Behaviour Research and Therapy 57 (2014) 21e28

Contents lists available at ScienceDirect

Behaviour Research and Therapy journal homepage: www.elsevier.com/locate/brat

Does dissociation moderate treatment outcomes of narrative exposure therapy for PTSD? A secondary analysis from a randomized controlled clinical trial Joar Øveraas Halvorsen a, *, Håkon Stenmark b, Frank Neuner c, Hans M. Nordahl a a b c

Department of Psychology, Norwegian University of Science and Technology, N-7491 Trondheim, Norway Centre on Violence, Traumatic Stress and Suicide Prevention, Mid-Norway, St. Olavs University Hospital, Schwacks gt. 1, N-7030 Trondheim, Norway Department of Clinical Psychology and Psychotherapy, Bielefeld University, 33501 Bielefeld, Germany

a r t i c l e i n f o

a b s t r a c t

Article history: Received 12 November 2013 Received in revised form 27 March 2014 Accepted 31 March 2014 Available online 12 April 2014

Dissociative symptoms, especially depersonalisation and derealisation, are often perceived as a contraindication for exposure-based treatments of posttraumatic stress disorder (PTSD) despite limited empirical evidence. The present paper examines whether derealisation and depersonalisation influence the treatment outcomes of narrative exposure therapy (NET) and treatment as usual (TaU) among severely traumatised asylum seekers and refugees. We performed a secondary analysis of a recently published randomized controlled multicentre trial comparing NET and TaU for the treatment of PTSD in asylum seekers and refugees. In order to investigate whether depersonalisation and derealisation moderate treatment outcomes, a number of moderated multiple, blockwise regression analyses were conducted. Missing data were handled with multiple imputation. The main finding from intention-totreat analyses is that derealisation and depersonalisation overall do not moderate the treatment outcomes of either NET or TaU. The treatment condition was the most stable predictor of residual gain scores across outcome measures, with NET being associated with lower residual gain scores indicating better treatment outcomes. The present study substantiates and extends previous research indicating that dissociative symptoms such as derealisation and depersonalisation do not moderate the treatment outcome of exposure-based treatments for PTSD. ClinicalTrials.gov identifier: NCT00218959. Ó 2014 Elsevier Ltd. All rights reserved.

Keywords: Narrative exposure therapy Posttraumatic stress disorder Depersonalisation Derealisation Moderators Treatment outcome

Narrative exposure therapy (NET), a recently developed standardised, short-term treatment for posttraumatic stress disorder (PTSD) in survivors of armed conflict, political violence and torture (Schauer, Neuner, & Elbert, 2005), is based on the principles of prolonged exposure therapy (Foa, Hembree, & Rothbaum, 2007) and testimony therapy (Cienfuegos & Monelli, 1983). Specifically, NET has two distinctive features: It uses the chronicity of testimony therapy and, instead of identifying the worst traumatic event as a target in therapy, the survivor constructs a narrative of his or her whole life, and is exposed to all the traumatic experiences in his or her life through imaginal reliving. Imaginal exposure for traumatic experiences is performed as in prolonged exposure therapy, however, there is no explicit focus on in-vivo exposure. Thus, the focus

* Corresponding author. Tel.: þ47 73597811. E-mail addresses: [email protected], [email protected] (J.Ø. Halvorsen). http://dx.doi.org/10.1016/j.brat.2014.03.010 0005-7967/Ó 2014 Elsevier Ltd. All rights reserved.

of NET is twofold. As with prolonged exposure therapy, one aim is to reduce the posttraumatic symptomatology by confronting the memories of the traumatic events. The second aim is to reconstruct the autobiographical memory of the traumatic events and create a consistent narrative or testimony as in testimony therapy. In line with the fact that exposure-based psychological treatments have the most and the methodological strongest evidence for its efficacy in the treatment of PTSD (Bisson et al., 2007; Institute of Medicine, 2008), NET has been found to be effective in treating both PTSD and comorbid disorders in a number of randomized controlled trials in a variety of refugee and asylum-seeking samples (see Robjant & Fazel, 2010 for a recent review) and is probably the treatment modality with the most empirical support to date for this specific patient group (Crumlish & O’Rourke, 2010). Although exposure therapy is a highly effective treatment for PTSD (Powers, Halpern, Ferenschak, Gillihan, & Foa, 2010), a substantial minority of patients either drop-out of treatment, present substantial residual symptoms after treatment or do not respond to

22

J.Ø. Halvorsen et al. / Behaviour Research and Therapy 57 (2014) 21e28

treatment at all (Bradley, Greene, Russ, Dutra, & Westen, 2005; Schottenbauer, Glass, Arnkoff, Tendick, & Gray, 2008). The same reservations are in place for NET. As evident from Stenmark, Catani, Neuner, Elbert, and Holen (2013), there are large variations in treatment responses to NET: Over 50% still satisfied the diagnostic criteria for PTSD at 6 months follow-up, while 36% did not achieve clinically significant symptom remission. Therefore, it is important to identify potential moderators of treatment outcomes (Kraemer, Frank, & Kupfer, 2006; Kraemer, Wilson, Fairburn, & Agras, 2002) in an effort to personalise treatments (Simon & Perlis, 2010) for PTSD. Dissociation has been suggested by a number of researchers and trauma therapists as an important moderator of treatment outcomes for PTSD. Despite controversy, the recently published DSM-5 (American Psychiatric Association, 2013) includes a dissociative subtype of PTSD marked by prominent depersonalisation (i.e., feeling as if oneself is not real) and derealisation (i.e., feeling as if the world is not real) symptoms. Whereas Friedman, Resick, Bryant, and Brewin (2011) concluded that sufficient evidence for a dissociative subtype of PTSD is lacking, others have argued that current research points toward the existence of such a subtype (Dalenberg & Carlson, 2012; Lanius, Brand, Vermetten, Frewen, & Spiegel, 2012). Recent research in both civilian (Steuwe, Lanius, & Frewen, 2012) and military samples (Wolf, Lunney, et al., 2012; Wolf, Miller, et al., 2012) indicates that derealisation and depersonalisation are salient features of PTSD in a subset of individuals with the disorder. The same pattern has also been found in a recent crosscultural epidemiologic survey (Stein et al., 2013). The inclusion of such a subtype of PTSD rests partly on demonstrating that these dissociative symptoms moderate treatment outcomes of already existing efficacious treatments for PTSD (Bryant, 2012; Resick, Bovin, et al., 2012). Several authors emphasise that this dissociative subtype of PTSD might be associated with treatment outcomes (Feeny & Danielson, 2004; Ginzburg & Neria, 2011; Lanius et al., 2010; Lanius et al., 2012; Steuwe et al., 2012; Wolf, Lunney, et al., 2012; Wolf, Miller, et al., 2012), notably with a poor response to ordinary cognitive behavioural treatment. In line with this, a survey among more than 200 practicing psychologists indicated that a majority experienced symptoms of dissociation as a significant contraindication to use exposure therapy for PTSD (Becker, Zayfert, & Anderson, 2004). Indeed, dissociative symptoms are associated with poorer treatment outcomes for in-patient dialectical behaviour therapy for borderline personality disorder (Kleindienst et al., 2011), cognitive behavioural treatment for panic disorder with agoraphobia (Michelson, June, Vives, Testa, & Marchione, 1998) and obsessivee compulsive disorder (Rufer et al., 2006), as well as for in-patient brief psychodynamic psychotherapy for affective, anxiety and somatoform disorders (Spitzer, Barnow, Freyberger, & Grabe, 2007). However, although the above mentioned theoretical assumptions and empirical studies indicate that dissociation is generally related to poorer treatment outcomes, the existing research on the influence of dissociation on treatment outcomes for PTSD is not as clear. Several clinical trials have examined whether dissociation is a predictor or a moderator of treatment outcomes of exposure-based treatments for PTSD. Overall, dissociation does not seem to be a predictor of treatment outcomes (Hagenaars, van Minnen, & Hoogduin, 2010; Jaycox, Foa, & Morral, 1998; Speckens, Ehlers, Hackmann, & Clark, 2006; Taylor, 2003). Secondary data analyses from two dismantling randomized controlled trials have investigated whether dissociation moderate treatment outcomes. In the first trial, Cloitre, Petkova, Wang, and Lu (2012) found that severity of dissociative symptoms at pretreatment did not moderate the treatment outcomes of skills training in affective and interpersonal regulation followed by

