions.

Table 20

Characteristics of included studies of other psychological interventions.

Four trials assessed the efficacy of a short-term manualized cognitive behavior treatment for people with PTSD and substance use disorders called Seeking Safety; three different active control approaches were designed to treat substance use disorders alone or to provide psychoeducation about women’s health issues.33,117,118 One of these three trials compared the addition (to treatment as usual) of a voluntary Seeking Safety intervention with a treatment as usual control group, which comprised incarcerated women enrolled in a residential substance use treatment program in a minimum security wing;33 the relatively large “dose” of treatment as usual along with the voluntary dose of Seeking Safety could bias results toward the null. Another active control involved treatment as usual in a substance use disorder clinic at a Veteran’s Administration outpatient mental health clinic.119 Three of the trials enrolled women generally in their 30s; one enrolled male veterans with a mean age of 54.119 Sample sizes ranged from 49 to 353;33,117,118 one of these was a pilot study (N=49) that may have been underpowered.33 One trial enrolled a sample of incarcerated women;33 two enrolled community-based samples of women seeking substance abuse treatment.117,118 Follow-up assessments were conducted at 3 and 6 months in all trials; one study each conducted additional assessments at 9 months117 or 12 months.118 In addition to assessing the effectiveness of Seeking Safety on symptoms of PTSD, all four trials assessed its effectiveness on substance use. One of the trials used less than half of the Seeking Safety model (only 12 of the 25 sessions/topics) and a large proportion of patients were either abstinent from substances at baseline or had very low levels of use, which could bias results toward the null.118

Two other trials describe an intervention called imagery rehearsal therapy.120,121 This approach is described as a “cognitive-behavioral technique” based on the notion that “waking activity can influence the content of night-time dreams.”120 Imagery rehearsal therapy targets trauma-related nightmares and, by doing so, attempts to reduce the severity of PTSD and improve the quality of sleep. Both trials were conducted in the United States. One trial of this approach versus waitlist involved women with a mean age of about 38 generally with moderate to very severe PTSD primarily associated with a history of sexual trauma (N=168).121 The other trial compared this approach with psychoeducation in male Vietnam-era combat veterans with a mean age of about 60 (N=124).120 Subjects were excluded if they had medical disorders known to affect sleep (e.g., narcolepsy, untreated sleep apnea). All subjects were screened for undiagnosed sleep apnea. Both trials allowed subjects to continue with psychotherapy and medication throughout the study. Both trials conducted follow-up assessments at 3 and 6 months after treatment ended.

Three trials assessed the effectiveness of narrative exposure therapy for PTSD among asylum seekers and refugees. Narrative exposure therapy is described as a “standardized short-term approach based on the principles of cognitive-behavioral exposure therapy by adapting the classical form of exposure therapy to meet the needs of traumatized survivors of war and torture.”127 All three trials were conducted by the same group of researchers. Sample sizes ranged from 32 to 277. Duration of treatment was usually 3 to 5 weeks. All three trials used the PDS to assess PTSD symptom severity. All samples contained males (25% to 69%) and females (31% to 75%) who were generally in their early to mid-30s. One trial compared narrative exposure therapy (n=17), supportive trauma counseling (n=14), and psychoeducation (n=12) in a Ugandan refugee settlement with Sudanese refugees.127 The second trial was also conducted in a Ugandan refugee settlement and compared narrative exposure therapy (n=111), trauma counseling (n=111), and a nontreatment symptom monitoring group (n=55) among Rwandan and Somalian refugees.125 The primary focus of this trial was to examine whether trained lay counselors can carry out effective treatment of PTSD in a refugee settlement as this might have important implications in resource-poor countries experiencing conflict. The third trial compared narrative exposure therapy (n=16) with treatment as usual (n=16) in a sample of asylum seekers living in Germany who were originally from Turkey, the Balkans, or Africa.126 Treatment as usual included “psychotherapy with a focus on stabilizing methods (n=6) and psychoactive medication (n=12).”

Four trials assessed brief eclectic psychotherapy, a manualized intervention that combines cognitive-behavioral and psychodynamic approaches for treating patients with PTSD. Three of the four compared brief eclectic psychotherapy with waitlist21,123 or minimal attention124; one compared it with EMDR.48 Three trials were conducted by the same research group in the Netherlands; one with police officers21 and the other two with heterogeneous group of subjects with a variety of index trauma types.48,123 One trial enrolled a diverse group of predominantly Swiss citizens (63.3%).124 Brief eclectic psychotherapy was conducted for 16 weeks in all four studies. Mean age was similar in all four trials (35 to 40 years of age). Twelve subjects (40.0%) of the Swiss sample were taking psychotropic medications, “mostly antidepressants.”

One trial compared trauma affect regulation (Trauma Affect Regulation: Guide for Education and Therapy [TARGET]) with present-centered therapy and with waitlist.122 The trial enrolled mothers with victimization-related PTSD, primarily low-income and ethnoracial minorities.

Five trials otherwise meeting criteria for this section were rated high risk of bias (Table 21), and thus were not included in our main data synthesis, and were only included in sensitivity analyses.

Table 21. Characteristics of other psychological intervention trials excluded from main analyses because of high risk of bias.

Table 21

Characteristics of other psychological intervention trials excluded from main analyses because of high risk of bias.

Seeking Safety

PTSD Symptom Reduction

Of the four Seeking Safety trials, one compared this approach, standard community treatment, and relapse prevention for women with both PTSD and substance use disorders (N=128).117 Women in the active treatment arm had a greater reduction in symptoms of PTSD than those in the standard community treatment arm (CAPS frequency and intensity, reduction from baseline to posttreatment −15.02 vs. −5.88, p<0.01), and subjects in the standard community treatment arm had worse PTSD severity at the end of treatment and at 3- and 6-month followup (as measured by a standardized composite score for PTSD severity).

All four trials of Seeking Safety found that the intervention reduced symptoms of PTSD; however, between-group differences were not statistically significant, and point estimates favored control groups rather than Seeking Safety for several of the trials.33,117119

Our meta-analysis of mean change from baseline in CAPS scores (reported by three of the trials) found no difference between Seeking Safety and active controls (WMD, 1.45; 95% CI, −2.5 to 5.4; I2=0%; three trials, N=477; Appendix F). Similarly, our meta-analysis of PTSD symptom reduction using any measure found no difference (SMD, 0.04; 95% CI, −0.12 to 0.2; I2=0%; four trials, N=594; Appendix F).

For followup at the end of treatment, all three trials comparing Seeking Safety with relapse prevention reported improvement in PTSD symptoms for both groups, but they found no between-group difference. This was maintained in all three trials at 3- and 6-month followup, and at 9-month117 and 12-month118 followup as well.

Overall, we concluded that evidence is insufficient to determine the efficacy of Seeking Safety for reduction of PTSD symptoms. One trial found Seeking Safety to be efficacious compared with standard care.117 Overall, evidence was limited to one trial designed to assess efficacy, consistency was unknown, and findings were imprecise.

Four trials of Seeking Safety compared with active controls (e.g., relapse prevention) found no differences, providing evidence of moderate strength supporting similar effectiveness for PTSD symptom reduction for people with PTSD and substance use disorders.

Loss of PTSD Diagnosis

The trial that compared Seeking Safety with standard community treatment did not report on achieving loss of diagnosis. The trial of Seeking Safety compared with relapse prevention (N=49) reported loss of PTSD diagnosis.33 At 3-month followup, 39 percent of the women in Seeking Safety and 43 percent of the women in the relapse prevention group met criteria for PTSD. At 6 months, the figures were 53 percent (of women available for followup) in both groups. Their analysis indicated no significant difference in the odds of meeting criteria for PTSD between the two conditions across all points in time (odds ratio for Seeking Safety vs. relapse prevention = 1.22; 95% CI, 0.48 to 3.13). We concluded that evidence is insufficient to support the efficacy of Seeking Safety for achieving loss of PTSD diagnosis (no studies available addressing efficacy); one trial found no difference between Seeking Safety and relapse prevention, but consistency is unknown and findings were imprecise.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

The trial that compared Seeking Safety with standard community treatment reported that subjects in the intervention arm had a greater reduction in substance use or abuse than those in standard community treatment (p<0.001).117 This effect was maintained at the 6-month followup (p<0.05) but not at the 9-month assessment (p=0.06).

Three Seeking Safety trials reported outcome data on substance use or abuse and found no between-group differences for the active treatment arms in the respective studies. One study sample comprised incarcerated women with no access to substances33 and two studies enrolled those in community-based substance use or abuse treatment programs.117,118 Substance use outcome measures included abstinence33,118 and substance use severity.33,117 One trial reported no statistically significant differences between Seeking Safety and relapse prevention117 but did not provide a statistical measure. Another trial reported no between-group differences on several measures of substance use or abuse; Anxiety Stress Index (ASI) composite score for drug (p=0.71), ASI composite score for alcohol (p=0.48), and weeks abstinent (p=0.20).33 Abstinence rates were not significantly different for Seeking Safety and Women’s Health Education (WHE) at 12-month followup.118 Overall, evidence did not support a difference in effectiveness between Seeking Safety interventions and active controls for reducing substance use for people with PTSD.

The trial conducted in male veterans reported better drug use outcomes for those in the Seeking Safety group than in the treatment as usual group (p<0.05), but found no difference between groups for alcohol use outcomes (alcohol use decreased equally in both groups).119

Imagery Rehearsal Therapy

PTSD Symptom Reduction

Both trials assessing imagery rehearsal therapy reported measures of PTSD symptoms. The trial (N=168) with a waitlist control reported that the intervention was more effective than waitlist for reducing symptoms of PTSD as measured by the CAPS (mean change −32.3 vs. −11.3, p=0.001).121 We determined that evidence from this one trial was insufficient to determine the efficacy of imagery rehearsal therapy for reducing PTSD symptoms.

In the trial comparing this intervention and an active comparator (psychoeducation), the authors reported no significant between-group difference in CAPS scores (mean change 7.3 vs. 4.6, Chi-square=0.20).120 The evidence was insufficient to determine whether imagery rehearsal therapy and psychoeducation differ in reducing PTSD symptoms; consistency is unknown (single study) and findings were imprecise.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Both trials assessed the effectiveness of imagery rehearsal therapy for reducing depression symptoms; one used the BDI120 and the other used the HAM-D.121 The trial comparing imagery rehearsal therapy with a waitlist (N=168) found the intervention to be more effective than waitlist for reducing symptoms of depression (HAM-D, effect sizes reported as Cohen d, 0.57 vs. 0.33, p=NS between groups).121 This trial also assessed symptoms of anxiety using the HAM-A. Anxiety symptoms improved in the therapy group (d=0.39) and worsened in the waitlist group (d=−0.16, p=0.04). Evidence was insufficient to determine the efficacy of imagery rehearsal therapy for reducing depression or anxiety symptoms because of unknown consistency (single study), imprecision, and small difference in effect sizes between the intervention and waitlist.

The trial comparing imagery rehearsal therapy with psychoeducation, reported no statistically significant difference between groups for reducing symptoms of depression (BDI scores) at the end of treatment or at any follow-up assessment.120 Mean change scores were as follows: 1 month, −2.69 vs. −1.2 (p=NS); 3 months, −2.05 vs. 0.25 (p=NS); and 6 months, −1.83 vs. −0.14 (p=NS).

Quality of Life

Both trials of imagery rehearsal therapy reported the SF-36 among outcome measures. Neither study found the therapy to be more effective than the comparator for improving quality of life. One trial did not report data;121 the other reported mean change scores for the SF-36 Physical Component at 1 month (2.31 vs. −1.69, p=NR), 3 months (0.55 vs. −2.57, p=NR), and 6 months (−1.37 vs. 1.32, p=NR); it also reported data for the SF-36 Mental Component at 1 month (2.64 vs. −1.68, p=NR), 3 months (1.29 vs. −0.52, p=NR), and 6 months (2.46 vs. 0.26, p=NR).120 Evidence from these two trials was insufficient to determine the efficacy or comparative effectiveness of imagery rehearsal therapy because of unknown consistency (only one study reported data) and imprecision.

Narrative Exposure Therapy

PTSD Symptom Reduction

All three trials assessing narrative exposure therapy used the PDS to measure PTSD symptom reduction. All three compared narrative exposure therapy with an inactive comparator;125127 two compared it with at least one other active intervention not directed at treating PTSD.125,127 All trials found that this intervention reduced symptoms of PTSD more than inactive comparators.

Our meta-analysis (Figure 20) found about a 10-point greater improvement in change from baseline to end of treatment for narrative exposure therapy than for inactive control groups for PDS score (corresponding Cohen’s d −1.25; 95% CI, −1.92 to −0.58, Appendix F). Analyses removing each individual study one at a time did not yield any significant differences in findings (Appendix F).

Figure 20 is titled “Mean change from baseline to end of treatment in PTSD symptoms (measured by PDS) for narrative exposure therapy compared with inactive controls. The figure displays a forest plot reporting weight mean difference in PTSD symptoms as measured by the PDS, narrative exposure therapy compared with inactive controls. Data for Figure 20 are presented in Appendix F. This figure depicts an overall weighted mean difference of −10.20 (95% CI −13.05 to −7.35), favoring narrative exposure therapy over inactive controls. Overall I2 = 19.0%.

Figure 20

Mean change from baseline to end of treatment in PTSD symptoms (measured by PDS) for narrative exposure therapy compared with inactive controls. Note: Timing of outcome assessment: after 5 to 17 sessions (Neuner, 2010), 3 weeks (Neuner, 2008), 3 to 4 (more...)

One trial reported a reduction (but no data) in PTSD symptoms for subjects in the intervention group at 6 months after the end of treatment;126 another reported that the intervention was significantly better in reducing symptoms of PTSD than no treatment (i.e., monitoring group) from baseline to 6-month followup (d=1.4 and 0.08, respectively, p<0.001).125 One year post-treatment data were reported by one trial; subjects in the narrative exposure group had better improvement on the PDS than those in the inactive treatment group (d=1.6 and −0.09, respectively, p<0.01).127

Overall, we concluded that evidence of moderate strength supports the efficacy of narrative exposure therapy for reducing PTSD symptoms, based on consistent, direct, and precise evidence from three trials.

Loss of PTSD Diagnosis

All three trials of narrative exposure therapy and an inactive control reported data on achieving loss of PTSD diagnosis.125127 Two of these also had at least one other active intervention not directed at treating PTSD.125,127 All three trials found point estimates favoring narrative exposure therapy.

Our meta-analysis (Figure 21) found that 15 percent more subjects were no longer diagnosed with PTSD at the end of treatment for narrative exposure therapy than for inactive comparator groups.

Figure 21 is titled “Loss of PTSD diagnosis for narrative exposure therapy compared with inactive controls.” The figure displays a forest plot reporting risk difference in loss of PTSD diagnosis, narrative exposure therapy compared with inactive controls. Data for Figure 21 are presented in Appendix F. This figure is described further in the “Loss of PTSD Diagnosis” section as follows: “Our meta-analysis found that 15 percent more subjects were no longer diagnosed with PTSD at the end of treatment for narrative exposure therapy than for inactive comparator groups.” Overall risk difference was 0.15 (95% CI 0.01 to 0.30). Overall I2 = 48.1%.

Figure 21

Loss of PTSD diagnosis for narrative exposure therapy compared with inactive controls. Note: Neuner, 2010, reported loss of diagnosis at 6-months; Neuner, 2008, at 9-months; and Neuner, 2004, at one-year.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Two trials evaluated the effectiveness of narrative exposure therapy on coexisting psychiatric conditions;126,127 one used the HSCL-25 Depression scale126 the other used the SRQ-20.127 One trial reported greater improvement in depression for subjects treated with narrative exposure therapy than for those receiving treatment as usual (HSCL-25 Depression scale, between-group effect size, Cohens d = 0.54, p=NR).126 The other trial found no significant differences among narrative exposure therapy, trauma counseling, or psychoeducation on the SRQ-20 (reductions from 2.2 to 3.3 across groups, p=NS).127 Using this same measure, narrative exposure therapy was no more effective than trauma counseling or psychoeducation in reducing the number of cases classified as suffering from a severe mental illness (p=0.08) at 1-year followup.

Overall, we concluded that evidence is insufficient to determine the efficacy of narrative exposure therapy for preventing or reducing coexisting psychiatric conditions. Evidence from two trials was inconsistent and imprecise.

One trial (N=32) evaluated the effectiveness of narrative exposure therapy on pain as assessed by part C of the Composite International Diagnostic Interview (CIDI-C) Pain score.126 Whether all of the subjects in this sample had pain was unclear, but the authors stated that 79 percent of their sample reported physical torture experiences. The between-group effect size of narrative exposure therapy and treatment as usual for the CIDI Pain score was d=0.65 (for CIDI-C pain score, a significant time by treatment interaction was found, p=0.034, but no significant main effect of time, p=0.46, or treatment, p=0.35). We concluded that evidence is insufficient to determine the efficacy of narrative exposure therapy for preventing or reducing pain. Evidence was imprecise and consistency is unknown.

Quality of Life

One trial evaluated quality of life using the Psychological Health subscale from the Medical Outcomes Study Short Form 12.127 Narrative exposure therapy was more effective for improving quality of life than trauma counseling (effect sizes pre- to posttreatment: −0.6 and 0.1, respectively, p<0.01) but not more effective than psychoeducation (p=0.54). We concluded that evidence is insufficient to determine the efficacy of narrative exposure therapy for improving quality of life. Evidence was imprecise and consistency is unknown.

