Trauma-focused and personality disorder treatment for posttraumatic stress disorder and comorbid cluster C personality disorder: a randomized clinical trial

ABSTRACT

Background: Posttraumatic stress disorder (PTSD) is associated with high rates of cluster C personality disorders (PD), which may negatively affect PTSD treatment. It is unknown whether concurrent treatment for PTSD and comorbid PD leads to superior treatment effects, compared to standard trauma-focused treatment.

Objective: The objective was to test the efficacy of adding personality disorder treatment (group schema therapy; GST) to individual trauma-focused treatment (imagery rescripting; ImRs).

Method: A two-arm randomized clinical trial (1:1 allocation ratio) was conducted between 2018 and 2023 at two sites of a mental health institution in the Netherlands. Raters were blind to treatment allocation. Adult outpatients with PTSD and comorbid cluster C personality disorders were randomized to receive either ImRs (12–18 sessions) or ImRs + GST (12–18 ImRs + 52–58 GST). The main outcome was PTSD severity one year after start of treatment measured with the Clinician-Administered PTSD Scale for DSM-5.

Results: Of 130 patients (mean [SD] age = 40.6 [11.2], 110 [85%] females), 66 were assigned to ImRs and 64 to ImRs + GST. At 12 months, there were large decreases in PTSD severity (d_ImRs= 2.42, 95%CI = 1.97–2.87; d_{ImRs + GST}= 2.44, 95%CI = 1.99–2.90), but there was no significant difference between conditions (d = 0.02, 95%CI = −0.33–0.36, p = .944). Reductions in personality disorder symptoms and all other secondary outcomes were observed in both conditions. There were no significant differences between conditions on any of the secondary outcomes at 12 months.

Conclusion: The more intensive concurrent trauma-focused and personality disorder treatment (ImRs + GST) was not superior to trauma-focused treatment alone (ImRs) for patients with PTSD and comorbid CPD. This suggests that trauma-focused treatment is the preferred primary treatment in patients presenting with both internalizing personality disorder and PTSD, reserving the stepping up to more intensive psychotherapy aimed at the personality disorder as a second line of treatment.

Trial registration: ClinicalTrials.gov identifier: NCT03833531

HIGHLIGHTS

• Concurrent trauma-focused and personality disorder treatment was not superior to only trauma-focused treatment for patients with posttraumatic stress disorder (PTSD) and comorbid cluster C personality disorders.

• Large reductions in PTSD severity and medium-to-large reductions in all secondary outcomes, including personality disorder symptoms, were observed in both treatment arms.

• These findings are remarkable, given the higher therapy dosage and specialized treatment for personality disorder comorbidity in the combined treatment arm.

1. Introduction

Posttraumatic stress disorder (PTSD) (American Psychiatric Association, 2013) is associated with high rates of comorbid personality disorders (PD) (Friborg et al., 2013). Meta-analytical evidence suggests that comorbid PD significantly attenuates the reduction of PTSD symptoms after PTSD treatment, although effect sizes are still large for patients with and without PD (Snoek et al., 2021). The authors propose that the integration of PTSD- and PD-specific interventions could enhance treatment effects, in line with earlier recommendations by Rizvi and Harned to move away from single-disorder treatment and concurrently treat comorbid disorders to reflect clinical realities and enhance treatment outcomes (Rizvi & Harned, 2013). This raises the question whether concurrent treatment of comorbid PD alongside PTSD treatment may lead to superior treatment outcomes, compared to standard trauma-focused treatment (TFT). Little is known about the optimal treatment for those with PTSD and PD and no evidence-based guideline for concurrent treatment exists. Some have argued for trauma-focused treatment in most if not all cases of comorbidity (van Minnen et al., 2012), while others have argued for an integrated (Ingenhoven, 2015) approach.