narrative storytelling (STAIReNST) and the constituent parts of the manual. In the second trial, comparing the different elements of cognitive processing therapy (i.e., the full manual, cognitive therapy only and written trauma account only) in the treatment of PTSD, severity of dissociative symptoms at pre-treatment did not influence treatment outcomes when averaged across treatment conditions (Resick, Suvak, Johnides, Mitchell, & Iverson, 2012). However, patients with more severe dissociative symptoms, especially depersonalisation symptoms, had better outcomes if they received the full manual as compared to cognitive therapy only, whereas patients with less severe dissociative symptoms had better treatment responses to cognitive therapy only compared to the full manual. Thus, these results indicate that therapeutic tasks with elements of exposure therapy might be especially indicated in patients with severe dissociative symptoms. Of note, both Hagenaars et al. (2010) and Cloitre et al. (2012) found that higher levels of dissociation at baseline was associated with more severe PTSD-symptoms at both pre- and posttreatment. The comorbidity between depression and depersonalisation and derealisation is high (Hunter, Sierra, & David, 2004) and the dissociative subtype of PTSD has higher comorbidity with depression as compared to “classical” PTSD (Steuwe et al., 2012). As such, it is important to examine whether dissociative symptoms influence treatment outcomes for comorbid depressive symptoms. Moreover, as pinpointed by Hagenaars et al. (2010) treatment efficacy concerns both improvement and drop-out. Dissociation has been found to predict drop-out from CBT treatment for OCD (Rufer et al., 2006). However, whereas Hagenaars et al. found that baseline dissociation was not related to drop-out from exposure therapy for PTSD, Cloitre et al. (2012) found that patients with high as compared to low dissociation were less likely to drop-out of treatment. Thus, it is also important to investigate whether dissociation predicts drop-out. Dissociation is a multidimensional phenomenon (Briere, Weathers, & Runtz, 2005) and as such Bryant (2007) underlined that research on dissociative phenomena should be based on specific symptoms rather than the global construct of dissociation. Furthermore, according to Wolf (2013), derealisation and depersonalisation “reflect more pathological forms of dissociative phenomena that are distinct from other types of dissociation” (p. 2). Thus, the present paper set out to examine whether derealisation and depersonalisation moderate treatment outcomes of NET and treatment as usual (TaU) among severely traumatised asylum seekers and refugees. The present exploratory analysis aims to extend previous research in several ways. First, to our knowledge, this is the first paper to examine whether dissociative symptoms moderate treatment outcomes of both NET and TaU. Furthermore, no other studies have investigated the role of dissociation in treatment outcomes in this specific patient population, i.e. severely traumatised asylum seekers and refugees. In addition, whereas most other studies have used global constructs of dissociation, we set out to examine whether specific symptoms of dissociation, i.e. derealisation and depersonalisation, moderate treatment outcomes independently. Method The present paper is based on exploratory secondary analyses from a recently published randomized controlled multicentre trial comparing NET and TaU for the treatment of PTSD in asylum seekers and refugees (Stenmark et al., 2013), and the details will only be briefly reviewed herein. The main finding of the trial was

