Cpp153a 126.144

Clin. Psychol. Psychother. 5, 126±144 (1998) Michelle L. Van Etten1 and Steven Taylor2* 1Department of Psychiatry, University of Michigan, USA 2Department of Psychiatry, University of British Columbia, Vancouver, Canada A meta-analysis was conducted on 61 treatment outcome trials for post- traumatic stress disorder (PTSD). Conditions included drug therapies (TCAs, carbamazepine, MAOIs, SSRIs, and BDZs), psychological therapies (behaviour therapy, Eye-Movement Desensitization and and dynamic therapy), and control conditions (pill placebo, wait-list controls, supportive psychotherapies, and non-saccade EMDR control).
Psychological therapies had significantly lower drop-out rates than pharmacotherapies (14% versus 32%), with attrition being uniformly low across all psychological therapies. In terms of symptom reduction, psychological therapies were more effective than drug therapies, and both were more effective than controls. Among the drug therapies, the SSRIs and carbamazepine had the greatest effect sizes, although the latter was based upon a single trial. Among the psychological therapies, behaviour therapy and EMDR were most effective, and generally equally so. The most effective psychological therapies and drug therapies were generally equally effective. Differences across treatment conditions were generally evident across symptom domains, with little matching of symptom domain to treatment type. However, SSRIs had some advantage over psychological therapies in treating depression.
Follow-up results were not available for most treatments, but available data indicates that treatment effects for behaviour therapy and EMDR are maintained at 15-week follow-up. # 1998 John Wiley & Sons, Ltd.
intrusive thoughts) (2) avoidance of trauma-related Post-traumatic stress disorder (PTSD) is character- stimuli and numbing of general responsiveness, ized by three clusters of symptoms, which arise and (3) persistent hyperarousal (e.g. hypervigilance, after the person is exposed to a traumatic stressor. exaggerated startle response: American Psychiatric The clusters are (1) recurrent reexperiencing of Association (APA), 1994). PTSD is often chronic, the traumatic event (e.g. flashbacks, nightmares, and persists for at least 1 year after the trauma in approximately 50% of cases (Davidson et al., 1996).
The most common precipitating events are combat trauma, physical and sexual assault, natural dis- *Correspondence to: Steven Taylor, Department of Psychiatry, 2255 Westbrook Mall, University of British Columbia, Van- asters, and motor vehicle accidents (Breslau et al., couver, B.C., Canada, V6T 2A1. Email: [email protected].
1991; Davidson et al., 1991; Norris, 1992).
Contract grant sponsor: Medical Research Council of Canada.
Community-based studies indicate that PTSD has Contract grant sponsor: National Institute on Drug Abuse.
a lifetime prevalence of 1 to 14%, depending on diagnostic methods and type of population, and not ence is thought to be so intense that it causes fear- surprisingly occurs at a much higher rate (3 to 58%) conditioning to a wide range of stimuli (e.g. sights, in people who are at risk for exposure to traumatic sounds, odours, and bodily sensations associated events (e.g. combat veterans, victims of natural with the trauma). Such stimuli can serve as disasters or criminal violence: APA, 1994).
reminders of the trauma (retrieval cues), thus Several forms of treatment have been applied to activating the fear structure and thereby producing PTSD. Many treatments seem promising, although hyperarousal and intrusive recollections of the the literature currently provides no clear indication trauma. Avoidance and numbing symptoms are as to the method(s) of choice. Drug therapies used in thought to arise from mechanisms for deactivating treating PTSD include tricyclic antidepressants the structure (Foa et al., 1992).
(TCAs), agents with anticonvulsant and mood- According to contemporary cognitive models, stabilizing properties (e.g. carbamazepine), benzo- PTSD can be reduced by exposing the person to diazepines (BDZs), monoamine oxidase inhibitors corrective information, which modifies the fear (MAOIs), and serotonin specific reuptake inhibitors structure. Behavioural and cognitive-behavioural (SSRIs). Drug therapies are based on the assumption treatments are seen as effective means of producing that exposure to trauma causes neurochemical this change. An important ingredient in these aberrations in mechanisms controlling arousal and treatments is exposure to fear-evoking but objec- other aspects of emotional processing, and that tively harmless stimuli. Some behavioral interven- medications correct these aberrations. Changes in tions also include cognitive restructuring, in which the opioid, noradrenergic, dopaminergic, seraton- the meaning of the trauma is examined. Training in ergic, and hypothalamic±pituitary±adrenal axis anxiety management skills is also provided (Foa systems have all been implicated in PTSD (van der et al., 1989). Throughout this article we will use the Kolk, 1987; Friedman, 1991; Sutherland and David- term `behaviour therapy' to include behavioral and son, 1994). It is beyond the scope of the present article cognitive-behavioural treatments. We will examine to offer a more detailed discussion of the neural these treatments as a class of interventions rather structures, circuits, and neurotransmitters impli- than evaluating specific types of treatment. This is cated in the various biochemical models of PTSD; because there are insufficient trials to separately see Sutherland and Davidson (1994) for details.
examine each form of behavioural and cognitive- Behavioural therapies (e.g. imaginal exposure) behavioural therapy. Our approach is similar to and cognitive-behavioural therapies (e.g. stress other meta-analytic efforts to evaluate the efficacy inoculation training: Veronen et al., 1978) for PTSD of broad classes of interventions (e.g. Lipsey and were developed from conditioning and cognitive Wilson, 1993).
theories. In an early formulation, Mowrer's (1960) Eye-movement desensitization and reprocessing two-factor model was used to account for combat- (EMDR) is a recent and controversial treatment that related PTSD (Keane et al., 1985). According to this entails imaginal exposure to traumatic images while formulation, exposure to trauma produces a con- systematic saccadic eye movements are produced.
ditioned fear or anxiety response to trauma-related Saccades are typically induced by tracking a stimuli. Escape and avoidance of trauma-related therapist's finger as it is moved rapidly from side stimuli are negatively reinforced (i.e. reinforced to side (Shapiro, 1991). Coping statements also are because they provide short-term relief from distress), introduced while the scene is being imagined.
and thereby prevent habituation from occurring.
Treatment typically takes one to four sessions.
Contemporary cognitive theories of PTSD are Recently, Shapiro (1995) suggested that eye move- consistent with neo-conditioning models (Rescorla, ments in EMDR can be replaced with other lateral, 1988), and emphasize expectations and appraisals stereotypic, motor movements. Shapiro (1995) about the meaning of aversive experiences. Such postulated that exposure to trauma produces models include the emotional processing model neuronal changes and disruption of a physiological (Foa and Kozak, 1986; Foa et al., 1989) and similar balance between excitatory and inhibitory systems approaches (e.g. Chemtob et al., 1988; Litz and in the brain, which prevents appropriate processing Keane, 1989; Litz, 1992). These models propose that of traumatic memories. EMDR purportedly restores PTSD symptoms arise from a fear structure stored this balance and reverses the neural pathology, and in long-term memory. The structure consists of a in so doing, allows appropriate reprocessing and network of information about stimuli, their mean- integration of the traumatic memories (Shapiro, ings, and responses to those stimuli (e.g. autonomic 1991, 1995). The theoretical underpinnings of arousal, escape, avoidance). The traumatic experi- EMDR have been criticized by several writers Clin. Psychol. Psychother. 5, 126±144 (1998) (e.g. Lilienfeld, 1996; McNally, 1996). It may be that observer-rated scales. Observer-rated scales typi- EMDR is an effective treatment, but not for the cally yield larger effect sizes than self-report scales (e.g. Lambert et al., 1986; Taylor, 1995). If observer- Another therapy used for PTSD includes relax- rated scales were more likely to be used in studies ation training, which is aimed at reducing hyper- of some treatments (e.g. drug therapies) than in arousal (e.g. Vaughan et al., 1994). Other treatments others (e.g. behaviour therapy), then the compari- include hypnotherapy and psychodynamic psycho- son between treatments will be confounded by therapy, which are aimed at uncovering and differences in assessment method.
resolving unconscious conflicts arising from the Otto et al. (1996) did not include uncontrolled traumatic events (e.g. Brom et al., 1989).
trials, and thereby excluded many studies from their Despite the many treatments used for PTSD, analysis. For a given type of treatment (e.g.
none have been established as treatments of choice, behaviour therapy), the effect sizes of these trials and there has been only one previous attempt to can be compared with those of controlled trials in quantify the relative efficacy of these interventions. order to determine the comparability of controlled Otto et al. (1996) reported an effect-size analysis for and uncontrolled trials. If the mean effect sizes for 14 treatment-outcome studies representing 20 trials controlled and uncontrolled trials do not differ, then of drug therapy or psychological therapy. Studies uncontrolled trials can be included, thereby increas- were published or presented between 1991 and 1994. ing statistical power (Hunter and Schmidt, 1990).
