Experiences of war and organised violence in combination with post-migration stressors Reference Bogic, Ajdukovic, Bremner, Franciskovic, Galeazzi and Kucukalic1 leave asylum seekers and refugees in Western countries at relatively high risk of developing post-traumatic stress disorder (PTSD). 2,Reference Fazel, Wheeler and Danesh3 PTSD in adult refugees has been shown to pose a burden not only for individuals and their families, Reference Van Ee, Mooren, Kleber, Pat-Horenczyk, Brom, Chemtob and Vogel4 but also for communities at large. Reference Kivling-Bodén and Sundborn5 Psychological treatment for refugees with chronic PTSD, although imperative, is a great clinical challenge. Reference Gorst-Unsworth and Goldenberg6 According to evidence-based guidelines for the treatment of adults with chronic PTSD, trauma-focused cognitive–behavioural therapy (TFCBT) and eye movement desensitisation and reprocessing (EMDR) therapy should be offered to all patients with this disorder. 7,Reference Tol, Barbui and Van Ommeren8 With refugee patients the experience-based recommendation is often made that trauma-focused treatment should be preceded or even replaced by stabilisation. 7,Reference Cloitre, Courtois, Ford, Green, Alexander and Briere9 Trauma-focused therapy per se, especially for refugees living in unstable conditions, has been suggested to cause unmanageable distress Reference Nickerson, Bryant, Silove and Steel10 and to be inappropriate and ineffective. 7 Nevertheless, systematic reviews of the psychological treatment of refugees have shown TFCBT and narrative exposure therapy (NET) to be safe and efficacious with refugees in various social conditions. Reference Nickerson, Bryant, Silove and Steel10–Reference Robjant and Fazel12 However, no full, high-quality randomised trials of EMDR therapy with refugees have yet been conducted. To determine the safety and efficacy of EMDR therapy in traumatised refugees, we designed a trial in which adult asylum seekers and refugees with chronic PTSD were randomly assigned to either EMDR therapy or stabilisation. In line with a pilot study Reference Ter Heide, Mooren, Kleijn, De Jongh and Kleber13 and evidence-based guidelines, our first hypothesis was that EMDR therapy would not differ from stabilisation in the occurrence of harms (defined as symptom increase and drop-out related to symptom increase). Our second hypothesis was that EMDR therapy would be more efficacious than stabilisation in reducing trauma-related symptoms (PTSD, anxiety and depression) and improving quality of life.
Method
The trial was performed at Foundation Centrum '45, a highly specialised Dutch centre for diagnostics and treatment of psychotrauma resulting from persecution, war and violence. Centrum '45 receives national referrals of patients considered too complex to be treated in their own municipalities. Participants were enrolled at two out-patient teams for refugees (in the towns of Oegstgeest and Diemen). Patients judged eligible for participation were asked by their intake therapists if they wished to receive any information about the study and, upon consent, were informed about the study by a research associate. Both treatments were presented as aimed at diminishing PTSD symptoms: EMDR through desensitisation of traumatic memories, stabilisation through enhancement of coping with PTSD symptoms and stressful circumstances in the here-and-now. Those willing to participate signed an informed consent form and were then interviewed with the MINI International Neuropsychiatric Interview (MINI) Reference Sheehan, Lecrubier, Harnett-Sheehan, Amorim, Janavs and Weiller14 to formally check inclusion and exclusion criteria. Data collection took place from September 2009 until August 2012. The trial was approved by the medical ethics committee of the University of Leiden. Trial registration: NARCIS (Dutch National Academic Research and Collaborations Information System) OND1324839; ISRCTN20310201.
Study entry criteria
Refugees who applied for treatment at Centrum '45 were eligible for participation if they were at least 18 years of age, met the criteria for a PTSD diagnosis according to the DSM-IV-TR, 2 and asked for individual therapy to diminish their PTSD symptoms. Patients who had at some point claimed asylum in The Netherlands – irrespective of whether their claim had been met or rejected or was still under consideration – were defined as ‘refugee’. Patients were excluded if they had disorders that acutely threatened their mental or physical health (i.e. depression with high suicidal intent or psychotic features, psychotic disorder, bipolar disorder and severe self-harm or eating disorders) or that interfered with their ability to participate (i.e. alcohol or substance dependence and cognitive disorders). No restrictions were placed on either refugee status or language proficiency. No other psychotherapeutic treatment could take place during the study, and psychotropic medication had to be kept stable from 2 months before treatment until the post-treatment assessment. For those participants who developed high suicidal intent, a psychotic disorder or another serious psychiatric disorder during the study, a psychiatric consultation was prescribed during which the necessity of prescribing or changing psychotropic medication (using a medication protocol) and of discontinuation of the intervention were to be evaluated.