J.Ø. Halvorsen et al. / Behaviour Research and Therapy 57 (2014) 21e28

that NET is superior to TaU in reducing PTSD symptom severity, but not severity of depressive symptoms. Participants The sample consists of 81 participants recruited among refugees and asylum-seekers referred to psychological treatment at outpatient clinics in the Mid-Norway health region. The inclusion criteria were (1) a primary PTSD diagnosis according to the DSM-IV (American Psychiatric Association, 1994) criteria and (2) 18 years of age. The exclusion criteria were (1) psychotic disorders, (2) current severe substance abuse, or (3) severe suicidal ideations. The majority of the sample was male (69%) and most of the participants were from Iraq (27%), Afghanistan (15%) or African countries (26%). The mean age of the sample was 35.55 (SD ¼ 11.05), and most completed primary and secondary school (59%). Mean time spent in Norway was 55.99 months (SD ¼ 50.58) and a substantial minority were asylum seekers at pre-treatment (38%). Based on The Life Events Checklist of the CAPS, which assesses a total of 16 potentially traumatic life events, the participants reported experiencing multiple traumatic life experiences (M ¼ 8.11, SD ¼ 2.51). Of note, a substantial minority reported having survived one or more instances of torture (43%). There were no significant differences between participants randomized to NET or TaU on any of the demographic variables measured (Stenmark et al., 2013). Instruments The following instruments were used: Clinician-administered PTSD scale The Clinician-Administered PTSD Scale (CAPS; Blake et al., 1995) is widely recognised as the gold standard for assessment of PTSD. The CAPS assesses all 17 core symptoms of PTSD, in addition to 5 associated features where two items measure feelings of guilt (guilt over omission or commission and survivor guilt) and three items measure dissociative symptoms (derealisation, depersonalisation and reduction in awareness of surroundings). The CAPS contains separate frequency and intensity rating scales for each of the symptoms. The frequency and intensity of each symptom are rated on a five-point Likert scale (0e4), and these ratings can be summed to create a nine-point (0e8) severity score for each symptom. The CAPS has excellent psychometric properties and has consistently been found to have an inter-rater reliability at the 0.90 level and above (Weathers, Keane, & Davidson, 2001). Based on the Jacobson and Truax (1991) criteria, and in line with Hien et al. (2009), we defined clinically significant change as a reduction in CAPS total score of 30 points, which is approximately equivalent to 2 SDs below the baseline mean. In line with previous research (Armour, Karstoft, & Richardson, 2014; Wolf, Lunney, et al., 2012; Wolf, Miller, et al., 2012), derealisation and depersonalisation were measured using the relevant/ specific items from the CAPS. Wolf, Miller, et al. reported high intraclass correlation coefficients for the dissociation items of the CAPS (ICC ¼ 0.79). Hamilton Rating Scale for Depression The Hamilton Rating Scale for Depression (HRSD; Hamilton, 1960) is a widely used clinician rating scale for assessing the severity of depression. The scale consists of 17 items measuring depression severity. The items are scored on either a three-point (0e2) or a five-point scale (0e4). HRSD has adequate psychometric properties (Trajkovic et al., 2011), including inter-rater reliability (Morriss, Leese, Chatwin, Baldwin, & Thread Study Group, 2008; Trajkovi c et al., 2011).

23

Procedure The method and design of the randomized controlled multicentre trial the present paper is based on has been reported in detail elsewhere (Stenmark et al., 2013), and therefore only a brief description is given here. The Regional Committee for Medical Research Ethics in MidNorway approved the project. Participants were block randomized to treatment conditions, in which 2/3 of the participants were allocated to NET and 1/3 allocated to TaU. Assessments took place before treatment, one month after treatment and a follow-up at six months. Therapists and assessors Twenty-four experienced mental health professionals (psychologists, psychiatrists, psychiatric nurses and clinical social workers) situated at 11 different outpatient clinics in Mid-Norway health region were recruited as therapists and assessors. The professionals were trained in NET and the application of the assessor instruments included in the study during a 5-day workshop. Subsequently, the professionals participated in 2-day workshops every six months to maintain their skills. In addition, the professionals received individual supervision after the 1st, 4th, 6th, and 9th sessions of therapy for each patient. Assessors and therapist were independent of each other, i.e. the professionals could not be both therapist and assessor for the same patient. The assessors were not aware of allocation, and in order to maintain blindness we always aimed for assessments to be undertaken by assessors from different clinics where the patients received their treatment. Furthermore, the therapists were instructed not to reveal the type of treatment their patients were given. Despite these measures, it appeared that in a substantial minority of cases (20%), the patients revealed information about their treatment to the assessors. As reported by Stenmark et al. (2013), a statistical analysis showed no significant differences of these post-tests from the other assessments. Treatment Treatment consisted of 10 sessions lasting 90 min in both conditions. NET was performed according to the manual as outlined by Schauer et al. (2005). Treatment adherence and competence was monitored and ensured through (1) the individual supervision and (2) a self-report measure, where the therapists had to report after each session whether they had used the main ingredients of NET, such as psychoeducation, the life-line exercise and prolonged exposure to memories of traumatic experiences. No major deviations from the NET-protocol as described in Schauer et al. (2005) were identified. In the TaU condition, the therapists were instructed to use any intervention that they would normally use, except for the interventions specific to NET. Based on information gathered through supervision and therapists self-report, it seems TaU mainly consisted of help with problems such as sleep difficulties, depressive symptoms, problems related to the asylum procedure and practical matters. During the 10 TaU sessions, an average of 86 min was spent on talking about traumatic events, mainly to give the therapist an overview of the patients’ history. Data analyses As our primary aim of the present paper is to examine whether derealisation and depersonalisation moderate treatment outcomes, we utilised multiple, blockwise regression analyses with interaction terms (Frazier, Tix, & Barron, 2004; Warner, 2013). The main multiple, blockwise regression analyses were performed with CAPS residual gain scores at post-treatment and follow-up as outcome/ dependent variables. In the first block, we entered treatment arm

24

J.Ø. Halvorsen et al. / Behaviour Research and Therapy 57 (2014) 21e28

(0 ¼ TaU, 1 ¼ NET), in the second block total HRSD score at pretreatment, in the third block derealisation and depersonalisation total scores at pre-treatment, and in the fourth and last block interaction terms between treatment and HRSD, treatment and depersonalisation, and treatment and derealisation. As recommended by Frazier et al. (2004), continuous variables were standardised before computation of interaction terms. Residual gain scores (Steketee & Chambless, 1992) were chosen as outcome/ dependent variables to reflect change in symptoms and to control for both initial differences and measurement errors inherent in repeated administration of the same instruments. As secondary analyses, we performed moderated multiple, blockwise regression analyses with HRSD residual gain scores at post-treatment and follow-up as dependent variables. We also performed a binary logistic regression analysis with drop-out before planned treatment termination as an outcome variable. We identified three influential outliers (i.e., 3 SD from the mean). All had high scores on depersonalisation at pre-treatment, and were removed prior to final modelling. We also divided the sample into three subgroups according to the severity of both derealisation and depersonalisation and examined Hedges g effect sizes between the groups. No or low derealisation or depersonalisation was defined as a total score of 0. Moderate derealisation or depersonalisation was defined as a total score of 1e3, and severe derealisation or depersonalisation was classified as a total score of 4 on either scale. In addition, we examined rates of clinically significant change in CAPS total scores from pre-treatment to follow-up based on baseline severity of derealisation and depersonalisation. Missing data Missing data were handled with multiple imputation (Graham, 2009, 2012; Schafer & Graham, 2002). The Little’s MCAR (Missing Completely At Random) test was not significant ðc2 1248 ¼ 618:37; p ¼ 1:00Þ, indicating that the data could be considered to be missing completely at random. We utilised itemlevel imputation (Gottschall, West, & Enders, 2012) to generate/ produce 20 imputed data sets. Multiple imputation has been found to perform very well with large multiple regression models with large portions of missing data, even with small sample sizes (Graham, 2009). Whereas it is often considered sufficient to perform five to ten imputations (Schafer, 1999), Graham, Olchowski, and Gilreath (2007) recommended performing many more imputations beyond what is usually considered sufficient. As missing data at post-test and follow-up ranged from 25.9% to 30.9%, based on recommendations by Graham et al. (2007), we generated 20 imputed data sets. We constrained minimum and maximum allowable imputed values in accordance with the range of values on the instruments, i.e., items on the CAPS range from 0 to 4 and imputed values were constrained to be in this range. However, as recommended by Graham (2009), we set no constraints on rounding. As IBM PASW Statistics 18 does not provide pooled outcomes for all statistics (e.g., standard deviations), we calculated some of these outcomes manually. The main results will be reported in line with recommendations made by Graham (2012). Results Table 1 presents means and standard deviations at pretreatment on the measures relevant for this paper in the two treatment arms. At baseline, there were no significant differences between the two treatment conditions on any of the measures, except for derealisation in which the TaU-condition had a significantly higher mean score compared to the NET-condition (t ¼ 2.87, p  0.05).