Nearly all research on PTSD and PD comorbidity has focused on borderline PD (Harned, 2014). However, cluster C PD (CPD, i.e. avoidant, obsessive-compulsive and dependent PD) comorbidity rates are among the highest in PTSD. Friborg and colleagues report a mean weighted proportion of .63 for CPD comorbidity based on 199 cases, with .23, .08, and .20 for avoidant, dependent, and obsessive-compulsive PD based on 1076 cases, respectively. For comparison, the mean weighted proportion of borderline PD was .22 based on 1401 cases (Friborg et al., 2013). Moreover, CPD are associated with high levels of impairment, economic costs, interpersonal problems, suicidality, all-cause mortality and low quality of life (de Reus & Emmelkamp, 2012; Disney, 2013; Lampe & Malhi, 2018). Yet, they are underrecognized and poorly studied, although some evidence suggests that schema therapy is effective (Bamelis et al., 2014) and more research is currently being conducted (Baljé et al., 2016; Daniëls et al., 2023; Groot et al., 2022). From a clinical perspective, the pervasive behavioral and experiential avoidance, passivity, control mechanisms, and hypersensitivity to rejection and criticism exhibited by patients with CPD may complicate PTSD treatment, in which exposure to fear-eliciting stimuli is an important element. For example, avoidance of fear-eliciting stimuli may result in difficulties in building a strong patient-therapist working alliance, suboptimal treatment engagement and poor treatment attendance. Indeed, higher levels of in-session avoidance behavior predicted higher dropout rates in cognitive processing therapy for PTSD (Shayani et al., 2023). These findings are important because attachment style is highly relevant to assessment and treatment. For example, it is more difficult for a therapist to establish and maintain a relationship with an individual who is distrustful of others, who is hypersensitive to criticism and rejection, and who relies on avoidant coping strategies. This may lead to attenuated treatment effects and higher dropout rates. Thus, the concurrent addition of schema therapy to TFT, which is aimed at breaking through such long-standing maladaptive mechanisms, may increase the effectiveness of TFT.

We present the first randomized clinical trial (RCT) directly comparing stand-alone TFT with and without personality disorder treatment in a sample of patients with PTSD and comorbid CPD. Specifically, we tested whether individually delivered, PTSD-focused imagery rescripting (ImRs, not to be confused with imagery rehearsal therapy, a cognitive behavioral approach to treating nightmares) (Krakow & Zadra, 2006) combined with group schema therapy (GST) was more effective than stand-alone ImRs. While ImRs is currently not included among the first-line recommended psychological interventions for PTSD in most guidelines, ImRs has been shown to be an evidence-based, effective and tolerable treatment for PTSD, specifically childhood-related PTSD (Boterhoven de Haan et al., 2020; Raabe et al., 2022). Additionally, ImRs is commonly utilized in schema therapy to address various adverse experiences linked to PD pathology, and its application aligns well with the principles of schema therapy. Opting for ImRs alongside GST eliminates the necessity for patients to familiarize themselves with an extra therapeutic technique and its underlying rationale, thereby streamlining their involvement in both ImRs and GST simultaneously. Finally, GST has been shown to be effective in treating CPD (Arntz et al., 2022; Jacob & Arntz, 2013; Rameckers et al., 2021). Therefore, based on the attenuated PTSD treatment effects in the case of PD comorbidity, and the recommendations to concurrently treat comorbidity, our primary hypothesis was that PTSD symptom severity reductions would be larger after concurrent ImRs + GST compared to ImRs only. Further, it was hypothesized that concurrent ImRs + GST would outperform ImRs on a range of secondary outcome variables: self-report PTSD severity, CPD symptoms, quality of life, functioning, general psychiatric symptom severity, treatment response, PTSD remission and dropout. The current study is part of a project in which the effect of adding PD treatment to TFT is investigated in CPD and borderline PD in two distinct RCTs (Snoek et al., 2020).

2. Methods

2.1. Study design

A single-blinded RCT with two arms was conducted at two sites of the Sinai Center, a mental health institution specialized in PTSD treatment in The Netherlands. The present study was registered under NCT03833531 on ClinicalTrials.gov and approved by the regional Medical Ethics Committee (reference number 2017.335) and the study design has been published elsewhere (van den End et al., 2021). The study was conducted in accordance with the Dutch Medical Research Involving Human Subjects Act. There were some deviations to the original research protocol, for example due to measures taken by the Dutch government in response to the 2020 COVID-19 pandemic (see supporting information).

2.2. Patients

Patients were eligible for the study if they were aged 18–65 years and had a primary diagnosis of post-traumatic stress disorder (PTSD) according to DSM-5 criteria diagnosed during the regular intake procedure. In addition, at least three symptoms of avoidant PD, four symptoms of dependent PD, or three symptoms of obsessive-compulsive PD were required. The reason for including patients with a subthreshold number of symptoms for any particular CPD was the recognition that some patients do not meet the full diagnostic criteria for a specific CPD, but satisfy the general criteria for PD and exhibit symptoms from various categories of CPD (often diagnosed as other specified personality disorder or mixed personality disorder). Finally, in the case of psychotropic medication use, a stable regimen for at least three weeks was required. Exclusion criteria were current psychosis, severe aggression, treatment-interfering substance and eating disorders or somatic problems, IQ < 70, and insufficient mastery of the Dutch language.

