Post-traumatic stress disorder (PTSD) is a prevalent and disabling disorder. Recognition of neurobiological abnormalities associated with this condition suggests the potential efficacy of medication in its treatment. Nevertheless, questions regarding the efficacy of medications remain, despite general endorsement by clinical practice guidelines of selective serotonin reuptake inhibitors (SSRIs) as first-line agents in treating PTSD. This paper reviews evidence from randomized controlled trials (RCTs) for the efficacy of acute and long-term pharmacotherapy for PTSD, including the treatment of refractory PTSD. In addition, we conducted a systematic meta-analysis to compare the efficacy of different medications in treating PTSD. The effects of methodological study features (including year of publication, duration, number of centres) and sample characteristics (proportion of combat veterans, gender composition) were also tested. The largest body of evidence for short- and long-term efficacy of medication currently exists for SSRIs, with promising initial findings for the selective noradrenergic reuptake inhibitor venlafaxine and the atypical antipsychotic risperidone. Treatment effect was predicted by number of centres and recency of the study, with little evidence that sample characteristics predicted response. Evidence for the effectiveness of benzodiazepines is lacking, despite their continued use in clinical practice. Finally, the α1 antagonist prazosin and the atypical antipsychotics show some efficacy in treatment-resistant PTSD. Adequately powered trials that are designed in accordance with best-practice guidelines are required to provide conclusive evidence of clinically relevant differences in efficacy between agents in treating PTSD, and to help estimate clinical and methodological predictors of treatment response.
It is estimated that as many as 80–100% of all people are exposed to traumatic events during their lifetimes (Breslau et al. 1998; Frans et al. 2005). Depending on the nature of the trauma, approximately 5–9% of the general population go on to develop post-traumatic stress disorder (PTSD), a condition characterized by the experience of re-experiencing/intrusion, hyperarousal and avoidance/numbing symptoms (Breslau et al. 1998; Frans et al. 2005; Kessler et al. 2005). This constellation of symptoms satisfy the criteria for PTSD when they extend beyond a month after exposure to the trauma and cause clinically significant functional disability, as conceptualized in current psychiatry diagnostic systems such as DSM-IV (APA, 1994).
PTSD is frequently chronic and associated with significant morbidity, poor quality of life, and high personal, social and economic costs. It additionally represents a risk factor for developing other mood and anxiety disorders, as well as substance use disorders. It has been estimated that the US economy alone loses in the region of 3 billion dollars annually due to PTSD-related loss in productivity (Brunello et al. 2001).
PTSD is characterized by a range of neurobiological disruptions that may be responsive to medication, including changes in the hypothalamus–pituitary–adrenal axis, as well as alterations in the serotonergic, and noradrenergic neurotransmitter systems. Conversely, reports of the efficacy of selective serotonin reuptake inhibitors (SSRIs), in treating PTSD implicates involvement of the serotonin system in its aetiology. Indeed, on the basis of both open-label and controlled trials of these agents the majority of clinical practice guidelines have recommended SSRIs as first-line agents in treating PTSD. The SSRIs paroxetine and sertraline are currently the only medications approved by the Federal Drug Agency (FDA) in the USA and the European Medicines Agency (EMA) for the treatment of PTSD.
Despite the general acceptance of SSRIs as first-line medication interventions for treating PTSD, a recent analysis of clinical practice treatment guidelines revealed considerable variability in conclusions regarding their efficacy (Stein et al. 2009). In addition, not all patients with PTSD respond to the SSRIs, leading to the need for augmentation or combination treatment strategies, and to interest in agents such as tiagabine that employ novel mechanisms of action. A comprehensive review of the efficacy of medication in treating PTSD is therefore warranted.
Accordingly, a narrative review was conducted of the effectiveness of medication in reducing PTSD symptoms, as reported by randomized controlled trials (RCTs) of pharmacotherapy for PTSD. In addition, a quantitative estimate of treatment efficacy was obtained by synthesizing trial data as part of a meta-analysis. Finally, a review of pharmacotherapy RCTs for treatment-resistant patients was also conducted.
This review addresses the following questions:
Is medication effective in treating PTSD?
Are some agents more effective than others?
How long should medications be administered?
Are augmentation strategies effective in treating resistant PTSD?