Only randomized controlled trials were included. On The purpose of the present study was to further measures of PTSD symptoms, general anxiety, and investigate the comparative efficacy of PTSD treat- depression, the drug therapies with the largest effect ments, using a broader range of treatments than sizes were fluoxetine and amitriptyline. Behavioural those examined by Otto et al. (1996). We also therapies tended to have larger effect sizes than these intended to circumvent the methodological con- drug therapies, and were associated with less cerns inherent in the latter study. We used meta- attrition. Follow-up data were not examined, and analysis to empirically evaluate the relative efficacy of treatments for PTSD. Our aims were (1) to iden- These findings were based on a small number of tify which classes of treatment are more effective trials, and so the results should be regarded with than wait-list controls and placebo; (2) to determine caution. Moreover, there are several major methodo- whether some classes of treatment are more effective logical concerns with Otto et al.'s (1996) study. They than others; and (3) to determine whether treatment computed each effect size by subtracting the mean gains are maintained at follow-up.
of the posttreatment treatment group from the mean of the posttreatment control group, and then dividing by the standard deviation of the control group. The problem with this approach is that it METHOD ignores pretreatment differences between treatment and control groups. The trials typically consisted of Inclusion and Exclusion Criteria small numbers of participants (e.g. Ns of 8 to 16). English-language articles published, unpublished, With such small samples it is likely that random or presented at conferences from 1984 to 1996 were assignment of participants to treatment versus located from Medline, the PILOTS Database, control groups would often fail to equate groups Psychological Abstracts, Current Contents, confer- on pretreatment severity. This means that some of ence programs, recent journal issues, and secondary the effect sizes may actually represent pretreatment sources (e.g. narrative reviews, book chapters), and differences rather than differences in the efficacy of by contacting PTSD researchers. Articles were treatment and control conditions. Moreover, Otto included if the following criteria were met: (1) all et al. compared treatments against different types of participants were diagnosed with PTSD according controls. Drug therapies were compared to pill to DSM III, DSM III-R, or DSM-IV criteria, as placebo whereas psychological therapies were typi- assessed by structured or unstructured clinical inter- cally compared to waiting-list controls. Thus, the views. (2) Five or more participants were included in comparison of treatments was confounded by the each trial. (3) Sufficient information was provided to compute effect sizes (or necessary additional data A further concern with the Otto et al. (1996) was supplied by the authors). (4) Outcome was study is that they computed effect sizes across presented in terms of self-report or observer-rated scales, thus combining data from self-report and measures for one or more of the following variables: Clin. Psychol. Psychother. 5, 126±144 (1998) intrusions, avoidance, total PTSD severity, depres- with one condition consisting of phenelzine sion, and anxiety. These variables were selected followed by pill placebo. Here, we included only because they are the ones most commonly used the phenelzine condition.
to assess outcome in treatments of PTSD. (5) The Of the trials included in the meta-analysis, six outcome measures had acceptable levels of were TCA treatments, which included desipramine reliability and validity, as reported in the outcome (n (number of trials) ˆ 2; mean dose (i.e. mean dose at the end of treatment) ˆ 200 mg/day), A total of 41 studies were located, yielding 68 imipramine (n ˆ 2; mean dose ˆ 242 mg/day), treatment-outcome trials. Three trials from three amitriptyline (n ˆ 1; mean dose ˆ 175 mg/day), different studies were of various inpatient treat- and trazadone (n ˆ 1; mean dose ˆ 300 mg/day).
ments that were sufficiently heterogeneous and/or Although carbamazepine is structurally similar to poorly described so as to prevent interpretation TCAs, we classified it separately because it appears of the data as a distinct treatment class. These to have different pharmacologic properties to were therefore excluded from the analysis, leaving conventional TCAs. In addition to its anti-seizure 65 trials. In 61 of the these trials, most participants effects, it is thought to reduce problems of impulse had chronic PTSD, and four trials included only control (Coccaro and Siever, 1995), which in turn people with acute PTSD. Chronic PTSD is defined raises the question of whether it plays an important as duration of symptoms of 3 months or longer role in the reduction of unwanted, intrusive, (APA, 1994). The four trials based on acute PTSD trauma-related thoughts (Lipper, 1990). One trial included two behaviour therapy trials, an assess- of carbamazepine that was suitable for inclusion ment-only trial, and a trial of relaxation training. was located (mean dose ˆ 661 mg/day).
Seven MAOI treatments were included, consisting commencement of each outcome trial ranged from of phenelzine (n ˆ 6; mean dose ˆ 60 mg/day) and 3 to 12 weeks in the studies of acute PTSD, brofaromine (n ˆ 1; mean dose ˆ 150 mg/day). BDZ compared to approximately 6 years in studies of treatment consisted of a single trial of alprazolam chronic PTSD. Compared to the effect sizes for (mean dose ˆ 3.75 mg/day). Four SSRIs trials chronic PTSD, the effect sizes for the four trials of included fluoxetine (n ˆ 2; mean dose ˆ 60 mg/ acute PTSD were statistical outliers. The large effect day), fluvoxamine (n ˆ 1; mean dose ˆ 150 mg/ sizes for acute PTSD may reflect spontaneous day), and sertraline (n ˆ 1; mean dose ˆ 105 mg/ remission, which is more common in acute than day). All patients in all drug trials were on chronic PTSD (Foa, 1994; Rothbaum et al., 1992). medication when assessed at posttreatment.
Thus, our meta-analysis consisted of 61 trials from Thirteen behavioural therapy trials were included.
39 studies of chronic PTSD. These are listed in They generally entailed some type of exposure Table 1. All treatments were provided in individual therapy (n ˆ 11), with some of these using imaginal format with the exception of one behaviour therapy exposure (n ˆ 4) and others using both imaginal and trial, which used a combination of group and in-vivo exposure (n ˆ 7). Some behavioural therapies individual treatment (Frueh et al., 1996). An also included stress-inoculation training (SIT: n ˆ 3).
appendix listing the studies that were excluded, As mentioned earlier, we examined behavioural and reasons for exclusion, is available on request.
therapies as a group (which included cognitive- Of the 61 trials included in the meta-analysis, behavioural treatments), rather than separately 36 were from studies in which two or more condi- examining each `type' of behavioural intervention.
tions (e.g. TCA versus placebo) were compared. This was because there were insufficient trials to Five studies used crossover designs, where partici- conduct a more fine-grained analysis. Thus, our pants completed one condition followed by a meta-analysis was directed toward examining be- waiting period, and then completed another con- haviour therapy as a class of interventions, which is dition (Shestatzky et al., 1988; Reist et al., 1989; similar to the way in which other meta-analyses Braun et al., 1990; Pitman et al., 1996a; Rothbaum have examined classes or groups of interventions et al., 1996). To avoid the problem of confounding (see, for example, Lipsey and Wilson's (1993) meta- within- and between-subject variance, and to avoid analysis of very broad classes of psychological and possible problems of carry-over effects from one educational interventions).
treatment to another, we included only the first EMDR therapies were also examined as a class of active treatment trial (i.e. drug or psychological therapies, consisting of 11 trials. Although EMDR treatment) from each crossover study. To illustrate, has been modified since it was first described by Shestatzky et al. (1988) used a cross-over design Shapiro (1989), the initially proposed elements of Clin. Psychol. Psychother. 5, 126±144 (1998) Table 1. Trials included in meta-analysis BDI, IES, IPAT, PTSD Index HAM-A, HAM-D, PTSD Sx Doses for drug trials refer to the mean dose attained by the end of treatment. `Ð' refers to data not reported (for % dropouts) or not used (for measures).
*Psychotherapy trial reporting treatment fidelity check.
{Psychotherapy trial reporting level of therapist training.