Brief Eclectic Psychotherapy

PTSD Symptom Reduction

Three trials reported measures of PTSD symptom reduction for brief eclectic psychotherapy compared with an inactive comparator.21,123,124 In all three trials (using different outcome measures), brief eclectic psychotherapy was effective in reducing symptoms of PTSD. One reported greater reduction in symptoms measured by the SI-PTSD Reexperiencing score (Cohen’s d = 0.45), Avoidance score (d = 0.52), and Hyperarousal score (d = 0.39) (for Cohen’s d, 0.2 would indicate a small effect size; 0.5 a medium effect size).123 Another used the CAPS (change from baseline to end of treatment: −17.8 vs. −7).124 The third study reported change in the frequency of symptoms within each symptoms cluster and found that brief eclectic psychotherapy was effective in eliminating reexperiencing symptoms (p<0.01 at end of treatment and at 3-month followup), was not effective in reducing the number of avoidance symptoms (<3 avoidance symptoms) at the end of treatment, but was effective at 3-month followup (p<0.001), and was effective in reducing the number of hyperarousal symptoms (<2 avoidance symptoms) (p<0.01 at end of treatment and p<0.05 at 3-month followup).21 Based on these three trials, we concluded that consistent, direct evidence supports the efficacy of brief eclectic psychotherapy for reducing PTSD symptoms, likely with a small to medium effect size (low SOE). However, the evidence was imprecise. Each trial reported different outcome measures, and data were not sufficient to determine the effect size accurately.

The trial comparing brief eclectic psychotherapy with EMDR reported that both treatments were equally effective in reducing PTSD symptom severity, but that EMDR resulted in faster recovery.48 The study reported improvement in PTSD symptoms in both groups using the IES-R and the SI-PTSD, but greater improvement from baseline to the first assessment for those treated with EMDR than for those treated with brief eclectic psychotherapy (SI-PTSD, mean estimated between-group difference 10.80; 95% CI 6.37 to 15.23)48 The between-group difference was no longer significant at the second assessment, conducted after both groups had completed treatment.

Due to unknown consistency (with data from a single trial), risk of bias, and imprecision, we graded the evidence as insufficient to determine the comparative effectiveness of brief eclectic psychotherapy and EMDR.

Remission (No Longer Having Symptoms)

One trial (N=30) reported data on symptom remission. At the end of treatment, 2 of 16 subjects (12.5%) in the group receiving brief eclectic psychotherapy were described as being in complete remission based on a total CAPS score of <20.124 At 6-month followup, 3 subjects (18.8%) were fully remitted. None of the subjects in the waitlist group achieved complete remission. We concluded that evidence from this single trial was insufficient to determine the efficacy of brief eclectic psychotherapy for remission of PTSD symptoms. Consistency is unknown (single study) and findings were imprecise.

Loss of PTSD Diagnosis

All three trials reported that brief eclectic psychotherapy was more effective than waitlist in reducing the proportion of subjects who continued to meet criteria for PTSD at the end of treatment and at followup. One trial (N=30), using a definition of CAPS<50 found more subjects receiving the intervention than on the waitlist lost their diagnosis of PTSD (2 subjects, 12.5% vs. 0 subjects, 0%).124 The other two trials used the SI-PTSD to determine PTSD diagnosis. One trial (N=24) reported that 83.3 percent of subjects receiving brief eclectic psychotherapy and 25 percent on a waitlist (p<0.05) no longer met criteria for PTSD at the end of treatment.123 The other trial (N=42) reported that 91 percent of subjects receiving the intervention and 50 percent on a waitlist (p<0.01) lost their diagnosis at the end of treatment; these changes were essentially maintained at 3-month followup (96% versus 35%, p<0.01).21 We concluded that evidence supports the efficacy of brief eclectic psychotherapy for achieving loss of PTSD diagnosis; however, findings from three trials (total N=96) were inconsistent (ranging from a small effect to a large effect) and imprecise (low SOE).

The trial comparing brief eclectic psychotherapy with EMDR reported that both treatments had similar benefits for achieving loss of diagnosis, but that EMDR resulted in earlier benefits.48 The study reported more improvement (a higher rate of achieving loss of PTSD diagnosis) among completers for the EMDR group than for the brief eclectic psychotherapy group at the first assessment (92.2% vs. 52.3%, p<0.001), but found no significant difference between groups at the second assessment (93.7% vs. 85.7%, p=0.30), conducted after both groups had completed treatment. These results included 51 out of 70 and 44 out of 70 subjects (not accounting for missing data from dropouts) in the EMDR and brief eclectic psychotherapy groups at the first assessment and 48 out of 70 and 42 out of 70 at the second assessment, respectively. Due to unknown consistency (with data from a single trial), risk of bias, and imprecision, we graded the evidence as insufficient to determine the comparative effectiveness of brief eclectic psychotherapy and EMDR.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

All three trials comparing brief eclectic psychotherapy with waitlist reported on reduction of depression and anxiety. Two used the Hospital Anxiety and Depression Scale (HADS) as an outcome measure. Both reported that brief eclectic psychotherapy was more effective than the waitlist in reducing symptoms of depression at the end of treatment and at later followup (Cohen’s d = 1.0 for both time points in one trial124 and d=0.38 for the other123). One trial used the SCL-90 as a multidimensional indicator of psychopathology and reported that brief eclectic psychotherapy was more effective than waitlist in reducing symptoms of depression at the end of treatment (data NR, p<0.01); this change was maintained at the 3-month followup.21

Two trials reported that brief eclectic psychotherapy was more effective than waitlist in reducing symptoms of anxiety as assessed by the HADS (Cohen’s d = 0.8, p<0.05; and d = 0.9, p<0.05 for one trial at the end of treatment and at followup124; for the other trial d = 0.54123). The trial using the SCL-90 reported that brief eclectic psychotherapy was more effective than waitlist in reducing symptoms of anxiety at the end of treatment and at 3-month followup (data NR, p-values of < 0.05 and <0.01).21

Evidence (low SOE) supports the efficacy of brief eclectic psychotherapy for reducing depression or anxiety symptoms. Although these trials (total N=96) support efficacy, the evidence was somewhat inconsistent and imprecise; effect sizes and outcomes not reported in one trial and ranged from a medium to a very large effect in the other two.

The trial comparing brief eclectic psychotherapy with EMDR reported measures of depression and anxiety symptoms (using the HADS depression and the HADS anxiety).48 Similar to findings for other outcomes (e.g. PTSD symptoms), the study reported greater improvement from baseline to the first assessment for those treated with EMDR than for those treated with brief eclectic psychotherapy, but no significant difference between groups at the second assessment (see Appendix D for detailed data).

Due to unknown consistency (with data from a single trial), risk of bias, and imprecision, we graded the evidence as insufficient to determine the comparative effectiveness of brief eclectic psychotherapy and EMDR.

Return to Work or Active Duty

Two trials reported outcomes related to work—one reported the percentage of subjects on sick leave;123 the other reported the percentage who had returned to work.21 The former trial (N=24) found fewer subjects on sick leave for the brief eclectic psychotherapy group compared with those on the waitlist, but the difference was not statistically significant (d=0.33, p=0.06).123 The other trial (N=42) reported a statistically significant difference between the groups at the end of treatment—86 percent of the intervention group and 60 percent of the waitlist group had returned to work (p<0.05).21 Together, evidence from these two trials suggests that brief eclectic psychotherapy is efficacious for improving return to work; SOE is low, primarily because of imprecision.

Trauma Affect Regulation

PTSD Symptom Reduction

The trial comparing trauma affect regulation, present-centered therapy, and waitlist reported greater improvement in PTSD symptoms for those treated with trauma affect regulation than those in the waitlist group (CAPS mean change from baseline: −23.6 vs. −6.2, p<0.001).122

For this outcome and the others (below) from this trial, due to unknown consistency (with data from a single trial), risk of bias, and imprecision, we graded the evidence as insufficient to determine the efficacy of trauma affect regulation.

Remission

The trial reported that more people in the trauma affect regulation group than in the waitlist group achieved full remission at posttreatment (21% vs. 0%, p<0.001).

Loss of Diagnosis

The trial reported that more people in the trauma affect regulation group than in the waitlist group achieved loss of diagnosis at posttreatment (35% vs. 11%).

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

The trial reported greater improvement in depression symptoms and anxiety symptoms for the trauma affect regulation group than for the waitlist group (BDI: −4.4 vs. −0.3, p<0.01; STAI: −6.7 vs. −0.4, p=0.19).

Key Question 2. Comparative Effectiveness of Different Pharmacological Treatments for Adults With PTSD

For this question, we included placebo-controlled trials (indirect evidence) and head-to-head trials (direct evidence) of pharmacotherapies. First, we evaluated the evidence of efficacy for the included medications (compared with placebo) and then assessed the direct evidence and conducted network meta-analysis to utilize both the indirect and direct evidence to inform a determination of the comparative effectiveness of pharmacotherapies. In the bulleted text below we summarize the main overall key points and then the key points for each medication class and report the strength of evidence (SOE) where appropriate.

The primary outcomes of interest for determining whether treatments are effective for adults with PTSD are improving PTSD symptoms, inducing remission, and losing PTSD diagnosis; we focus more on these outcomes than on other outcomes in the key points. We also comment on other outcomes of interest, such as prevention or reduction of coexisting medical or psychiatric conditions, quality of life, disability or functional impairment, and return to work or active duty. The findings in these key points are primarily based on meta-analyses of the trials that we rated low or medium risk of bias described later in the detailed synthesis sections of the chapter. Those trials are cited in the detailed synthesis and related tables. In the detailed synthesis section for each treatment, we provide section headers for each outcome reported (PTSD symptoms, remission, loss of PTSD diagnosis, prevention or reduction of coexisting medical or psychiatric conditions, quality of life, disability or functional impairment, and return to work or active duty). If an outcome does not appear, no trial reported data on it.

Key Points: Overall—Efficacy

  • Evidence supports the efficacy of fluoxetine, paroxetine, sertraline, topiramate, and venlafaxine for improving PTSD symptoms (moderate SOE).
  • Evidence suggests that risperidone may have some, albeit minimal, benefit for reduction of PTSD symptoms (low SOE).
  • Evidence was insufficient to determine efficacy for other medications.
  • For the medications with evidence of efficacy, the mean size of the effect for improving symptoms was small or medium (mean change from baseline in the Clinician-Administered PTSD Scale [CAPS] compared with placebo ranged from −4.9 to −15.5 for the medications with moderate SOE).
  • However, paroxetine and venlafaxine also had evidence of efficacy for inducing remission, with numbers needed to treat (NNTs) of ~8 (moderate SOE).

Table 22 summarizes the efficacy and SOE for all included medications for improving PTSD symptoms, inducing remission, and achieving loss of PTSD diagnosis.

Table 22. Summary of efficacy and strength of evidence of pharmacologic treatments for adults with PTSD, by drug class.

Table 22

Summary of efficacy and strength of evidence of pharmacologic treatments for adults with PTSD, by drug class.

Key Points: Overall—Comparative Effectiveness

  • Very few head-to-head trials were identified.
    • One four-arm trial enrolling veterans with comorbid alcohol dependence compared desipramine with paroxetine (N=88) and found similar reduction in PTSD symptoms (CAPS, mean change from baseline from −33.2 to −36.4) and depression symptoms, but found greater improvement in alcohol use outcomes for those treated with desipramine than those treated with paroxetine (low SOE).132
    • Evidence from one large, multicenter (59 sites) trial comparing venlafaxine ER, sertraline, and placebo (N=538) found no statistically significant difference between venlafaxine and sertraline (moderate SOE).133
  • Our network meta-analysis of 28 trials (4,817 subjects) found paroxetine and topiramate to yield greater improvement in PTSD symptoms than most other medications (low SOE; primarily based on indirect evidence).
    • When compared with other medications with moderate SOE supporting efficacy, paroxetine was more effective than sertraline (weighted mean difference [WMD], −7.6; 95% credible interval [CrI], −12 to −2.8), but was not significantly different from fluoxetine, topiramate, or venlafaxine (low SOE).
    • When compared with other medications with moderate SOE supporting efficacy, topiramate was more effective than fluoxetine (WMD, 8.6; 95% CrI, 2.4 to 14.9), sertraline (WMD, 11; 95% CrI, 5.7 to 16.6), and venlafaxine (WMD, −8.8; 95% CrI, −15 to −2.5), but was not significantly different from paroxetine (low SOE).

Key Points: Alpha Blockers

  • Evidence was insufficient to determine efficacy of prazosin for improving outcomes for adults with PTSD. Improvement in PTSD symptoms was greater for subjects treated with prazosin than for those who received placebo, but the difference did not reach statistical significance and findings were imprecise (CAPS mean change from baseline compared with placebo, WMD, −8.86; 95% CI, −22.06 to 4.33; standardized mean difference [SMD], −0.40; 95% CI, −0.97 to 0.16, two trials, N=50).

Key Points: Anticonvulsants

  • Consistent, direct, fairly precise evidence from three trials supported the efficacy of topiramate for reduction of PTSD symptoms (CAPS mean change from baseline compared with placebo, WMD, −15.53; 95% CI, −19.40 to −11.65; SMD −0.96; 95% CI −1.89 to −0.03; N=142, moderate SOE). Evidence was insufficient to determine the efficacy of topiramate for improving other outcomes for adults with PTSD.
  • Evidence was insufficient to determine the efficacy of divalproex, lamotrigine, or tiagabine. Consistency is unknown (with either zero or one trials contributing data for each medication) and findings were imprecise.

Key Points: Atypical Antipsychotics

  • Evidence from two small trials (total N=34) was insufficient to determine whether olanzapine is efficacious for improving PTSD symptoms, inducing remission, or for improving other outcomes for adults with PTSD.
  • Existing evidence suggested that risperidone has little or no clinically significant benefit for reduction of PTSD symptoms on average (CAPS mean change from baseline compared with placebo: WMD, −4.60; 95% CI, −9.01 to −0.20; SMD, −0.26; 95% CI, −0.52 to −0.00, four trials, N=419, low SOE). Although subjects treated with risperidone had a statistically significant reduction in PTSD symptoms compared with those receiving placebo, trials had medium risk of bias, the magnitude of difference was small and likely not clinically significant, and findings were imprecise.

Key Points: Benzodiazepines

  • No studies with low or medium risk of bias (insufficient SOE).

Key Points: Serotonin and Norepinephrine Reuptake Inhibitors (SNRIs)

  • Consistent, direct, precise evidence supports the efficacy of venlafaxine ER for improving PTSD symptoms (CAPS mean change from baseline: WMD, −7.2; 95% CI, −11.0 to −3.3; two trials, N=687), inducing remission (risk difference 0.12; 95% CI, 0.05 to 0.19; NNT 9), improving quality of life, and improving disability or functional impairment (moderate SOE).
  • No trials that assessed desvenlafaxine or duloxetine (insufficient SOE).

Key Points: Selective Serotonin Reuptake Inhibitors (SSRIs)

  • Evidence was insufficient to determine the efficacy of citalopram.
  • PTSD symptoms: consistent, direct, and precise evidence from trials supports the efficacy of fluoxetine, paroxetine, and sertraline for improving PTSD symptoms (moderate SOE). The magnitude of benefit is in the small to medium range—CAPS mean change from baseline compared with placebo from −4.9 (95% CI, −7.4 to −2.4; seven trials, N=1,085) for sertraline to −12.6 (95% CI, −15.7 to −9.5; two trials, N=886) for paroxetine (Cohen’s d from −0.25 to −0.49).
  • Remission (no longer having symptoms): consistent, direct, precise information from two trials (N=346) supports the efficacy of paroxetine for achieving remission—best evidence found 12.9 percent more subjects treated with paroxetine than with placebo achieved remission,134 NNT 8 (moderate SOE). Evidence was insufficient to determine the efficacy of fluoxetine or sertraline for achieving remission because of unknown consistency and imprecision.
  • Depression symptoms: both fluoxetine and paroxetine improve depression symptoms for adults with PTSD (moderate SOE). Evidence for sertraline does not support its efficacy for improving depression symptoms for adults with PTSD (low SOE).
  • Anxiety symptoms: greater improvement in anxiety symptoms for subjects treated with fluoxetine than for subjects who received placebo (moderate SOE). Evidence was insufficient to determine the efficacy of paroxetine (no trials reported) or sertraline (two trials, N=377, with inconsistent and imprecise findings).
  • Disability or functional impairment: insufficient evidence to determine the efficacy of fluoxetine and sertraline. For paroxetine, consistent, direct, and precise findings support its efficacy (mean change from baseline in the Sheehan Disability Scale (SDS): WMD, −2.3; 95% CI, −3.3 to −1.4; two trials, N=886, moderate SOE).
  • Achieving loss of PTSD diagnosis, improving quality of life, or return to work or active duty: evidence was insufficient for all SSRIs.

Key Points: Tricyclic Antidepressants (TCAs)

  • Insufficient SOE to determine efficacy. We found no studies with low or medium risk of bias.

Key Points: Other Second-Generation Antidepressants

  • Insufficient SOE to determine efficacy of bupropion, mirtazapine, nefazodone, or trazodone.