The full search strategy to retrieve eligible trials and list of inclusion criteria can be found in the Supplementary material (available online). Briefly, inclusion of studies in this review was restricted to all placebo-controlled RCTs of pharmacotherapy for adults diagnosed with PTSD. Eligible RCTs were identified in February, 2010 by systematically searching the following databases: Medline, EMBASE, the Cochrane Central Register of Controlled Trials, the Cochrane Collaboration Depression, Anxiety and Neurosis Controlled Trials Register, and the National PTSD Center PILOTS database. The selection procedure for trials that were eligible for inclusion in the review and subsequent meta-analysis is presented in Fig. 1. Study characteristics and outcomes are provided in a separate table (Table 1).
–, No data provided.
Significant differences in PTSD symptom severity in favour of medication is indicated by an upward pointing arrow (↑) and non-significant differences by a double-headed horizontal arrow (↔). Symptom severity data was collected for the CAPS, or from another primary efficacy measures where CAPS was not employed.
Sample included in maintenance trial.
CGI-I, Clinical Global Impressions Scale – Improvement item; DGRP-I, Duke Global Rating of PTSD – Improvement item; SPRINT-I, Short PTSD Rating Interview – Improvement item; CAPS, Clinician-Administered PTSD Scale.
Trials include unpublished industry-funded short-term RCTs of paroxetine (SKB627, SKB650) and sertraline (Pfizer 588) (NICE, 2005).
Narrative review of the findings of RCTs of pharmacotherapy for PTSD
Monoamine oxidase inhibitors (MAOIs)
The MAOIs were one of the first class of agents to be tested in RCTs for treating PTSD. The monoamine oxidase A enzyme is a deaminator of both norepinephrine and serotonin, neurotransmitters that have been implicated in PTSD (Baker et al. 1995). Two RCTs of phenelzine, a MAOI, provide mixed evidence for its efficacy in treating this patient population. No differences were detected in a 5-wk cross-over RCT (Shestatzky et al. 1988), with over half (7/13) of the participants withdrawing during the course of the study while receiving phenelzine. A placebo-controlled comparison of phenelzine and the tricyclic antidepressant (TCA) imipramine in 60 male combat veterans reported a significant decrease after 3 wk of treatment with phenelzine in the score on the primary PTSD symptom severity scale, i.e. the self-rated Impact of Event Scale (IES; Kosten et al. 1991).
Reversible inhibitors of monoamine oxidase A (RIMAs)
The development of RIMAs overcame many of the factors limiting the clinical utility of MAOIs, such as their potential for serious drug-related adverse events and stringent dietary restrictions that reduced the likelihood of compliance. Nevertheless, results from two placebo-controlled trials of the RIMA brofaromine have been disappointing. In the first, Katz et al. (1994) observed no evidence of reductions in the total score of the gold-standard observer-rated measure of symptom severity, the Clinician-Administered PTSD scale (CAPS; Blake et al. 1995), following 14 wk of treatment. A significant treatment effect did emerge, however, when restricting the sample of 60 patients to those diagnosed for at least 1 yr. A larger 12-wk study of 118 patients failed to detect difference on the CAPS following a similar dosing regimen (Baker et al. 1995).
The SSRIs represent the most frequently investigated class of agents in placebo-controlled trials of PTSD, with a total of 18 RCTs conducted to date.
Paroxetine is registered by the FDA for the short-term treatment of PTSD. All three published randomized placebo-controlled trials of this medication have reported favourable results. In the first two trials, improvements in symptom severity on CAPS were detected after 4 wk of paroxetine, with medication effective for all three symptom clusters (Marshall et al. 2001; Tucker et al. 2001). Almost one third (29.4%) of patients on paroxetine went into remission after 12 wk of treatment in Tucker et al. (2001). A large placebo-controlled comparison of fixed doses of paroxetine (20 mg/d vs. 40 mg/d) failed to detect a difference in treatment response as a function of dosage or comorbid depression (Marshall et al. 2001). Marshall et al. (2007) reported significant differences in treatment response on the improvement item of the Clinical Global Impressions scale (CGI-I) after 10 wk of treatment with paroxetine, with one third (14/21) of patients in a sample of 52 mostly Hispanic adults responding to treatment (Guy, 1976).
Six RCTs of fluoxetine have been conducted to date. This includes the first published placebo-controlled trial of a SSRI for the treatment of PTSD, a 5-wk trial in which significant reductions in PTSD symptom severity were observed in a sample of 64 patients on CAPS (van der Kolk et al. 1994). However, this effect was not observed in a subsample from a VA site.