2.3. Randomization and blinding

Patients were randomized to one of two conditions by computerized block randomization (allocation ratio 1:1, n = 6 per block) after screening for inclusion and exclusion criteria and obtaining informed consent. In the case of both CPD and borderline PD comorbidity during the screening phase, patients were given the choice to participate in one of two trials. The other trial compared eye movement desensitization and reprocessing with versus without dialectical behavior therapy for patients with PTSD and comorbid borderline PD. Sequentially numbered, sealed envelopes containing the allocation information were prepared by an independent administrative assistant unfamiliar with the study and opened together with the patient. Assessors were blind to treatment allocation. Blinding patients and therapists was not possible for these psychological treatments.

2.4. Procedures

2.4.1. Therapists and assessors

All therapists had at least three years of experience in evidence-based trauma-focused therapy and completed at least 100 h of accredited training in cognitive–behavioral therapy. Additionally, all ImRs therapists received training from an expert in ImRs and all schema therapists completed at least 25 h of accredited schema therapy training. To ensure treatment fidelity in addition to formal adherence and competence analysis, therapists were required to attend biweekly supervision sessions by experienced ImRs and/or schema therapists in which audiotaped (ImRs) and videotaped (ST) material was discussed. Some therapists provided both GST and ImRs but they did not provide ImRs to patients for whom they served as GST therapist.

Clinical interviews were conducted by adequately trained and supervised graduate-level research assistants who attended weekly meetings in which audiotaped material was scored and discussed to ensure interrater reliability. Questionnaires were provided digitally.

2.4.2. Treatments and timing of measurements

Patients in the ImRs + GST condition started with four pre-treatment sessions. In these sessions, the schema mode case conceptualization was created. T0 immediately followed the pre-treatment. GST started after T0 and ImRs was introduced for patients in this condition after approximately six weeks (up to a maximum of three months depending on therapist availability). Patients in the ImRs only condition received 12–18 weekly sessions of ImRs starting after T0. For both conditions, T1 followed three months after T0. Moreover, T2 was linked to the end of trauma-focused treatment in both conditions. T3, T4 and T5 took place 3, 6 and 9 months after T2, respectively. See Figure S1 in the supporting information for a visual representation of the study design and timing of measurements.

ImRs is a standalone intervention for PTSD. It is aimed at changing emotional and cognitive aspects of aversive memories by changing the course of events within the traumatic memory by using imagination. A detailed description of the ImRs procedure has been published elsewhere (van den End et al., 2021). The ImRs protocol was based on the protocol used in a previous study on ImRs and consisted of a minimum of twelve and a maximum of 18 75 min individual sessions (Raabe et al., 2015). After 12 sessiTreatment was prolonged with a maximum of six sessions when indicated by a multidisciplinary team based on symptom severity, patient preference and therapist judgment. Clinicians were instructed to complete treatment within six months.

Schema therapy is aimed at identifying and meeting core emotional needs that have not been met in early life (i.e. secure attachment, autonomy, freedom to express, spontaneity and play, and realistic limits) (Young et al., 2003). The GST procedure was preregistered (van den End et al., 2021) and the protocol was based on existing manuals (Farrell et al., 2014; Vreeswijk et al., 2012), and consisted of four individual pre-treatment sessions (45 min) and 40 weekly GST sessions (90 min). The goal of the pre-treatment sessions was to acquaint patients with the schema and schema mode concepts and to make an individualized caseconceptualization according to schema therapy principles for every individual group member. Additionally, 12–18 sessions (75 min) of group schema therapy were added in 2–3 blocks divided over a year in which experiential components of the GST protocol were emphasized. These blocks were denoted as schema-focused psychomotor therapy in the original research protocol to emphasize the experiential focus of these group schema therapy sessions. To be sure, these experiential GST blocks consisted of evidence-based GST interventions, such as chair work, mode awareness exercises, and imagery work. Finally, patients were offered three optional individual booster sessions posttreatment (i.e. schema therapy or ImRs). The GST protocol did not contain ImRs sessions targeting PTSD A-criterion traumata. Treatment protocols are available upon request.