Detailed Synthesis: Placebo-Controlled Trials of Alpha-Blockers

Characteristics of Trials

We found two studies that met our inclusion criteria for this section (Table 23). Both trials were conducted within VA Medical Centers in the United States and compared prazosin with placebo. Both enrolled subjects with moderate to severe PTSD. Both enrolled all or a large majority of male subjects; average age was similar (53 to 56 years). Trial durations were 20 weeks135 and 8 weeks.136 Further details describing the included trials are provided in Appendix D.

Table 23. Characteristics of included placebo-controlled trials of alpha-blockers.

Table 23

Characteristics of included placebo-controlled trials of alpha-blockers.

Results for Placebo-Controlled Trials of Alpha-Blockers

PTSD Symptom Reduction

Both trials reported numerically greater improvements in CAPS for subjects treated with prazosin than for those receiving placebo.135,136 Similarly, our meta-analyses found greater improvement in PTSD symptoms for subjects treated with prazosin, but the difference did not reach statistical significance and findings were imprecise (mean reduction in CAPS: WMD, −8.86, 95% CI, −22.06 to 4.33; SMD −0.40, 95% CI, −0.97 to 0.16, two trials, N=50, Appendix F). Overall, the evidence from two trials was insufficient to determine efficacy of prazosin for improving outcomes for adults with PTSD, primarily because of imprecision.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

One of the included studies (N=40) reported the Hamilton Depression Scale (HAM-D) to assess depression.136 The study found that patients treated with prazosin had a greater reduction in depression symptoms than those administered placebo, but the difference between groups was not statistically significant (−5.6 vs. −0.6, p = 0.08). We concluded that evidence is insufficient for determining whether prazosin is effective for improving depression symptoms for adults with PTSD; consistency is unknown (single study) and findings were imprecise.

Detailed Synthesis: Placebo-Controlled Trials of Anticonvulsants/Mood Stabilizers

Characteristics of Trials

Table 24 summarizes the five trials that met inclusion criteria. Further details are provided in Appendix D. The trials enrolled subjects with moderate to severe PTSD. Three were conducted in the United States; one in Iran;137 and one in Brazil.138 Sample sizes ranged from 35 to 232. Treatment duration ranged from 8 to 12 weeks. Two of the included studies focus on combat-related PTSD;137,139 three enrolled a heterogeneous group of subjects with a variety of index trauma types (e.g., physical and sexual assault/violence, witnessing harm or death, combat, natural disaster, childhood sexual abuse, childhood physical abuse, motor vehicle accident).138,140,141 The trials generally recruited middle-aged adults, with mean ages ranging from ~40 to ~55 years. Three trials enrolled at least two-thirds female subjects;138,140,141 two enrolled all or nearly all males. The primary outcome for all five trials was some version of the CAPS.

Table 24. Characteristics of included placebo-controlled trials of anticonvulsants, by drug.

Table 24

Characteristics of included placebo-controlled trials of anticonvulsants, by drug.

We rated three trials otherwise meeting criteria for this section as high risk of bias (Table 25). We excluded them from our main data synthesis, and used them only in sensitivity analyses. Appendix E provides additional rationale for risk of bias assessments. Briefly, the trials deemed high risk of bias only analyzed subjects who completed treatment (did not use an intention-to-treat analysis) or had very high attrition or differential attrition rates.

Table 25. Characteristics of placebo-controlled trials of anticonvulsants excluded because of risk of bias.

Table 25

Characteristics of placebo-controlled trials of anticonvulsants excluded because of risk of bias.

Results of Placebo-Controlled Trials of Anticonvulsants/Mood Stabilizers

PTSD Symptom Reduction

All five of the included studies reported CAPS. Our meta-analyses (Figure 22 and Appendix F) found topiramate to improve PTSD symptoms more than placebo (WMD, −15.53; SMD, −0.96; 95% CI, −1.89 to −0.03). A sensitivity analysis adding the topiramate trial rated high risk of bias144 did not significantly change the results (WMD, −15.29; 95% CI, −19.00 to −11.57; I2 0%, four trials, N=182, Appendix F). Overall, we concluded that evidence of moderate strength supports the efficacy of topiramate for reducing PTSD symptoms; evidence was consistent, direct, and fairly precise.

Figure 22 is titled “Change in CAPS for anticonvulsants compared with placebo.” The figure displays a forest plot reporting the standardized mean difference in CAPS stratified by divalproex compared with placebo, tiagabine compared with placebo, and topiramate compared with placebo. Data for Figure 22 are presented in Appendix F. This figure is described further in the “PTSD symptom reduction” section as follows: “Our meta-analyses found topiramate to improve PTSD symptoms more than placebo (WMD, −15.53; SMD, −0.96; 95% CI −1.89 to −0.03).” “Evidence was insufficient to determine the efficacy of divalproex or tiagabine for improvement in PTSD symptoms. Consistency is unknown (with just 1 study contributing data for each medication) and findings were imprecise.”

Figure 22

Change in CAPS for anticonvulsants compared with placebo. Note: Timing of outcome assessment: 8 weeks (Davis, 2008), 12 weeks (Davidson, 2007; Akuchekian, 2004; Tucker, 2007; Yeh, 2011).

Evidence was insufficient to determine the efficacy of divalproex or tiagabine for improvement in PTSD symptoms. Consistency is unknown (with just one study contributing data for each medication) and findings were imprecise.

Three studies, one each for divalproex, tiagabine, and topiramate, reported data from the Treatment Outcome PTSD scale (TOP-8) as an additional measure of PTSD symptoms. The studies examining tiagabine and divalproex did not find a difference between those receiving medication and those receiving placebo. The study of tiagabine simply stated that the reduction in TOP-8 score was not significant compared to baseline.140 The study of divalproex reported a TOP-8 mean change from baseline of −4.0 for those receiving divalproex and −3.9 for those receiving placebo (p=NS).139

The only study of topiramate reporting TOP-8 found greater improvement in symptoms for those receiving topiramate than for those receiving placebo (mean percentage change from baseline: −67.9 vs. −41.6, p=0.023).141

Remission (No Longer Having Symptoms)

Two studies reported PTSD remission rates; one study of tiagabine140 and one of topiramate.141 Both of the studies defined remission as a CAPS score less than 20. Neither study found a statistically significant difference between anticonvulsants and placebo. The former (N=232) reported similar remission rates for tiagabine and placebo (16% vs. 14%, p=0.88). The latter (N=40) reported higher remission rates for those treated with topiramate than those who received placebo, although the difference was not statistically significant (42% vs. 21%, p=0.295). Overall, evidence was insufficient to determine the efficacy of either tiagabine or topiramate for inducing remission, largely due to unknown consistency (with just one study contributing data for each medication) and imprecision.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Three studies reported depression symptoms—one assessed divalproex139 and two assessed topiramate.138,141 All three used different outcome measures. None of the studies reported statistically significant reductions in depression for an anticonvulsant compared with placebo, although all point estimates favored anticonvulsants. The trial comparing divalproex with placebo (N=85) reported no significant difference for mean improvement in Montgomery-Asberg Depression Rating Scale (MADRS) scores (−5.1 vs. −4.5).139 One trial of topiramate (N=40) reported no significant difference between topiramate and placebo for HAM-D score (mean percentage change from baseline: −50.7 vs. −33.3, p=0.25);141 the other (N=35) found no significant difference between topiramate and placebo for Beck Depression Inventory (BDI) score (mean change from baseline: −8.5 vs. −3.9, p=0.72).138

Overall, evidence is insufficient to determine the efficacy of any of the anticonvulsants for reducing depression symptoms; just one study reported each outcome measure (MADRS, HAM-D, and BDI), consistency is unknown, and results were imprecise.

Two trials reported on anxiety. Both used the Hamilton Anxiety Scale (HAM-A), and neither found statistically significant reductions in anxiety. The first (N=85) reported similar changes for divalproex and placebo (mean change from baseline: −15.1 vs. −16.5, p = NS).139 The other (N=40) found no statistically significant difference between topiramate and placebo (mean percentage change from baseline: −53.9 and −40.0, with p=0.33).141 Overall, evidence is insufficient to determine the efficacy of any of the anticonvulsants for preventing or reducing anxiety. Consistency is unknown (one trial each for divalproex and topiramate) and findings were imprecise.

Disability or Functional Impairment

Two studies, one of tiagabine (N=232) and one of topiramate (N=40), reported the SDS.140,141 Both trials reported similar changes between subjects treated with medication and those treated with placebo (see Appendix D for details). Overall, evidence is insufficient to determine the efficacy of any of the anticonvulsants for improving disability or functional impairment. Consistency is unknown (one trial each for tiagabine and topiramate) and findings were imprecise.

Detailed Synthesis: Placebo-Controlled Trials of Atypical Antipsychotics

Characteristics of Trials

Table 26 summarizes the characteristics of the seven trials meeting our inclusion criteria. Further details describing the included trials are provided in Appendix D.

Table 26. Characteristics of included placebo-controlled trials of atypical antipsychotics, by drug.

Table 26

Characteristics of included placebo-controlled trials of atypical antipsychotics, by drug.

Most of the trials compared risperidone with placebo. Two compared olanzapine with placebo. All included trials were conducted in the United States. Sample sizes ranged from 15 to 65. Duration of treatment ranged from 5 weeks to 6 months. Most trials enrolled a majority of males with combat-related trauma;146150 one enrolled females with childhood abuse-related trauma;151 and one enrolled males and females with mixed types of trauma, 53 percent of which had rape as the index trauma.145 One trial exclusively enrolled subjects with PTSD and concurrent psychotic features.148 Subjects with a history of schizophrenia, bipolar disorder, or recent substance abuse/dependence were frequently excluded.145,148,150,151 The majority of trials permitted cointerventions. Three trials permitted continuation of antidepressants and anxiolytics as long as doses were stable prior to enrollement;148,150,151 one also allowed the continuation of lithium and carbamazepine in one subject.148 One trial required all subjects to be nonresponsive to 12 weeks of SSRI therapy (4 weeks at optimal doses) and subjects continued on their SSRI during the trial.146 One trial required intolerance, nonresponse, or inadequate response to antidepressants prior to enrollment, but allowed patients to take any medications deemed appropriate during the trial (including adrenergic drugs, antidepressants, anxiolytics, and mood stabilizers).149 Two trials did not address use of cointerventions.145,147 Mean age was similar across most trials, generally ranging from 43 to 54. Mean age in one trial was slightly lower, at 27.151 Two trials enrolled a majority of females, 93 to 100 percent.145,151

The primary outcome for most trials was some version of the CAPS (CAPS total, CAPS-1, CAPS-2).145149,151 The primary outcomes in one trial were reduction in irritability using the Overt Aggression Scale-Modified for Outpatients (OAS-M) and PTSD symptoms using the PTSD Checklist – Military Version (PCL-M).150

We rated two trials152,153 otherwise meeting criteria for this section as rated high risk of bias (Table 27). We excluded them from our main data synthesis and used them only in sensitivity analyses.

Table 27. Characteristics of placebo-controlled trials of atypical antipsychotics excluded from main analyses because of high risk of bias.

Table 27

Characteristics of placebo-controlled trials of atypical antipsychotics excluded from main analyses because of high risk of bias.

Results of Placebo-Controlled Trials of Atypical Antipsychotics

PTSD Symptom Reduction

For olanzapine, one trial146 (N=19) found some benefit for reduction of PTSD symptoms as measured by CAPS compared with placebo (−14.8 vs. −2.67, p<0.05). Another trial145 (N=15) found no statistically significant difference between olanzapine and placebo on three different measures of PTSD symptoms: Davidson Trauma Scale (−34.2 vs. −39.8, p=NR), Treatment Outcome PTSD Scale (−6.7 vs. −11.3, p=NR), and Short Post-Traumatic Stress Disorder Rating Interview (−13.6 vs. −14.3, p=NR).145 Overall, evidence from these two trials (total N=34) was insufficient to determine whether olanzapine is efficacious for improving PTSD symptoms; consistency is unknown (with just one trial reporting each outcome) and findings were imprecise. Our meta-analysis found no statistically significant difference between olanzapine and placebo for PTSD symptom reduction (SMD, −0.14; 95% CI, −1.80 to 1.53; N=34; Appendix F).

For risperidone, four trials assessed PTSD symptom reduction. Our meta-analysis found a statistically significant reduction in PTSD symptoms compared with placebo, as measured by improvement in CAPS (four trials, N=419; WMD, −4.60; 95% CI, −9.01 to −0.20; I2 22.3%; SMD, −0.26; 95% CI, −0.52 to −0.00; Appendix F). Although the finding was statistically significant, the magnitude of difference was small and it is unclear whether it is clinically significant. Some suggest that a reduction of 15 points on the CAPS constitutes a clinically significant reduction.40 However, the value representing a clinically significant reduction has not been validated and is somewhat uncertain. Our sensitivity analysis (adding high risk of bias trials) found a similar, but slightly reduced magnitude of difference, but the difference was no longer statistically significant. (five trials, N=444; WMD, −4.00, 95% CI, −8.48 to 0.49; I2 23.1%; Appendix F).

One trial (N=16) found a statistically significant change in the PCL-M for risperidone compared to placebo (−10.0 vs. −0.50, p=0.02).150

Overall, evidence suggests little or no clinically significant benefit for reducing PTSD symptoms for risperidone (low SOE). Although subjects treated with risperidone had a greater reduction in PTSD symptoms than those who received placebo that was statistically significant, the magnitude of difference was small (−4.60 points on CAPS) and possibly not clinically significant.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

For olanzapine, one trial (N=19) found a greater reduction of depression symptoms than for placebo measured by the change in the Center for Epidemiologic Studies Depression Scale (−5.25 vs. −4.88, p<0.03).146 Overall, evidence from this trial was insufficient to determine whether olanzapine is efficacious for improving depression symptoms.

For risperidone compared with placebo, two trials assessed reduction of coexisting psychiatric conditions.147,148 One assessed depression symptoms and anxiety symptoms;147 the other assessed psychosis.148

The first trial (N=65) did not find a statistically significant reduction in depression symptoms as measured by the HAM-D (−3.7 vs. −1.4, p>0.05). However, it did report greater reduction in anxiety for those treated with risperidone than for those treated with placebo using the HAM-A (−7.4 vs. −2.0, p<0.001).147 The other trial (N=40) examined the effect of risperidone compared with placebo on psychosis. All patients included in the trial had current psychotic features and had a baseline score of ≥60 on the total Positive and Negative Syndrome Scale for psychosis (PANSS). The trial found greater reduction in psychosis for subjects treated with risperidone than for those treated with placebo (−10.0 vs. −2.3, p≤0.05).148

Overall, evidence from two trials (total N=105) was insufficient to determine whether risperidone is efficacious for improving coexisting psychiatric conditions. Individual trials had medium risk of bias, consistency is unknown (one trial contributing data for each outcome), and findings were imprecise.

Disability or Functional Impairment

One trial (N=15) found no difference in disability, as measured by the SDS, between olanzapine and placebo (−7.7 vs. −8.0, p>0.05).145 Evidence is insufficient to determine whether olanzapine is efficacious for reducing disability in patients with PTSD.

Detailed Synthesis: Placebo-Controlled Trials of Benzodiazepines

Characteristics of Trials

We found no studies with low or medium risk of bias meeting our inclusion criteria. We identified one trial otherwise meeting criteria for this section that we rated as high risk of bias (Table 28). Thus, we did not include it in our main our data synthesis.154 The identified study was a 12-week randomized, double-blind crossover trial of alprazolam and placebo (N=16). The study was conducted in Israel and enrolled a heterogeneous group of subjects with a variety of index trauma types (e.g., military combat stress, industrial accident, automobile accident, terrorist bomb in bus). Appendix E provides additional rationale for risk of bias assessments. Briefly, the trial had high attrition, high risk of measurement bias, and high risk of selection bias. In addition, it did not report information needed to determine comparability of treatment groups.

Table 28. Characteristics of placebo-controlled trials of benzodiazepines excluded from main analyses because of high risk of bias.

Table 28

Characteristics of placebo-controlled trials of benzodiazepines excluded from main analyses because of high risk of bias.

PTSD Symptom Reduction, Remission, and Other Outcomes

With no low or medium risk of bias studies identified, evidence is insufficient to determine the efficacy of benzodiazepines for improving outcomes for adults with PTSD.

Detailed Synthesis: Selective Serotonin Reuptake Inhibitors (SSRIs)

Characteristics of Trials

Table 29 summarizes the characteristics of the 16 trials meeting our inclusion criteria. Further details describing the trials are provided in Appendix D.

Table 29. Characteristics of included placebo-controlled trials of selective serotonin reuptake inhibitors, by drug.

Table 29

Characteristics of included placebo-controlled trials of selective serotonin reuptake inhibitors, by drug.