In a second trial in civilian subjects (n=54) differences in treatment response between the placebo and medication groups only reached significance for subjects classified as very much improved (Connor et al. 1999). Using a composite measure, almost half (41%) of the subjects on fluoxetine were regarded as displaying minimal levels of symptoms and non-disability by study endpoint. No differences were observed on any of the self- or clinician-rated symptom severity measures after 12 wk of fluoxetine treatment in a small sample (n=12) of combat veterans with high levels of comorbid depression (Hertzberg et al. 2000). In a larger sample of 54 predominantly male participants a statistically significant effect of medication emerged after 6 wk on the observer-rated TOP-8 symptom severity scale (Connor & Davidson, 1999; Martenyi et al. 2002b).
A subsequent placebo-controlled trial comparing treatment with 20 or 40 mg/d fluoxetine in a sample composed primarily of women did not detect superiority of medication after 12 wk between any of the comparison groups (n=411) (Martenyi et al. 2007). Similar results were reported for a study comparing 8 wk treatment with fluoxetine or eye movement desensitization and reprocessing (EMDR) (van der Kolk et al. 2007). No patients on fluoxetine in this study were asymptomatic (CAPS score <20) at 6 months follow-up.
Sertraline is licensed by the FDA for the short- and long-term treatment of PTSD. There have been a total of seven published placebo-controlled RCTs of sertraline conducted to date. Brady and colleagues (2000) detected differences on CAPS after only 2 wk sertraline treatment in 187 outpatients, with 70% of the reductions on CAPS and IES apparent after 4 wk. Using a similar study design, Davidson et al. (2001b) reported significant differences on all primary severity measures after 12 wk in an intent-to-treat sample of 202 outpatients.
In contrast, however, two small 10-wk sertraline RCTs, one in a sample of Israeli military veterans (n=42) and the other in predominantly female outpatients (n=35), both failed to detect an effect of medication on total CAPS score (Tucker et al. 2003; Zohar et al. 2002). However, Tucker and colleagues (2003) also observed negative effects in the citalopram arm of their study, suggesting the study had insufficient power to detect treatment effects. No differences on intent-to-treat analyses were observed for either drinking or PTSD severity outcomes in a 12-wk fixed-dose study of sertraline in 94 subjects with comorbid PTSD and alcohol abuse/dependence (Brady et al. 2005). Additionally, a 12-wk placebo-controlled comparison of sertraline and the serotonin norepinephrine reuptake inhibitor (SNRI) venlafaxine in 538 mixed-trauma subjects did not detect differences between sertraline and placebo on the CAPS-SX, the primary outcome measure (Davidson et al. 2006b).
In the final RCT of sertraline for PTSD, an equivalent number of treatments responders on the CGI-I were observed in the sertraline (36.9%) and placebo (41.5%) groups in a sample of 169 combat veterans after 12 wk (Friedman et al. 2007). The majority of patients treated with sertraline experienced treatment-emergent adverse events (86%), with diarrhoea and headaches being most common. Combat trauma was associated with a significantly smaller placebo response than non-combat trauma.
A single published RCT of citalopram (described above), failed to demonstrate an effect on PTSD symptom severity after 10 wk compared to either sertraline or placebo in 33 outpatients (Tucker et al. 2003).
Despite being one of the most established classes of antidepressants, only three relatively small controlled trials of TCAs for PTSD have been published to date. In the earliest (cross-over) study, the TCA desipramine did not reduce PTSD symptom severity after 4 wk of treatment in 18 male US war veterans with high levels of comorbid psychopathology, including affective disorder and substance abuse (Reist et al. 1989).
A subsequent 8-wk flexible-dose, parallel group study demonstrated superiority of amitriptyline over placebo in 46 veterans, although only for self-rated PTSD severity scales (Davidson et al. 1990). Finally, an 8-wk comparison of the TCA imipramine and phenelzine detected a significant reduction on IES after 5 wk imipramine treatment (Kosten et al. 1991). However, a relatively larger proportion of patients on imipramine rather than phenelzine dropped out due to treatment-related side-effects (17.4% vs. 5.3%, respectively).