Table 1 Clinical characteristics of the sample at pre-treatment (N ¼ 78).

CAPS total score HRSD total score Derealisation Depersonalisation

NET

TaU

Total sample

M (SD)

M (SD)

M (SD)

83.31 18.41 0.87 0.47

83.17 19.52 2.15 0.86

83.26 18.82 1.35 0.61

(15.58) (6.90) (1.46) (1.29)

(16.59) (5.83) (2.10) (1.46)

(15.85) (6.5) (1.82) (1.36)

Note. CAPS ¼ Clinician-administered PTSD scale. HRSD ¼ Hamilton rating scale for depression. NET ¼ Narrative exposure therapy. TaU ¼ Treatment as usual.

Moderators of treatment outcome for PTSD The results of the moderated multiple, blockwise regression analysis with CAPS residual gain score at post-treatment are presented in Table 2. In the final block, only treatment predicted CAPS residual gain score at post-treatment. The NET condition was associated with lower residual gain score, indicating lower symptom severity at post-treatment compared to the TaU condition. IBM PASW Statistics 18 does not provide a pooled outcome for the Fvalues, but all 20 models were significant at p  0.009. The Durbine Watson values for the 20 models ranged from 2.208 to 2.350, indicating that the residuals were independent. Furthermore, there seems to be no substantial multicollinearity within our data as the variance inflation factor (VIF) did not exceed 3.655 and the lowest Tolerance statistics was 0.274. Table 3 presents the results of the moderated multiple, blockwise regression analysis with CAPS residual gain score at follow-up. As for the foregoing analysis, in the final block only treatment condition predicted CAPS residual gain score at follow-up, with NET being associated with a lower residual gain score indicating better outcomes. All 20 models were significant at p  0.034. The Durbine Watson value ranged from 1.876 to 1.996 for the 20 models, and the highest VIF value was 3.655 whereas the lowest Tolerance statistics was 0.274. Figs. 1 and 2 depicts CAPS mean scores at pre-treatment, posttreatment and follow-up for NET and TaU based on severity of derealisation or depersonalisation at pre-treatment, respectively. Table 2 Pooled outcome of moderated multiple, blockwise regression analysis with CAPS residual gain scores at post-treatment as the dependent variable (N ¼ 78).

Block 1 Constant Treatment Block 2 Constant Treatment HRSD Block 3 Constant Treatment HRSD Derealisation Depersonalisation Block 4 Constant Treatment HRSD Derealisation Depersonalisation Treatment  HRSD Treatment  Derealisation Treatment  Depersonalisation

B

SE

t

p

FMI

.116 .187

.081 .102

1.436 1.839

.151 .066

.019 .16

.603 .215 .025

.158 .095 .007

3.812 2.257 3.495

.000 .024 .000

.021 .017 .024

.660 .283 .022 .058 .005

.159 .099 .007 .028 .036

4.139 2.848 3.015 2.039 .152

.000 .004 .003 .041 .879

.028 .019 .023 .034 .043

.723 .271 .022 .057 .055 .001 .008 .184

.270 .100 .013 .038 .053 .101 .113 .125

2.680 2.698 1.692 1.515 1.034 .007 .067 1.475

.007 .007 .091 .130 .301 .995 .947 .140

.016 .018 .020 .021 .012 .016 .045 .045

Note. CAPS ¼ Clinician-administered PTSD scale. HRSD ¼ Hamilton rating scale for depression. B ¼ Unstandardised regression coefficient. SE ¼ Standard error. FMI ¼ Fraction of missing data.

J.Ø. Halvorsen et al. / Behaviour Research and Therapy 57 (2014) 21e28

25

Table 3 Pooled outcome of moderated multiple, blockwise regression analysis with CAPS residual gain scores at follow-up as the dependent variable (N ¼ 78).

Block 1 Constant Treatment Block 2 Constant Treatment HRSD Block 3 Constant Treatment HRSD Derealisation Depersonalisation Block 4 Constant Treatment HRSD Derealisation Depersonalisation Treatment  HRSD Treatment  Derealisation Treatment  Depersonalisation

B

SE

t

p

FMI

.119 .191

.070 .088

1.699 2.172

.089 .030

.009 .005

.485 .212 .019

.140 .084 .006

3.457 2.524 2.960

.001 .012 .003

.022 .006 .028

.523 .257 .017 .038 .003

.143 .089 .006 .026 .032

3.666 2.894 2.563 1.491 .089

.000 .004 .010 .136 .929

.019 .006 .028 .036 .029

.481 .246 .012 .041 .036 .044 .011 .119

.246 .090 .012 .034 .048 .092 .102 .112

1.954 2.732 1.025 1.222 .744 .474 .113 1.063

.051 .006 .305 .222 .457 .636 .910 .288

.033 .005 .041 .019 .012 .033 .036 .037

Note. CAPS ¼ Clinician-administered PTSD scale. HRSD ¼ Hamilton rating scale for depression. B ¼ Unstandardised regression coefficient. SE ¼ Standard error. FMI ¼ Fraction of missing data.

Moderators of treatment outcome for depression Table 4 displays the results of the moderated multiple, blockwise regression analysis with HRSD residual gain score at posttreatment as the dependent variable. As can be seen from Table 4, treatment condition and derealisation at pre-treatment were significant predictors of depression at post-treatment. More specifically, both NET and derealisation were associated with better treatment outcomes at post-treatment. However, the interaction term treatment  derealisation was not significant, indicating that derealisation at pre-treatment did not moderate treatment outcomes more in one treatment arm compared to the other. As such, derealisation seems to be a more generic predictor of treatment outcomes among traumatised asylum seekers and refugees rather than a specific moderator of outcomes in specific psychological treatments. Three of the 20 models were not significant, but the remaining 17 models were all significant at the p  0.05 level. DurbineWatson values of the 20

Fig. 1. Change in CAPS total score from pre-treatment to follow-up for treatment as usual (TaU; n ¼ 30) and narrative exposure therapy (NET; n ¼ 51) based on severity of derealisation (DR) at pre-treatment.