The vast majority were conducted in the United States; one in Israel;170 one in Iran;169 one in the United States and Canada;134 and one in Europe, Israel, and South Africa.159 Sample sizes ranged from 12 to 563. Duration of treatment ranged from 5 to 12 weeks. The majority of trials enrolled a heterogeneous group of subjects with a variety of index trauma types (e.g., sexual abuse, nonsexual abuse, combat, injury, motor vehicle accident, natural disaster); five trials enrolled a majority of subjects with combat-related PTSD;159,168171 one enrolled 60 percent exposed to war;164 and one enrolled just under half of subjects with combat-related PTSD.162 One study enrolled subjects with PTSD and coexisting alcohol dependence.166 Mean age was very similar across trials, ranging from 36 to 46. Nine trials enrolled two-thirds or more female subjects.113,134,155,157,161,163,165,167,172

The primary outcome for the majority of trials was some version of the CAPS (CAPS, CAPS-2, or CAPS-Sx); five trials identified other primary outcomes, including TOP-8,159,161 DTS,171 Duke Global Rating for PTSD,157 IES,169 or SPRINT.164

We rated two trials171,172 otherwise meeting criteria for this section as high risk of bias (rationale in Appendix E) (Table 30). We excluded them from our main data synthesis and used them only for sensitivity analyses.

Table 30. Characteristics of placebo-controlled trials of SSRIs excluded from main analyses because of high risk of bias.

Table 30

Characteristics of placebo-controlled trials of SSRIs excluded from main analyses because of high risk of bias.

Results of Placebo-Controlled Trials of SSRIs

PTSD Symptom Reduction

Our meta-analyses found fluoxetine, paroxetine, and sertraline to improve PTSD symptoms, as measured by improvement in CAPS, more than placebo (Figure 23). We found no statistically significant difference between citalopram and placebo, and results favored placebo (one trial). Magnitude of benefit ranged from a difference of −4.9 for sertraline to −12.6 for paroxetine compared with placebo. The meta-analyses for change in CAPS did not have significant statistical heterogeneity for any of the individual medications compared with placebo (I2=0% for each). Analyses removing each individual study one at a time did not result in any significant differences in findings (Appendix F). Sensitivity analyses adding the trials rated high risk of bias with available data did not significantly change the results (Appendix F).

Figure 23 is titled “Mean change from baseline in CAPS for SSRIs compared with placebo.” The figure displays a forest plot reporting the standardized mean difference in CAPS stratified by citalopram compared with placebo, fluoxetine compared with placebo, paroxetine compared with placebo, and setraline compared with placebo. Data for Figure 23 are presented in Appendix F. This figure is described further in the “PTSD symptom reduction” section as follows: “Our meta-analyses found fluoxetine, paroxetine, and sertraline to improve PTSD symptoms, as measured by improvement in CAPS, more than placebo. We found no statistically significant difference between citalopram and placebo, and results trended in favor of placebo (1 trial). Magnitude of benefit ranged from a difference of −4.9 for sertraline to −12.6 for paroxetine compared with placebo. The meta-analyses for change in CAPS did not have significant statistical heterogeneity for any of the individual medications compared with placebo (I2=0% for each).”

Figure 23

Mean change from baseline in CAPS for SSRIs compared with placebo. Note: Timing of outcome assessment: 10 weeks (Tucker, 2003; Zohar, 2002), 12 weeks (Martenyi, 2007; Martenyi, 2002; Marshall, 2001; Tucker, 2001; Brady, 2005; Brady, 2000; Davidson, 2001; (more...)

The second most frequently reported measure of PTSD symptoms was the Davidson Trauma Scale (DTS). Our meta-analyses of mean change from baseline in DTS found similar results as those for CAPS, with fluoxetine, paroxetine, and sertraline improving PTSD symptoms more than placebo (Figure 24). Magnitude of benefit ranged from −7.7 (95% CI, −12.9 to −2.4, four trials, N=916) for sertraline to −12.2 (95% CI, −15.8 to −8.7, two trials, N=886) for paroxetine compared with placebo. Sensitivity analyses adding the trials rated high risk of bias with available data did not significantly change the results (Appendix F).

Figure 24 is titled “Mean change from baseline in DTS for SSRIs compared with placebo.” The figure displays a forest plot reporting the weighted mean difference in DTS scores stratified by fluoxetine compared with placebo, paroxetine compared with placebo, and setraline compared with placebo. Data for Figure 24 are presented in Appendix F. This figure is described further in the “PTSD symptom reduction” section as follows: Our meta-analyses of mean change from baseline in DTS found similar results as those for CAPS, with fluoxetine, paroxetine, and sertraline improving PTSD symptoms more than placebo. Magnitude of benefit ranged from −7.7 (95% CI −12.9 to −2.4, 4 trials, N=916) for sertraline to −12.2 (95% CI −15.8 to −8.7, 2 trials, N=886) for paroxetine compared with placebo.”

Figure 24

Mean change from baseline in DTS for SSRIs compared with placebo. Note: Timing of outcome assessment: 12 weeks for all included studies.

Fewer studies reported other measures of PTSD symptoms, such as the IES, TOP-8, SI-PTSD, or Duke Global rating. Overall, findings on these measures were consistent with those of the CAPS and DTS. Five trials that compared citalopram with placebo (one trial) and/or sertraline with placebo measured improvement in PTSD symptoms with the Impact of Event Scale (IES).155,165,167169 One trial found no statistically significant difference between citalopram and placebo (WMD, 7.8; 95% CI, −4.8 to 20.5, N=35). Our meta-analysis found greater improvement in symptoms measured with the IES for subjects treated with sertraline than for those who received placebo (WMD, −3.96; 95% CI, −6.0 to −1.9; I2=0%, five trials, N=667; Appendix F). Using the TOP-8 as a measure of PTSD symptoms, one trial159,160 found greater improvement with fluoxetine compared with placebo (WMD, −2.3; 95% CI, −3.5 to −1.1, N=301); our meta-analysis of trials comparing paroxetine with placebo found greater improvement with paroxetine (WMD, −3.3; 95% CI, −4.2 to −2.4; I2=0%, two trials134,163 contributing three comparisons, N=886, Appendix F). A single included trial reported each of the following outcomes as a measure of PTSD symptoms: SI-PTSD (aka SIP),157 Duke Global rating,157 SPRINT,164 and the Mississippi Scale for Combat-related PTSD168—details for these outcomes are available in Appendix D.

Our meta-analyses using any measure of PTSD symptom reduction found effect sizes (i.e., Cohen’s d; SMD) of −0.31 (95% CI, −0.44 to −0.17), −0.49 (95% CI, −0.61 to −0.37), and −0.25 (95% CI, −0.42 to −0.07) for fluoxetine, paroxetine, and sertraline, respectively (Appendix F).

Overall, we determined that the evidence from trials is consistent, direct, and precise and supports the efficacy of fluoxetine, paroxetine, and sertraline for improving PTSD symptoms (moderate SOE). The magnitude of benefit is in the small to medium range. Evidence was insufficient to determine the efficacy of citalopram for improving PTSD symptoms.

Remission (No Longer Having Symptoms)

Four trials reported remission using varying definitions of remission; one of fluoxetine,113 two of paroxetine,134,164 and one of sertraline.133 Three of the trials defined remission as a score of less than 20 on some version of the CAPS. For all four trials, point estimates favored SSRIs. All four trials reported that greater proportions of subjects treated with SSRIs achieved remission than subjects who received placebo; differences between groups ranged from 3 percent to 19 percent, but often did not reach statistical significance. Some of the trials were underpowered to detect anything but a very large difference for remission.

One trial of fluoxetine randomized 88 subjects to eye movement desensitization and reprocessing (EMDR), fluoxetine, or placebo and defined remission (i.e., percent asymptomatic) as CAPS total score less than 20. It reported that a greater percentage of subjects treated with fluoxetine achieved remission than subjects who received placebo, but findings did not achieve statistical significance (13% vs. 10%, p=0.72, 58 subjects total in the fluoxetine and placebo groups). The evidence was insufficient to determine the efficacy of fluoxetine for achieving remission because of unknown consistency and imprecision.

For paroxetine, two trials reported remission. Both trials reported a similar between-group difference in the percentage of subjects achieving remission. One enrolled subjects refractory to prolonged exposure therapy (N=23). It defined remission as a SPRINT score less than 6.164 It found 33 percent (3 out of 9) of subjects in the paroxetine group and 14 percent (2 out of 14) of the placebo group achieved remission (p=0.34). The difference did not reach statistical significance; the trial was underpowered to detect anything but a large difference for this outcome. The other trial (N=323)134 defined remission as a CAPS-2 total score less than 20 and found that a significantly greater proportion of paroxetine-treated subjects achieved remission than placebo subjects at week 12 (29.4% vs. 16.5%, p=0.008). The difference (12.9% difference between paroxetine and placebo) would translate to a NNT of 8. With consistent, direct, precise information from two trials, we determined that evidence supports the efficacy of paroxetine for achieving remission (moderate SOE).

The fourth trial (N=538) reporting remission randomized subjects to sertraline, venlafaxine, or placebo.133 It defined remission as CAPS-SX17 score less than 20. At week 12, remission rates were numerically greater for patients treated with sertraline than for patients who received placebo, but the difference did not reach statistical significance (24.3% vs. 19.6%, p=NS, 352 subjects total in the sertraline and placebo arms). The evidence was insufficient to determine the efficacy of fluoxetine for achieving remission because of unknown consistency and imprecision.

Loss of PTSD Diagnosis

A single trial comparing EMDR (N=29), fluoxetine (N=30), and placebo (N=29) found no statistically significant difference between the three groups for the percentage of subjects achieving loss of diagnosis after 8 weeks of treatment (76% vs. 73% vs. 59%, respectively, p=0.23 for fluoxetine compared with placebo).113 The evidence was insufficient to determine the efficacy of fluoxetine for achieving loss of diagnosis because of imprecision and unknown consistency.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Many of the included trials reported reduction of depression symptoms using various outcome measures: just two measures, HAM-D and MADRS, were reported with sufficient data to conduct meta-analyses—six trials reported the HAM-D133,162,165168 and five reported MADRS.134,160,161,163,170 Most of the trials assessed fluoxetine, paroxetine, or sertraline. Evidence from one trial (N=35)155 was insufficient to determine efficacy of citalopram for reducing comorbid depression in adults with PTSD.

Our meta-analysis of trials reporting MADRS found greater improvement in depression symptoms for subjects treated with fluoxetine than for those who received placebo (MADRS, mean change from baseline: WMD −2.4, 95% CI, −3.7 to −1.1; I2=0%, two trials, N=712, Appendix F) and for subjects treated with paroxetine than for those who received placebo (MADRS, mean change from baseline: WMD, −5.7; 95% CI, −7.1 to −4.3; I2=0%, two trials, N=886, Appendix F). Overall, consistent, direct, precise evidence provided moderate SOE that both fluoxetine and paroxetine improve depression symptoms for adults with PTSD.

For sertraline, data were available for meta-analysis of five trials reporting the HAM-D. Our meta-analysis found no statistically significant difference between sertraline and placebo (HAM-D, mean change from baseline: WMD, −0.77; 95% CI, −2.1 to 0.55; I2=25%, five trials, N=1,010, Appendix F). Point estimates favored sertraline for three of the individual trials; they favored placebo for the other two. One trial reporting MADRS found greater improvement in depression symptoms for subjects treated with sertraline than for those who received placebo (WMD, −3.2; 95% CI, −5.2 to −1.2; N=42).170 Taken together, primarily because of lack of consistency and imprecision, the evidence for sertraline does not support its efficacy for improving depression symptoms for adults with PTSD (low SOE).

Our meta-analyses using any measure for depression symptom reduction found effect sizes (i.e., Cohen’s d; SMD) of −0.04 (95% CI, −0.77 to 0.70), −0.20 (95% CI, −0.40 to −0.00), −0.49 (95% CI, −0.64 to −0.34), and −0.13 (95% CI, −0.32 to 0.06) for citalopram, fluoxetine, paroxetine, and sertraline, respectively (Appendix F).

Four of the included trials assessed anxiety symptoms using the HAM-A.159,161,167,168 Two trials compared fluoxetine with placebo159,161 and two compared sertraline with placebo.167,168 Our meta-analysis found greater improvement in anxiety symptoms for subjects treated with fluoxetine than for those who received placebo (WMD, −2.1; 95% CI, −3.2 to −0.9; I2=0%, two trials, N=712, Appendix F). Evidence for fluoxetine was consistent, direct, and precise (moderate SOE). The two trials that compared sertraline with placebo reported mixed results; one favored sertraline and one favored placebo. Meta-analysis of the two trials had substantial heterogeneity (WMD, 0.19; 95% CI, −3.14 to 3.51; I2=68.3%, 2 trials, N=377). Overall, evidence from these two trials was inconsistent and imprecise; thus, evidence is insufficient to determine the efficacy of sertraline for reducing anxiety symptoms in subjects with PTSD.

Quality of Life

Three included trials assessed quality of life; two trials of sertraline used the Q-LES-Q133,165 and one of fluoxetine used the SF-36.160 Results of our meta-analysis found no statistically significant difference between sertraline and placebo (mean change in Q-LES-Q: WMD, 4.9; 95% CI, −0.88 to 10.7, two trials, N=539). The analysis found substantial statistical heterogeneity (I2=72.6%). This could be explained by differences in the enrolled populations—one trial enrolled male and females with comorbid alcohol dependence. Overall, because of inconsistency and imprecision, evidence was insufficient to determine the efficacy of sertraline for improving quality of life.

The SF-36 was reported as an outcome in a subgroup analysis of subjects with combat-related PTSD in one trial (N=144 of the 301 from the main trial).160 It reported greater improvement in the mental health subscore of the SF-36 for those treated with fluoxetine than for those who received placebo (15.5 vs. 0.33, p<0.001) and no difference between groups for the physical functioning subscore (8.62 vs. 8.07, p=0.891). This evidence was insufficient to determine the efficacy of fluoxetine for improving quality of life.

Disability or Functional Impairment

Four trials assessed disability using the Sheehan Disability Scale (SDS); one trial of fluoxetine,157 two trials of paroxetine,134,163 and one of sertraline.133 Evidence from one trial each for fluoxetine and sertraline provided insufficient evidence to determine their efficacy for reducing disability (detailed results provided in Appendix D and Appendix F). For paroxetine, consistent, direct, precise findings support its efficacy for improving disability or functioning (mean change from baseline in SDS: WMD, −2.3; 95% CI, −3.3 to −1.4; two trials, N=886, Appendix F, moderate SOE).

Detailed Synthesis: Serotonin and Norepinephrine Reuptake Inhibitors (SNRIs)

Characteristics of Trials

Table 31 summarizes the characteristics of the two trials meeting our inclusion criteria. Further details describing the included trials are provided in Appendix D. Both trials evaluated venlafaxine extended release. One was a 12-week trial conducted in the United States133 and one was a 24-week multinational collaboration of 56 outpatient psychiatric clinic sites in South America, Europe, Mexico, and South Africa.173 The U.S.-based trial had an active comparator arm (sertraline) as well as a placebo comparison. Both trials enrolled a heterogeneous group of subjects with a variety of index trauma types (e.g., sexual abuse, nonsexual abuse, combat, injury, motor vehicle accident, natural disaster), and very few subjects with combat-related PTSD (9% to 12%). The primary outcome for both trials was the CAPS.

Table 31. Characteristics of included placebo-controlled trials of SNRIs.

Table 31

Characteristics of included placebo-controlled trials of SNRIs.

Results of Serotonin and Norepinephrine Reuptake Inhibitors (SNRIs)

PTSD Symptom Reduction

Both trials reported similar improvement in CAPS. Our meta-analysis found venlafaxine to improve PTSD symptoms more than placebo (WMD −7.15, 95% CI, −11.02 to −3.28, Figure 25; SMD −0.28, 95% CI, −0.43 to −0.13, Appendix F).

Figure 25 is titled “Mean change from baseline in CAPS for venlafaxine ER compared with placebo.” The figure displays a forest plot reporting weight mean difference in CAPS, venlafaxine ER compared with placebo. Data for Figure 25 are presented in Appendix F. This figure is described further in the “PTSD Symptom Reduction” section as follows: “Our meta-analysis found venlafaxine to improve PTSD symptoms more than placebo (weighted mean difference −7.15, 95% CI −11.02 to −3.28).” There was no considerable statistical heterogeneity (I2=0.0%).

Figure 25

Mean change from baseline in CAPS for venlafaxine ER compared with placebo. Note: Timing of outcome assessment: 24 weeks (Davidson, 2006), 12 weeks (Davidson, 2006).

The U.S.-based trial also reported the Davidson Trauma Scale (DTS) total score, finding greater improvement in symptoms with venlafaxine than with placebo (−42.9 vs. −34.6, p=0.015).

Remission (No Longer Having Symptoms)

Both trials reported remission, defined as CAPS-Sx total score of 20 or less, and both found that more subjects receiving venlafaxine achieved remission than those receiving placebo. Our meta-analysis for remission at 12 weeks found that 12 percent more subjects receiving venlafaxine achieved remission than those receiving placebo (Figure 26). This would translate to a NNT of 9.

Figure 26 is titled “Percentage of subjects achieving remission for venlafaxine ER compared with placebo.” The figure displays a forest plot reporting risk difference in achieving remission, venlafaxine ER compared with placebo. Data for Figure 26 are presented in Appendix F. This figure depicts a risk difference of 0.12 (95% CI 0.05 to 0.19), favoring venlafaxine ER. There was no considerable statistical heterogeneity (I2=0.0%).

Figure 26

Percentage of subjects achieving remission for venlafaxine ER compared with placebo. Note: Timing of outcome assessment: 24 weeks (Davidson, 2006), 12 weeks (Davidson, 2006).

One of the trials173 also reported that about 13 percent more subjects in the venlafaxine group than the placebo group achieved remission at 24 weeks (50.9% vs. 37.5%).