BDI, Beck Depression Inventory; CAPS, Clinician Administered PTSD Scale; Covi Anx, Covi Anxiety Inventory; DTS, Davidson Self-Rating Trauma Scale; EMDR, Eye Movement Desensitization and Reprocessing; HAM-A, Hamilton Anxiety Scale; HAM-D, Hamilton Depression Scale; IES, Impact of Event Scale; IPAT, IPAT Anxiety Scale; MISS, Mississippi Scale for PTSD; M-PTSD, Modified PTSD Scale; PE, Prolonged Exposure; PENN, Penn Inventory for PTSD; PSS-I, PTSD Symptom ScaleÐInterview Form; Raskin, Raskin Depression Scale; RAST, Rape Aftermath Symptom Test; SCL Anx & Dep, Anxiety and depression scales from the SCL-90-R; SIT, Stress Innoculation Training; SI-PTSD, Structured Interview for PTSD; SRRS, Stress Response Rating Scale; STAI-T, Trait version of State-Trait Anxiety Inventory.
treatment have remained unchanged: i.e. imaginal rather than an effect size defined as the posttreat- exposure with concomitant lateralized move- ment difference between a treatment and control ments, along with coping statements (Shapiro, trial. Thus, we were able to include uncontrolled 1995; personal communication, 1 November 1996). studies in the meta-analysis, thereby increasing the Relaxation therapy included biofeedback-guided number of trials and statistical power to detect relaxation (n ˆ 1). Other treatment trials included differences between treatments. (Note also that we hypnotherapy (n ˆ 1) and psychodynamic therapy were able to determine whether the effect sizes of controlled studies differed from those of uncon- Control groups included pill placebo (n ˆ 4), trolled studies). The different formulae for effect wait-list control (n ˆ 5), a non-saccade control for sizes can differ in the magnitudes of obtained EMDR studies (n ˆ 1), and supportive psycho- effects, and the effect size used in the present study therapy (n ˆ 5). The non-saccade condition is an tends to be larger than effect sizes computed EMDR control in that it includes no eye saccades according to other methods (Abramowitz, 1997a).
nor any other oscillating stimulation. The support- Accordingly, one should not interpret in isolation ive psychotherapy condition included three trials an effect size for a given treatment. The meaning of where participants received standard supportive the effect size is determined by comparing it to the therapy at a VA medical centre, one trial described effect sizes for other treatments.
as a supportive control in which subjects received Regardless of the method of computing effect general therapist support and some teaching in sizes, a further concern that we will examine is the general problem-solving, and one trial in which possibility of inflated effect sizes due to the `file participants met weekly with a social support drawer' problem (Rosenthal, 1979). We use the service team. Four of the five supportive psycho- procedures outlined by Hunter and Schmidt (1990, therapy trials were described by their authors as pp. 512±513) to determine whether this was a control conditions. These trials can therefore be problem for the treatments examined in the present study. The file drawer problem is a publication bias in which studies obtaining significant findings and large effect sizes are published, whereas findings obtaining null results are unpublished. To deter- Effect sizes were calculated according to Cohen's mine the likelihood of this bias, the fail-safe N is (1988) d statistic. For each trial the magnitude of computed (Orwin, 1983), which is the number of change from pre- to posttreatment was defined unpublished trials obtaining zero effect sizes that are required to reduce an obtained mean effect size ‰…SD2pre ‡ SD2post†a2Š. The magnitude of change to a trivial level.
from pre-treatment to follow-up was defined by If a large number of unpublished trials are replacing Mpost with Mfollow-up, and SDpost with required, then it is unlikely that the obtained effect SDfollow-up. For the outcome measures used in is biased by the file drawer problem. The number of the present study, positive effect sizes represent unpublished or unobtained trials obtaining zero improvements in PTSD and other symptoms (i.e. effect sizes is defined by k[(dk/dc) 7 1], where reductions in problem severity), whereas negative k ˆ the number of obtained trials, dk ˆ mean ob- effect sizes indicate a worsening of symptoms. tained effect size, dc ˆ the trivial value to which the Effect sizes were based on completer analyses rather obtained effect would be reduced. Note that dc than end-point or intent-to-treat analyses. In other cannot equal 0, because dk/dc would be undefined.
words, effect sizes were based on pre- and post- Orwin (1983) suggested that a small effect size treatment data for participants completing each (i.e. 0.200) would qualify as a trivial value. How- trial. This was because most trials only reported ever, such an effect size is considered non-trivial by some meta-analysts (e.g. Lipsey and Wilson, 1993).
There are a number of different formulae for Accordingly, we defined a trivial effect size as computing effect sizes (for examples, see Smith et al., 0.050.
1980; Wolf, 1986) and none has been established as a gold standard. We selected the above-mentioned effect size because the same or very similar effect Assessment size formulae are commonly used (e.g. Christensen Previous et al., 1987; Taylor, 1996; Abramowitz, 1997a,b) interviewer-rated scales consistently yield larger and because it provides an effect size for each trial, effect-sizes than self-report scales (e.g. Lambert Clin. Psychol. Psychother. 5, 126±144 (1998) et al., 1986; Taylor, 1995). This may reflect the treatment conditions on important variables greater sensitivity of interviewer-rated scales, or it (e.g. treatment duration) before comparing effect may be an artifact reflecting interviewer bias (i.e. sizes for treatment outcome. In the aggregate, the interviewer typically knows whether or not the treatment and control conditions (as listed in assessment is a pre- or posttreatment evaluation, Table 2) did not differ in the mean duration of and therefore may be biased by expecting compara- their trials, F(13, 47) ˆ 1.86, p 4 0.05 (M ˆ 8.2 tively lower symptom scores at the posttreatment weeks, SD ˆ 4.7 weeks). Note, however, that assessment). Systematic bias in computing effect EMDR trials tended to be shorter than those of sizes can occur if some types of treatments behaviour therapy (M ˆ 3.7 versus 10.1 weeks, are evaluated with interviewer-rated measures respectively; t(22) ˆ 4.66, p 5 0.001), and consisted (e.g. drug therapies), while others are evaluated of fewer treatment sessions (M ˆ 4.6 versus 14.8, with self-report measures (e.g. psychological thera- respectively; t(22) ˆ 5.51, p 5 0.001). We will return pies). Thus, we calculated treatment effect sizes to consider these differences later in this article.
separately for interviewer-rated and self-rated out- The effect sizes of studies that included a control group were compared to effect sizes of The outcome measures used in each study are uncontrolled studies. There was no significant presented in Table 1. As the table shows, scores for difference for either self-report, F(1, 40) ˆ 2.77, self-reported intrusions, avoidance and total PTSD p 4 0.1, or observer-rated measures of total PTSD symptoms were obtained from the Impact of Event symptoms, F(1, 32) ˆ 0.62, p 4 0.1. Among the Scale (Horowitz et al., 1979), the Mississippi Scale psychological therapies, 75% reported the level of for Combat-related PTSD (Keane et al., 1988), and therapist training. Studies were coded as having various DSM-tailored measures such as the PTSD adequate therapist training if they specifically Symptom Checklist (Richards et al., 1994), PTSD reported adequate years of therapist experience Index (Lipper, 1990), and Modified PTSD Scale (e.g. over 5 years) or formal training with a senior (Saunders et al., 1990). Self-reported depression was colleague experienced in the treatment modality. In typically assessed by the Beck Depression Inventory the aggregate (i.e. across treatment conditions), (Beck et al., 1979), and self-reported anxiety was effect sizes did not significantly differ on either typically assessed by the State-Trait Anxiety In- self-report or observer-rated measures for trials that ventory (Spielberger et al., 1970). Scores from other reported therapist training versus those that did not measures were also included in each symptom report on this variable, ps 4 0.1.
domain if the alternative measure adequately Three types of trauma were classified: combat- represented the domain. For example, single flash- related, rape or assault-related, and a category back or nightmare scores were not included in the reflecting a mix of various trauma or another intrusion score as they represent only a portion of trauma. Across the 59 trials that reported trauma type, 51% involved combat-related trauma only, The table also shows that scores for observer- 19% rape or assault-related trauma only, and 30% a rated intrusions, avoidance, and total PTSD symp- mix of trauma or other trauma. Within each toms were typically obtained from the Clinician treatment condition (for conditions with three or Administered PTSD Scale (Blake et al., 1995) or from more trials), mean effect sizes did not significantly the Structured Interview for PTSD (Davidson et al., differ across trauma types, ps 4 0.1.
1989). Observer-rated depression was obtained We intended to examine the relationship between from the Hamilton Rating Scale for Depression degree of psychiatric comorbidity and effect size.
(Hamilton, 1960), and observer-rated anxiety was However, there were insufficient data for such typically assessed by from the Hamilton Rating analysis. Of the 61 trials, only 21% reported a quantifiable level of comorbid anxiety disorders apart from PTSD. A total of 33% of trials reported comorbidity data on mood disorders, and 16% of trials reported comorbidity data for substance use disorders. Thus, we did not examine the relation- ship between effect size and comorbidity.