Protocol adherence and therapist competence were rated by three independent, graduate-level research assistants who were trained in the use of the rating instruments by the first author. A random selection of audiotaped ImRs sessions was rated with the ImRs Therapist Adherence and Competence Protocol (Raabe et al., 2015). Sessions were rated on the presence and quality of 37 elements of ImRs and averaged, resulting in an adherence (number of present elements) and competence (1 = not provided according to the protocol to 4 = intervention was provided well by the therapist) score. GST delivery was rated with the Group Schema Therapy Rating Scale – Revised (Zarbock et al, 2014) by averaging adherence (number of present elements) and competence (0 = very poor to 6 = excellent) scores for 28 items on general therapist behavior, limited reparenting, group climate, structure, and therapist team quality.

2.5. Outcomes

A range of clinical and demographic variables were assessed at baseline, including psychiatric comorbidity (Structured Clinical Interview for DSM-5 Disorders: Clinician Version; First et al., 2016) childhood trauma (Childhood Trauma Questionnaire; Bernstein et al., 2003) potentially traumatic events (Life Event Checklist-5; Weathers et al., 2013) depression severity (Bech Depression Inventory-II; Beck et al., 1996) emotion regulation difficulties (Difficulties in Emotion Regulation Scale; Gratz & Roemer, 2008) and dissociation (Dissociative Experiences Scale-II) (Carlson & Putnam, 1993). All outcome variables were pre-registered (van den End & Thomaes, 2018).

Information on psychometric properties and Cronbach’s alpha values for the primary and secondary outcomes can be found in the supporting information. The primary outcome variable was PTSD severity at T4 (6 months after ImRs) measured with the Clinician-Administered PTSD Scale for DSM-5 (CAPS-5) (Weathers et al., 2018). The CAPS-5 is a structured diagnostic interview assessing the occurrence and severity of 20 PTSD symptoms in four symptom clusters (B, C, D and E) on a 5-point scale, resulting in a total score ranging from 0 to 80. The CAPS-5 is widely regarded as the gold standard to assess the occurrence and severity of PTSD in adults. Secondarily, self-reported PTSD symptom severity was measured with the PTSD Checklist for DSM-5 (PCL-5), a 20-item questionnaire (total score range 0–80) (Blevins et al., 2015).

CPD symptoms were assessed with the Structured Clinical Interview for DSM-5 Personality Disorders (SCID-5-PD) (Arntz et al., 2017). To allow the effect of therapy to be reflected in the scoring, a symptom was scored as absent at T4 if the symptom was not present in the past 6 months. Quality of life, functioning and general psychiatric symptom severity were measured with the EuroQol visual analogue scale (EQ VAS, a 1-item self-report scale ranging from 0 to 100) (Herdman et al., 2011) the World Health Organization Disability Assessment Schedule (WHODAS 2.0, a 36-item self-report questionnaire resulting in an averaged score ranging from 1 to 5; Üstün et al., 2010) and the Outcome Questionnaire-45 (OQ-45, a 45-item self-report questionnaire resulting in a summed score ranging from 0 to 180; de Jong et al., 2007) respectively. Treatment response was pre-registered as a PTSD severity reduction of baseline SD_pooled≥ 1.0. PTSD remission was defined as failure to reach the DSM-5 PTSD criteria threshold at T4 in the original study design. However, due to constraints in the number of variables that we were able to include to achieve a stable multiple imputation model (see statistical analysis), we defined PTSD remission as a CAPS-5 total score below 12 points (the minimally required score to satisfy the diagnostic threshold) at T4. We included PTSD remission defined by the DSM-5 algorithm for the nonimputed dataset (see statistical analysis). Finally, treatment completion was defined as >8 (75%) ImRs sessions attended.

2.6. Statistical analysis

All analyses were performed in SPSS version 23. A sample of n = 126 patients was required to demonstrate a significant and medium (Cohen, 1988) effect size difference (Cohen’s d or equivalent = 0.5) between conditions (van den End et al., 2021). The main analyses were intention-to-treat analyses. Missing data were handled by performing multiple imputation (50 imputations) using fully conditional specification (Collins et al., 2001; Huque et al., 2018).