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Both trials assessed depression using the HAM-D, and both reported a statistically significant reduction with venlafaxine compared with placebo. Our meta-analysis found about a 2-point greater reduction in HAM-D with venlafaxine than with placebo (WMD, −2.08; 95% CI, −3.12 to −1.04; I2=0%, Appendix F).

Quality of Life

Both trials reported a statistically significant improvement in quality of life with venlafaxine compared with placebo, as measured by the Quality of Life Enjoyment and Satisfaction Questionnaire-Short Form (Q-LES-Q-SF). Our meta-analysis found a 3.4-point greater improvement in Q-LES-Q-SF with venlafaxine than with placebo (WMD, 3.42; 95% CI, 1.58 to 5.26; I2 = 0%, Appendix F).

Disability or Functional Impairment

Both trials reported a statistically significant improvement in functional impairment with venlafaxine than with placebo, as measured by the Sheehan Disability Scale (SDS). Our meta-analysis found about a 2-point greater improvement in SDS with venlafaxine than with placebo (WMD, −2.06; 95% CI, −3.28 to −0.84; I2 = 0%, Appendix F). Similarly, our meta-analyses found greater improvement in Global Assessment of Functioning (GAF) with venlafaxine than with placebo (WMD, 3.41; 95% CI, 1.41 to 5.40, I2 = 0%, Appendix F).

Detailed Synthesis: Placebo-Controlled Trials of Tricyclic Antidepressants

We did not find any trials comparing tricyclic antidepressants with placebo or other medications that had low or medium risk of bias. We rated three trials otherwise meeting criteria for this section as high risk of bias (Table 32). Appendix E provides additional rationale for risk of bias assessments. Briefly, the trials only analyzed subjects who completed treatment (did not use an intention-to-treat analysis) and/or had very high dropout rates.

Table 32. Characteristics of placebo-controlled trials of tricyclic antidepressants excluded because of high risk of bias.

Table 32

Characteristics of placebo-controlled trials of tricyclic antidepressants excluded because of high risk of bias.

Overall evidence was insufficient to make conclusions about the efficacy of any tricyclic antidepressants for treating PTSD in adults (insufficient SOE).

Detailed Synthesis: Placebo-Controlled Trials of Other Second-Generation Antidepressants

Characteristics of Trials

Table 33 summarizes the characteristics of the two trials that met our inclusion criteria. Further details describing the included trials are provided in Appendix D.

Table 33. Characteristics of included placebo-controlled trials of other second-generation antidepressants.

Table 33

Characteristics of included placebo-controlled trials of other second-generation antidepressants.

Of the two included placebo-controlled trials, one assessed bupropion (N=30)178 and one assessed mirtazapine (N=29).179 Both were conducted in the United States. Treatment duration was 8 weeks. Both enrolled a heterogeneous group of subjects with a variety of index trauma types (e.g., military combat or war trauma, childhood sexual abuse, physical abuse, rape, motor vehicle accident, witnessing a trauma, death or suicide of a loved one). The trials generally recruited middle-aged adults, with mean ages ranging from ~43 to ~50 years. The trial of bupropion recruited subjects from a VA Medical Center and from the community.178 The trial of mirtazapine recruited subjects by advertisements or from the clinical practice of the investigators.179

We rated one trial comparing nefazodone with placebo29 otherwise meeting criteria for this section as high risk of bias (Table 34). We excluded it from our main data synthesis and used it only for sensitivity analyses.

Table 34. Characteristics of placebo-controlled trials of other second-generation antidepressants excluded from main analyses because of high risk of bias.

Table 34

Characteristics of placebo-controlled trials of other second-generation antidepressants excluded from main analyses because of high risk of bias.

Results of Placebo-Controlled Trials of Other Second-Generation Antidepressants

PTSD Symptom Reduction

Both included trials reported measures of PTSD symptoms.178,179 For all but one of the measures of PTSD symptoms, neither trial found statistically significant differences between medication and placebo groups.

The trial comparing bupropion with placebo found no statistically significant difference between groups for improvement in PTSD symptoms, assessed by mean reduction in CAPS (−12.33; SD, 24.12 vs. −16.99; SD, 11.26) or DTS (−13.22; SD, 21.62 vs. −10.6; SD, 29.20).178 Both groups improved.

The trial comparing mirtazapine with placebo reported three measures of PTSD symptoms: SPRINT, SIP, and DTS.179 For SPRINT and DTS, results were not statistically significantly different between groups (−9.3 vs. −5.6, p=0.20 and −20.7 vs. −11.2, p=0.20, respectively). However, the trial reported statistically significant differences between groups for the SIP, favoring those treated with mirtazapine (−17.3 vs. −6.5, p=0.04).

Overall, we found insufficient evidence to determine the efficacy of either bupropion or mirtazapine for improving PTSD symptoms. Evidence was limited to one trial for each medication, consistency was thus unknown, and findings were imprecise.

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Both trials reported measures of reduction of depression and/or anxiety.178,179 Neither reported prevention or reduction of comorbid medical conditions. The trial of bupropion reported similar decreases in depression scores for the intervention group and the control group (mean change in BDI: −3.22 vs. −3.61, p=NS). The trial of mirtazapine found a greater reduction in depressive symptoms using the Hospital Depression Scale (HADS-D), but the difference between mirtazapine-treated subjects and those receiving placebo was not statistically significant (−2.2 vs. −0.5, p=0.08).

The trial of mirtazapine reported greater reduction in anxiety for subjects treated with mirtazapine than for those receiving placebo, using the Hospital Anxiety Scale (HADS-A) (−2.8 vs. −1.2, p<0.05).179

Overall, we found insufficient evidence to determine the efficacy of either bupropion or mirtazapine for prevention or reduction of comorbid medical or psychiatric conditions for adults with PTSD. Evidence was limited to one small medium-risk-of-bias trial for each medication, consistency was thus unknown, and findings were imprecise.

Detailed Synthesis: Head-to-Head Pharmacotherapy Trials

Characteristics of Trials

Table 35 summarizes the three trials that met inclusion criteria. Further details are provided in Appendix D. The trials enrolled subjects with severe to extreme PTSD symptomatology. All were conducted in the United States. Sample sizes ranged from 59 to 538. Treatment duration ranged from 10 to 12 weeks. One of the included trials enrolled veterans with comorbid alcohol dependence;132 the other two enrolled a heterogeneous group of subjects with a variety of index trauma types. The trial enrolling veterans randomized subjects to paroxetine plus naltrexone, paroxetine plus placebo, desiprimine plus naltrexone, or desiprimine plus placebo.132 The primary outcome for all five trials was some version of the CAPS.

Table 35. Characteristics of included head-to-head pharmacotherapy trials.

Table 35

Characteristics of included head-to-head pharmacotherapy trials.

We rated three trials otherwise meeting criteria for this section as high risk of bias (Table 36). We excluded them from our main data synthesis. Appendix E provides additional rationale for risk of bias assessments. Briefly, the trials deemed high risk of bias only analyzed subjects who completed treatment (did not use an intention-to-treat analysis) and/or had very high overall and differential attrition rates.

Table 36. Characteristics of head-to-head pharmacotherapy trials excluded because of high risk of bias.

Table 36

Characteristics of head-to-head pharmacotherapy trials excluded because of high risk of bias.

Results of Head-to-Head Pharmacotherapy Trials

PTSD Symptom Reduction

All three included trials assessed PTSD symptom reduction. Outcome measures included versions of the CAPS,132,133,155 the DTS,133 and the IES.155 The four-arm trial enrolling veterans with PTSD and comorbid alcohol dependence (N=88) reported similar improvements in PTSD symptoms for all treatment groups (CAPS, mean change from baseline: −33.5 vs. −33.2 vs. −35.7 vs. −36.4, p NS). With evidence from one trial, unknown consistency, and imprecise findings, we concluded that evidence of low strength indicates no difference between desipramine and paroxetine for reducing PTSD symptoms in adults with PTSD and coexisting alcohol dependence.

The trial comparing venlafaxine ER, sertraline, and placebo (N=538) reported similar improvements in PTSD symptoms for both active treatment arms using the CAPS-SX17 (mean change from baseline: −41.5 vs. −39.4, p=0.49) and the DTS (−42.9 vs. −38.9, p=0.25).133 Results favored venlafaxine ER and differences between venlafaxine ER and placebo reached statistical significance (CAPS-SX17 −41.5 vs. −34.2, p=0.015; DTS −42.9 vs. −34.6, p=0.015), whereas those between sertraline and placebo did not (CAPS-SX17 −39.4 vs. −34.2, p=0.081; DTS −38.9 vs. −34.6, p=0.203). Although evidence is from a single trial, and consistency is unknown, direct and precise findings suggest no significant difference between venlafaxine ER and sertraline for reducing PTSD symptoms (moderate SOE).

The trial comparing sertraline, citalopram, and placebo (N=58) found greater improvement in CAPS for those treated with sertraline than for citalopram or placebo, but differences did not reach statistical significance (−41.8 vs. −30.7 vs. −38.7, p=NS). It also reported no statistically significant differences between groups for change in IES, although the greatest numerical reduction was seen in the placebo group (−29.1 vs. −19.3 vs. −33.2, p=NS).155 Evidence was from a single trial, consistency is unknown, and findings were imprecise (insufficient SOE).

Remission (No Longer Having Symptoms)

Just one of the included head-to-head trials reported remission—the trial comparing venlafaxine ER, sertraline, and placebo (N=538).133 It defined remission as a CAPS-SX17 score of ≤20. By week 12, 30.2 percent of subjects treated with venlafaxine, 24.3 percent of subjects treated with sertraline, and 19.6 percent of those receiving placebo achieved remission (p<0.05 for venlafaxine ER vs. placebo; p=NS for sertraline vs. placebo or venlafaxine vs. sertraline).

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

All three trials assessed reduction of depression symptoms. Outcome measures included the HAM-D132,133 and the BDI.155 The four-arm trial enrolling veterans with PTSD and coexisting alcohol dependence (N=88) reported similar improvements in PTSD symptoms for all treatment groups (HAM-D, mean change from baseline: −3.9 vs. −2.7 vs. −2.6 vs. −4.2, p=NS). With evidence from one trial, unknown consistency, and imprecise findings, we concluded that evidence of low strength indicates no difference between desipramine and paroxetine for reduction of depression symptoms in adults with PTSD and coexisting alcohol dependence.

The trial comparing venlafaxine ER, sertraline, and placebo (N=538)133 reported similar findings for reduction of depression as it reported for reduction of PTSD symptoms, with results favoring the venlafaxine ER group, but reaching statistically significant differences only for the comparison between venlafaxine and placebo (mean change from baseline in HAM-D: −7.09 vs. −6.42 vs. −5.54, p=0.38 for venlafaxine vs. sertraline, p=0.04 for venlafaxine vs. placebo, p=0.24 for sertraline vs. placebo).

The trial comparing sertraline, citalopram, and placebo (N=58) reported similar reduction in depression symptoms for all groups (BDI, mean change from baseline: −13.4 vs. −16.1 vs. −15.6, p=NR).155 It also reported change in systolic and diastolic blood pressure, but it was not clear if any enrolled subjects had hypertension; the reported information is not useful to inform the question of whether treatments reduced coexisting hypertension. Evidence was from a single trial, consistency is unknown, and findings were imprecise.

The four-arm trial enrolling veterans with PTSD and coexisting alcohol dependence (N=88) also reported alcohol use outcomes, finding greater reduction in the percentage of heavy drinking days (p=0.009) and drinks per drinking days (p=0.027) for subjects receiving desipramine than for those receiving paroxetine.132 The data were not reported for drinking outcomes (shown in figure only for drinks per week—all groups ended up less than 20 standard drinks per week, from baselines above 70 drinks per week, and it appears that the desipramine groups ended up in the 0 to 10 drinks per week range and the paroxetine groups ended up in the 10–20 range at the 12-week endpoint). Overall, the trial provides some evidence of a slightly greater benefit for drinking outcomes for those treated with desipramine than for those treated with paroxetine. We concluded that evidence was of low strength; consistency is unknown (single study), and findings were imprecise.

Quality of Life

One trial assessed quality of life using the Q-LES-Q.133 The trial comparing venlafaxine ER, sertraline, and placebo (N=538) reported similar findings for improvement in quality of life as it reported for other outcomes, with results favoring the venlafaxine ER group, but reaching statistically significant differences only for the comparison between venlafaxine and placebo (Appendix D).

Disability or Functional Impairment

One trial assessed disability, using the Sheehan Disability Scale (SDS) and the Global Assessment of Functioning (GAF).133 The trial comparing venlafaxine ER, sertraline, and placebo (N=538) reported similar findings for improvement in disability as it reported for other outcomes, with results from the SDS showing benefit for the venlafaxine ER group, but reaching statistically significant differences only for the comparison between venlafaxine and placebo (Appendix D). For the GAF none of the between-group differences were statistically significant. Although evidence is from a single trial, and consistency is unknown, direct and precise findings suggest no significant difference between venlafaxine ER and sertraline for reducing PTSD symptoms (moderate SOE).

Network Meta-Analysis of Pharmacotherapy Trials

We conducted network meta-analyses using Bayesian methods for the PTSD symptoms outcome, measured by mean change from baseline in CAPS compared with placebo. The analysis included 28 trials and 13 active treatments (4,817 subjects) incorporating both direct and indirect evidence from the trials included in the previous sections of this Key Question (KQ). A network diagram illustrates the number of subjects contributing to each comparison; thickness of lines connecting each drug-drug or drug-placebo is directly proportional to the number of trials with available data for that comparison (Figure 27).

Figure 27 is titled “Evidence network: comparisons, and number of subjects for each, included in network meta-analysis.” The figure is a network diagram showing the number of subjects in the included 28 trials and 13 active treatments (4,817 subjects). The figure is described further in the section entitled “Network Meta-Analysis of Pharmacotherapy Trials” as follows: “A network diagram illustrates the number of subjects contributing to each comparison; thickness of lines connecting each drug-drug or drug-placebo is directly proportional to the number of trials with available data for that comparison.” The diagram has “placebo” in the center of the diagram with several lines or arms (12) extended from the center. In a clockwise direction, the first line represents trials comparing venlafaxine and placebo, 687 subjects included. The second line represents trials comparing fluoxetine and placebo, 835 subjects included. The third line represents trials comparing risperidone and placebo, 419 subjects included. The fourth line represents trials comparing topiramate and placebo, 142 subjects included. The fifth line represents trials comparing olanzapine and placebo, 19 subjects included. The sixth line represents trials comparing tiagabine and placebo, 232 subjects included. The seventh line represents trials comparing divalproex and placebo, 85 subjects included. The eighth line represents trials comparing paroxetine and placebo, 886 subjects included; an additional line to the left of this line represents a trial comparing paroxetine to desipramine, 44 subjects included. The ninth line represents trials comparing buproprion and placebo, 28 subjects included. The tenth line represents trials comparing prazosin and placebo, 44 subjects included. The eleventh line represents trials comparing citalopram and placebo, 35 subjects included; an additional line to the left of this line connects with sertraline and represents trials comparing citalopram and sertraline, 48 subjects included. The twelfth and final line represents trials comparing sertraline and placebo, 1085 subjects included; an additional line to the left of this line connects with venlafaxine and represents trials comparing sertraline and venlafaxine, 352 subjects included.

Figure 27

Evidence network: comparisons, and number of subjects for each, included in network meta-analysis.

Findings from our network meta-analysis are presented in Figure 28, showing the difference between each pair of treatments for change from baseline in total CAPS score (WMD and 95% credible interval [CrI] for each comparison). Our network meta-analysis found paroxetine and topiramate to be more effective for reducing PTSD symptoms than most other medications included in the analysis. When compared with other medications with moderate SOE of efficacy for improving PTSD symptoms (based on our findings in previous sections of this KQ), paroxetine was more effective than sertraline (WMD −7.6, 95% CrI −12 to −2.8), but was not significantly different from fluoxetine, topiramate, or venlafaxine. When compared with other medications with moderate SOE of efficacy, topiramate was more effective than fluoxetine (WMD 8.6, 95% CrI 2.4 to 14.9), sertraline (WMD 11, 95% CrI 5.7 to 16.6), and venlafaxine (WMD −8.8, 95% CrI −15 to −2.5), but was not significantly different from paroxetine. Results of our sensitivity analysis adding in studies rated as high risk of bias were similar to those for the main analysis (Appendix F).