Follow-up data were not reported for drug therapies. For the control conditions, follow-up A series of preliminary analyses were conducted to data were reported for only one supportive therapy determine whether it was necessary to match trial. For several psychological therapies, follow-up Clin. Psychol. Psychother. 5, 126±144 (1998) data were available only on few outcome measures. 1990). Confidence intervals were computed around Only behaviour therapy and EMDR provided weighted means.
sufficient follow-up data. Analysis of follow-up data was therefore restricted to these two treatment Attrition conditions. Across these conditions, the duration Table 2 shows the proportions of dropouts for between posttreatment and follow-up did not differ each treatment condition. With regard to the three significantly, F(1, 17) ˆ 1.68, p 4 0.1 (grand M ˆ 15 classes of treatment conditions (i.e. drugs treat- weeks; behaviour therapy: M ˆ 18 weeks, SD ˆ 12; ments, psychological therapies, and control con- ditions), the table shows that attrition was significantly greater (i.e. confidence intervals did not overlap) for drug therapies (M ˆ 31.9%) com- pared to psychological therapies (14.0%) and con- trols (16.6%). Within the drug therapy conditions, the dropout rates tended not to differ from one Rather than conducting multiple comparisons another. No differences in dropout rates were (e.g. t-tests) between the 14 treatment conditions, observed within the psychological therapy con- the simplest and most efficient method of evaluat- ditions or the control conditions, or between the ing the comparative efficacy of treatments is to psychological therapies and controls.
compute confidence intervals. Given the small number of trials in each condition for each outcome measure, we chose to use 90th percentile confidence Effect Sizes at Posttreatment: PTSD Symptoms intervals. Confidence intervals have methodological Intrusions (Self-report) advantages over the conventional use of p-values Table 2 shows that for the three classes of (see Cohen, 1990). Accordingly, 90% confidence treatment conditions (drugs treatments, psycho- intervals were calculated for all conditions across all logical therapies, and control conditions), drug measures. Conditions with non-overlapping confi- therapies and psychological therapies were gener- dence intervals differ significantly at p 5 0.10. Some ally equally effective (i.e. their confidence intervals treatment conditions consisted of only a single trial, overlapped), and only psychological therapies were in which case it is not possible to compute significantly superior to controls. Among the drug confidence intervals. However, the effect sizes of therapies, all treatments were superior to WLCs.
these trials could be compared to the confidence Both SSRIs and carbamazepine were superior to intervals of other trials, to determine whether the supportive therapy controls. Only the effect size for obtained effect size fell within the confidence the single carbamazepine was significantly greater interval. If the effect size falls within the interval, than that of pill placebo. SSRIs and carbamazepine then the two conditions do not significantly differ.
had comparable effect sizes, and both were superior For each dependent measure, comparisons were to all other drug therapies.
as follows: (1) general comparisons between the All psychological therapies were significantly overall drug therapy, psychological therapy, and superior to WLCs, and none were significantly control groups; (2) comparison of treatments to better than pill placebo. Behaviour therapy was as controls and to one another (e.g. SSRIs versus effective as other psychological therapies, and controls; SSRIs versus TCAs); (3) comparisons EMDR was superior to relaxation and dynamic across treatment types (e.g. SSRIs versus EMDR); therapy. Across drug therapies and psychological (4) where relevant, comparisons between control therapies, SSRIs were superior to all psycho- conditions (e.g. WLCs and pill placebo versus logical therapies except EMDR and behaviour supportive psychotherapy). Following the recom- therapy. Carbamazepine was similarly superior mendations of Hunter and Schmidt (1990), Wolf to all psychological therapies except behaviour (1986), and others, we made these comparisons by therapy.
computing weighted means. That is, means were computed by weighting the effect size of each trial Intrusions (Observer-rated) by the number of participants completing that trial.
Only some drug therapies (one TCA trial, SSRIs, This procedure gives greater weighting to the effect and one BDZ trial), two of the psychological sizes of larger trials, which are likely to be more therapies (behaviour therapy and EMDR), and reliable estimates of treatment efficacy than the two control conditions (WLCs and one supportive effect sizes of small trials (Hunter and Schmidt, psychotherapy) reported results for observer-rated Clin. Psychol. Psychother. 5, 126±144 (1998) Table 2. Dropout proportions and pre±post effect sizes for measures of PTSD symptoms Effect size ˆ (Mpre 7 Mpost)/SDpooled, where SDpooled ˆ ‰…SD2pre ‡ SD2post†a2Š. All means are weighted by sample size. See text for details. 90%CI ˆ 90th percentile confidence interval arounded weighted mean. Note that `Ð' refers to data missing or not reported. For the 90%CIs `Ð' appears when there was only one effect size. Within each row, total number of trials may differ across outcome domains (intrusions, avoidance, and global severity) because some trials did not assess all domains.
BDZ, benzodiazepines; Behav Tx, behaviour therapy; Carbmz, carbamazepine; Dynamic, psychodynamic psychotherapy; EMDR, eye movement desensitization and reprocessing; MAOI, monoamine oxidase inhibitors; No Sacc, no saccade control (control for EMDR); Pill Plac, pill placebo; Relaxat'n, relaxation training; SSRI, selective serotonin reuptake inhibitors; Sup Psych, supportive psychotherapy; TCA, tricyclic antidepressants; WLC, waiting-list control.
measures. Among these trials, the drug therapies more effective than controls. Among the drug and the psychological therapies were generally therapies, only the SSRIs and carbamazepine were equally effective, although there was a trend for more effective than all control conditions. The SSRIs psychological therapies to have larger effect sizes had a significantly greater mean effect size than all (Table 2). Drug and psychological therapies were other drug therapies. Among the psychological more effective than controls. Within the drug therapies, all but relaxation were significantly therapies, only SSRIs were more effective than all more effective than all control groups. EMDR and controls, and SSRIs were also more effective than all behaviour therapy were equally effective, but only other drug therapies. Within the psychological EMDR was significantly superior to all other therapies, EMDR and behaviour therapy demon- psychological therapies. Behaviour therapy and strated comparable effect sizes, but only EMDR was EMDR were equally effective as the SSRIs.
significantly more effective than all controls. EMDR was more effective than the most effective drug Total PTSD Symptoms (Observer-rated) As seen in Table 2, outcome data for total PTSD symptoms were reported by all drug therapies and all controls except the non-saccade control.
Table 2 shows that psychological therapies were However, these data were reported by only two more effective than both drug therapies and control psychological therapies (behaviour therapy and conditions. All drug therapies were more effective EMDR). Psychological therapies and drug therapies than pill placebo and WLCs, and none were more were equally effective, although there was a trend effective than the single supportive therapy control. for psychological therapies to have larger effect SSRIs were significantly more effective than all sizes. Drug and psychological therapies were other drug therapies. All psychological therapies more effective than controls. Among the drug were more effective than pill placebo and WLCs, therapies, only the SSRIs and carbamazepine were but again, not more effective than the single more effective than all controls, and both were also supportive psychotherapy trial. Among the psycho- more effective than all other drug therapies. Among logical therapies, behaviour therapy and EMDR the psychological therapies, only behaviour therapy were equally effective, but only EMDR was was more effective than all controls, and behaviour superior to relaxation and dynamic therapy. therapy was also more effective than EMDR. The SSRIs, EMDR, behaviour therapy and hypnother- best psychological therapy, behaviour therapy, was significantly more effective than carbamazepine, and there was a trend for behaviour therapy to be Only some of the drug therapies (one TCA trial, SSRIs, and one BDZ trial), two of the psychological therapies (behaviour therapy and EMDR), and two control conditions (WLCs and one supportive Effect Sizes at Posttreatment: psychotherapy) reported on avoidance. Drug thera- Anxiety and Depression pies and psychological therapies were more effec- Anxiety (Self-report) tive than controls, and were equally effective to one Psychological therapies were more effective than another (Table 2). However, there was a trend for drug therapies, and both were more effective than psychological therapies to have a larger effect size. controls (Table 3). All psychological therapies were Among the drug therapies, all were significantly more effective than all controls. All drug therapies more effective than supportive psychotherapy, and except TCAs were superior to all controls. The SSRI there was a trend for SSRIs to also be more effective trial tended to be superior to all other drug than WLCs. Among the psychological therapies, therapies. Within the psychological therapies, the EMDR and behaviour therapy were equally effec- largest effect sizes were observed in the EMDR and tive, but only EMDR was more effective than the behaviour therapy conditions. Of these conditions, control conditions. SSRIs, behaviour therapy, and only behaviour therapy was more effective than relaxation therapy. Across drug therapies and psychotherapies, SSRIs and behaviour therapy were equal in efficacy. The SSRI trial was compar- As seen in Table 2, psychological therapies were able to behaviour therapy, but more effective than more effective than drug therapies, and both were EMDR and other psychological therapies. EMDR Clin. Psychol. Psychother. 5, 126±144 (1998) Table 3. Pre±post effect sizes for measures of anxiety and depression See footnote to Table 2 for a definition of statistics and acronyms.
yielded a larger effect size than the non-saccade the relaxation condition. The single SSRI trial was significantly more effective than the best psycho- logical therapies, EMDR and behaviour therapy.