Trial design
A two-arm design was used in which participants were randomly assigned to either 12 h (9 sessions) of EMDR therapy or 12 h (12 sessions) of stabilisation as usual. To create maximum ecological validity, the recommended session length of EMDR and stabilisation was preserved, 7 and conditions were equated on number of treatment hours rather than number of sessions. Twelve treatment hours is considered a minimum to reach improvement in multiply traumatised patients. 7
Blocked, simple randomisation was conducted with the latest two participants who had satisfied the inclusion criteria at the same study site forming a block. Participants were assigned to their experimental group through flipping a coin: the outcome (EMDR therapy for heads, stabilisation for tails) was assigned to the participant lowest in the alphabet. An independent research associate who was not otherwise involved in the inclusion process performed randomisation. As both the EMDR condition and the stabilisation condition contained active treatment elements, the design was complemented with a naturalistic waitlist condition to control for time (see online supplement DS1). Primary outcome measures consisted of the change in PTSD symptom severity and diagnosis, both clinician-rated (Clinician-Administered PTSD Scale, CAPS) Reference Blake, Weathers, Nagy, Kaloupek, Gusman and Charney15 and self-administered (Harvard Trauma Questionnaire, HTQ). Reference Mollica, Caspi-Yavin, Lavelle, Tor, Yang and Chan16 Secondary outcome measures were changes in anxiety and depression (Hopkins Symptom Checklist, HSCL-25) Reference Mollica, Wyshak, de Marnette, Tu, Yang and Khuon17 and quality of life (World Health Organization Quality of Life Assessment, WHOQOL-BREF). 18 Harms were defined as an increase in PTSD symptoms of at least ten points on the CAPS Reference Schnurr, Friedman, Foy, Shea, Hsieh and Lavori19 and premature termination of the study treatment because of symptom increase. All instruments were administered before treatment (T 1), 2 weeks post-treatment (T 2) and at a 3-month follow-up (T 3). After T 2, participants converted to care as usual.
Interventions
EMDR therapy is a trauma-focused treatment that consists of several steps including treatment planning, preparing the patient for trauma-focused treatment, desensitisation and reprocessing, and evaluation. During desensitisation and reprocessing (which is considered the main active element) a focus on a traumatic image and the thoughts, sensations, feelings and memories that it elicits, is combined with an attention-demanding task such as tracking the therapist's fingers with the eyes. Reference Shapiro20 The EMDR condition started with three 60 min sessions dedicated to treatment planning and preparation (including discussing study course and use of interpreters, discussing the patient's explanatory model and subsequently providing psychoeducation on PTSD and EMDR, and making a timeline of traumatic experiences and symptoms). Traumatic memories that were expected to lead to the greatest remission of PTSD symptoms were selected for desensitisation. The preparatory sessions were followed by six 90 min desensitisation sessions, using the Dutch version of the EMDR protocol. Reference De Jongh and Ten Broeke21 Stabilising interventions were proscribed. The EMDR condition was performed by seven clinical psychologists, one physician/psychotherapist and two psychotherapists. Only therapists who had earlier in their career completed an accredited advanced EMDR course participated. Their average experience in EMDR therapy was 5.3 years (s.d. = 2.9). They received monthly supervision by a licensed EMDR supervisor.
The stabilisation condition consisted of 12 sessions of 60 min stabilisation as usual. In phase-oriented treatment for PTSD, the first phase or stabilisation phase is aimed at enhancing safety, control over symptoms and socio-psychological competencies through interventions such as emotion regulation and relational skills building, stress management and cognitive restructuring; processing of traumatic memories is left until the second phase. Reference Cloitre, Courtois, Ford, Green, Alexander and Briere9 Stabilisation as usual, rather than a structured form of stabilisation, was chosen as a control condition to reflect the regular non-structured stabilisation offered in European mental healthcare centres for refugees. Reference Neuner, Kurreck, Ruf, Odenwald, Elbert and Schauer22,Reference Stenmark, Catani, Neuner, Elbert and Holen23 Therapists were asked to select stabilising interventions to match their patient's needs. Exposure to traumatic memories was proscribed. Stabilisation was performed by three clinical psychologists, five psychotherapists, one physician/systemic therapist, one psychiatrist, one social psychiatric nurse, two psychiatrists in training and one psychotherapist in training. Those therapists participated who regularly conducted stabilising interventions in their usual care. Their mean number of years of experience in working with traumatised refugees was 9.9 (s.d. = 5.5), which did not differ from EMDR therapists' experience (mean 9.00, s.d. = 5.5; t(22) = −0.41, P = 0.75). Stabilisation therapists received monthly supervision from a registered cognitive–behavioural and family therapy supervisor/trainer with a specialisation in trauma therapy.