The possibility that limbic hypersensitization or kindling might underlie increased arousal to traumatic stimuli in PTSD suggests that anticonvulsants might be effective in treating this disorder (Post et al. 2003). A 10-wk RCT of lamotrigine was conducted in war veterans to test this hypothesis (Hertzberg et al. 1999). Although a larger proportion of participants responded to treatment in the medication group than placebo group (2/5 vs. 1/4, respectively), the small sample (n=15) precluded estimation of a treatment effect size.
More recently, 232 patients from 38 centres in the USA were treated for 12 wk with the selective GABA reuptake inhibitor tiagabine (Davidson et al. 2007). No differences were observed in any of the efficacy outcomes assessed. Similar lack of efficacy was observed in 40 outpatients on most outcomes (with the exception of scores on the self-rated TOP-8) after 12 wk of treatment with topiramate (Tucker et al. 2007). No effect of treatment was detected on any outcome measure after 8 wk of placebo-controlled treatment with the selective GABA inhibitor divalproex in 85 US military veterans with PTSD (Davis et al. 2008).
Three small randomized controlled trials have evaluated the effectiveness of antipsychotic medication as monotherapy in treating PTSD. An initial trial of the atypical antipsychotic olanzapine was unable to detect a treatment response in 15 mostly female patients with non-combat PTSD over 10 wk (Butterfield et al. 2001). Patients treated with olanzapine gained significantly more weight than those given placebo by study end (11.5 lb vs. 0.9 lb, respectively).
Two small studies provide preliminary evidence that risperidone may be effective in treating PTSD in female patients. In a trial of 21 women with PTSD from childhood physical, sexual, verbal and emotional abuse, 8 wk risperidone treatment resulted in a significant reduction of symptom severity on CAPS-2 (Reich et al. 2004). Almost half of the subjects were receiving concurrent antidepressants or benzodiazepines. A similar superiority of risperidone over placebo was detected on CAPS after 10 wk treatment in 20 women exposed to domestic violence and sexual assault (Padala et al. 2006).
We were only able to identify a single benzodiazepine RCT for PTSD. In a small controlled trial, 16 outpatients were administered alprazolam or placebo for two 5-wk periods in a cross-over fashion, with medication effects only observed for depression among the 10 study completers (Braun et al. 1990).
A total of five placebo-controlled RCTs of medications with novel mechanisms of action have been conducted. In the first placebo-controlled trial inositol, a glucose isomer, was administered over 4 wk in a cross-over fashion to 17 outpatients with mixed trauma, with no discernible effect of medication on the primary outcome (IES) among the 13 subjects who completed the trial (Kaplan et al. 1996).
The second trial compared the effectiveness of 8 wk treatment with mirtazapine or placebo in 26 subjects with PTSD, the majority of whom had comorbid depression (Davidson et al. 2003). A greater number of treatment responders were observed in the medication group, with mirtazapine also demonstrating an antidepressant effect. Davis et al. (2004) detected superiority of 12 wk nefazodone to placebo in 42 combat veterans on the continuous total CAPS score. However, the high dropout rate (46%) among those treated with medication raises questions regarding the tolerability of this agent.
A 6-month study of the SNRI venlafaxine assessed its effectiveness in patients sampled from 56 outpatient centres outside of the USA (Davidson et al. 2006a). An improvement was observed on CAPS-SX total score, as well as on measures of quality of life, functional disability, resilience to stress and comorbid depression. A subsequent trial by the same group detected significant reductions in symptom severity following venlafaxine treatment in 538 outpatients on the CAPS-SX total score. Differences were observed as early as 2 wk following the initiation of treatment (Davidson et al. 2006b).
Meta-analysis of pharmacotherapy for PTSD
This brief review highlights the inconsistency of the evidence for the efficacy of medication in treating PTSD. Of the 37 short-term studies included in this review that conducted between-group comparisons, only 12 detected a significant reduction in PTSD symptom severity on CAPS or an alternate primary outcome measure (see Table 1). Poor sensitivity to treatment effects may be partially due to variation in study methodology and the clinical characteristics of patient groups, as well as insufficient power to detect a treatment effect in small studies. The quantitative synthesis or meta-analysis of treatment outcome data across studies allows one to maximize power in detecting an effect.