Fig. 2. Change in CAPS total score from pre-treatment to follow-up for treatment as usual (TaU; n ¼ 30) and narrative exposure therapy (NET; n ¼ 51) based on severity of depersonalisation (DP) at pre-treatment.

models ranged from 1.827 to 2.031, while the VIF values did not exceed 2.698 and the Tolerance statistic was 0.371. None of the variables investigated predicted HRSD residual gain score at follow-up, and as such none of the 20 models were significant. DurbineWatson values ranged from 1.558 to 1.780, the VIF values were 2.822 and the Tolerance statistic 0.354. Visual inspection of the normal probability plots indicated that the residuals were normally distributed, and visual inspection of the scatter plots indicated that the assumption of linearity was met with no substantial heteroscedasticity within our data. Thus, the assumptions underlying linear regression analyses are deemed to be met. We also performed moderated multiple, blockwise regression analyses without excluding the three outliers. Overall, these analyses did not differ from the analyses excluding the outliers. However, when using HRSD residual gain scores at post-treatment as the dependent variable, the interaction between treatment and depersonalisation approached significance (B ¼ 0.472; SE ¼ 0.244; t ¼ 1.934; p ¼ 0.053). Visual inspection of the 20 scatterplots for Table 4 Pooled outcome of moderated multiple, blockwise regression analysis with HRSD residual gain scores at post-treatment as the dependent variable (N ¼ 78). B Block 1 Constant Treatment Block 2 Constant Treatment CAPS Block 3 Constant Treatment CAPS Derealisation Depersonalisation Block 4 Constant Treatment CAPS Derealisation Depersonalisation Treatment  CAPS Treatment  Derealisation Treatment  Depersonalisation

SE

t

p

FMI

.188 .336

.183 .230

1.024 1.462

.306 .144

.44 .030

.500 .337 .008

.612 .229 .007

.817 1.471 1.177

.414 .141 .239

.030 .030 .029

.426 .512 .011 .168 .100

.595 .236 .007 .067 .084

.717 2.168 1.522 2.488 1.189

.474 .030 .128 .013 .235

.034 .023 .028 .054 .035

1.033 .474 .021 .200 .064 .279 .165 .480

.908 .234 .011 .087 .126 .224 .263 .295

1.138 2.024 1.905 2.296 .511 1.248 .629 1.626

.255 .043 .057 .022 .609 .212 .530 .104

.027 .025 .020 .052 .038 .029 .051 .070

Note. CAPS ¼ Clinician-administered PTSD scale. HRSD ¼ Hamilton rating scale for depression. B ¼ Unstandardised regression coefficient. SE ¼ Standard error. FMI ¼ Fraction of missing data.

26

J.Ø. Halvorsen et al. / Behaviour Research and Therapy 57 (2014) 21e28

HRSD residual gain score at post-treatment by depersonalisation at pre-treatment with case markers for treatment condition all indicated the same pattern; severe depersonalisation at pre-treatment is associated with higher HRSD residual gain scores at posttreatment in the NET condition. Effect sizes Within-group Hedges g effect sizes according to the severity of derealisation and depersonalisation at pre-treatment are presented in Table 5. Although there are some notable differences in effect sizes across the subgroups, these differences must be interpreted with caution due to the low number of patients in the subgroups with moderate and severe symptoms. However, the effect sizes seem to underscore that even patients with moderate and severe dissociative symptoms can achieve substantial symptom remission on par with patients without dissociative symptoms at pretreatment.

Table 6 Pooled outcome of binary logistic regression analyses with drop-out before planned treatment termination as the dependent variable (N ¼ 78). B

SE

ta

p

Exp(B) FMI

Treatment .177 .606 .292 .771 1.193 CAPS .011 .030 .367 .717 .989 HRSD .093 .087 1.069 .285 1.098 Derealisation .219 .210 1.043 .297 1.245 Depersonalisation .031 .327 .095 .925 .970 Treatment  CAPS .250 .643 .389 .697 1.284 Treatment  HRSD 1.080 .680 1.589 .112 .340 Treatment  Derealisation .644 .651 .989 .323 .525 Treatment  Depersonalisation .290 .722 .402 .688 1.337 Constant 2.230 2.296 .971 .331 .108

.003 .008 .028 .003 .001 .005 .019 .006 .004 .001

CAPS ¼ Clinician-administered PTSD scale. HRSD ¼ Hamilton rating scale for depression. B ¼ Unstandardised regression coefficient. SE ¼ Standard error. FMI ¼ Fraction of missing data. a t-values were not given in the pooled outcomes in IBM PASW Statistics 18 and the values shown in the table above were calculated manually (t ¼ B/SE).

approximately 43% and 50% achieved clinically significant change in their CAPS total scores, respectively.

Clinically significant change We also estimated the percentage of patients pooled across the 20 imputed datasets that achieved clinically significant change in their CAPS total scores from pre-treatment to follow-up based on the severity of derealisation and depersonalisation at baseline. In the TaU-condition, none of the patients with no or low derealisation at pre-treatment achieved clinically significant change from pre-treatment to follow-up. Of the patients with moderate derealisation at pre-treatment, approximately 9% achieved clinically significant change, whereas approximately 42% of the patients with severe derealisation at pre-treatment achieved clinically significant symptom remission from pre-treatment to follow-up. Approximately 45% of patients receiving NET with no or low derealisation at pre-treatment achieved clinically significant change. Of the patients receiving NET with moderate and severe derealisation at pre-treatment, approximately 63% and 50% achieved clinically significant symptom remission from pre-treatment to follow-up, respectively. In the TaU-condition, 14% of the patients with no or low depersonalisation at pre-treatment achieved clinically significant change, whereas approximately 19% and 25% of patients with moderate or severe depersonalisation achieved clinically significant symptom remission, respectively. Of the patients receiving NET, approximately 50% with no or low depersonalisation at pre-treatment achieved clinically significant symptom remission from pre-treatment to follow-up. Of the patients with moderate or severe depersonalisation at pre-treatment,