For both conditions, therapist manuals were designed with information on the study methods (such as study design and rules for drop-out), study treatment (such as pre- and proscribed elements), the medication protocol and camera use (manuals available from the authors on request). To assess treatment integrity, treatment sessions were videotaped. For the EMDR condition, a detailed treatment fidelity scale was put together consisting of the scale used in a Dutch EMDR trial Reference Nijdam, Gersons, Reitsma, De Jongh and Olff24 and additional prescribed, proscribed and non-specific elements. For the stabilisation condition, a brief treatment fidelity scale was designed containing prescribed, proscribed and non-specific elements. All interventions were delivered in Dutch when possible and translated by registered interpreters (physically present or by telephone) when necessary.
Measures
PTSD severity and diagnosis were measured by the CAPS and the HTQ. The CAPS yields frequency and intensity scores (ranging from 0 to 4) for all PTSD symptoms according to the DSM-IV-TR in the past week. A symptom was considered present if its frequency was rated as at least 1 and its intensity as at least 2. Reference Weathers, Ruscio and Keane25 To capture full PTSD severity, interviewers referred to clusters of war or persecution experiences rather than to one index traumatic event. The HTQ consists of three parts: one on traumatic events, one on DSM-IV trauma symptoms, and one on other trauma symptoms. Scores for the symptom parts range from 1 (not at all) to 4 (extremely). Anxiety and depression were measured using the HSCL-25, which uses the same scale. Quality of life was measured by the WHOQOL-BREF, which measures four domains of quality of life (physical, psychological, social relationships, and environment) on a scale of 1–5 (a higher score indicating a higher quality of life). All of these instruments have good psychometric properties and are widely used in transcultural research. Reference Weathers, Keane and Davidson26–Reference Skevington, Lotfy and O'Connell29
Interpreters were used whenever the participant did not speak Dutch and the instrument was not available in the participant's native language. To maximise participant understanding of the questionnaires while minimising dependency on individual interpreters' skills, where possible we used questionnaires in the participant's native language that had been carefully translated by our institute (HTQ, HSCL) Reference Kleijn, Hovens and Rodenburg30 or by the WHOQOL Group (WHOQOL-BREF). 18 Interviews were administered by trained Master's students in psychology who were kept masked to treatment condition by having limited access to participant data and by asking participants not to reveal treatment content. They received monthly supervisions of their CAPS ratings using videotaped interviews. Participants received a gift coupon at T 2 and a box of chocolates at T 3.
Statistical analysis
Sample size was calculated with the power analysis program G*Power version 2 for Windows (Erdfelder, Faul and Buchner at the University of Trier, Germany, www.psycho.uni-duesseldorf.de/aap/projects/gpower/). Power calculations were based on outcomes of our pilot study, which resulted in a medium effect size between EMDR and stabilisation on the HTQ. Reference Ter Heide, Mooren, Kleijn, De Jongh and Kleber13 For the main study, a sample size of 36 patients per condition was needed (using a power of 0.80, a two-sided significance level of 0.05 and three repeated measures) to detect a medium between-treatment effect size at T 3. In response to reviews of our pilot study, in the final analyses we used a statistically more advanced variation of the planned analysis strategy (Bayesian latent growth modelling instead of repeated measures analysis), which the sample size allowed for. Reference Van de Schoot, Broere, Perryck, Zondervan-Zwijnenburg and Van Loey31
Treatment fidelity, interrater reliability and demographic and clinical variables were analysed with SPSS version 20.0 for Windows. Chi-squared and t-tests were conducted to check for demographical and clinical differences between participants and those who refused to participate as well as between the two treatment conditions. Mean scale and subscale scores were computed, allowing for a maximum number of three missing values in the HTQ and HSCL and following the questionnaire manual rules for missing values for the WHOQOL-BREF. Chi-squared tests (for treatment condition, gender, refugee status, drop-out, use of an interpreter and work status) and t-tests (for age, distance from home to treatment centre, and CAPS score at baseline) were conducted to explore relationships between missing values and demographic and clinical variables.