A number of meta-analyses of medication treatment for PTSD have been conducted to date, synthesizing data from trials of individual agents (Adamou et al. 2007; Mooney et al. 2004), as well as comparing treatment efficacy across agents (Penava et al. 1996; Stein et al. 2006; Stewart & Wrobel, 2009; Van Etten & Taylor, 1998). Although the findings of these meta-analyses have in general supported the use of pharmacotherapy, and SSRIs in particular, in treating PTSD, their conclusions have been weakened by methodological shortcomings (listed in Table 2). Accordingly, we conducted a meta-analysis of placebo-controlled RCTs of PTSD in adults that was restricted to between-group comparisons of outcome data from validated scales. This meta-analysis draws on an existing Cochrane review of pharmacotherapy for PTSD, which is currently in the process of being updated (Stein et al.2006).
AMED, Allied and Complementary Medicine Database; CCDAN-TR, Cochrane Collaboration Depression, Anxiety and Neurosis Trial Registry; CINAHL, Cumulative Index to Nursing and Allied Health Literature; PILOTS, Published International Literature on Traumatic Stress; SIGLE, System for Information on Grey Literature in Europe.
Total number of pharmacotherapy trials included in study, with the specific number of trials included in meta-analysis in parentheses.
CAPS, Clinician-Administered PTSD Scale; IES, Impact of Events Scale; DTS, Davidson Trauma Scale.
Effect size estimates and 95% confidence intervals reported using standardized (*) or non-standardized (†) mean difference metrics.
Primary outcomes included the reduction in total symptom severity on the Clinician-Administered PTSD Scale and the number of subjects rated as ‘much improved’ or ‘very much improved’ on the improvement item of the CGI-I (or closely related measure). Secondary outcomes included the efficacy of medication in alleviating the severity of PTSD symptom clusters, as assessed by the respective subscales of the observer-rated CAPS and the self-rated Davidson Trauma Scale (DTS; Davidson et al. 1997). Medication acceptability was estimated by calculating the total proportion of participants who withdrew from RCTs due to treatment-emergent adverse events.
Weighted mean differences (WMD) for continuous measures and relative risks for categorical outcomes were obtained from a random-effects model. Treatment response on the CGI-I was converted into a number needed to treat (NNT) for each medication agent (see notes in Table 3 for the exact procedure).
Effect sizes reported as (non-standardized) mean differences.
Heterogeneity classified as minimal, moderate and large, based on an I2 statistic of <30%, 30–50 %, and >50%, respectively.
The number needed to treat (NNT) was calculated from the risk ratio estimates of treatment response, defined as ‘much improved’ or ‘very much improved’ on the Clinical Global Impressions Improvement item (or related scale). The NNT is based on a high and low estimate of response in the control group of 0.25 and 0.4, respectively, calculated by rounding the limits of the interquartile range for the placebo group response rate in the included studies to the nearest 5 percentage points. Baseline risk was calculated using the metannt command of the Metan package in Stata 11 (www.stata.com).
Efficacy analyses (detailed below) were conducted using the metafor package in R statistical software, and employing the DerSimonian–Laird estimator of heterogeneity (DerSimonian & Laird, 1986; Viechtbauer, 2010). Differences in the efficacy of classes of medication were assessed by means of Deeks' stratified test of heterogeneity (Deeks et al. 2001). Egger's regression test of funnel plot asymmetry was employed in order to determine whether there was evidence of possible publication bias (Egger et al. 1997). A mixed-model meta-regression was conducted to determine the degree to which methodological (gender distribution and proportion of patients with combat trauma) and clinical (duration of trial in weeks, number of sites, year of publication, pharmaceutical funding) differences between trials might have influenced the reduction of PTSD symptom severity.
A total of 37 short-term RCTs (4–24 wk), containing data for 5008 patients treated with medication for an average of 10 wk, were included in the review. In addition, we identified five published RCTs that included a maintenance component (Connor et al. 2006; Davidson et al. 2001a, 2005; Marshall et al. 2007; Martenyi et al. 2002a). The greatest number of trials assessed the effectiveness of SSRIs (n=20). There was little evidence of publication bias across the 23 studies that were included in the meta-analysis (z=1.12, p=0.27; see Supplementary Fig. S2).
Medication treatment resulted in a significant reduction in PTSD symptom severity, with a reduction of about 6 points on CAPS total score relative to placebo [n=23, mean difference (MD) −6.10, 95% CI −7.98 to −4.23, n=4112] (see Fig. 2). Of the four SSRI agents for which there was data (citalopram, fluoxetine, paroxetine, sertraline), evidence of efficacy was only available for paroxetine (n=4, MD −10.65, 95% CI −14.16 to −7.14, n=1100) and sertraline (n=8, MD −4.35, 95% CI −6.76 to −1.93, n=1260). Paroxetine was significantly more effective than sertraline and fluoxetine in reducing symptom severity using a fixed-effects model (χ2=10.37, p=0.001 and χ2=3.08, p=0.08, respectively).