Predictors of drop-out None of the variables examined predicted drop-out from treatment before planned termination (see Table 6). Discussion The main result from this exploratory secondary analysis is that depersonalisation and derealisation do not seem to substantially moderate the treatment outcomes of either NET or TaU among severely traumatised asylum seekers and refugees. This is contrary to prevalent theoretical assumptions (Feeny & Danielson, 2004; Ginzburg & Neria, 2011; Lanius et al., 2010; Lanius et al., 2012; Steuwe et al., 2012; Wolf, Lunney, et al., 2012; Wolf, Miller, et al., 2012), but in line with most empirical treatment research to date (Cloitre et al., 2012; Hagenaars et al., 2010; Resick, Suvak, et al., 2012; Speckens et al., 2006; for a recent review see van Minnen, Harned, Zoellner, & Mills, 2012). The present results substantiate and extend previous findings. Although we replicate what seems to be a rather consistent finding, that dissociative phenomena do not moderate the treatment outcome of exposure-based treatments for PTSD in any significant degree, our results extend this finding to NET and among severely traumatised asylum seekers and refugees in which a substantial minority were torture survivors. Thus, dissociative symptoms do not seem to be an overall contraindication for exposure based treatments for PTSD among this specific patient group.

Table 5 Within-group Hedges g effect sizes based on level of derealisation and depersonalisation at pre-treatment (N ¼ 81). CAPS pre-post

CAPS pre-FU

HRSD pre-post

HRSD pre-FU

TaU

Low derealisation (n ¼ 11) Moderate derealisation (n ¼ 10) Severe derealisation (n ¼ 9) Low depersonalisation (n ¼ 20) Moderate depersonalisation (n ¼ 6) Severe depersonalisation (n ¼ 4)

0.49 0.43 1.51 0.55 1.03 1.23

0.50 0.66 1.27 0.65 0.82 1.09

0.40 0.31 1.93 0.46 1.04 0.64

0.85 0.05 0.76 0.55 0.36 0.57

NET

Low derealisation (n ¼ 33) Moderate derealisation (n ¼ 11) Severe derealisation (n ¼ 7) Low depersonalisation (n ¼ 42) Moderate depersonalisation (n ¼ 4) Severe depersonalisation (n ¼ 5)

1.06 2.40 1.36 1.38 0.85 0.97

1.24 2.43 0.98 1.46 1.41 0.83

0.52 1.37 1.02 0.80 0.16 0.35

0.36 1.23 0.51 0.60 0.23 0.22

Note. CAPS ¼ Clinician-administered PTSD scale. HRSD ¼ Hamilton rating scale for depression. TaU ¼ Treatment as usual. NET ¼ Narrative exposure therapy.

J.Ø. Halvorsen et al. / Behaviour Research and Therapy 57 (2014) 21e28

The most consistent finding in the present study is that NET was associated with superior outcomes compared to TaU, except for depressive symptoms at follow-up. No other stable predictors or moderators of treatment outcomes emerged from the present analyses. While several of the variables entered in the third block of the models emerged as significant predictors, they were no longer significant following the insertion of the interaction terms in the fourth block. When interaction terms are not significant, we need to decide whether to retain the interaction terms in the models or remove them (Frazier et al., 2004; Warner, 2013). Due to the theoretically proposed importance of dissociative symptoms in predicting treatment outcomes, we retained the interaction terms in the analyses although they were non-significant. There are some nuances to this general finding, specifically in relation to depressive symptoms at post-treatment. When using HRSD residual gain score at post-treatment as the dependent variable, two variables emerged as significant predictors of outcomes. Both NET and derealisation predicted more depressive symptom remission at post-treatment. It is especially interesting to note that derealisation might be related to better treatment outcomes in depressive symptomatology. However, none of the interaction terms were significant, indicating that the aforementioned variables are general predictors of treatment outcomes abreast treatment modalities rather than specific moderators of specific treatment modalities (Kraemer et al., 2002; Simon & Perlis, 2010). It is difficult to speculate why derealisation might be associated with less depressive symptoms at post-treatment. One reasonable explanation is that comorbidity is a predictor for better treatment outcomes (Olatunji, Cisler, & Tolin, 2010). However, it is important to note that patients with more severe derealisation at baseline also had higher HRSD total scores at baseline. Thus, the finding that derealisation predicted greater reductions in depression severity might be explained by the fact that these patients had more room for improvement in their depressive symptoms than other patients with lower symptom severity at baseline. Neither NET nor derealisation were significant predictors when HRSD residual gain score at follow-up was used as the dependent variable, indicating that these effects are either transient/temporary or that the present findings are unstable. In addition, these findings might indicate that other variables not investigated in the present analyses are important for depressive symptomatology in the long term after trauma-focused treatments for PTSD in asylum seekers and refugees. The present paper extends previous research in several ways. As mentioned, this is the first study to our knowledge to investigate whether dissociative symptoms moderate treatment outcome in a sample of severely traumatised asylum seekers and refugees, of which a substantial minority has been tortured, residing in a western country. Furthermore, the present paper has some methodological strengths (compared to earlier studies), especially with respect to handling missing data with multiple imputation and the use of multiple, blockwise regression analyses with interaction terms. An additional strength of the present paper is that our analyses are based on the specific symptoms of derealisation and depersonalisation rather than the global construct of dissociation. In addition, we did not limit our analyses to only PTSD, but also analysed the impact of these dissociative symptoms on depressive symptomatology after treatment. And whereas all previous studies investigating the impact of dissociation on treatment outcomes in PTSD consisted of samples with a vast majority of females, from 73% (Speckens et al., 2006) to 100% (Cloitre et al., 2012; Resick, Suvak, et al., 2012), the present sample consisted of a majority of males. However, the present paper also has several noteworthy limitations. Secondary analyses are prone to Type I errors and significant associations should therefore be interpreted with caution. On