Data were then converted to Mplus version 7 (Muthén and Muthén at University of California, www.statmodel.com/). Bayesian estimation was used in all analyses with the default settings in Mplus with regard to prior specifications. Reference Gelman, Carlin, Stern and Rubin32 Bayesian analysis enables full intent-to-treat analysis as missing data are automatically imputed. For the burn-in and convergence criteria we used a minimum of 20 000 iterations after which the Gelman-Rubin convergence criterion Reference Gelman, Carlin, Stern and Rubin32 was used to monitor convergence with a cut-off value of 0.01. Additionally, to ensure convergence was reached, we checked the trace-plots manually. In Bayesian statistics, credibility intervals are used to indicate the 95% probability that the estimate will lie between the lower and upper value of the interval. A treatment effect can be assumed to be present when the credibility interval does not include zero. A difference in treatment effect between conditions can be assumed to be present when credibility intervals between conditions do not overlap. As a measure of effect size between conditions we divided the difference between the linear slopes by the square root of the linear slope's variance (which is equal for the two conditions). Reference Feingold33
Finally, to analyse individual changes in post-traumatic stress scores over time and to identify predictors in the separate treatment conditions, we applied a multigroup latent growth model to the data. Latent growth modelling (LGM) enables an examination of individual growth trajectories for each condition, allowing participants to have a different starting point (i.e. a random intercept model) and a different growth rate (i.e. a random slope model). The R-squared statistic provides the proportion of variance in post-traumatic stress symptoms that is explained by the latent growth factors. The posterior predictive P-value was checked as an indication for model fit (outcome files available from the authors on request).
Results
Participants
Participant flow during the trial is depicted in Fig. 1. The flow diagram includes all patients who met inclusion criteria at intake, i.e. before they were informed about the study and formally interviewed. Reference Spinazzola, Blaustein and Van der Kolk34 Although the inclusion of 72 participants was planned, 2 participants who terminated their participation before the first treatment session, unaware of which condition they had been assigned to, were replaced as it was clear that no post-treatment data could be obtained for them.
Fig. 1 CONSORT flow diagram.
EMDR, eye movement desensitisation and reprocessing; IQR, interquartile range.
a. 34 substance or alcohol dependence, 7 self-harm, 3 cognitive disorder, 5 eating disorder, 10 serious suicidal ideations, 16 psychotic disorder, 1 bipolar disorder.
b. 10 did not want any help at the institute, 30 found participation too much hassle, 18 did not want trauma-focused treatment, 3 did not want stabilisation, 2 did not want treatment for post-traumatic stress disorder (PTSD), 6 refused for various study-related reasons.
c. 3 did not show up for 4 consecutive appointments; 1 thought the travel distance too great; 2 did not want to continue trauma-focused therapy.
d. 6 did not show up for 4 consecutive appointments; 1 developed high suicidal intent; 1 wanted to change to trauma-focused therapy.
Chi-squared and t-tests revealed no significant demographic or clinical differences between participants and those who refused to participate (data available from the authors on request). Table 1 shows demographic and clinical characteristics for the two groups at baseline (see online Table DS1 for a version of this table that details a wider range of demographic characteristics). The EMDR group was found to contain significantly fewer female participants than the stabilisation group. Types of traumatic experiences most frequently reported in the HTQ were being close to death (60/72, 83%), murder of family or friend (54/72, 75%) and threatened with torture (52/72, 72%). Drop-out numbers for the two groups were comparable, with 6 EMDR participants (16.7%) and 8 stabilisation participants (22.2%) discontinuing the intervention (χ2(1,n = 72) = 0.36, P = 0.55). One participant, in the stabilisation group, terminated treatment prematurely because of symptom increase (attempted suicide). In both groups, asylum seekers and those in the country illegally were no more likely to drop out of treatment than participants with a refugee status (EMDR therapy: χ2(1,n = 36) = 0.05, P = 0.83; stabilisation: χ2(1,n = 36) = 0.05, P = 0.83).