As seen in Table 3, among the psychological Depression (Observer-rated) therapies, only one trial of behaviour therapy No psychological therapies reported observer- reported on this measure, and only one pill placebo rated depression, and only two pill placebo trials trial was available among controls. The behaviour reported outcomes among the control conditions.
therapy trial demonstrated a significantly greater All drug therapies except the BDZ trial were more effectiveness than the drug therapies, and both were effective than pill placebo. Among the drug more effective than pill placebo. Among the drug therapies, the SSRIs were more effective than therapies, the carbamazepine trial and SSRI trial carbamazepine, and both SSRIs and carbamazepine were more effective than all other drug therapies.
had larger effect sizes than MAOIs and TCAs.
Table 3 shows that the drug therapies and Posttreatment: Summary of Results psychological therapies were generally comparable Psychological therapies tended to be more effective in efficacy, although there was a trend for psycho- than drug therapies, and both tended to be more logical therapies to have larger effect sizes. Both effective than controls. Of the drug therapies, were more effective than the control conditions. All the SSRIs and carbamazepine were the most drug therapies except TCAs were superior to all effective, although the carbamazepine condition controls. The one SSRI trial tended to be more was based on only a single trial. SSRIs tended to effective than all other drug therapies. All psycho- be more effective in treating intrusions than logical therapies were more effective than all control avoidance symptoms according to self-report but conditions. Behaviour therapy and EMDR were not observer-rated measures. Other drug therapies equally effective, and both were more effective than generally demonstrated small effect sizes in relation Clin. Psychol. Psychother. 5, 126±144 (1998) to control conditions, and affected only a few of the Effect Sizes at Follow-up symptom domains. Of the psychological therapies, behaviour therapy and EMDR were the most No drug therapy, pill placebo, or wait-list control effective, with the two being generally equally results were available for follow-up, and only single efficacious, although behaviour therapy was signifi- trials of other psychological therapies and the cantly more effective than all treatments, including supportive psychotherapy trial provided follow-up EMDR, SSRIs, and carbamazepine, on observer- data, and only across some symptom measures.
rated total PTSD symptoms. No differences in Therefore the only conditions which provided comparative treatment efficacy were discernible adequate trials for calculating follow-up effect between behaviour therapy and EMDR across the sizes across most symptom domains were beha- viour therapy and EMDR. Table 4 shows the follow- The less effective psychological therapies demon- up effect sizes for PTSD symptoms, and Table 5 strated moderate yet consistent effect sizes across shows the results for measures of anxiety and symptom domains in relation to control conditions, depression. As noted in the section titled `Prelimi- with relaxation being the least effective, followed by nary analyses', the mean duration between post- dynamic therapy and lastly, hypnotherapy. How- treatment and follow-up did not differ significantly ever, these conditions were all based upon single across conditions (mean follow-up duration ˆ 15 trials, and therefore should be interpreted with weeks).
caution. Supportive psychotherapies tended to Across all self-report and observer-rated measures demonstrate efficacy comparable to the less effec- of PTSD symptoms, depression and anxiety, both tive psychological therapies and drug therapies.
behaviour therapy and EMDR demonstrated a The best psychological therapies, behaviour maintenance of treatment effects at follow-up, therapy and EMDR, tended to be as effective as the with no differences between the two conditions at best drug therapies, SSRIs and carbamazepine, follow-up on any measures. Differences in effect size across PTSD outcomes domains. However, as noted from posttreatment to follow-up were nonsignificant above, behaviour therapy was more effective than all for all measures across both conditions, except that treatments on observer-rated total PTSD symptoms. EMDR demonstrated a significant increase in effect SSRIs may have been more effective than behaviour size for observer-rated total PTSD symptoms at therapy and EMDR in treating depression, but this follow-up, making it equal to behaviour therapy, was based upon a single SSRI trial, and results were whereas EMDR was less effective than behaviour only available for self-reported depression.
therapy for this measure at posttreatment.
Table 4. Effect sizes at follow-up (i.e. symptom reductions from pretreatment to 15 week follow-up) for PTSD See footnote to Table 2 for a definition of statistics and acronyms.
Table 5. Effect sizes at follow-up for measures of anxiety and depression See footnote to Table 2 for a definition of statistics and acronyms.
Clin. Psychol. Psychother. 5, 126±144 (1998) Although dropout rates were high for drug Recall that fail-safe N is defined as the number of therapies, the SSRIs demonstrated the greatest unpublished null trials (i.e. those obtaining zero treatment efficacy of all drug therapies (other than effect sizes) required to reduce an obtained mean carbamazepine, for which there was only one trial), effect size to a trivial magnitude (0.050). An average with large effect sizes apparent across all symptom of self-reported and observer-rated total PTSD domains. Effects were perhaps stronger for intru- sive symptoms and depressive symptoms than for obtained effect size) in calculating fail-safe N avoidance symptoms, but all effect sizes were large calculations. For the posttreatment data, the necess- relative to control conditions. However, posttreat- ary number of unpublished null trials in each active ment assessments for drug therapies occurred prior treatment condition were as follows: TCA (41), to medication discontinuation, so it is not clear carbamazapine (23), MAOI (66), SSRI (95), BDZ whether treatment effects maintain when medi- (10), behaviour therapy (220), EMDR (139), relaxa- cations are withdrawn.
tion therapy (8), hypnotherapy (18), and dynamic These results suggest that although SSRIs may therapy (17). These results suggest that for most not be the treatment of choice for PTSD, given the conditions, there would need to be a large number higher dropout rate, they may still be an acceptable of unpublished null trials to reduce obtained mean and useful treatment if the patient can be retained effect sizes to trivial levels. It seems unlikely, in treatment, and may be particularly useful for especially for SSRIs, EMDR, and behaviour therapy, patients whose intrusive and depressive symp- that there would be so many unpublished trials toms predominate their experience. Unfortunately, finding zero effects for these treatments, and so we follow-up data were unavailable for these trials, so conclude that these findings were unlikely to have it is unclear as to whether effects were maintained been biased by the `file-drawer' problem. However, over time, or even immediately after medication significantly fewer unpublished null trials were discontinuation. Further research is warranted to indicated to suggest bias in results for the carba- investigate the maintenance of SSRI effects in PTSD mazepine, BDZ, relaxation, hypnotherapy, and patients. Carbamazepine also appeared to be an dynamic therapy conditions. Therefore these con- effective treatment for PTSD, but results were based ditions should be viewed cautiously as they may be upon a single trial and should therefore be Only one BDZ trial was available for this meta- analysis. This study used alprazolam at a moder- ately high dose (1.5 to 6 mg/day) and generally failed to demonstrate treatment efficacy (Braun et al., 1990). Effect sizes were small if not trivial relative to The purpose of the present study was to empirically control conditions. Although not included in our evaluate the relative efficacy of various treatments meta-analysis due to failure to meet our 100% PTSD for PTSD. We found that behaviour therapy and diagnosis rule, another study also supports the EMDR were the most effective psychological inefficacy of BDZs for PTSD. Gelpin et al. (1996) therapies for PTSD. Effect sizes for these therapies reported that clonazepam and alprazolam failed to were large relative to control conditions, indicating benefit acute trauma survivors' PTSD and anxiety strong treatment effects, and dropout rates were symptoms over controls. In another study not low, indicating good treatment acceptance. The included in the meta-analysis due to lack of a SSRIs and carbamazepine were the most effective standard posttreatment assessment time across drug therapies, with similarly large effect sizes, but patients, alprazolam did not alter auditory startle dropout was fairly high (Table 2). Even SSRIs, response in PTSD patients (Bloch et al., 1996). This is characteristically more acceptable to patients due to enlightening given the high frequency of BDZ a more tolerable side-effect profile, were tolerated prescriptions by general practitioners for anxiety poorly by the PTSD patients. An average of 36% of disorders. If medications are to be prescribed for patients treated with SSRIs discontinued treat- PTSD treatment, it appears as though SSRIs are a ment prematurely. It may be that PTSD patients, more effective treatment than BDZs. The more characteristically hyperaroused, may be particularly narrow side-effect profile, non-addictiveness, and sensitive to side-effects occasionally associated with greater overdose threshold of SSRIs over BDZs also SSRIs (e.g. agitation), and may discontinue second- support this point (Canadian Pharmaceutical Clin. Psychol. Psychother. 5, 126±144 (1998) As noted, behaviour therapy and EMDR were the sizes of EMDR tended to be larger than those of most effective psychological therapies, and both control conditions, such as pill placebo and suppor- were as effective as SSRIs. Effect sizes were large tive psychotherapy.