Pharmacotherapy was more likely to result in a global clinical response on the CGI-I than placebo [n=16, relative risk (RR) 1.4, 95% CI 1.17–1.66, n=1821]. A larger proportion of patients were responders on this scale in the medication (57.2%) than placebo (40.2%) groups. The corresponding NNT indicates that, relative to patients in the placebo groups, approximately 7–10 patients have to be treated for an average of 11 wk with medication in order for an additional patient to respond to treatment.
Significant treatment effects were observed for all symptom clusters on CAPS. Comorbid depression but not anxiety were reduced following medication treatment (HAMD: n=10, MD −2.31, 95% CI −3.6 to −1.02, n=930). Significant improvements were also observed in functional disability following medication treatment (n=10, MD −1.87, 95% CI −2.72 to −1.02, n=1852). Finally, a larger number of patients on medication withdrew from treatment due to adverse events than on placebo (n=29, RR 1.38, 95% CI 1.10–1.72, n=4045), although the absolute proportion that withdrew was relatively small (9.5%).
Recently published trials were more likely to report smaller reductions in symptom severity that older trials (z=3.28, p=0.001), a finding that does not appear attributable to an increase in placebo response over time (rho=−0.21, p>0.05). Instead, it seems more likely a result of the fact that many of the recent studies are small trials of novel psychotropics, for which there is limited evidence of effectiveness, and which are under-powered to detect treatment effects. The effect of medication on CAPS increased as the number of centres involved in the RCTs grew larger (z=−2.31, p=0.02), possibly also related to increased power in the larger trials. Surprisingly, no effect on outcome was detected for the duration of trial in weeks, or proportion of females or veterans comprising the study sample.
Length of treatment
The available pharmacotherapy evidence base suggests that treatment effects may emerge as early as 2–4 wk for SSRIs and the SNRI venlafaxine. Less is known about how long patients should be treated to achieve a maximal response, and when it is safe to discontinue pharmacotherapy without risking relapse.
Davidson and colleagues administered 28 wk of placebo or sertraline to patients who completed a 12-wk placebo-controlled trial of sertraline, and who responded to a subsequent 24 wk of open-label treatment with this agent (50–200 mg/d) (Davidson et al. 2001a). Response to acute-treatment with sertraline was maintained in the majority of patients during open-label continuation, with more than half of the non-responders to sertraline (54%) during the acute phase responding during the continuation phase (Londborg et al. 2001). Among responders to continuation treatment, over a quarter (26%) of those subsequently randomized to placebo relapsed; this was 6.4 times larger than the relapse rate observed among those randomized to sertraline.
Responders to medication in a 12-wk double-blind, placebo-controlled trial of fluoxetine for PTSD were randomized to another 24 wk of treatment with fluoxetine or placebo (Martenyi et al. 2002a). Time to relapse was significantly longer in the medication than the placebo groups, with a lower proportion (5.8%) of the 131 responders to acute-phase fluoxetine treatment relapsing after continuation with medication than placebo (16.1%). Patients who were randomized to fluoxetine continued to improve significantly on the clinical and PTSD severity scores, as well as in anxiety and depression symptoms.
Davidson et al. (2005), randomized 62 subjects to 6 months continuation treatment with fluoxetine or placebo after the same period of open-label treatment (max. 60 mg/d) with fluoxetine. Although a relatively strong effect of medication was observed in preventing relapse during the maintenance phase of the study, relapse rates were high for participants who continued receiving a stable dose of fluoxetine (22%) as well as for those for whom medication was discontinued (50%).
The efficacy of maintenance treatment with the selective GABA reuptake inhibitor tiagabine was tested by randomizing 18 completers of 12-wk of open-label treatment with tiagabine (max. dose 16 mg) to ongoing medication treatment or placebo for an additional 12 wk (Connor et al. 2006). No differences in clinical response were detected.
The limited evidence reviewed above suggests that treatment with SSRIs may be beneficial over the long term. These agents also appear to be well-tolerated among those patients who have achieved stable doses, with no drug-related adverse events reported for more than 20% of subjects in any of the relapse prevention studies reviewed.