Table 1 Demographic and clinical characteristics before treatment a
| EMDR group (n = 36) |
Stabilisation group (n = 36) |
χ2 (d.f.) | t-test (d.f) | P | |
|---|---|---|---|---|---|
| Age, years: mean (s.d.) | 43.1(10.7) | 39.8 (11.9) | 1.26 (70) | 0.21 | |
| Women, n (%) | 6 (16.7) | 14 (38.9) | 4.43 (1) | 0.04 | |
| Types of traumatic experiences, HTQ: mean (s.d.) | 13.8 (5.5) | 13.7 (5.6) | 0.85 (70) | 0.93 | |
| Years with PTSD, mean (s.d.) | 7.9 (7.2) | 8.0 (6.5) | −0.41 (59) | 0.97 | |
| Comorbid depression, n (%) | 28 (77.8) | 28 (77.8) | 0.00 (1) | 1.00 | |
| On psychotropic medication, n (%) | 21 (58.3) | 21 (58.3) | 0.00 (1) | 1.00 | |
| CAPS symptom severity, mean (s.d.) | 74.7 (18.0) | 78.3 (18.3) | −0.83 (70) | 0.41 | |
| Use of interpreter during study treatment, n (%) | 20 (55.6) | 20 (55.6) | 0.00 (1) | 1.00 | |
EMDR, eye movement desensitisation and reprocessing; HTQ, Harvard Trauma Questionnaire; PTSD, post-traumatic stress disorder; CAPS, Clinician-Administered PTSD Scale.
a. See online Table DS1 for a table that details a wider range of demographic and clinical characteristics.
Treatment integrity and content
Treatment integrity was rated by four trained graduate-level research assistants. We randomly selected 12.5% of all treatment sessions for rating of treatment fidelity. Interrater agreement was determined for the first three ratings for both conditions. As this was consistently high (Cronbach's α = 0.95 for EMDR and 1.0 for stabilisation), interrater agreement was deemed to be satisfactory and was not monitored thereafter.
For the EMDR group, 36 out of 291 sessions were rated, a third of which were taken from the preliminary sessions and two-thirds of which from the EMDR protocol sessions. The mean treatment fidelity score for the preliminary sessions was 97.0 (s.d. = 3.0) and for the protocol sessions 87.8 (s.d. = 9.2) on a scale of 0–100. For EMDR treatment completers (n = 30), subjective unit of distress (SUD) scores decreased significantly from the start of treatment to the end of treatment (from a mean of 8.3, s.d. = 1.7 to a mean of 3.9, s.d. = 3.7, t(29) = 7.5, P<0.001), with only 11 participants reaching the desired SUD of 0–1. Mean number of targets treated was 1.6 (s.d. = 1.0, range 1–5), with most participants (19/30, 63%) staying with one target.
For the stabilisation group, 48 out of 387 sessions were rated. The mean treatment fidelity for the stabilisation condition was 88.7 (s.d. = 8.9) on a scale of 0–100. For these 48 sessions the main interventions were registered using an intervention menu. Reference Ter Heide, Mooren, Kleijn, De Jongh and Kleber13 The most frequently registered interventions were discussing and teaching of coping strategies, identification and validation of negative emotions, and active problem-solving by participant and therapist.
For three EMDR participants and three stabilisation participants a change of medication took place during the study. In most cases this entailed a new or changed prescription for antidepressants.
Reliability
To assess interrater reliability, 12.5% of all interviews (MINI and CAPS) were randomly selected, using stratification for time of assessment. Interrater reliability for the decision whether or not to include a patient in the study (using the MINI; 12.5%, 12/94) was excellent (a Cronbach's α of 1). Interrater reliability for the CAPS (12.5%, 25/198) was excellent for PTSD symptom severity (Cronbach's α = 0.95) and good for PTSD diagnosis (Cohen's κ = 0.78). Internal consistency for all scales was excellent, with a Cronbach's α of 0.86 for the CAPS; 0.88 for HTQ symptoms; 0.90 for the HSCL; and 0.85 for the WHOQOL-BREF.
Missing data
Out of a total database of 1944 total or mean scores (i.e. nine outcome measures administered three times with 72 participants), 186 scores (9.6%) were missing and automatically imputed for each Bayesian analysis. ‘Missingness’ was significantly related to drop-out, with those who ended participation prematurely being more likely to have missing data than those who completed the study (χ2(1,n = 72) = 12.85, P<0.001).
Outcomes
Primary outcomes
Table 2 describes PTSD diagnoses for the two groups at each assessment.