Continuous outcomes were analyzed with linear mixed models with age, baseline outcome score (all centered), sex, treatment, time, baseline outcome score*time interaction and, finally, treatment*time interaction as fixed effects. The inclusion of covariates may validly increase the precision of the estimates but increases the risk of overfitting the statistical model. We chose to limit the number of covariates in our analysis to these three covariates based on the literature. For transparency, we reported whether analyses of the primary and secondary outcomes without controlling for covariates altered the findings. The primary outcome was analyzed with a random intercept for subject and a scaled identity structure for the repeated part (Twisk, 2013). Secondary outcomes were analyzed similarly, depending on model convergence. Effect sizes were calculated following the formula provided by Hedges and Olkin (Hedges & Olkin, 1985) using estimated marginal means of the linear mixed models per outcome. The baseline pooled standard deviation was derived from the standard error of the estimated marginal means. The threshold for statistical significance was set at α ≤ .05 (two-tailed). Binary outcomes were analyzed with generalized linear models using a binomial distribution with a logit link. Age, sex, condition, and baseline CAPS-5 scores were entered as fixed effects. For the analysis of PTSD remission, we analyzed remission defined as CAPS-5 total score < 12 in the imputed dataset, as constraints in the number of variables in the multiple imputation model prevented imputation on an item-level basis. We performed an additional analysis of remission defined as failure to reach the DSM-5 PTSD threshold in the nonimputed dataset. In addition to the intention-to-treat analysis for the primary outcome, three sensitivity analyses were performed following the linear mixed model approach: with a non-imputed dataset, excluding TFT non-completers and excluding those with subclinical CPD at baseline.

3. Results

A total of 130 patients participated between June 2018 and May 2023 and were randomized to either individual ImRs (n = 66) or concurrent ImRs + GST (n = 64). See Figure 1 for the CONSORT flowchart. Table 1 shows that the sample consisted of predominantly Dutch, female participants, and a minority was currently employed. Almost everyone received prior psychological treatment. Patients had been exposed to high rates of sexual and physical violence, and the majority reported childhood trauma. Avoidant PD was the most common CPD, followed by obsessive-compulsive PD. For those with subclinical CPD (16%) the mean number of CPD symptoms was M (SD) = 4.05 (1.32), and M (SD) = 6.55 (2.58) symptoms of any PD. Therapist competence was high for both ImRs (14 ratings, M = 3.92, SD = 0.07) and GST (11 ratings, M = 4.59, SD = 0.58). Overall protocol adherence was satisfactory, with M (SD) = 27.93 (3.10) out of 37 elements present in ImRs and M (SD) = 23.2 (2.18) out of 28 elements present in GST.

Figure 1.

CONSORT flowchart. Abbreviation: ImRs = imagery rescripting; ST = schema therapy; CAPS-5 = Clinician-Administered PTSD Scale for DSM-5; T2 = post-ImRs; T4 = 6-months post-ImRs.

Table 1.

Demographic and clinical baseline variables for the study sample.