A large proportion of patients with PTSD fail to respond to treatment with pharmacotherapy. Nevertheless, few rigorous trials of pharmacotherapy in non-responders to first-line treatments for PTSD have been conducted (Table 4). In a recent review, Ipser et al. (2006) identified only four RCTs that assessed augmentation strategies for treating PTSD in populations who were currently receiving psychotropic medication, and who could be defined as treatment resistant according to lack of response on CGI-I or a validated measure of symptom severity (Bartzokis et al. 2005; Hamner et al. 2003; Raskind et al. 2003; Stein et al. 2002). These trials are briefly described below, as well as a subsequently published study of the antipsychotic risperidone that satisfied the inclusion criteria employed by Ipser et al. (2006) (Rothbaum et al. 2008).
CAPS, Clinician-Administered PTSD Scale; BDI, Beck Depression Inventory; CES-D, Center for Epidemiologic Studies Depression Scale; CGI-I, Clinical Global Impression scale – Improvement item; DTS, Davidson Trauma Scale; HAMA, Hamilton Anxiety scale; HAMD, Hamilton Depression scale; PANSS, Positive and Negative Syndrome Scale; PANSS-P, Positive and Negative Syndrome Scale (Positive subscale); PSQI, Pittsburgh Sleep Quality Index.
Doses reported as mean daily doses or dose ranges.
Dysregulation of the sympathetic system and the centrality of hyperarousal symptoms in PTSD suggest that agonists of adrenergic receptors might be effective treating refractory patients. A small 20-wk double-blind cross-over study of the efficacy of the α1-adrenergic antagonist prazosin in treating sleep disturbances in 10 patients with chronic PTSD reported clinically significant improvement following medication treatment in PTSD and sleep symptoms (Raskind et al. 2003).
Antipsychotic agents may be indicated as augmentation agents in treating PTSD, given the presence of psychotic symptoms in complicated cases of PTSD, and the possibility that the disorder is characterized by dysregulation of the dopaminergic system (Seedat et al. 2003). In the first published augmentation RCT for PTSD, olanzapine was administered for the treatment of chronic PTSD in 21 war veterans who did not respond to a minimum of 12 wk prior treatment with a SSRI (Stein et al. 2002). Olanzapine significantly reduced symptom severity on CAPS and sleep disturbance after 8-wk double-blind treatment, despite the absence of psychotic symptoms in any of the patients.
The efficacy of risperidone in treating PTSD has been assessed in three RCTs. No effect of medication was observed on CAPS following 5-wk of treatment with risperidone in 37 Vietnam veterans diagnosed with PTSD and psychotic symptoms (Hamner et al. 2003). A subsequent placebo-controlled trial of risperidone was conducted in 65 veterans with PTSD who were participating in a 5-wk psychotherapy residential programme (Bartzokis et al. 2005). Patients with psychotic symptoms were excluded, and most of the study participants were being treated with antidepressants (88%). Superiority of medication was observed after 16 wk in reducing PTSD symptoms on total CAPS score, as well as anxiety and negative and positive psychotic symptoms.
In one of the few trials to assess pharmacotherapy for treatment-resistant PTSD in a non-veteran population, Rothbaum and colleagues (2008) administered add-on risperidone for 8 wk to 20 patients who failed to achieve a 75% reduction in total score on CAPS after 8 wk open-label treatment with sertraline. No effect of medication was detected at endpoint on CAPS, with approximately one third (4/11) of the patients treated with risperidone dropping out due to possible treatment-related adverse events.
The majority of controlled trials of pharmacotherapy in combating anxiety disorders add a course of antipsychotics to ongoing treatment with SSRIs (Ipser et al. 2006). In general, findings from trials of PTSD appear to support the efficacy of this strategy, at least with respect to combat-related traumas. There is less evidence regarding the management of PTSD in civilian populations. Evidence that prazosin is efficacious in treating sleep disturbances associated with refractory PTSD is consistent with findings from an RCT of this agent when used as monotherapy in treating sleep-related symptoms in PTSD patients (Taylor et al. 2008).
The results of a meta-analysis are only as valid as the quality of the individual trials from which it is composed. With this in mind the authors conducted an exhaustive search for eligible trials, and restricted the results of this review to RCTs, widely regarded as the most rigorous study design. Moreover, the test of publication bias provided little evidence that articles that reported positive results were disproportionally represented in the meta-analysis.