across all PTSD symptom domains for these Regarding the question of active ingredients in treatments in relation to controls, and treatments EMDR, even the utility of the actual eye-movements were generally statistically comparable in efficacy, in EMDR is unclear, since other stimuli are used with some minor exceptions. For example, beha- clinically. Only one non-saccade trial was available viour therapy was more effective than EMDR and for comparison with the EMDR trials in our meta- SSRIs on observer-rated total PTSD symptoms at analysis (Devilly and Spence, 1996), therefore posttreatment. However, by follow-up, the differ- permitting little comment on this issue. When all ences between behaviour therapy and EMDR were eye-movement conditions in the meta-analysis were nonsignificant (SSRIs did not report follow-up compared to this one fixed-eye condition, EMDR data). Because attrition from posttreatment to was more effective than the fixed-eye control across follow-up was comparable across EMDR and measures. However, when the fixed-eye control was behaviour therapy conditions, the `catching up' of compared to the EMDR condition within the same EMDR to behaviour therapy cannot be attributed to study (i.e. Devilly and Spence, 1996), the fixed-eye differential attrition of poor responders. No condition was comparable to EMDR across differences in treatment efficacy between behaviour measures. However, that study may have been therapy and EMDR were noted across the specific compromised by insufficient treatment duration PTSD symptom domains. Both treatments main- because only two EMDR sessions for multiply tained effects at follow-up, and were equally traumatized sessions were used. Thus, there may effective on all measures at follow-up.
not have been sufficient time to treat each traumatic The efficacy of EMDR in PTSD treatment is a memory.
controversial finding in light of recent discussions in Pitman and colleagues (1996b) also included a the literature critically questioning the validity of fixed-eye control (that included hand-tapping) in EMDR as a treatment for anxiety disorders (Herbert their EMDR study. This was not included in the and Mueser, 1992; Lohr et al., 1995; Lilienfeld, 1996). meta-analysis because it was tested within a The results of the present study suggest that EMDR crossover design for which we already included is effective for PTSD, and that it is more efficient the EMDR treatment phase. Pitman and colleagues than other treatments. Despite its apparent efficacy, report that the fixed-eye condition performed as what works in EMDR and the mechanism for how well if not better than the EMDR trial. Another it works remains unclear. That is, we know little study not included in the meta-analysis (due to the about the active ingredients in EMDR and the fact that not all patients had full PTSD) reported mechanisms by which these ingredients result in that a fixed-eye control demonstrated large treat- decreased PTSD symptoms. EMDR theory suggests ment effect sizes comparable to that of the EMDR that eye movements or other oscillatory move- condition, although the eye movement condition ments during trauma imagining somehow result in was more efficient (Renfrey and Spates, 1994).
brain alterations which then allow more appropriate This study also demonstrated that a condition processing of the trauma, thereby reducing PTSD using light bars versus a therapist's finger to symptoms. This still does not inform us of what stimulate eye-movements was also comparable to changes occur in what part of the brain, how the other conditions, suggesting that alternate oscillatory movements are involved in those means of eye-movement production may be com- changes, how that leads to `reprocessing' of the parable to the procedures originally described for trauma, and how such reprocessing results in EMDR. Clarification of the mechanisms by which decreased PTSD symptoms. Some might argue symptoms change and the active ingredients in that EMDR works through exposure and desensiti- EMDR is now critical given its apparent efficacy.
zation, similar to behaviour therapy. However, this Without such clarification, the acceptability of is unlikely to be the case given that EMDR provides EMDR within the professional community is likely significantly less trauma exposure than behaviour to remain controversial.
therapy and is demonstrating comparable effects, Recent efforts in treatment outcome research have which suggests that another treatment component been made to examine both separate and combined specific to EMDR is active. Expectancy may play a treatments for various disorders. In a recent meta- role, but this has not been thoroughly examined, analysis of treatments for panic disorder, Gould and is inconsistent with our finding that the effect et al. (1995) reported that the combination of Clin. Psychol. Psychother. 5, 126±144 (1998) benzodiazepine treatment with behaviour therapy Marcus, Pitman, Rothbaum, Scheck, Shapiro and is less effective than behaviour therapy alone. Wilson for providing unpublished data.
Future research is needed to determine whether a similar finding holds for PTSD. That is, whether combined benzodiazepine therapy and psycho- REFERENCES logical therapy is less effective than the latter alone. Another direction for future research Abramowitz, J. S. (1997a). Effectiveness of psycho- suggested by the results of this meta-analysis is logical and pharmacological treatments for obsessive- compulsive disorder: A quantitative review. Journal of the study of the effectiveness of combined psycho- Consulting and Clinical Psychology, 65, 44±52.
logical therapies (e.g. combined behaviour therapy Abramowitz, J. S. (1997b). Variants of exposure and and EMDR) versus individual psychological thera- response prevention in the treatment of obsessive- pies. Finally, we focused our analysis on studies of compulsive disorder: A meta-analysis. Behaviour fully diagnosed chronic PTSD in adults. Future studies examining the comparability of treatment American Psychiatric Association. (1994). Diagnostic and Statistical Manual of Mental Disorders, 4th edn.
efficacy in subclinical PTSD and full PTSD in children would also enhance our understanding of Baker, D. G., Diamond, B. I., Gillette, G., Hammer, M., the treatment of psychological trauma.
Katzelnick, D., Keller, T., Mellman, T. A., Pontius, E., A limitation of this paper is the relatively low Rosenthal, M., Tucker, P., van der Kolk, B. A. and number of trials available per treatment condition, Katz, R. (1995). A double-blind, randomized, placebo- and our corresponding use of 90% rather than 95% controlled, multi-center study of brofaromine in the treatment of post-traumatic stress disorder. Psychophar- confidence intervals. Yet despite the small number of trials, the number of trials was sufficient to detect Beck, A. T., Rush, A. J., Shaw, B. and Emery, G. (1979).
differences in effect sizes across various treatment Cognitive Therapy of Depression. New York: Guilford.
conditions. Even so, more research trials on PTSD Blake, D. D., Weathers, F. W., Nagy, L. M., Kaloupek, treatment outcome would likely increase the power D. G., Gusman, F. D., Charney, D. S. and Keane, T.
of the meta-analysis and would also permit more (1995). The development of a clinician-administered fine-grained analyses. Future research should PTSD Scale. Journal of Traumatic Stress, 8, 75±90.
Bloch, M., Peri, T., Bonne, O. and Shalev, A. Y. (1996).
address this issue. Another limitation that was Effect of alprazolam on auditory startle in PTSD and true for our study and is also common to many panic disorder. Unpublished manuscript, Department of meta-analyses is that we were only able to base our Psychiatry, Hadassah University Hospital, Jerusalem, results on completer-analyses. It remains to be seen whether our conclusions hold up for intent-to-treat Braun, P., Greenberg, D., Dasberg, H. and Lerer, B. (1990).
analyses. Although effect sizes should be smaller Core symptoms of post-traumatic stress disorder unimproved by alprazolam treatment. Journal of Clinical for intent-to-treat analyses, we expect that most of these findings would maintain. One exception Breslau, N., Davis, G. C., Andreski, P. and Peterson, E.
is that the effect sizes for drug treatments (1991). Traumatic events and post-traumatic stress may diminish more than those for psychological disorder in an urban population of young adults.
therapies, because the former had more dropouts, Archives of General Psychiatry, 48, 216±222.
and dropouts may have poorer treatment response. Brom, D., Kleber, R. J. and Defares, P. B. (1989).
Brief Psychotherapy for post-traumatic stress disorders.
To conclude, this meta-analysis provides new Journal of Consulting and Clinical Psychology, 57, 607±612.
information about the relative efficacy of treatments Burstein, A. (1984). Treatment of post-traumatic stress for PTSD. Our results support the use of behaviour disorder with imipramine. Psychosomatics, 25, 681±687.
therapy, EMDR, and SSRIs. It remains to be seen Canadian Pharmaceutical Association (1996). Compendium whether the efficacy of treatment can be improved of Pharmaceuticals and Specialties, 31st edn. Ottawa, by using these interventions in combination.