	Conditionb
Variablesa	ImRs	ImRs + GST
Sex
Female	56 (85%)	54 (84%)
Male	10 (15%)	10 (16%)
Age (years)	40.1 (11.2)	41.2 (11.2)
Country of origin
The Netherlands	40 (61%)	38 (59%)
Other/unknownc	26 (39%)	26 (41%)
Education level
Primary education	7 (11.3%)	2 (3.2%)
Lower secondary	9 (14.5%)	12 (19.4%)
Upper secondary	30 (48.4%)	21 (33.9%)
Short cycle tertiary	1 (1.6%)	0 (0%)
Bachelor or equivalent	12 (19.4%)	21 (33.9%)
Master or doctoral level	3 (4.8%)	6 (9.7%)
In a Relationship	39 (59%)	38 (59%)
Has dependants	27 (41%)	29 (45%)
Employment status
Primarily homemaker	4 (6%)	8 (13%)
Primarily currently employed or in education	26 (39%)	15 (23%)
Unemployed and not receiving benefits	1 (2%)	1 (2%)
Receiving (partial) disability benefits	18 (27%)	31 (48%)
Receiving (partial) social benefits	17 (26%)	9 (14%)
Received previous individual outpatient psychological therapy	50 (76%)	51 (80%)
Received previous group outpatient psychological therapy	20 (31%)	14 (23%)
Previous psychiatric hospitalization	11 (17%)	9 (14%)
Current regular drug use	17 (26%)	9 (14%)
Clinician-rated primary PTSD diagnosis	66 (100%)	64 (100%)
CAPS-5 baseline	57 (86%)	60 (94%)
Total score	39.3 (11.0)	42.1 (9.4)
Index trauma
Life threat	39 (68%)	34 (57%)
Serious injury	31 (54%)	38 (63%)
Sexual violence	32 (56%)	37 (62%)
Childhood trauma	40 (74%)	52 (88%)
Age during childhood trauma	8.4 (4.5)	8 (5.2)
PCL-5 baseline	53.1 (13.2)	55.8 (11.6)
CTQ
Physical neglect	10.7 (3.8)	12.2 (4.2)
Emotional neglect	17.8 (5.4)	19.3 (4.5)
Sexual abuse	10.2 (6.3)	10.5 (6.2)
Emotional abuse	14.7 (6.5)	17.4 (6.2)
Physical abuse	10.3 (6.0)	12.7 (5.7)
LEC-5
No of trauma types	4.6 (1.9)	5.2 (2.1)
Physical violence	48 (91%)	56 (95%)
Sexual violence	43 (81%)	52 (88%)
Repeated exposure	44 (83%)	55 (93%)
BDI-II total score	29.1 (11.2)	35.0 (11.2)
OQ-45 total score	96.9 (18.6)	105.3 (19.0)
EQ VAS	55.7 (18.9)	51.9 (20.8)
WHODAS 2.0 total score	2.5 (0.6)	2.7 (0.7)
DERS total score	113.5 (19.4)	118.7 (19.5)
DES-II total score	14.2 (9.6)	20.6 (15.4)
SCID-5 Comorbid disorders
Current depressive episode	38 (58%)	32 (50%)
Persistent depressive disorder	33 (50%)	25 (39%)
Substance and/or alcohol use disorder	11 (17%)	9 (14%)
Any anxiety disorder	35 (53%)	32 (50%)
Any eating disorder	9 (14%)	8 (13%)
Obsessive-compulsive disorder	7 (11%)	6 (9%)
SCID-5 Comorbid PD
Any CPD	56 (85%)	51 (84%)
Subclinical CPD	10 (15%)	10 (16%)
Avoidant	43 (65%)	34 (56%)
Dependent	2 (3%)	2 (3%)
Obsessive-compulsive	17 (26%)	22 (36%)
Borderline	8 (12%)	3 (5%)
Paranoid	3 (5%)	4 (7%)
Schizoid	0 (0%)	1 (2%)
No. of ImRs sessions
Intent-to-treat sample	9.2 (6.1)	9.3 (6.4)
ImRs completers	13.4 (2.8)	13.7 (3.1)
ImRs noncompleters	2.3 (2.7)	2.0 (2.5)
No. of GST sessions
Intent-to-treat sample	–	33.3 (18.7)
GST completers	–	47.6 (7.3)
GST noncompleters	–	18.0 (14.4)

Notes: CAPS-5 = Clinician Administered PTSD Scale for DSM-5; PCL-5 = PTSD Checklist – DSM-5; CTQ = Childhood Trauma Questionnaire; LEC-5 = Life Events Checklist – DSM-5; BDI-II = Beck Depression Inventory – II; OQ-45 = Outcome Questionnaire – 45; EQ VAS = EuroQol Visual Analog Scale; WHODAS 2.0 = World Health Organisation Disability Assessment Schedule – 2.0; DERS = Difficulties in Emotion Regulation Scale; DES-II = Dissociative Experiences Scale – II; SCID-5 = Structured Clinical Interview for DSM-5.

^aThe amount of available data varies per instrument. A measurement may have consisted of multiple occasions, for example due to limited patient capacity. Incomplete data are available for: education level (n = 124), previous group outpatient psychological therapy (n = 127), CAPS-5 (n = 117), age during childhood trauma (n = 76), PCL-5 (n = 106), CTQ (n = 110), LEC-5 (n = 112), BDI-II (n = 109), OQ-45 (n = 110), EQ VAS (n = 107), WHODAS 2.0 (n = 107), DERS-II (n = 107), DES-II (n = 108), SCID-5 comorbid PD (n = 127).

^bData are n (%) or M (SD). Percentages are calculated based on the available data.

^cCountry of origin in the ‘other’ category was diverse and included Afghanistan, Bosnia-Herzegovina, Eritrea, Indonesia, Iran, Belgium, Morocco, Türkiye, the Dutch Caribbean, and Suriname.