Table 2 Post-traumatic stress disorder (PTSD) diagnosis and clinically significant change in PTSD severity in the eye movement desensitisation and reprocessing (EMDR) therapy and stabilisation groups
| n/N (%) | ||||
|---|---|---|---|---|
| EMDR group | Stabilisation group | χ2 (d.f.) | P | |
| CAPS diagnosis | 30/36 (83) | 32/36 (89) | ||
| T 1 | 21/33 (64) | 20/29 (69) | 0.47 (1) | 0.50 |
| T 2 | 26/32 (81) | 22/31 (71) | 0.08 (1) | 0.78 |
| T 3 | 0.92 (1) | 0.34 | ||
| CAPS severity change T 1–T 3 | 0.23 (2) | 0.89 | ||
| Deterioration (⩾−10 points) | 7/32 (21.9) | 8/31 (25.8) | ||
| No change (<10 points to >−10 points) | 12/32 (37.5) | 10/31 (32.3) | ||
| Improvement (⩾10 points) | 13/32 (40.6) | 13/31 (41.9) | ||
CAPS, Clinician-Administered PTSD Scale.
Numbers indicate that outcomes displayed a slightly quadratic development over time for both groups. Between T 1 and T 3, the majority of assessment completers in both groups achieved a clinically significant improvement in PTSD severity (defined as improving at least ten points on the CAPS, Table 2). Reference Schnurr, Friedman, Foy, Shea, Hsieh and Lavori19
In the LGM analyses, best model fit was obtained when including a quadratic slope. Table 3 shows the results of the intent-to-treat analyses for primary outcomes. Participants in both groups initially achieved a clinically significant improvement in clinician-rated PTSD severity, which was partly lost after T 2. The EMDR group significantly improved in self-reported PTSD symptoms according to DSM-IV. No significant differences between the two groups were found in either linear or quadratic slopes and effect sizes between the groups were small (for within- and between-treatment effect sizes based on unimputed data, see online Table DS2).
Table 3 Intent-to-treat analyses of the effects of treatment on post-traumatic stress disorder (PTSD), anxiety, depression and quality of life in the eye movement desensitisation and reprocessing therapy (EMDR) and stabilisation group
| Credibility interval | Credibility interval | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Intercept | Slope | Lower 2.5% | Upper 2.5% | Effect size a | Q slope | Lower 2.5% | Upper 2.5% | Effect size a | |
| Primary outcomes | |||||||||
| CAPS, severity | |||||||||
| EMDR therapy group | 74.71 | −11.68 | −24.91 | 1.12 | −0.04 | 4.66 | −1.70 | 10.50 | −0.03 |
| Stabilisation group | 78.30 | −12.75 | −27.26 | 1.61 | 4.33 | −2.55 | 10.98 | ||
| HTQ, DSM-IV | |||||||||
| EMDR therapy group | 3.06 | −0.40* | −0.74 | −0.04 | 0.20 | 0.15 | −0.01 | 0.30 | −0.29 |
| Stabilisation group | 3.19 | −0.25 | −0.57 | 0.06 | 0.06 | −0.08 | 0.20 | ||
| HTQ, total | |||||||||
| EMDR therapy group | 2.85 | −0.31 | −0.63 | 0.02 | 0.29 | 0.12 | −0.02 | 0.27 | −0.38 |
| Stabilisation group | 2.90 | −0.11 | −0.41 | 0.18 | 0.02 | −0.11 | 0.14 | ||
| Secondary outcomes | |||||||||
| HSCL, anxiety | |||||||||
| EMDR therapy group | 2.86 | −0.09 | −0.43 | 0.25 | 0.09 | 0.02 | −0.13 | 0.16 | −0.15 |
| Stabilisation group | 3.04 | −0.01 | −0.45 | 0.43 | −0.03 | −0.23 | 0.17 | ||
| HSCL, depression | |||||||||
| EMDR therapy group | 2.95 | −0.20 | −0.55 | 0.16 | −0.03 | 0.06 | −0.09 | 0.21 | 0.03 |
| Stabilisation group | 2.97 | −0.22 | −0.59 | 0.18 | 0.07 | −0.09 | 0.23 | ||
| WHOQOL-BREF, physical | |||||||||
| EMDR therapy group | 2.35 | 0.07 | −0.36 | 0.49 | 0.07 | −0.03 | −0.23 | 0.16 | −0.10 |
| Stabilisation group | 2.34 | −0.01 | −0.44 | 0.43 | 0.01 | −0.18 | 0.21 | ||
| WHOQOL-BREF, psychological | |||||||||
| EMDR therapy group | 2.34 | 0.00 | −0.47 | 0.47 | 0.07 | −0.03 | −0.22 | 0.17 | −0.15 |
| Stabilisation group | 2.35 | −0.09 | −0.61 | 0.44 | 0.04 | −0.19 | 0.27 | ||
| WHOQOL-BREF, social relationships | |||||||||
| EMDR therapy group | 2.71 | 0.05 | −0.48 | 0.58 | −0.28 | −0.09 | −0.33 | 0.15 | 0.22 |
| Stabilisation group | 2.56 | 0.38 | −0.18 | 0.95 | −0.20 | −0.46 | 0.05 | ||
| WHOQOL-BREF, environment | |||||||||
| EMDR therapy group | 2.99 | −0.24 | −0.61 | 0.14 | −0.52 | 0.06 | −0.11 | 0.24 | 0.38 |
| Stabilisation group | 2.68 | 0.22 | −0.17 | 0.63 | −0.08 | −0.26 | 0.10 | ||
CAPS, Clinician-Administered PTSD Scale; HTQ, Harvard Trauma Questionnaire; HSCL, Hopkins Symptom Checklist; WHOQOL-BREF, World Health Organization Quality of Life Assessment.