Carlson, J. G., Chemtob, C. M., Rusnak, K., Hedlund, N. L.
and Muraoka, M. Y. (in press). Eye-movement desensi- tization and reprocessing (EMDR) treatment for com- bat-related post-traumatic stress disorder. Journal of This research was supported in part by a grant from Chemtob, C., Roitblat, H. L., Hamada, R. S., Carlson, J. G.
the Medical Research Council of Canada and a and Twentyman, C. T. (1988). A cognitive action theory of post-traumatic stress disorder. Journal of Anxiety training grant from the National Institute on Drug Abuse. Thanks to Drs Baker, Blanchard, Burstein, Christensen, H., Hadzi-Pavlovic, D., Andrews, G. and Carlson, Devilly, Foa, Frueh, Lazrove, Lipper, Lohr, Mattick, R. (1987). Behaviour therapy and tricyclic Clin. Psychol. Psychother. 5, 126±144 (1998) medication in the treatment of obsessive-compulsive Journal of Consulting and Clinical Psychology, 59, disorder: A quantitative review. Journal of Consulting and Clinical Psychology, 55, 701±711.
Foa, E. B., Steketee, G. and Rothbaum, B. O. (1989).
Coccaro, E. F. and Siever, L. J. (1995). The neuropsycho- Behavioural/cognitive conceptualizations of post- pharmacology of personality disorders. In: F. E. Bloom traumatic stress disorder. Behaviour Therapy, 20, and D. J. Kupfer (Eds), Psychopharmacology: The Fourth Generation of Progress. New York: Raven, pp. 1567±1579.
Foa, E. B., Zinbarg, R. and Rothbaum, B. O. (1992).
Cohen, J. (1988). Statistical Power Analysis for the Beha- Uncontrollability and unpredictability in post-traumatic vioural Sciences, 2nd edn. Hillsdale, NJ: Erlbaum.
stress disorder: An animal model. Psychological Bulletin, Cohen, J. (1990). Things I have learned (so far). American Forbes, D., Creamer, M. and Rycroft, P. (1994). Eye Cooper, N. A. and Clum, G. A. (1989). Imaginal flooding movement desensitization and reprocessing in post- as a supplementary treatment for PTSD in combat traumatic stress disorder: A pilot study using assess- veterans: A controlled study. Behaviour Therapy, 20, ment measures. Journal of Behaviour Therapy and Experimental Psychiatry, 25, 113±120.
Davidson, J. (1987). A pilot study of phenelzine in the Friedman, J. J. (1991). Biological approaches to the treatment of post-traumatic stress disorder. British diagnosis and treatment of post-traumatic stress dis- Journal of Psychiatry, 150, 252±255.
order. Journal of Traumatic Stress, 4, 67±91.
Davidson, J., Foa, E. B., Blank, A. S., Brett, E. A., Frueh, B. C., Turner, S. M., Beidel, D. C., Mirabella, R. F. and Fairbank, J., Green, B. L., Herman, J. L., Keane, T. M., Jones, W. J. (1996). Trauma management therapy: A Kilpatrick, D. L., March, J. S., McNally, R. J., Pitman, preliminary evaluation of a multicomponent behavioral R. K., Resnick, H. S. and Rothbaum, B. O. (1996). Post- treatment for chronic combat-related PTSD. Behaviour traumatic stress disorder. In: T. A. Widiger, A. J. Frances, H. A. Pincus, R. Ross, M. B. First and W. W. Davis Gelpin, E., Bonne, O., Peri, T., Brandes, D. and Shalev, A. Y.
(Eds), DSM-IV Sourcebook, Vol. 2. Washington, DC: (1996). Treatment of recent trauma survivors with American Psychiatric Association, pp. 577±605.
benzodiazepines: A prospective study. Unpublished Davidson, J. R. T., Hughes, D., Blazer, D. G. and George, manuscript, Department of Psychiatry, Hadassah Uni- L. G. (1991). Post-traumatic stress disorder in the versity Hospital, Jerusalem, Israel.
community: An epidemiological study. Psychological Gould, R. A., Otto, M. W. and Pollack (1995). A meta- analysis of treatment outcome for panic disorder.
Davidson, J., Kudler, H., Smith, R., Mahorney, S. L., Clinical Psychology Review, 15, 819±844.
Lipper, S., Hammett, E., Saunders, W. B. and Cavenar, Hamilton, M. (1959). The assessment of anxiety states by J. O. (1990). Treatment of post-traumatic stress disorder rating. British Journal of Medical Psychology, 32, 50±55.
with amitriptyline and placebo. Archives of General Hamilton, M. (1960). A rating scale for depression.
Journal of Neurology, Neurosurgery and Psychiatry, 23, Davidson, J., Smith, R. and Kudler, H. (1989). Validity and reliability of the DSM-III criteria for post-traumatic Herbert, J. D. and Mueser, K. (1992). Eye movement stress disorder: Experience with a structured interview.
desensitization: A critique of the evidence. Journal Journal of Nervous and Mental Disease, 177, 336±341.
of Behaviour Therapy and Experimental Psychiatry, 23, Devilly, G. J. and Spence, S. (1996). The clinical and statistical efficacy of eye movement desensitization Hertzberg, M. A., Feldman, M. E., Beckham, J. C. and and reprocessing: Treating PTSD within a veteran Davidson, R. T. (1996). Trial of trazadone for post- population. Unpublished manuscript, Department of traumatic stress disorder using a multiple baseline Psychology, University of Queensland, Australia.
group design. Journal of Clinical Psychopharmacology, 16, Foa, E. B. (1994, July). Psychopathology and treatment of PTSD in rape victims. Paper presented at the 102nd Hickling, E. J. and Blanchard, E. B. (1997). The private Annual Convention of the American Psychological psychologist and manual-based treatments: Post-trau- matic stress disorder secondary to motor vehicle Foa, E. B. and Kozak, M. J. (1986). Emotional processing accidents. Behaviour Research and Therapy, 35, 191±203.
of fear: Exposure to corrective information. Psychological Horowitz, M., Wilner, N. and Alvarez, W. (1979). Impact of Event Scale: A measure of subjective stress.
Foa, E. B., Freund, B. F., Hembree, E., Dancu, C. V., Psychological Medicine, 41, 209±218.
Franklin, M. E., Perry, K. J., Riggs, D. S. and Molnar, Hunter, J. and Schmidt, F. (1990). Methods of Meta C. (1996). Efficacy of short term behavioural treatments of PTSD in sexual and nonsexual assault victims.
Jensen, J. A. (1994). An investigation of eye movement Unpublished manuscript, Medical College of Pennsyl- desensitization and reprocessing (EMD/R) as a treat- vania at Eastern Pennsylvania Psychiatric Institute, ment for post-traumatic stress disorder (PTSD) symp- toms of Vietnam combat veterans. Behaviour Therapy, Foa, E. B., Rothbaum, B. O., Riggs, D. S. and Murdock, T. B. (1991). Treatment of post-traumatic Kauffman, C. D., Reist, C., Djenderedjian, A., Nelson, J. N.
stress disorder in rape victims: A comparison between and Haier, R. J. (1987). Biological markers of affective cognitive-behavioural procedures and counseling.
disorders and post-traumatic stress disorder: A pilot Clin. Psychol. Psychother. 5, 126±144 (1998) study with desipramine. Journal of Clinical Psychiatry, T. (1996). Open trial of fluvoxamine treatment for combat-related post-traumatic stress disorder. Journal of Keane, T. M., Fairbank, J. A., Caddell, J. M. and Zimering, Clinical Psychiatry, 57 (Suppl 8), 66±72.
R. T. (1989). Implosive (flooding) therapy reduces McNally, R. J. (1996). Review of `Eye movement desensi- symptoms of PTSD in Vietnam combat veterans.
tization and reprocessing: Basic principles, protocols, and procedures' by F. Shapiro. Anxiety, 2, 153±155.
Keane, T. M., Caddell, J. M. and Taylor, K. L. (1988).
Mollica, R. F., Wyshak, G., Lavelle, J., Truong, T., Tor, S.
Mississippi scale for combat-related post-traumatic and Yang, T. (1990). Assessing symptom change in stress disorder: Three studies in reliability and validity.
Southeast Asian refugee survivors of mass violence and Journal of Consulting and Clinical Psychology, 56, 85±90.
torture. American Journal of Psychiatry, 147, 83±88.
Keane, T. M., Zimering, R. T. and Caddell, J. M. (1985). A Montgomery, R. W. and Ayllon, T. (1994). Eye move- behavioural formulation of post-traumatic stress dis- ment desensitization across subjects: Subjective and order. The Behaviour Therapist, 8, 9±12.
physiological measures of treatment efficacy. Journal Kosten, T. R., Frank, J. B., Dan, E., McDougle, C. J. and of Behaviour Therapy and Experimental Psychiatry, 25, Giller, E. L. (1991). Drug therapy for post-traumatic stress disorder using phenelzine or imipramine. Journal Mowrer, O. H. (1960). Learning Theory and Behaviour.
of Nervous and Mental Disease, 179, 366±370.