ImRs noncompletion (<9 sessions ImRs) was n (%) = 25 (37.9%) in the ImRs only condition and n (%) = 24 (37.5%) in ImRs + GST. Patients attended M (SD) = 9.3 (6.2) sessions. Of 49 non-completers, ten were classified as responders at T4 in the non-imputed dataset. When excluding early responders (i.e. those whose CAPS-5 score reduction was >1 pooled baseline SD but with <75% of TFT sessions attended), a series of post-hoc independent samples t-tests and chi-square tests revealed that noncompletion was associated with PCL-5 (n = 106, p = .027, mean difference = 7.47), but not with OQ-45, CAPS-5, age, and sex (all ps > .10). GST noncompletion was n (%) = 31 (48.4%) and the mean number of attended GST sessions was M (SD) = 33.3(18.7).

See Tables 2 and 3 for the results of the analyses testing the primary and secondary hypotheses. Figure 2 displays the symptom change over time for each condition on the CAPS-5, PCL-5, and SCID-5-P CPD symptoms. Additional figures and statistics of all the non-imputed and imputed analyses can be found in Tables S1–S27 and Figure S2. First, there was a large reduction in CAPS-5 total scores in both conditions at the primary endpoint T4 (d_ImRs= 2.42, 95% CI = 1.97–2.87; d_{ImRs + GST}= 2.44, 95%CI = 1.99–2.90) as well as at T2 (ImRs: d_ImRs= 1.77, 95% CI = 1.37–2.18; d_{ImRs + GST}= 1.69, 95%CI = 1.29–2.09). The effect of time was significant at T4 (p < .0001) and at T2 (p < .0001). There was no difference between ImRs and ImRs + GST on the primary outcome at T4 (d = 0.02, 95%CI = −0.33–0.36) or T2 (d = 0.08; 95% CI = −0.26–0.43). Both T0–T2 (B = 0.15, SE = 2.12, p = .944) and T0–T4 (B = −0.81, SE = 2.10, p = .698) estimates from the condition*time interaction term were not significant. Analysis of the non-imputed data, treatment completers, analysis after exclusion of subclinical CPD cases, and analysis without controlling for covariates did not change the results in terms of significance and effect size.

Table 2.

Imputed estimated marginal means for all continuous outcomes per condition.