Figure 28 is titled “Results of network meta-analysis comparing improvement in PTSD symptoms (change in CAPS total score).” The figure displays 11 forest plots of studies that completed head to head comparisons of PTSD medications to determine each medications effectiveness in improving PTSD symptoms based on CAPS. Negative change scores favor Drug A and positive change scores favor Drug B. The first forest plot compares buproprion (Drug A) versus other drugs (Drug B); buproprion versus citalopram (−2.10, 95% CI, −21.00 to 16.80); buproprion versus desipramine (13.30, 95% CI, −6.70 to 33.90); buproprion versus divalproex (3.38, 95% CI, −13.0 to 20.60); buproprion versus fluoxetine (11.90, 95% CI, −2.00 to 26.30); buproprioun versus olanzapine (17.00, 95% CI, −0.71 to 34.80); buproprion versus paroxetine (17.00, 95% CI, 2.81 to 31.40); buproprion versus prazosin (13.30, 95% CI, −6.10 to 32.40); buproprion versus risperidone (8.91, 95% CI, −5.10 to 23.10); buproprion versus setraline (9.43, 95% CI, −4.60 to 23.40); buproprion versus tiagabine (5.18, 95% CI, −10.00 to 20.60); buproprion versus topiramate (20.40, 95% CI, 6.07 to 35.00); buproprion versus venlafaxine (11.70, 95% CI, −2.80 to 26.00). The second forest plot compares citalopram (Drug A) to other drugs (Drug B); citalopram versus desipramine (15.40, 95% CI, −3.90 to 35.10); citalopram versus divalproex (5.48, 95% CI, −11.00 to 22.30); citalopram versus fluoxetine (14.00, 95% CI, 0.17 to 28.10); citalopram versus olanzapine (19.10, 95% CI, 1.48 to 35.70); citalopram versus paroxetine (19.10, 95% CI, 1.48 to 32.60); citalopram versus prazosin (15.30, 95% CI, −2.80 to 34.70); citalopram versus risperidone (11.00, 95% CI, −2.30 to 24.80); citalopram versus setraline (11.50, 95% CI, −1.20 to 24.70); citalopram versus tiagabine (7.28, 95% CI, −7.60 to 22.30); citalopram versus topiramate (22.50, 95% CI, 8.66 to 36.40); citalopram versus venlafaxine (13.80, 95% CI, 0.47 to 27.50). The third forest plot compares desipramine (Drug A) to other drugs (Drug B); desipramine versus divalproex (−9.90, 95% CI, −28.00 to 8.59); desipramine versus fluoxetine (−1.40, 95% CI, −17.00 to 13.70); desipramine versus olanzapine (3.64, 95% CI, −16.00 to 22.50); desipramine versus paroxetine (3.73, 95% CI, −11.00 to 17.60); desipramine versus prazosin (−0.07, 95% CI, −21.00 to 19.90); desipramine versus risperidone (−4.40, 95% CI, −20.00 to 10.60); desipramine versus setraline (−3.90, 95% CI, −19.00 to 10.90); desipramine versus tiagabine (−8.10, 95% CI, −25.00 to 8.39); desipramine versus topiramate (7.11, 95% CI, −8.90 to 22.50); desipramine versus venlafaxine (−1.60, 95% CI, −17.00 to 13.80). The fourth forest plot compares divalproex (Drug A) to other drugs (Drug B); divalproex versus fluoxetine (8.51, 95% CI, −2.70 to 19.10); divalproex versus olanzapine (13.60, 95% CI, −1.50 to 28.80); divalproex versus paroxetine (13.70, 95% CI, 2.72 to 24.40); divalproex versus prazosin (9.87, 95% CI, −7.50 to 26.60); divalproex versus risperidone (5.53, 95% CI, −5.30 to 16.20); divalproex versus setraline (6.05, 95% CI, −4.60 to 16.40); divalproex versus tiagabine (1.80, 95% CI, −11.00 to 14.00); divalproex versus topiramate (17.10, 95% CI, 5.62 to 28.10); divalproex versus venlafaxine (8.31, 95% CI, −2.80 to 19.30).The fifth forest plot compares fluoxetine (Drug A) to other drugs (Drug B); fluoxetine versus olanzapine (5.07, 95% CI, −7.00 to 17.20); fluoxetine versus paroxetine (5.17, 95% CI, −0.70 to 10.90); fluoxetine versus prazosin (1.36, 95% CI, −13.00 to 15.70); fluoxetine versus risperidone (−3.00, 95% CI, −8.40 to 2.70); fluoxetine versus setraline (−2.50, 95% CI, −7.40 to 2.57); fluoxetine versus tiagabine (−6.70, 95% CI, −15.00 to 1.95); fluoxetine versus topiramate (8.55, 95% CI, 2.37 to 14.90); fluoxetine versus venlafaxine (−0.20, 95% CI, −6.10 to 5.67). The sixth forest plot compares olanzapine (Drug A) to other drugs (Drug B); olanzapine versus paroxetine (0.10, 95% CI, −12.00 to 11.90); olanzapine versus prazosin (−3.70, 95% CI, −21.00 to 14.90); olanzapine versus risperidone (−8.00, 95% CI, −20.00 to 3.94); olanzapine versus setraline (−7.50, 95% CI, 19.00 to 3.89); olanzapine versus tiagabine (−12.00, 95% CI, −25.00 to 1.48); olanzapine versus topiramate (3.47, 95% CI, −9.00 to 15.50); olanzapine versus venlafaxine (−5.30, 95% CI, −18.00 to 6.57). The seventh forest plot compares paroxetine (Drug A) to other drugs (Drug B); paroxetine versus prazosin (−3.80, 95% CI, −18.00 to 10.20); paroxetine versus risperidone (−8.10, 95% CI, −13.00 to −2.60); paroxetine versus setraline (−7.60, 95% CI, −12.00 to −2.80); paroxetine versus tiagabine (−12.00, 95% CI, −20.00 to −3.60); paroxetine versus topiramate (3.38, 95% CI, −2.70 to 9.64); paroxetine versus venlafaxine (−5.40, 95% CI, −11.00 to 0.39). The eighth forest plot compares prazosin (Drug A) to other drugs (Drug B); prazosin versus risperidone (−4.30, 95% CI, −18.00 to 9.65); prazosin versus setraline (−3.80, 95% CI, −18.00 to 10.10); prazosin versus tiagabine (−8.10, 95% CI, −23.00 to 7.62); prazosin versus topiramate (7.18, 95% CI, −7.40 to 21.30); prazosin versus venlafaxine (−1.60, 95% CI, −16.00 to 12.60). The ninth forest plot compares risperidone (Drug A) to other drugs (Drug B); risperidone versus setraline (0.52, 95% CI, −4.30 to 5.11); risperidone versus tiagabine (−3.70, 95% CI, −12.00 to 4.31); risperidone versus topiramate (11.50, 95% CI, 5.46 to 17.50); risperidone versus venlafaxine (2.77, 95% CI, −2.80 to 8.21). The 10th forest plot compares setraline (Drug A) to other drugs (Drug B); setraline versus tiagabine (−4.20, 95% CI, −12.00 to 3.75); setraline versus topiramate (11.00, 95% CI, 5.66 to 16.60); setraline versus venlafaxine (2.26, 95% CI, −2.20 to 6.81). The eleventh forest plot compares tiagabine (Drug A) to other drugs (Drug B); tiagabine versus topiramate (15.30, 95% CI, 6.46 to 23.90); tiagabine versus venlafaxine (6.50, 95% CI, −2.10 to 14.80). The twelfth forest plot compares topiramate (Drug A) to other drugs (Drug B); topiramate versus venlafaxine (−8.80, 95% CI, −15.00 to −2.50).

Figure 28

Results of network meta-analysis comparing improvement in PTSD symptoms (change in CAPS total score).

It appears that paroxetine and topiramate were found to be more effective than most other medications mainly due to the magnitude of effects and the precision in the individual trials that compared each of them with placebo. Two trials (total N=886) contributed data for paroxetine compared with placebo—the effect sizes in those trials were greater on average (WMD −12.6, 95% CI, −15.7, −9.5) than those for most other medications. Three trials (total N=142) contributed data for topiramate compared with placebo—the effect sizes in those three trials were greater on average (WMD, −15.5, 95% CI, −19.4 to −11.7) than those for all other medications.

Three head-to-head comparisons contributed data, but the majority of evidence in the network meta-analysis was indirect evidence (from placebo-controlled trials). Thus, we consider the findings to be of low SOE. Indirect comparisons, in general, have to be interpreted cautiously because the validity of results is based on assumptions that cannot be completely verified—particularly the similarity of study populations.

Key Question 3. Psychotherapy Versus Pharmacotherapy for Adults With PTSD

This Key Question (KQ) focused on studies that directly compared a psychological treatment with a pharmacological treatment.

Key Points

  • Just one trial (N=88) included a head-to-head comparison of a psychotherapy (eye movement desensitization and reprocessing [EMDR]) and a pharmacotherapy (paroxetine). We concluded that the head-to-head evidence was insufficient to draw any firm conclusions about comparative effectiveness because of risk of bias, unknown consistency (with data from just one study), and imprecision.
  • The trial found that EMDR- and fluoxetine-treated subjects had similar improvements in PTSD symptoms, rates of remission, and loss of PTSD diagnosis at the end of treatment. At 6-month followup, those treated with EMDR had higher remission rates and greater reductions in depression symptoms than those who received fluoxetine.

Detailed Synthesis

Characteristics of Trials

We found one medium risk of bias trial meeting our inclusion criteria. Table 37 summarizes the characteristics of the trial. Further details are provided in Appendix D.

Table 37. Characteristics of included studies directly comparing psychotherapy with pharmacotherapy.

Table 37

Characteristics of included studies directly comparing psychotherapy with pharmacotherapy.

The included trial was conducted in the United States, and randomized subjects to 8 weeks of fluoxetine, EMDR, or placebo.113 The results related to placebo comparisons are included in KQs 1 and 2. Participants were a heterogenous group of males and females with a variety of index trauma types (described as child sexual abuse, child physical abuse, adult sexual assault, adult physical assault, domestic violence, other adult victimization, traumatic loss, war/terrorism/violence, and injury/accident). All were studied in an outpatient specialty mental health setting and were followed for 6 months after treatment ended. The primary outcome was reduction in PTSD symptoms according to the total Clinician-Administered Post Traumatic Stress Disorder Scale (CAPS) score. Secondary outcomes included depression as measured by the Beck Depression Inventory (BDI).

One trial (N=21) comparing paroxetine with cognitive behavioral therapy (CBT)182 otherwise meeting criteria for this section was rated high risk of bias (Table 38), and thus was not included in our data synthesis.

Table 38. Characteristics of trials directly comparing psychotherapy with pharmacotherapy excluded from main analyses because of high risk of bias.

Table 38

Characteristics of trials directly comparing psychotherapy with pharmacotherapy excluded from main analyses because of high risk of bias.

Results for Psychotherapy Versus Pharmacotherapy

PTSD Symptom Reduction

The CAPS total score (1 month) was used to assess PTSD symptoms at study entry and followup. The CAPS total score (1 week) was used to assess PTSD symptoms at pre- and posttreatment. After 8 weeks of treatment, the CAPS total score (1 week) was not statistically significantly different between those treated with EMDR and those treated with fluoxetine (32.55 vs. 42.67, respectively, p=0.13, intention-to-treat analysis, adjusted for baseline). At 6-month posttreatment followup, the CAPS total score (1 month) was significantly lower in the EMDR-treated group than in the fluoxetine-treated group (25.79 vs. 42.12, p<0.005, intention-to-follow analysis including 85% of randomized subjects, adjusted for baseline). Effect sizes for PTSD symptom reduction favoring EMDR over fluoxetine were larger among participants with adult-onset vs. child-onset traumas.

Remission

Remission rates favored EMDR-treated subjects compared with fluoxetine-treated subjects at end of treatment, but the difference was not statistically significant (28% vs. 13%, p=0.17, intention-to-treat analysis). Remission rates were higher for those treated with EMDR than for those treated with fluoxetine at posttreatment followup (58% vs. 0%, p<0.001, intention-to-follow analysis including 85% of randomized subjects).

Loss of PTSD Diagnosis

The percentages of subjects no longer meeting diagnostic criteria for PTSD were similar for EMDR compared with fluoxetine at end of treatment (76% vs. 73%, p=0.82, intention-to-treat analysis); results at followup found no statistically significant difference (88% vs. 73%, p=0.20, intention-to-follow analysis including 85% of randomized subjects).

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

The mean (SD) BDI scores were similar for EMDR compared to fluoxetine at baseline (16.2 vs. 18.2, p=NS). At end of treatment, point estimates favored of those treated with EMDR, but differences were not statistically significant (9.10 vs. 13.00, p=0.08, intention-to-treat analysis). At followup, depression symptom scores were lower in the EMDR-treated group than in the fluoxetine-treated group (5.25 vs. 14.00, p<0.001, intention-to-follow analysis including 85% of randomized subjects).

Key Question 4. Combinations of Psychological Treatments and Pharmacological Treatments Compared With Either One Alone

For this question, we included studies that randomized subjects to the combination of a psychological and a pharmacological treatment compared with either one alone. The intention was to inform whether clinicians should start treatment with combinations of therapies at the outset as opposed to starting with a single modality.

Key Points

  • Overall, evidence was insufficient to determine whether combinations of psychological treatments and pharmacological treatments are better than either one alone when initiating treatment.
  • Two trials provided limited information related to this Key Question (KQ). Although both trials used prolonged exposure therapy as the psychological treatment, the trials differed in type of trauma population included and the timing of initiating the other intervention.
  • The trial most relevant for this KQ (N=37) found greater improvement in PTSD symptoms (Clinician-Administered PTSD Scale [CAPS] −51.1 vs. −29.8, p = 0.01) and greater likelihood of remission for those treated with both prolonged exposure and paroxetine than those treated with prolonged exposure plus placebo.183 Evidence was limited by unknown consistency (single trial), attrition, and imprecision.

Detailed Synthesis

Characteristics of Trials

Table 39 summarizes the characteristics of the two trials meeting our inclusion criteria. Further details describing the trials are provided in Appendix D. Both were conducted in the United States.

Table 39. Characteristics of included trials assessing combinations of treatments compared with either one alone.

Table 39

Characteristics of included trials assessing combinations of treatments compared with either one alone.

One trial (N=37) compared 10 weeks of prolonged exposure (10 sessions) plus paroxetine with prolonged exposure plus placebo in adult survivors of the World Trade Center attack of September 11, 2001, with chronic PTSD.183 After 10 weeks of prolonged exposure plus paroxetine or placebo, subjects were offered 12 additional weeks of randomized treatment; 13 subjects in each group began the additional 12 weeks (11 in each group completed it). Adequacy of prior PTSD treatment was not systematically documented, but 15 subjects had been previously medicated for PTSD—9 of these reported prior SSRI treatment—and 20 had previously received psychotherapy, but none reported an adequate course of at least 10 sessions of trauma-focused CBT. The primary outcomes were CAPS score and remission status at weeks 5 and 10.

The other trial (N=65) enrolled subjects with chronic PTSD for 10 weeks of open-label sertraline, followed by randomization to 5 additional weeks of sertraline alone or sertraline plus 10 sessions of twice weekly prolonged exposure.184 The trial provides limited information about whether to start treatment with combinations of therapies at the outset as opposed to starting with a single modality, primarily because all subjects were treated with 10 weeks of sertraline prior to randomization. This trial was therefore more relevant for the question of whether prolonged exposure adds benefit for people who have been treated with (and responded to) sertraline. Subjects had a variety of types of index traumas including sexual assault (24), nonsexual assault (16), death of another (14), motor vehicle accident (6), combat exposure (1), house fire (1), airplane crash (1), discovering a parent after a nonfatal overdose (1), and a police officer who felt he came close to shooting an unarmed suspect (1). The main outcomes were the Structured Interview for PTSD (SIP) (for PTSD symptoms), the Beck Depression Inventory (BDI), and the State–Trait Anxiety Inventory-State-Anxiety (STAI-S).

We rated one trial otherwise meeting criteria for this section as high risk of bias. We excluded it from our main data synthesis (Table 40). It compared sertraline plus CBT with sertraline alone.185 We excluded it from our main data synthesis.

Table 40. Characteristics of trials assessing combinations of treatments compared with either one alone excluded from main analyses because of high risk of bias.

Table 40

Characteristics of trials assessing combinations of treatments compared with either one alone excluded from main analyses because of high risk of bias.

Results for Combinations of Psychological Treatments and Pharmacological Treatments Compared With Either One Alone

PTSD Symptom Reduction

Both trials reported some measure of PTSD symptoms. The trial comparing prolonged exposure (10 sessions) plus paroxetine with prolonged exposure plus placebo in adult survivors of the World Trade Center attack reported greater improvement in symptoms for those treated with both paroxetine and prolonged exposure (CAPS, mean change from baseline to week 10: −51.1 vs. −29.8, p=0.01). The evidence from this single trial is insufficient to determine whether the combination is better than prolonged exposure alone for improving PTSD symptoms (due to risk of bias, unknown consistency with data from a single study, and attrition—with 13 subjects completing the trial in each group out of 19 and 18, respectively).

The trial comparing sertraline plus prolonged exposure with sertraline alone provides limited information related to this KQ, primarily because all subjects were treated with 10 weeks of sertraline prior to randomization. The trial therefore is more relevant for the question of whether prolonged exposure adds benefit for people who have been treated with (and responded to) sertraline. For subjects enrolled in the trial, PTSD symptoms decreased from a mean of 35.9 on the Structured Interview for PTSD (SIP), indicating moderate to severe PTSD, to a mean of 15.3 at the point of randomization, indicating mild (rarely and/or not bothersome) PTSD. After randomization, the sertraline plus prolonged exposure group had greater improvement than the group that continued only sertraline (SIP, within-group mean reduction from baseline 5.9 with p<0.001 vs. −0.3, p = NS), although the difference between groups was not statistically significant.