27

the other hand, the data in the present paper was primarily collected to investigate the differential effectiveness between NET and TaU and thus no power analysis for moderation was conducted a priori. Furthermore, the unequal sample sizes between the treatment arms might lower power (Frazier et al., 2004). Thus, the present study might not have enough power to detect significant interaction effects. It is also important to note that the variance in symptom severity of derealisation and depersonalisation is modest and the majority of patients scored in the lower range of symptom severity. The limited variance may be a risk factor for false-negative findings. However, the variance in symptom severity in the present study is on par or higher than in some previous published studies (e.g., Hagenaars et al., 2010). The use of single items to assess derealisation and depersonalisation is also a major limitation, although in line with previous research (Armour et al., 2014; Wolf, Lunney, et al., 2012; Wolf, Miller, et al., 2012). Despite these potential limitations, the present paper substantiates and extends previous research indicating that dissociative symptoms should not be regarded as a universal contraindication either for exposure-based treatments for PTSD specifically or for PTSD-treatment generally. References American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Arlington, VA, US: American Psychiatric Publishing, Inc. American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders: DSM-5Ô (5th ed.). Arlington, VA, US: American Psychiatric Publishing, Inc. Armour, C., Karstoft, K. I., & Richardson, J. D. (2014). The co-occurrence of PTSD and dissociation: differentiating severe PTSD from dissociative-PTSD. Social Psychiatry and Psychiatric Epidemiology. Becker, C. B., Zayfert, C., & Anderson, E. (2004). A survey of psychologists’ attitudes towards and utilization of exposure therapy for PTSD. Behaviour Research and Therapy, 42, 277e292. Bisson, J. I., Ehlers, A., Matthews, R., Pilling, S., Richards, D., & Turner, S. (2007). Psychological treatments for chronic post-traumatic stress disorder: systematic review and meta-analysis. British Journal of Psychiatry, 190, 97e104. Blake, D. D., Weathers, F. W., Nagy, L. M., Kaloupek, D. G., Gusman, F. D., Charney, D. S., et al. (1995). The development of a Clinician-Administered PTSD Scale. Journal of Traumatic Stress, 8, 75e90. Bradley, R., Greene, J., Russ, E., Dutra, L., & Westen, D. (2005). A multidimensional meta-analysis of psychotherapy for PTSD. American Journal of Psychiatry, 162, 214e227. Briere, J., Weathers, F. W., & Runtz, M. (2005). Is dissociation a multidimensional construct? Data from the Multiscale Dissociation Inventory. Journal of Traumatic Stress, 18, 221e231. Bryant, R. A. (2007). Does dissociation further our understanding of PTSD? Journal of Anxiety Disorders, 21, 183e191. Bryant, R. A. (2012). Simplifying complex PTSD: comment on Resick et al. (2012). Journal of Traumatic Stress, 25, 252e253. Cienfuegos, A. J., & Monelli, C. (1983). The testimony of political repression as a therapeutic instrument. American Journal of Orthopsychiatry, 53, 43e51. Cloitre, M., Petkova, E., Wang, J., & Lu, F. (2012). An examination of the influence of a sequential treatment on the course and impact of dissociation among women with PTSD related to childhood abuse. Depression and Anxiety, 29, 709e717. Crumlish, N., & O’Rourke, K. (2010). A systematic review of treatments for posttraumatic stress disorder among refugees and asylum-seekers. Journal of Nervous and Mental Disease, 198, 237e251. Dalenberg, C., & Carlson, E. B. (2012). Dissociation in posttraumatic stress disorder part II: how theoretical models fit the empirical evidence and recommendations for modifying the diagnostic criteria for PTSD. Psychological Trauma: Theory, Research, Practice, and Policy, 4, 551e559. Feeny, N. C., & Danielson, C. K. (2004). PTSD, dissociation, and treatment. In S. Taylor (Ed.), Advances in the treatment of posttraumatic stress disorder: Cognitive-behavioral perspectives (pp. 223e241). New York, NY, US: Springer Publishing Co. Foa, E. B., Hembree, E. A., & Rothbaum, B. O. (2007). Prolonged exposure therapy for PTSD: Emotional processing of traumatic experiences. Therapist guide. New York, NY, US: Oxford University Press. Frazier, P. A., Tix, A. P., & Barron, K. E. (2004). Testing moderator and mediator effects in counseling psychology research. Journal of Counseling Psychology, 51, 115e134. Friedman, M. J., Resick, P. A., Bryant, R. A., & Brewin, C. R. (2011). Considering PTSD for DSM-5. Depression and Anxiety, 28, 750e769.

28

J.Ø. Halvorsen et al. / Behaviour Research and Therapy 57 (2014) 21e28

Ginzburg, K., & Neria, Y. (2011). Mental health interventions for survivors of torture: current status and future directions. Zeitschrift für Psychologie/Journal of Psychology, 219, 187e189. Gottschall, A. C., West, S. G., & Enders, C. K. (2012). A comparison of item-level and scale-level multiple imputation for questionnaire batteries. Multivariate Behavioral Research, 47, 1e25. Graham, J. W. (2009). Missing data analysis: making it work in the real world. Annual Review of Psychology, 60, 549e576. Graham, J. W. (2012). Missing data: Analysis and design. New York, NY: Springer New York. Graham, J. W., Olchowski, A. E., & Gilreath, T. D. (2007). How many imputations are really needed? Some practical clarifications of multiple imputation theory. Prevention Science, 8, 206e213. Hagenaars, M. A., van Minnen, A., & Hoogduin, K. A. L. (2010). The impact of dissociation and depression on the efficacy of prolonged exposure treatment for PTSD. Behaviour Research and Therapy, 48, 19e27. Hamilton, M. (1960). A rating scale for depression. Journal of Neurology, Neurosurgery and Psychiatry, 23, 56e61. Hien, D. A., Wells, E. A., Jiang, H., Suarez-Morales, L., Campbell, A. N. C., Cohen, L. R., et al. (2009). Multisite randomized trial of behavioral interventions for women with co-occurring PTSD and substance use disorders. Journal of Consulting and Clinical Psychology, 77, 607e619. Hunter, E. C. M., Sierra, M., & David, A. S. (2004). The epidemiology of depersonalisation and derealisation: a systematic review. Social Psychiatry and Psychiatric Epidemiology, 39, 9e18. Institute of Medicine. (2008). Treatment of posttraumatic stress disorder: An assessment of the evidence. Washington, D.C: The National Academies Press. Jacobson, N. S., & Truax, P. (1991). Clinical significance: a statistical approach to defining meaningful change in psychotherapy research. Journal of Consulting and Clinical Psychology, 59, 12e19. Jaycox, L. H., Foa, E. B., & Morral, A. R. (1998). Influence of emotional engagement and habituation on exposure therapy for PTSD. Journal of Consulting and Clinical Psychology, 66, 185e192. Kleindienst, N., Limberger, M. F., Ebner-Priemer, U. W., Keibel-Mauchnik, J., Dyer, A., Berger, M., et al. (2011). Dissociation predicts poor response to dialectial behavioral therapy in female patients with borderline personality disorder. Journal of Personality Disorders, 25, 432e447. Kraemer, H. C., Frank, E., & Kupfer, D. J. (2006). Moderators of treatment outcomes. JAMA: The Journal of the American Medical Association, 296, 1286e1289. Kraemer, H. C., Wilson, G. T., Fairburn, C. G., & Agras, W. S. (2002). Mediators and moderators of treatment effects in randomized clinical trials. Archives of General Psychiatry, 59, 877e883. Lanius, R. A., Brand, B., Vermetten, E., Frewen, P. A., & Spiegel, D. (2012). The dissociative subtype of posttraumatic stress disorder: rationale, clinical and neurobiological evidence, and implications. Depression and Anxiety, 29, 701e 708. Lanius, R. A., Vermetten, E., Loewenstein, R. J., Brand, B., Schmahl, C., Bremner, J. D., et al. (2010). Emotion modulation in PTSD: clinical and neurobiological evidence for a dissociative subtype. The American Journal of Psychiatry, 167, 640e 647. Michelson, L., June, K., Vives, A., Testa, S., & Marchione, N. (1998). The role of trauma and dissociation in cognitive-behavioral psychotherapy outcome and maintenance for panic disorder with agoraphobia. Behaviour Research and Therapy, 36, 1011e1050. van Minnen, A., Harned, M. S., Zoellner, L., & Mills, K. (2012). Examining potential contraindications for prolonged exposure therapy for PTSD. European Journal of Psychotraumatology, 3. Morriss, R., Leese, M., Chatwin, J., Baldwin, D., &, Grp, T. S.. (2008). Inter-rater reliability of the Hamilton Depression Rating Scale as a diagnostic and outcome measure of depression in primary care. Journal of Affective Disorders, 111, 204e 213. Olatunji, B. O., Cisler, J. M., & Tolin, D. F. (2010). A meta-analysis of the influence of comorbidity on treatment outcome in the anxiety disorders. Clinical Psychology Review, 30, 642e654.