Lambert, M., Hatch, D., Kingston, M. and Edwards, B.
Nagy, L. M., Morgan, C. A., Southwick, S. M. and (1986). Zung, Beck, and Hamilton rating scales as Charney, D. S. (1993). Open prospective trial of measures of treatment outcome: A meta-analytic fluoxetine for post-traumatic stress disorder. Journal of comparison. Journal of Consulting and Clinical Psychol- Clinical Psychopharmacology, 13, 107±113.
Norris, F. H. (1992). Epidemiology of trauma: frequency Lazrove, S., Triffleman, E., Kite, L., McGlashan, T. and and impact of different potentially traumatic events on Rounsaville, B. (1996). The use of EMDR as treatment different demographic groups. Journal of Consulting and for chronic PTSD: Encouraging results of an open trial.
Unpublished manuscript, Department of Psychiatry, Orwin, R. (1983). A fail-safe N for effect size. Journal of Educational Statistics, 8, 147±159.
Lerer, B., Bleich, A., Kotler, M., Garb, R., Hertzberg, M.
Otto, M. W., Penava, S. J., Pollack, S. J. and Smoller, and Levin, B. (1987). Post-traumatic stress disorder in J. W. (1996). Cognitive-behavioural and pharmacologic Israeli combat veterans. Archives of General Psychiatry, perspectives on the treatment of post-traumatic stress disorder. In: M. H. Pollack, M. W. Otto and Lilienfeld, S. (1996). EMDR treatment: Less than meets the J. F. Rosenbaum (Eds), Challenges in Clinical Practice: eye? The Skeptical Inquirer, 20, 25±31.
Pharmacologic and Psychosocial Strategies. New York: Lipper, S. (1990). Carbamazepine in the treatment of post- traumatic stress disorder: Implications for the kindling Pitman, R. K., Orr, S. P., Altman, B., Longpre, R. E., hypothesis. In: M. E. Wolf and A. D. Mosnaim (Eds), Poire, R. E., Macklin, M. L., Michaels, M. J. and Post-traumatic Stress Disorder: Etiology, Phenomenology, Steketee, G. S. (1996a). Emotional processing and and Treatment. Washington, DC: American Psychiatric outcome of imaginal flooding therapy in Vietnam veterans with chronic post-traumatic stress disorder.
Lipsey, M. and Wilson, D. (1993). The efficacy of Comprehensive Psychiatry, 37, 409±418.
psychological, educational, and behavioural treatment: Pitman, R. K., Orr, S. P., Altman, B., Longpre, R. E., Poire, Confirmation from meta-analysis. American Psycho- R. E. and Macklin, M. L. (1996b). Emotional processing during eye-movement desensitization and reprocessing Litz, B. T. (1992). Emotional numbing in combat-related therapy of Vietnam veterans with chronic post- post-traumatic stress disorder: A critical review and traumatic stress disorder. Comprehensive Psychiatry, 37, reformulation. Clinical Psychology Review, 12, 417±432.
Litz, B. T. and Keane, T. M. (1989). Information processing Reist, C., Kauffman, C. D., Haier, R. J., Sangdahl, C., in anxiety disorders: Application to the understanding DeMet, E. M., Chicz-DeMet, A. and Nelson, J. N.
of post-traumatic stress disorder. Clinical Psychology (1989). A controlled trial of desipramine in 18 men with post-traumatic stress disorder. American Journal of Lohr, J., Kleinknecht, R. A., Tolin, D. F. and Barrett, R. H.
(1995). The empirical status of the clinical application of Renfrey, G. and Spates, C.R. (1994). Eye movement eye movement desensitization and reprocessing. Journal desensitization: A partial dismantling study. Journal of Behaviour Therapy and Experimental Psychiatry, 26, of Behaviour Therapy and Experimental Psychiatry, 25, Marcus, S., Marquis, P. and Sakai, C. (1996). Eye move- Rescorla, R. A. (1988). Pavlovian conditioning: It's not ment desensitization and reprocessing: A clinical out- what you think it is. American Psychologist, 43, come study for post-traumatic stress disorder.
Unpublished manuscript, Kaiser Adult Psychiatry, Richards, D. A., Lovell, K. and Marks, I. M. (1994).
Post-traumatic stress disorder: Evaluation of a Marmar, C. R., Schoenfeld, F., Weiss, D. S., Metzler, T., behavioural treatment program. Journal of Traumatic Zatzick, D., Wu, R., Smiga, S., Tecott, L. and Neylan, Clin. Psychol. Psychother. 5, 126±144 (1998) Rosenthal, R. (1979). The `file drawer' problem and Sutherland, S. M. and Davidson, J. R. T. (1994). Drug tolerance for null results. Psychological Bulletin, 86, therapy for post-traumatic stress disorders. Psychiatric Clinics of North America, 17, 409±423.
Rothbaum, B. O. (in press) A controlled study of EMDR Taylor, S. (1995). Assessment of obsessions and for PTSD. Bulletin of the Menninger Clinic.
compulsions: Reliability, validity, and sensitivity Rothbaum, B. O., Foa, E. B., Riggs, D., Murdock, T. and to treatment effects. Clinical Psychology Review, 15, Walsh, W. (1992). A prospective examination of post- traumatic stress disorder in rape victims. Journal of Taylor, S. (1996). Meta-analysis of cognitive-behavioural treatments for social phobia. Journal of Behaviour Rothbaum, B. O., Ninan, P. T. and Thomas, L. (1996).
Therapy and Experimental Psychiatry, 27, 1±9.
Sertraline in the treatment of rape victims with post- van der Kolk, B. (1987). The drug treatment of post- traumatic stress disorder. Journal of Traumatic Stress, 9, traumatic stress disorder. Journal of Affective Disorders, Saunders, B. E., Arata, C. M. and Kilpatrick, D. G.
van der Kolk, B. A., Dreyfuss, D., Michaels, M., Shera, D., (1990). Development of a crime-related post-traumatic Berkowitz, R., Fisler, R. and Saxe, G. (1994). Fluoxetine stress disorder scale for women within the Symptom in post-traumatic stress disorder. Journal of Clinical Checklist-90-Revised. Journal of Traumatic Stress, 3, Vaughan, K., Armstrong, M. S., Gold, R., O' Connor, N., Scheck, M. M., Schaffer, J. A. and Gillette, C. S. (in press).
Jenneke, W. and Tarrier, N. (1994). A trial of eye Brief psychological intervention with young high-risk movement desensitization compared to image habitu- females: A comparison of eye-movement desensitiza- ation training and applied muscle relaxation in post- tion and reprocessing with active reflective listening.
traumatic stress disorder. Journal of Behaviour Therapy and Experimental Psychiatry, 25, 283±291.
Shapiro, F. (1989). Eye movement desensitization: A new Vaughan, K. and Tarrier, N. (1992). The use of image treatment for post-traumatic stress disorder. Journal habituation training with post-traumatic stress dis- of Behavior Therapy and Experimental Psychiatry, 20, orders. British Journal of Psychiatry, 161, 658±664.
Veronen, L. J., Kilpatrick, D. G. and Resick, P. A.
Shapiro, F. (1991). Eye movement desensitization and (1978). Stress inoculation training for victims of rape.
reprocessing procedure: From EMD to EMD/R:A, Paper presented at the annual meeting of the Associ- new treatment model for anxiety and related trauma.
ation for Advancement of Behaviour Therapy, Chicago, The Behavior Therapist, 5, 128±133.
Shaprio, F. (1995). Eye Movement Desensitization and Wilson, S. A., Becker, L. A. and Tinker, R. H. (1995). Eye Reprocessing: Basic Principles, Protocols, and Procedures.
movement desensitization and reprocessing (EMDR) treatment for psychologically traumatized individuals.
Shestatzky, M., Greenberg, D. and Lerer, B. (1988).
Journal of Consulting and Clinical Psychology, 63, A controlled trial of phenelzine in post-traumatic stress disorder. Psychiatry Research, 24, 149±155.
Wilson, S. A., Becker, L. A. and Tinker, R. H. (1997).
Smith, M. L., Glass, G. V. and Miller, T. I. (1980).
15-Month follow-up of EMDR treatment for psycho- The Benefits of Psychotherapy. Baltimore, MD: Johns logical trauma. Journal of Consulting and Clinical Spielberger, C. D., Gorsuch, R. L. and Lushene, R. E.
Wolf, F. M. (1986). Meta-analysis. Newbury Park, CA: (1970). Manual for the State-Trait Inventory. Palo Alto, Clin. Psychol. Psychother. 5, 126±144 (1998)

Source: http://www.emdr.nl/acrobat/VANETT.PDF