Remission (No Longer Having Symptoms)

The trial comparing prolonged exposure (10 sessions) plus paroxetine with prolonged exposure plus placebo in adult survivors of the World Trade Center attack reported more subjects in the former group achieving remission (intention-to-treat sample: 42.1% vs. 16.7%, modeled data adjusted for baseline values: OR 12.6; 95% CI, 1.23 to 129), defined as a CAPS score less than 20 and a CGI-C of 1 (very much improved). The findings are limited by the small sample size and missing data—data were available for 13 subjects in each group (out of 19 and 18, respectively). The very wide confidence interval reflects the limited precision of this estimate. Evidence is insufficient to determine whether the combination is better than prolonged exposure alone for remission (because of unknown consistency with data from a single study, missing data, attrition, and imprecision).

Prevention or Reduction of Comorbid Medical or Psychiatric Conditions

Both trials reported measures of reduction of comorbid psychiatric conditions. The trial comparing prolonged exposure (10 sessions) plus paroxetine with prolonged exposure plus placebo in adult survivors of the World Trade Center attack reported no statistically significant difference in reduction of depression symptoms between groups (HAM-D, −9.2 vs. −5.2, p=0.14; modeled data: OR, 0.74; 95% CI, 0.50 to 1.11). The findings are limited by the small sample size and missing data—data were available for 27 out of 37 subjects. The wide confidence interval reflects the limited precision of this estimate. We concluded that evidence is insufficient to determine whether the combination is better than prolonged exposure alone for reduction of depressive symptoms (because of unknown consistency with data from a single study, missing data, and imprecision).

The trial comparing sertraline plus prolonged exposure with sertraline alone found no statistically significant difference between treatment groups for reduction of depression symptoms (change from week 10 to week 15 for mean BDI, −3.2 vs. +0.3, p=NS) or anxiety (change from week 10 to week 15 for mean STAI-S, −3.9 vs. 0, p=NS); in both cases, point-estimates favored the sertraline plus prolonged exposure group. As described above, this trial provides limited information related to this KQ, primarily because all subjects were treated with 10 weeks of sertraline prior to randomization. Overall, evidence was insufficient to determine whether the combination is better than sertraline alone for reduction of depression or anxiety in people with PTSD (because of the study design, unknown consistency with data from a single study, and imprecision).

Quality of Life

The trial comparing prolonged exposure (10 sessions) plus paroxetine with prolonged exposure plus placebo in adult survivors of the World Trade Center attack reported greater improvement in quality of life for those in the combination treatment group (increase in Q-LES-Q: 20.8 vs. 9.4, p=0.02). The findings are limited by the small sample size and missing data/high risk of attrition bias—data were available for just 9 subjects in the combination group and 10 in the prolonged exposure plus placebo group (out of 19 and 18, respectively). In addition, evidence is from a single study. Thus, consistency is unknown and findings have not been reproduced. Thus, evidence is insufficient to determine whether the combination is better than prolonged exposure alone for improving quality of life.

Key Question 5. Are Any Treatment Approaches More Effective for Victims of Particular Types of Trauma?

This Key Question (KQ) evaluated whether any of the treatments are more effective than other treatments for victims of particular types of trauma, such as military/combat trauma, first responders, refugees, disaster victims, assault survivors, and those with exposure to childhood trauma or repeat victimization. For this question, we used two general sources of information: (1) included studies—subgroup analyses reported by individual studies that focus on subjects with a particular type of trauma or comparative effectiveness studies that compared two or more treatments within a group of subjects all with the same trauma type, and (2) subgroup analyses (stratified analyses by trauma type) of our meta-analyses of reduction in PTSD symptoms for the treatments found to be efficacious in KQs 1 and 2. For the latter, we only had sufficient data to conduct analyses for exposure-based therapy for female assault compared with other trauma types, CBT-mixed therapies for various trauma types, EMDR for female sexual assault compared with other trauma types, and SSRIs for combat trauma compared with mixed trauma (studies enrolling heterogeneous populations). There were often insufficient numbers of trials conducted in subjects with any particular type of trauma to conduct any meaningful stratified analyses and trials often enrolled heterogenous populations of subjects with a variety of index trauma types (e.g., sexual abuse, nonsexual abuse, combat, injury, motor vehicle accident, natural disaster).

Key Points

  • Overall, evidence was insufficient to make definitive conclusions about whether any treatment approaches are more effective for victims of particular types of trauma. Analyses were generally not powered to detect anything but large differences. In addition, many other factors (other than trauma type) vary across studies included in our subgroup analyses. Findings should be considered hypothesis-generating.
  • Subgroup analyses from one trial (N=169) that compared cognitive processing therapy, prolonged exposure, and waitlist found that cognitive processing therapy and prolonged exposure had similar effectiveness for participants with a history of child sexual abuse and participants whose sexual abuse occurred during adulthood.77
  • Subgroup analyses from one trial (N=88) that compared EMDR, fluoxetine, and placebo found that treatments were less effective for those with child-onset trauma.113 In addition, it found that EMDR was more effective than paroxetine at 6-month posttreatment followup for those with either child- or adult-onset trauma.
  • Our subgroup analyses (of our meta-analyses of reduction in PTSD symptoms stratified by trauma type) found no significant difference in efficacy of
    • fluoxetine, paroxetine, or sertraline for studies enrolling mixed trauma populations compared with those enrolling only subjects with combat-related trauma,
    • exposure therapy for studies enrolling females with assault or sexual abuse compared with those enrolling subjects with combat-related trauma or other trauma types, or
    • CBT-mixed therapies for studies enrolling subjects with a history of childhood sexual or physical abuse, females with assault or interpersonal violence, or refugees compared with those enrolling subjects with other trauma types.
  • Our subgroup analyses found a trend toward greater efficacy of EMDR for studies enrolling females with a history of sexual assault than for those enrolling subjects with other trauma types—EMDR was found to be efficacious for both groups, but found a large effect size for females with a history of sexual assault (SMD, −1.68; 95% CI, −2.23 to −1.13; two trials, N=71) and a small to medium effect size (that did not reach a statistically significant benefit) for those with other trauma types (SMD, −0.44; 95% CI, −1.03 to 0.15; two trials, N=46).
  • For first responders, disaster victims, or those with repeat victimization, we found no studies meeting our inclusion criteria addressing whether any treatment approaches are more or less effective, and data were insufficient to conduct any meaningful subgroup analyses (stratified analyses of our meta-analyses) or to perform meta-regression to explore whether any treatment approaches are more or less effective for these groups.

Detailed Synthesis: Trauma Type

Characteristics of Included Studies

Table 41 summarizes the characteristics of the two included studies. Both were randomized controlled trials or subgroup analyses of trials that have been described in previous parts of this report. Study treatment durations ranged from 6 to 8 weeks, with posttreatment follow-up periods from 6 to 9 months. Both studies enrolled subjects with severe PTSD symptoms at baseline. Both used the Clinician-Administered PTSD Scale (CAPS) as the primary outcome measure. Additional details describing the included studies can be found in Appendix D.

Table 41. Characteristics of studies that evaluated specific trauma types.

Table 41

Characteristics of studies that evaluated specific trauma types.

One study compared cognitive processing therapy, prolonged exposure, and waitlist in women with a history of sexual assault and conducted a subgroup analysis for those with a history of childhood sexual abuse (subgroup analysis used data from 121 of the 171 women randomized in the original trial).72,77 Forty-one percent of all study participants had been sexually abused as children. Participants had slightly more than 14 years of education on average. Time since rape ranged from 3 months to 33 years, with a mean of 8.5 years. Participants with a history of childhood sexual assault reported significantly greater criminal victimization histories than their counterparts with no childhood sexual assault history (number of times victimized, childhood physical abuse, robbery, kidnapping, rape prior to index rape, attempted rape, sexual assault, minor physical assault, and attempted murder). The two groups were similar in terms of age, race, education, and months since index rape.

The other study compared EMDR, fluoxetine, and placebo in subjects with a variety of trauma types including child sexual abuse, child physical abuse, child sexual and physical abuse, adult sexual assault, adult physical assault, domestic violence, other adult victimization, traumatic loss, war/terror/violence, and injury/accident.113 Participants had experienced mixed trauma exposure at least 1 year prior to intake. For 50 percent of enrolled subjects, trauma onset occurred prior to age 18. The authors reported subgroup analyses for those with child-onset trauma and those with adult-onset trauma.

Exposure to Childhood Trauma

For the study that compared cognitive processing therapy, prolonged exposure, and waitlist, the main analyses found that those who were randomized to one of the active treatments (i.e., cognitive processing therapy or prolonged exposure) had significant reductions in PTSD symptoms and comorbid depression compared with those who were assigned to a waitlist. The subgroup analysis comparing participants with a history of child sexual abuse to those whose sexual abuse occurred during adulthood found that cognitive processing therapy and prolonged exposure had similar effectiveness for both groups.77

For the study that compared EMDR, fluoxetine, and placebo, the main analyses found that EMDR- and fluoxetine-treated subjects had similar improvements in PTSD symptoms, rates of remission, and loss of PTSD diagnosis at the end of treatment. At 6-month followup, those treated with EMDR had higher remission rates and greater reductions in depression symptoms than those who received fluoxetine (see KQ 3 for additional details). At 6-month followup, more than twice the percentage of participants with adult-onset trauma than with child-onset trauma achieved asymptomatic functioning (75% versus 33%, respectively) in the EMDR group. No participants achieved this level of relief in the fluoxetine or placebo group. For most child-onset trauma participants, neither treatment produced complete remission of PTSD symptoms.

Our subgroup analyses for reduction of PTSD symptoms found no statistically significant difference in efficacy of CBT-mixed therapies for studies enrolling subjects with a history of childhood sexual or physical abuse (SMD, −0.95; 95% CI, −1.93 to 0.02; two trials, N=110) compared with those enrolling subjects with other trauma types (Appendix F). Confidence intervals were wide and overlapped in all cases.

Combat-Related Trauma

We found no studies meeting our inclusion criteria addressing whether any treatment approaches are more or less effective for those with combat-related trauma.

Our subgroup analyses for reduction of PTSD symptoms found no significant difference in efficacy of exposure therapy for studies enrolling subjects with combat-related trauma (just one trial with N=30) compared with those enrolling females with assault or sexual abuse or other trauma types (Appendix F). Confidence intervals were wide and overlapped (for sexual abuse/assault and other types of trauma).

Our subgroup analyses for reduction of PTSD symptoms (change from baseline in CAPS scores) found no statistically significant difference in efficacy of fluoxetine, paroxetine, or sertraline for studies enrolling mixed trauma populations compared with those enrolling only subjects with combat-related trauma (Appendix F). Confidence intervals were wide and overlapped (for mixed and combat-related trauma) in all cases. For example, pooled point estimates for fluoxetine were not significantly different for studies enrolling mixed trauma populations (WMD, −5.9; 95% CI, −10.1 to −1.6) compared with those enrolling only subjects with combat-related trauma (WMD, −9.1; 95% CI, −15.0 to −3.1). Point estimate sometimes favored mixed populations and sometimes favored combat-related trauma populations. The sertraline subgroup analysis for combat-related trauma studies (two trials, total N=211) did not find a statistically significant reduction in PTSD symptoms for those with combat-related trauma (WMD, −2.4; 95% CI, −9.5 to 4.7); however, the confidence interval is wide and overlaps the confidence interval for mixed populations (WMD, −5.8; 95% CI, −9.3 to −2.4).

First Responders

We found no studies meeting our inclusion criteria addressing whether any treatment approaches are more or less effective for first responders. In addition, data were insufficient to conduct any subgroup analyses (stratified analyses of our meta-analyses) or to perform meta-regression to explore whether any treatment approaches are more or less effective for first responders.

Refugees

We found no studies meeting our inclusion criteria addressing whether any treatment approaches are more or less effective for refugees.

Our subgroup analyses for reduction of PTSD symptoms found no statistically significant difference in efficacy of CBT-mixed therapies for studies enrolling refugees (SMD, −1.26; 95% CI, −3.16 to 0.64; two trials, N=64) compared with those enrolling subjects with other trauma types (Appendix F). Confidence intervals were wide and overlapped in all cases.

Disaster

We found no studies meeting our inclusion criteria addressing whether any treatment approaches are more or less effective for disaster victims. In addition, data were insufficient to conduct any subgroup analyses (stratified analyses of our meta-analyses) or to perform meta-regression to explore whether any treatment approaches are more or less effective for disaster victims.

Assault or Sexual Abuse

We found no studies meeting our inclusion criteria addressing whether any treatment approaches are more or less effective for assault survivors.

Our subgroup analyses for reduction of PTSD symptoms found no significant difference in efficacy of exposure therapy for studies enrolling females with assault or sexual abuse (SMD, −1.17; 95% CI, −1.47 to −0.88; four trials) compared with those enrolling subjects with combat-related trauma or other trauma types (Appendix F). Confidence intervals were wide and overlapped (for sexual abuse/assault and other types of trauma).

Our subgroup analyses for reduction of PTSD symptoms found no statistically significant difference in efficacy of CBT-mixed therapies for studies enrolling females with assault or interpersonal violence (SMD, −1.27; 95% CI, −2.16 to −0.37; four trials, N=252) compared with those enrolling subjects with other trauma types (Appendix F). Confidence intervals were wide and overlapped (for female assault/violence and other trauma types) in all cases.

Our subgroup analyses for reduction of PTSD symptoms found greater efficacy of EMDR for studies enrolling females with a history of sexual assault (two trials) than for those enrolling subjects with other trauma types (one trial each for mixed, public transportation workers, and combat-related trauma). Analyses found a large effect size (for benefit for EMDR) for females with a history of sexual assault (SMD, −1.68; 95% CI, −2.23 to −1.13; two trials, N=71) and a small to medium effect size (that did not reach a statistically significant benefit) for those with other trauma types (SMD, −0.44; 95% CI, −1.03 to 0.15; two trials, N=46, Appendix F). We also conducted sensitivity analyses including studies rated high risk of bias. The sensitivity analyses found no significant difference by trauma type. Although the summary effect size was in the same direction, point estimates moved closer together and confidence intervals overlapped (Appendix F).

Repeat Victimization

We found no studies meeting our inclusion criteria addressing whether any treatment approaches are more or less effective for those with repeat victimization. In addition, data were insufficient to conduct any subgroup analyses (stratified analyses of our meta-analyses) or to perform meta-regression to explore whether any treatment approaches are more or less effective for those with repeat victimization.

Limitations

We found insufficient data to conduct meaningful subgroup analyses (stratified analyses of our meta-analyses) or meta-regressions for many of the treatments with evidence of efficacy and for many of the trauma types. For example, for cognitive processing therapy, our meta-analysis for PTSD symptom reduction (CAPS scores) included four trials. All four trials found moderate to large benefits of cognitive processing therapy, but with differences in magnitude of benefit (from −20.7 to −51.1). We wanted to explore whether cognitive processing therapy is more or less efficacious for victims of particular types of trauma, but the four trials enrolled populations with three different trauma types (adult sexual assault,72 childhood sexual abuse,71 or military70,74). With just one trial each for three different trauma types and all finding moderate to large benefits, we can’t say with confidence if cognitive processing therapy works more or less for those with various trauma types. However, we observed a larger effect size for those with childhood sexual abuse than for military veterans (and those confidence intervals did not overlap).

As another example, we concluded that evidence supports the efficacy of topiramate for reduction of PTSD symptoms. We wanted to explore whether it is more or less efficacious for victims of particular types of trauma, but we found insufficient data to do so. Our meta-analysis for reduction of symptoms (measured by CAPS) for topiramate included three trials enrolling either mixed populations (two trials138,141) or those with combat-related trauma (one trial137)—and all three trials found similar results. Similarly, we found insufficient data for venlafaxine—with just two trials, both enrolling a heterogeneous group of subjects with a variety of index trauma types, and both with almost identical findings.

Frequently, trials enrolled heterogeneous populations of subjects with a variety of index trauma types (e.g., sexual abuse, nonsexual abuse, combat, injury, motor vehicle accident, natural disaster). Our ability to make definitive conclusions was limited by heterogeneity of enrolled populations. With individual patient data from trials, additional analyses might be possible. Further, there were often no trials for a given treatment enrolling an entire group of subjects with a particular trauma type. When there were some trials doing so, there were often insufficient numbers of trials (or with few total subjects) conducted in subjects with a particular type of trauma to conduct any meaningful subgroup analyses (stratified analyses of our meta-analyses).

Key Question 6. Adverse Effects of Treatments for PTSD

For this question, we evaluated the trials included in Key Questions (KQs) 1 through 4. In addition, we searched for nonrandomized controlled trials and observational studies (specifically, prospective cohort studies with an eligible comparison group, and case-control studies). We did not find any nonrandomized trials or observational studies meeting our inclusion criteria (e.g., prospective cohort studies or case-control studies with a sample size of at least 500; see the Methods section). Therefore, the results for this question are based on the trials included in KQs 1 through 4. Throughout this KQ we often describe risks of various adverse events—risks reported are absolute risk differences between intervention and control.

Key Points: General

  • Adverse events were often not collected using standardized measures and methods for systematically capturing adverse events were often not reported.
  • Overall, evidence was insufficient to determine comparative rates of adverse events for various interventions—very little head-to-head data were available.

Key Points: Psychological Treatments

  • The vast majority of trials reported no information about adverse effects.
  • With such a small proportion of trials reporting data, evidence was insufficient to draw conclusions about withdrawals due to adverse events, mortality, suicide, suicidal ideation, self-harmful behaviors, or other specific adverse events.