Powers, M. B., Halpern, J. M., Ferenschak, M. P., Gillihan, S. J., & Foa, E. B. (2010). A meta-analytic review of prolonged exposure for posttraumatic stress disorder. Clinical Psychology Review, 30, 635e641. Resick, P. A., Bovin, M. J., Calloway, A. L., Dick, A. M., King, M. W., Mitchell, K. S., et al. (2012). A critical evaluation of the complex PTSD literature: implications for DSM-5. Journal of Traumatic Stress, 25, 241e251. Resick, P. A., Suvak, M. K., Johnides, B. D., Mitchell, K. S., & Iverson, K. M. (2012). The impact of dissociation on PTSD treatment with cognitive processing therapy. Depression and Anxiety, 29, 718e730. Robjant, K., & Fazel, M. (2010). The emerging evidence for narrative exposure therapy: a review. Clinical Psychology Review, 30, 1030e1039. Rufer, M., Held, D., Cremer, J., Fricke, S., Moritz, S., Peter, H., et al. (2006). Dissociation as a predictor of cognitive behavior therapy outcome in patients with obsessiveecompulsive disorder. Psychotherapy and Psychosomatics, 75, 40e46. Schafer, J. L. (1999). Multiple imputation: a primer. Statistical Methods in Medical Research, 8, 3e15. Schafer, J. L., & Graham, J. W. (2002). Missing data: our view of the state of the art. Psychological Methods, 7, 147e177. Schauer, M., Neuner, F., & Elbert, T. (2005). Narrative exposure therapy: A short-term intervention for traumatic stress disorders after war, terror, or torture. Ashland, OH, US: Hogrefe & Huber Publishers. Schottenbauer, M. A., Glass, C. R., Arnkoff, D. B., Tendick, V., & Gray, S. H. (2008). Nonresponse and dropout rates in outcome studies on PTSD: review and methodological considerations. Psychiatry: Interpersonal and Biological Processes, 71, 134e168. Simon, G. E., & Perlis, R. H. (2010). Personalized medicine for depression: can we match patients with treatments? The American Journal of Psychiatry, 167, 1445e 1455. Speckens, A. E. M., Ehlers, A., Hackmann, A., & Clark, D. M. (2006). Changes in intrusive memories associated with imaginal reliving in posttraumatic stress disorder. Journal of Anxiety Disorders, 20, 328e341. Spitzer, C., Barnow, S., Freyberger, H. J., & Grabe, H. J. (2007). Dissociation predicts symptom-related treatment outcome in short-term inpatient psychotherapy. Australian and New Zealand Journal of Psychiatry, 41, 682e687. Stein, D. J., Koenen, K. C., Friedman, M. J., Hill, E., McLaughlin, K. A., Petukhova, M., et al. (2013). Dissociation in posttraumatic stress disorder: evidence from the World Mental Health Surveys. Biological Psychiatry, 73, 302e312. Steketee, G., & Chambless, D. L. (1992). Methodological issues in prediction of treatment outcome. Clinical Psychology Review, 12, 387e400. Stenmark, H., Catani, C., Neuner, F., Elbert, T., & Holen, A. (2013). Treating PTSD in refugees and asylum seekers within the general health care system. A randomized controlled multicenter study. Behaviour Research and Therapy, 51, 641e 647. Steuwe, C., Lanius, R. A., & Frewen, P. A. (2012). Evidence for a dissociative subtype of PTSD by latent profile and confirmatory factor analyses in a civilian sample. Depression and Anxiety, 29, 689e700. Taylor, S. (2003). Outcome predictors for three PTSD treatments: exposure therapy, EMDR, and relaxation training. Journal of Cognitive Psychotherapy, 17, 149e161. Trajkovi c, G., Star cevi c, V., Latas, M., Lestarevi c, M., Ille, T., Bukumiri c, Z., et al. (2011). Reliability of the Hamilton Rating Scale for Depression: a meta-analysis over a period of 49 years. Psychiatry Research, 189, 1e9. Warner, R. M. (2013). Applied statistics: From bivariate through multivariate techniques (2nd ed.). Thousand Oaks, CA, US: Sage Publications, Inc. Weathers, F. W., Keane, T. M., & Davidson, J. R. (2001). Clinician-administered PTSD scale: a review of the first ten years of research. Depression and Anxiety, 13, 132e 156. Wolf, E. J. (2013). The dissociative subtype of PTSD: rationale, evidence, and future directions. PTSD Research Quarterly, 24, 1e8. Wolf, E. J., Lunney, C. A., Miller, M. W., Resick, P. A., Friedman, M. J., & Schnurr, P. P. (2012). The dissociative subtype of PTSD: a replication and extension. Depression and Anxiety, 29, 679e688. Wolf, E. J., Miller, M. W., Reardon, A. F., Ryabchenko, K. A., Castillo, D., & Freund, R. (2012). A latent class analysis of dissociation and posttraumatic stress disorder: evidence for a dissociative subtype. Archives of General Psychiatry, 69, 698e705.