The results of our review support the efficacy of medication in treating PTSD over the short-term, with the largest body of evidence of efficacy for the SSRIs and venlafaxine. The agents appear relatively fast-acting, with response reported as early as the first 2–4 wk of treatment. Nevertheless, maximizing response to medication might require treatment for substantially longer, with RCTs of paroxetine and fluoxetine observing substantial improvements in clinical response beyond 12 wk of treatment (Martenyi et al. 2002a). No evidence of efficacy was available for benzodiazepines, despite their continued popularity in clinical practice (Cloos & Ferreira, 2009).
The current evidence-base supports the efficacy and tolerability of treatment with SSRIs over the longer-term. The finding that over a quarter of patients relapsed after discontinuation of fluoxetine after up to 26 wk of treatment (Davidson et al. 2001a) provides some support to the consensus that treatment of chronic PTSD with medication should be continued for at least a year (Bandelow et al. 2008). Of the two FDA- and EMA-approved agents for the treatment of PTSD, stronger support is available for the efficacy of paroxetine than sertraline.
War trauma is commonly perceived as being prognostic of a poorer response to treatment. The finding in a 12-wk fluoxetine trial of increased response in patients recently exposed to combat suggests that the salient characteristic with regards to medication response might be duration of PTSD, rather than combat trauma or gender, per se (Martenyi et al. 2002b). Indeed, in a comparison of the efficacy of 13 pharmacotherapy and 12 psychotherapy trials for combat-related PTSD, not only did Stewart & Wrobel (2009) observe a significantly larger effect for medication than placebo on PTSD symptom severity (t=−2.74, p=0.01), but the mean reduction in symptom severity was approximately twice as large for drug than psychotherapy treatments (−1 vs.−0.52 standardized units, respectively). The results of our regression analysis supports the contention that combat trauma or gender are not influential determinants of study outcome, and indicates instead that certain study design features, such as sample size, might be more important in determining the size of the treatment effects observed.
Neither of the dose-comparison studies of fluoxetine or paroxetine were able to detect significant differences in efficacy between higher and lower doses (Marshall et al. 2001; Martenyi et al. 2007). This is consistent with the general observation of a flat response curve for SSRIs, and suggests that it might be prudent to initiate medication treatment at the low end of the recommended dose range, with the aim of minimizing potential treatment-related adverse events. Future research should address the comparative efficacy of doses of venlafaxine, as there is evidence in studies of depression that the efficacy of venlafaxine is dose-dependent (Stahl et al. 2005).
There has recently been interest shown in the treatment of PTSD with medications that employ extra-serotonergic mechanisms of action, including the anticonvulsants, atypical antipsychotics and venlafaxine. With the exception of risperidone and venlafaxine, the results of monotherapy trials with these agents have been disappointing. Positive results from small underpowered studies of risperidone need to be followed up with larger placebo-controlled trials. Similarly, although medications such as prazosin and the atypical antipsychotics appear to hold promise as augmenting agents, much additional work is needed to determine how best to optimize remission, and how to manage PTSD symptoms in treatment-refractory patients. Such management may well include a role for psychotherapy, a topic that falls outside the scope of the current review.
Investigators are encouraged to design their clinical trials according to best-practice guidelines, and using gold-standard outcomes, such as CAPS, in order to facilitate future meta-analyses of the efficacy of particular agents. Additional information on remission in addition to treatment response would also be useful with regard to establishing the clinical significance of the findings from the RCTs. It is hoped that work on understanding the psychobiology of PTSD will ultimately translate into medications with novel targets.
Supplementary material accompanies this paper on the Journal's website.
The authors thank Carli Sager for assisting with trial selection and data extraction, and Sarah Dawson, the trial search coordinator of the Cochrane Collaboration Depression, Anxiety and Neurosis (CCDAN) review group, for assisting with database search queries. We are also grateful to the PIs who responded to requests for information required to determine the eligibility of trials for inclusion in the review. Drs Mark Hamner, Mark Pollack, Gina Manguno-Mire, and Jonathan Davidson were particularly helpful in this regard. Finally, Dr Michael Hertzberg kindly responded to queries concerning trial methodology.
Statement of Interest
Dan Stein has received research grants and/or consultancy honoraria from AstraZeneca, Eli-Lilly, GlaxoSmithKline, Lundbeck, Orion, Pfizer, Pharmacia, Roche, Servier, Solvay, Sumitomo, and Wyeth.