Key Points: Pharmacological Treatments

  • Mortality, suicidality, or self-harmful behaviors: evidence was insufficient to determine whether risk was increased with any of the medications.
  • Withdrawals due to adverse events: evidence was insufficient to determine whether rates differ between most medications and placebo, mainly because of imprecision. For fluoxetine, sertraline, and venlafaxine, evidence of low strength suggests similar rates (within 1% to 2%) for subjects treated with medication and those who received placebo. For paroxetine, evidence suggests a 4 percent higher rate of withdrawals due to adverse events with paroxetine than with placebo (risk difference [RD], 0.04; 95% confidence interval [CI], 0.00 to 0.07; moderate strength of evidence [SOE]).
  • Specific adverse events—focusing on medications with evidence of efficacy:
    • For topiramate, evidence was insufficient to determine whether the risk of specific adverse events is increased compared with placebo for adults with PTSD.
    • For fluoxetine, evidence suggests a small increase (~5%) in the risk of nausea (RD, 0.05; 95% CI, 0.00 to 0.09; low SOE), but was insufficient to determine whether the risk of other specific adverse events is increased.
    • For paroxetine, evidence suggests an increase (of 10% to 13%) in the risk of nausea, dry mouth, and somnolence (low SOE), but was insufficient to determine whether the risk of other specific adverse events is increased.
    • For sertraline, we found between 7 percent and 12 percent increases in the risk of nausea, diarrhea, fatigue, and decreased appetite (moderate SOE). Findings were insufficient to determine whether the risks of other adverse events are increased.
    • For venlafaxine, we found an increased risk (of 6% to 10%) of nausea, dry mouth, and dizziness for subjects treated with venlafaxine compared with those who received placebo (moderate SOE). Evidence suggests no difference in risk of headache or somnolence between subjects treated with venlafaxine compared with those who received placebo (low SOE). Findings were insufficient to determine whether the risks of other adverse events are increased.
  • Risk of bias of included studies, inconsistency, and imprecision all contributed to the insufficient SOE determinations for most adverse effects.
  • Study durations ranged from 8 to 24 weeks and were generally not designed to assess adverse events.

Detailed Synthesis: Psychological Treatments

Characteristics of Trials

The included trials are described in KQ 1 on efficacy and comparative effectiveness of pharmacological interventions. Very few of the included trials reported any information about adverse events. One of the 8 included trials of cognitive therapies and none of the 4 included trials of relaxation or stress inoculation training reported any information about adverse events. One trial of cognitive processing therapy reported only that no treatment-related adverse events occurred during the trial.74 Three of the 15 included trials of exposure therapies, 5 of the 23 trials of CBT-mixed interventions, 1 of the 7 trials of EMDR, and 2 of the 14 trials of other psychological interventions (trauma affect regulation, Seeking Safety, narrative exposure therapy, brief eclectic psychotherapy, imagery rehearsal therapy) reported some information about withdrawals due to adverse events or specific adverse events. Two of the 14 trials of other psychological interventions reported that no treatment-related adverse outcomes were observed during the trials.119,122 Additional details about the specific number of adverse events reported in each included trial are available in Appendix D.

Withdrawals Due to Adverse Events

Just 3 of the 23 included trials of CBT-mixed interventions50,55,64 and 1 of the 7 of EMDR111 reported any information about withdrawals due to adverse events. None of the trials of other psychological interventions reported withdrawals due to adverse events. With such a small proportion of trials reporting data, evidence was insufficient to draw conclusions. Any conclusions would be highly subject to reporting bias.

Mortality

Just three of the included trials from KQ 1 reported any information about mortality—one compared prolonged exposure (n=141) with present-centered therapy (n=143);91 one compared group exposure (n=180) with present-centered therapy (n=180);92 and one (n=190) compared prolonged exposure, prolonged exposure plus cognitive restructuring, and waitlist.25 The latter trial reported one postrandomization removal from the trial due to death caused by an unrelated medical condition, but the trial did not report which group the death was in.25 The trial that compared prolonged exposure with present-centered therapy reported two nonsuicide deaths in the present-centered group and none in the exposure group.91 The trial that compared group exposure with present-centered therapy reported four deaths in the present-centered group and none in the exposure group.92

Suicide, Suicidal Ideation, or Self-Harmful Behaviors

Four of the included trials from KQ 1 reported any information—one compared prolonged exposure (n=141) with present-centered therapy (n=143);91 one compared group exposure (n=180) with present-centered therapy (n=180);92 one (N=32) compared narrative exposure therapy with treatment as usual;126 and one (N=190) compared prolonged exposure, prolonged exposure plus cognitive restructuring, and waitlist.25 The latter trial reported that four participants had severe depression and suicidal ideation that required immediate intervention, but the trial did not report which groups they were in.25 The trial that compared prolonged exposure with present-centered therapy reported one suicide attempt in the exposure group and three in the present-centered group.91 The trial that compared group exposure with present-centered therapy reported no suicides in the group exposure arm and one in the present-centered arm.92 The trial that compared narrative exposure therapy with treatment as usual reported two hospital admissions for suicidal ideation in the narrative exposure therapy group and none in the treatment as usual group.126

Other Specific Adverse Events

No information about additional specific adverse events was reported by the vast majority of the psychological intervention trials. A few trials reported on hospitalizations (Appendix D), but with such a small proportion of trials reporting data, and those trials making different comparisons, evidence was insufficient to draw conclusions.

Detailed Synthesis: Pharmacological Treatments

Characteristics of Trials

The included trials are described in KQ 2 on efficacy and comparative effectiveness of pharmacological interventions.

Withdrawals Due to Adverse Events

Of the included trials, all but four reported data on withdrawals due to adverse events—one of olanzapine compared with placebo,145 two of fluoxetine compared with placebo,113,162 and one of sertraline compared with citalopram and placebo.155 Table 42 summarizes the results of our meta-analyses for withdrawals due to adverse events. When we included two rows in the table for any drug, the second row for the drug is a sensitivity analysis that included trials rated as high risk of bias. Additional details and forest plots are available in Appendix F. None of the differences between drug and placebo reached statistical significance for the main analyses or for the sensitivity analyses with the exception of paroxetine. Point estimates usually favored placebo (i.e., fewer withdrawals due to adverse events) or were on the line of no difference (i.e., equal proportion of withdrawals due to adverse events for those treated with drug or placebo).

Table 42. Results of meta-analyses for withdrawals due to adverse events: risk difference between each medication and placebo.

Table 42

Results of meta-analyses for withdrawals due to adverse events: risk difference between each medication and placebo.

Focusing on the medications with moderate SOE supporting efficacy (topiramate, fluoxetine, paroxetine, sertraline, and venlafaxine), all point estimates favored placebo except for the comparison between fluoxetine and placebo. Evidence was insufficient to determine whether withdrawals due to adverse events differ between topiramate and placebo, mainly because of imprecision. For fluoxetine, sertraline, and venlafaxine, evidence of low strength suggests similar rates (within 1% to 2%) of withdrawals due to adverse events for subjects treated with medication and those who received placebo. For paroxetine, evidence of moderate strength suggests a 4 percent higher rate of withdrawals due to adverse events with paroxetine than with placebo. Appendix G provides additional details for SOE grades.

Mortality

Just two of the included medication trials reported any information about mortality—one 12-week trial (N=411) that compared fluoxetine 20mg, fluoxetine 40mg, and placebo161 and one 12-week trial (N=538) that compared venlafaxine, sertraline, and placebo.133 One trial reported no deaths in any participants.161 The trial that compared venlafaxine, sertraline, and placebo reported that one patient randomized to the venlafaxine ER group died, secondary to acute coronary insufficiency.133 The investigators considered the death unrelated to study medication—the subject was an obese 62-year-old veteran who was a smoker with a history of treated type 2 diabetes, elevated cholesterol, and cardiac problems.133

Suicide, Suicidal Ideation, or Self-Harmful Behaviors

Just two of the included medication trials reported any information about suicidality or self-harmful behaviors—one 12-week trial (N=411) that compared fluoxetine 20mg, fluoxetine 40mg, and placebo161 and one 10-week trial (N=25) that compared paroxetine and placebo.164

The trial that compared fluoxetine 20mg (n=163), fluoxetine 40mg (n=160), and placebo (n=88) reported self-harm related events: one patient in the fluoxetine 40mg group experience self-harmful behaviors; one patient in the 20mg fluoxetine group experienced thoughts of self-mutilation, and four patients experienced suicidal ideation (one in the fluoxetine 20mg group and three in the 40mg group).161 Study authors considered two of these to be serious adverse events: one patient with thoughts of self-mutilation in the fluoxetine 20mg group and one with suicidal ideation in the fluoxetine 40mg group.

The trial that compared paroxetine with placebo reported one inpatient psychiatric hospitalization for suicidal ideation for a patient (with a previous history of suicidal ideation) who was taking paroxetine.164

Other Specific Adverse Events, By Medication

Limited information was reported for most of the medications to allow synthesis of any specific adverse events or to make definitive conclusions. We therefore focus here on the medications with moderate SOE supporting efficacy (see KQ 2)—topiramate, venlafaxine, fluoxetine, paroxetine, and sertraline—to conduct additional meta-analyses for specific adverse events. Additional details about the specific number of adverse events reported in each included trial are available in Appendix D.

Topiramate Compared With Placebo

Of the three trials that compared topiramate with placebo, two reported data on rates of some specific adverse events.138,141 The other only reported that the reason for the two dropouts in the topiramate group included drug side effects such as sexual dysfunction, light headedness, and dizziness.137 Table 43 summarizes the results of our meta-analyses (when both trials reported an outcome) and our risk difference calculations (when just one trial reported the outcome). Forest plots are available in Appendix F. Additional details about the specific number of adverse events reported in each trial are available in Appendix D.

Table 43. Results of meta-analyses and risk difference calculations for specific adverse events: topiramate compared with placebo.

Table 43

Results of meta-analyses and risk difference calculations for specific adverse events: topiramate compared with placebo.

One trial also reported mean change in weight, finding a greater mean reduction in weight for the topiramate group than for the placebo group, but the difference was not statistically significant (−1.8±3.3 kg vs. −1.1±2.8 kg, p=0.43).141

Overall, findings for topiramate were insufficient to determine whether the risk of any of the specific adverse events is increased compared with placebo for adults with PTSD. Just two trials (total N=75) contributed data; with most specific adverse events only reported by one trial (with N of either 35 or 40 subjects). Risk of bias, unknown consistency (as data were often from just one study), and imprecision all contributed to our determination that evidence was insufficient to draw conclusions. Data suggest that the risk of insomnia, taste perversion, dyspepsia, paresthesias, nervousness, and fatigue may be increased with topiramate.

SSRIs Compared With Placebo

Fluoxetine Compared With Placebo

Of the five trials that compared fluoxetine with placebo, three reported data on rates of some specific adverse events.160162 Table 44 summarizes the results of our meta-analyses (when multiple trials reported an outcome) and our risk difference calculations (when just one trial reported the outcome). Forest plots are available in Appendix F. Additional details about the specific number of adverse events reported in each trial are available in Appendix D.

Table 44. Results of meta-analyses and risk difference calculations for specific adverse events: fluoxetine compared with placebo.

Table 44

Results of meta-analyses and risk difference calculations for specific adverse events: fluoxetine compared with placebo.

The trial comparing fluoxetine 20mg (n=163), fluoxetine 40mg (n=160), and placebo (n=88) reported that no deaths occurred during 12 weeks of treatment.161 It reported the following “serious adverse events”: one patient experienced thoughts of self-mutilation in the fluoxetine 20mg group; two patients had anxiety, one had chest pain, one had suicidal ideation, and one had gastritis in the fluoxetine 40mg group; and one patient reported palpitations and one thyroid carcinoma in the placebo group.

Overall, findings for fluoxetine were insufficient to determine whether the risk of most of the specific adverse events is increased compared with placebo for adults with PTSD. Three trials (total N=776) contributed data; with most specific adverse events only reported by one trial. Evidence suggests a small increase (~5%) in the risk of nausea (low SOE). The one trial reporting diarrhea found a 24 percent increase for those treated with fluoxetine compared with those who received placebo, but data were limited to one trial (N=64), thus consistency is unknown, and findings were imprecise (with confidence interval ranging from 1% to 47%, insufficient SOE). Risk of bias, inconsistency, and imprecision all contributed to our insufficient SOE determinations for most adverse effects. Appendix G provides additional details supporting our SOE grades.

Paroxetine Compared With Placebo

Of the three trials that compared paroxetine with placebo, two reported specific data for a few adverse events.134,164 The third provided some narrative description of which adverse events occurred with an incidence of at least 10 percent and twice that of placebo, but did not report the actual data.163 Table 45 summarizes the results of our risk difference calculations. There were insufficient data to conduct meta-analyses—as none of the adverse events had data reported by more than one trial. Forest plots are available in Appendix F. Additional details about the specific number of adverse events reported in each trial are available in Appendix D.

Table 45. Results of risk difference calculations for specific adverse events: paroxetine compared with placebo.

Table 45

Results of risk difference calculations for specific adverse events: paroxetine compared with placebo.

The trial (N=563) that provided narrative description reported that the most commonly reported adverse events associated with paroxetine use (with an incidence of at least 10% and twice that of placebo) were asthenia, diarrhea, abnormal ejaculation, impotence, nausea, and somnolence.163 The majority of the treatment-emergent adverse events were rated as mild to moderate in severity and most occurred at the beginning of treatment. There were no unexpected adverse events, and serious adverse experiences were infrequent (9 of the 365 patients treated with paroxetine). One patient experienced an onset of severe headaches on day 2 of paroxetine treatment and discontinued participation.

Overall, findings for paroxetine were insufficient to determine whether the risk of most specific adverse events is increased compared with placebo for adults with PTSD. Three trials (total N=911) contributed information, but little data were reported. Evidence suggests an increase (of 10% to 13%) in the risk of nausea, dry mouth, and somnolence (low SOE). Risk of bias, lack of consistency, and imprecision all contributed to the insufficient SOE determinations for some adverse effects. Appendix G provides additional details supporting our SOE grades.

Sertraline Compared With Placebo

Of the eight trials that compared sertraline with placebo, seven reported data for specific adverse events.133,155,165,167170 Table 46 summarizes the results of our meta-analyses (when multiple trials reported an outcome) and our risk difference calculations (when just one trial reported an outcome). Forest plots are available in Appendix F. Additional details about the specific number of adverse events reported in each trial are available in Appendix D.

Table 46. Results of meta-analyses and risk difference calculations for specific adverse events: sertraline compared with placebo.

Table 46

Results of meta-analyses and risk difference calculations for specific adverse events: sertraline compared with placebo.

Overall, findings suggest increases in the risk of some specific adverse effects for people treated with sertraline. Evidence of moderate strength found between 7 percent and 12 percent increases in the risk of nausea, diarrhea, fatigue, and decreased appetite. Findings were insufficient to determine whether the risks of headache, insomnia, dry mouth, dizziness, somnolence, drowsiness, increased appetite, or constipation are increased for subjects treated with sertraline compared with those who received placebo. Risk of bias, inconsistency, and imprecision all contributed to the insufficient SOE determinations. Appendix G provides additional details supporting our SOE grades.

Venlafaxine Compared With Placebo

Of the two included trials that compared venlafaxine with placebo (total N=687), both reported data on rates of some specific adverse events.133,173 Table 47 summarizes the results of our meta-analyses (when both trials reported an outcome) and our risk difference calculations (when just one trial reported the outcome). Forest plots are available in Appendix F. Additional details about the specific number of adverse events reported in each trial are available in Appendix D.

Table 47. Results of meta-analyses and risk difference calculations for specific adverse events: venlafaxine compared with placebo.

Table 47

Results of meta-analyses and risk difference calculations for specific adverse events: venlafaxine compared with placebo.

Overall, findings suggest small increases in the risk of some specific adverse effects for people treated with venlafaxine and no difference between venlafaxine and placebo for others. Evidence of moderate strength found a small increased risk (risk difference of 6% to 10%) of nausea, dry mouth, and dizziness for subjects treated with venlafaxine compared with those who received placebo. Evidence suggests no difference in risk of headache or somnolence between subjects treated with venlafaxine compared with those who received placebo (low SOE). Findings were insufficient to determine whether the risks of insomnia, diarrhea, fatigue, decreased appetite, or constipation are increased for subjects treated with venlafaxine compared with those who received placebo. Risk of bias, inconsistency, and imprecision all contributed to the insufficient SOE determinations. Appendix G provides additional details supporting our SOE grades.

Detailed Synthesis: Head-to-Head Studies of Psychological and Pharmacological Interventions

One included trial (N=88) compared a psychotherapy (EMDR) and a pharmacotherapy (fluoxetine).113 It is described in KQ 3. The trial did not report any information about withdrawals due to adverse events, mortality, suicide, suicidal ideation, self-harmful behaviors, or other specific adverse events.

Detailed Synthesis: Combinations of Psychological Treatments and Pharmacological Treatments Compared With Either One Alone

Two included trials compared combinations with a psychological or pharmacological treatment alone. Both are described in KQ 4. Neither trial reported any data about withdrawals due to adverse events, mortality, suicide, suicidal ideation, self-harmful behaviors, or other specific adverse events. One reported that treatment-emergent adverse events were numerically greater in the prolonged exposure plus paroxetine group but did not differ significantly from those of the prolonged exposure plus placebo group, but the trial did not report any data.183