a. Cohen's d: 0.20 small, 0.50 medium, 0.80 large. Reference Cohen35
* P<0.05.
Secondary outcomes
No significant differences were found between the EMDR therapy and stabilisation group on any of the secondary outcome measures (Table 3). Neither intervention had a significant effect on anxiety, depression or quality of life. Quality of life, in fact, in both groups did not show uniform improvement.
Post hoc analyses
Gender
As the EMDR group contained significantly fewer female participants than the stabilisation group, we added gender as a covariate to the LGM model. This, however, led to a decreased model fit. We therefore analysed the effect of gender on the slopes of the primary outcome measures. Low R squares for all measures and conditions showed that gender had little influence on treatment effect (online Table DS3).
Refugee status
As some clinicians argue that EMDR therapy with asylum seekers is not possible because of their insecure living conditions, we also analysed the effect of refugee status on the primary outcome measures. We divided the groups into participants with no refugee status (i.e. asylum seekers and those staying in the country illegally) and those with temporary or permanent refugee status. The direction of the effect was that participants without a refugee status, regardless of treatment group, showed more PTSD symptom reduction than participants with a refugee status. Although not statistically significant, in the stabilisation group the effect size was medium (online Table DS3).
Discussion
Main findings
In this study, no differences in safety or efficacy were found between EMDR therapy and stabilisation as usual. As previously stated, some clinicians argue that trauma-focused treatment in refugees, especially those living in unstable circumstances, may be harmful. Reference Nickerson, Bryant, Silove and Steel10 However, in this study, the EMDR and stabilisation groups had comparable numbers of participants who dropped out of treatment and participants reporting symptom increase. In fact, drop-out numbers were relatively low compared with other PTSD outcome studies. Reference Hembree, Foa, Dorfan, Street, Kowalski and Tu36 Additionally, in the EMDR as well as the stabilisation group asylum seekers showed an improvement at least equal to that of refugees. Results are in line with an increasing body of evidence suggesting that trauma-focused therapy carries no risk of psychologically overwhelming refugee patients, even those in unstable conditions. Reference Nickerson, Bryant, Silove and Steel10 However, conclusions may not generalise to refugee patients who meet our exclusion criteria – notably those with psychotic disorders, substance dependence or severe suicidal ideations – although the justifiability of using these exclusion criteria in patients who are treated for these disorders has been called into question. Reference Van den Berg, De Bont, Van der Vleugel, De Roos, De Jongh and Van Minnen37
Comparison with a non-randomised waitlist condition suggested that treating refugees with EMDR therapy is more effective than not treating them. However, contrary to expectation, EMDR therapy was found to be no more effective than stabilisation. The effect of stabilisation was similar to effects found for unstructured stabilisation in other refugee samples in Western countries. Reference Neuner, Kurreck, Ruf, Odenwald, Elbert and Schauer22,Reference Stenmark, Catani, Neuner, Elbert and Holen23 However, the effect of EMDR therapy was lower than expected, with effect sizes for other trauma-focused therapies in refugee samples in Western countries ranging from 0.93 to 1.6 for NET Reference Neuner, Kurreck, Ruf, Odenwald, Elbert and Schauer22,Reference Stenmark, Catani, Neuner, Elbert and Holen23 and from 2.4 to 2.6 for exposure. Reference Nickerson, Bryant, Silove and Steel10,Reference Paunovic and Öst38 A primary explanation is that the number of trauma-focused sessions was lower in this study (i.e. 6 sessions of desensitisation and reprocessing) than in comparable studies (9 for NET Reference Neuner, Kurreck, Ruf, Odenwald, Elbert and Schauer22,Reference Stenmark, Catani, Neuner, Elbert and Holen23 and 20 for TFCBT Reference Paunovic and Öst38 ). Considering the high number of types of traumatic experiences (14 in both conditions), six sessions appear to have been insufficient to process all memories driving PTSD symptom severity. Additionally, it is possible that equalisation of number of treatment sessions rather than number of treatment hours would have resulted in differences in efficacy between the two groups, as it did in our pilot study. Reference Ter Heide, Mooren, Kleijn, De Jongh and Kleber13 A second explanation is that the study sample consisted of refugee patients who are relatively difficult to treat. Centrum '45 is a specialised institute that receives national referrals of patients who have insufficiently benefited from, or are expected to insufficiently benefit from, treatment within primary mental healthcare. This may be related to the complexity of their traumatic experiences (i.e. multiple, prolonged, interpersonal traumatic events often involving intentional and extreme cruelty) as well as the complexity of their present-day lives (such as being threatened with expulsion, having no financial means, being socially isolated, fearing the effects of ongoing conflict in the country of origin). Reference Steel, Chey, Silove, Marnane, Bryant and Van Ommeren39 Comparable studies included only participants who were fluent in the language of their resettlement country Reference Paunovic and Öst38 or who could be treated within general healthcare. Reference Stenmark, Catani, Neuner, Elbert and Holen23 A third possible explanation is that TFCBT (including NET) is indeed more effective than EMDR therapy in treating refugees with chronic PTSD. The culturally sensitive rationale Reference Benish, Quintana and Wampold40 and relatively simple protocol of NET might make this therapy easier to grasp for refugees and EMDR therapy may have some catching up to do in those respects. A randomised controlled trial is needed to clarify this issue.
Strengths and limitations
This study is the first full trial that meets all CONSORT criteria to test the safety and efficacy of EMDR therapy in refugees with chronic PTSD. A broad range of refugee patients were engaged in this study, including patients who needed interpreters and highly vulnerable patients who were homeless, stayed in the country illegally or were listed for forced return. Both groups were treated by highly experienced therapists. Bayesian estimation allowed for full intention-to-treat analysis.
The study also has several limitations. Although all instruments used have been extensively validated in refugee or transcultural samples, not all language versions were validated, which may have compromised measurement validity. In addition, a measure of positive expectancy for both therapists and patients would have been useful to explore the influence of treatment preference on treatment outcome. Asking refugees which treatment would be preferable under which circumstances would have yielded valuable information and would have enabled refugees to contribute as experts as well as participants. Finally, inequality in number of treatment sessions between the two groups may have led to a greater risk of treatment drop-out in the stabilisation condition than in the EMDR condition.
Clinical implications
On the basis of our study it may be concluded that therapists need not refrain from offering EMDR therapy to asylum seekers or refugees with chronic PTSD for fear of deterioration, although this conclusion may not necessarily generalise to refugees with comorbid untreated psychosis, substance dependence and high suicidal intent. Offering only a limited number of EMDR sessions, as was the case in this study, may not result in a satisfactory reduction of PTSD and comorbid symptoms. Whether EMDR therapy would show greater efficacy with refugees after a larger number of sessions, or with refugee patients who are referred to general rather than specialised mental healthcare, or when preceded by or combined with stabilising interventions, remains to be tested in future trials. This study adds to an increasing body of evidence that directly targeting traumatic memories of refugees carries no harm.
Funding
This study was jointly funded by ZonMW, The Netherlands organisation for health research and development, and Foundation Centrum ‘45, partner in Arq Psychotrauma Expert Group. R.v.d.S. was supported by a grant from The Netherlands organization for scientific research: NWO-VENI-451-11-008.
Acknowledgements
The authors thank all participants and colleagues at Foundation Centrum '45 who contributed to this study; Berthold Gersons, Jeroen Knipscheer and Geert Smid who served as advisors; Mariëlle Zondervan-Zwijnenburg who provided statistical assistance; and Paula Schnurr and Mirjam Nijdam who commented on earlier versions of this paper.
eLetters
No eLetters have been published for this article.