	ImRs		ImRs + GST		ImRs versus ImRs + GST
	Estimated marginal means (95%CI)a	Cohen’s db	Estimated marginal means (95% CI)	Cohen’s d	Cohen’s d change over time between conditions
CAPS-5 total score
Baseline	40.7 (38.4, 43.1)c		40.7 (38.3, 43.1)
T2	23.6 (21.2, 26.0)	1.8 (1.4, 2.2)	24.4 (22.0, 26.8)	1.7 (1.3, 2.1)	0.1 (−0.3, 0.4)
T4	17.3 (14.8, 19.7)	2.4 (2.0, 2.9)	17.1 (14.7, 19.5)	2.4 (2.0, 2.9)	0.0 (−0.3, 0.4)
PCL-5 total score
Baseline	54.6 (51.8, 57.5)		54.5 (51.6, 57.4)
T1	42.7 (39.8, 45.6)	1.0 (0.6, 1.4)	51.2 (48.3, 54.1)	0.3 (−0.1, 0.6)	0.7 (0.4, 1.1)
T2	28.3 (25.4, 31.2)	2.2 (1.8, 2.7)	32.9 (30.0, 35.8)	1.8 (1.4, 2.2)	0.4 (0.0, 0.7)
T3	29.8 (26.9, 32.8)	2.1 (1.7, 2.5)	31.0 (28.0, 34.1)	2.0 (1.6, 2.4)	0.1 (−0.24, 0.45)
T4	27.9 (24.9, 30.8)	2.3 (1.8, 2.7)	30.0 (27.1, 33.0)	2.1 (1.6, 2.5)	0.2 (−0.2, 0.5)
T5	26.5 (23.6, 29.5)	2.4 (1.9, 2.8)	27.2 (24.2, 30.2)	2.3 (1.9, 2.7)	0.1 (−0.3, 0.4)
OQ-45 total score
Baseline	101.4 (96.9, 106.0)		101.4 (96.8, 106.0)
T1	90.0 (85.3, 94.6)	0.6 (0.3, 1.0)	100.9 (96.2, 105.6)	0.0 (−0.3, 0.3)	0.6 (0.2, 0.9)
T2	72.5 (67.9, 77.1)	1.5 (1.1, 1.9)	77.1 (72.4, 81.9)	1.3 (0.9, 1.7)	0.2 (−0.1, 0.6)
T3	71.5 (66.8, 76.1)	1.5 (1.2, 2.0)	73.9 (69.2, 78.6)	1.5 (1.1, 1.8)	0.1 (−0.2, 0.5)
T4	72.5 (67.9, 77.2)	1.5 (1.1, 1.9)	76.3 (71.6, 81.0)	1.3 (1.0, 1.7)	0.2 (−0.1, 0.5)
T5	64.4 (59.7, 69.1)	2.0 (1.5, 2.4)	67.5 (62.8, 72.3)	1.8 (1.4, 2.2)	0.2 (−0.2, 0.5)
WHODAS 2.0 mean score
Baseline	2.5 (2.3, 2.7)		2.7 (2.5, 2.9)
T2	2.1 (1.9, 2.4)	0.4 (0.1, 0.8)	2.2 (2.0, 2.5)	0.5 (0.2, 0.9)	0.1 (−0.2, 0.5)
T4	2.0 (1.8, 2.3)	0.6 (0.2, 0.9)	2.2 (1.9, 2.5)	0.5 (0.2, 0.9)	0.0 (−0.3, 0.4)
T5	1.9 (1.7, 2.1)	0.7 (0.3, 1.0)	2.0 (1.8, 2.2)	0.8 (0.4, 1.2)	0.1 (−0.2, 0.5)
EQ VAS
Baseline	53.8 (49.9, 57.7)		53.8 (49.9, 57.7)
T2	62.3 (58.0, 66.6)	0.5 (0.2, 0.9)	64.2 (59.8, 68.5)	0.6 (0.3, 1.0)	0.1 (−0.2, 0.5)
T4	61.9 (57.3, 66.5)	0.5 (0.2, 0.9)	60.1 (55.5, 64.6)	0.4 (0.0, 0.7)	0.1 (−0.2, 0.5)
T5	64.3 (59.4, 69.2)	0.7 (0.3, 1.0)	64.0 (58.4, 69.5)	0.6 (0.3, 1.0)	0.0 (−0.3, 0.4)
SCID-5-PD mean CPD criteria
T0	5.9 (5.6, 6.2)		5.9 (5.6, 6.2)
T4	1.9 (1.5, 2.3)	2.9 (2.4, 3.4)	1.7 (1.3, 2.1)	3.0 (2.5, 3.5)	−0.1 (−0.5, 0.2)

Notes: CAPS-5 = Clinician-Administered PTSD Scale for DSM-5; PCL-5 = PTSD Checklist – DSM-5; OQ-45 = Outcome Questionnaire – 45; WHODAS 2.0 = World Health Organisation Disability Assessment Schedule – 2.0; EQ VAS = EuroQol Visual Analog Scale; SCID-5-PD = Structured Clinical Interview for DSM-5 Personality Disorders.

^aEstimates are in their original scale and based on final linear mixed models adjusted for condition, age, sex, baseline outcome score and the interaction of baseline outcome score with time.

^bCohen’s d was calculated from the estimated data using the pooled standard deviation at baseline as the denominator. Confidence intervals for Cohen’s d were calculated manually.

^cConfidence intervals that do not include zero indicate statistical significance. Exact p-values of the linear mixed models can be found in Table 3 and the supporting information.

Table 3.

Results of pooled analysis of multiple imputed data for the primary and secondary hypotheses at the primary endpoint T4.

	Estimatea	t	p
Continuous variables
Primary outcome
CAPS-5 total
Timeb	−23.61 (1.50)	−15.79	< .0001
Time × conditionc	0.15 (2.12)	0.07	.944
Secondary outcomes
PCL-5 total
Time	−24.48 (1.80)	−13.57	< .0001
Time × condition	−2.24 (2.54)	−0.88	.377
SCID-5 CPD total
Time	−4.22 (0.26)	−16.24	< .0001
Time × condition	−0.19 (0.37)	−0.52	.600
OQ-45 total
Time	−25.11 (2.80)	−8.98	< .0001
Time × condition	−3.78 (3.96)	−0.95	.341
WHODAS 2.0 mean
Time	−0.46 (0.16)	−2.83	.005
Time × condition	−0.01 (0.22)	−0.04	.967
EQ VAS
Time	56.34 (6.28)	8.97	< .0001
Time × condition	1.85 (4.00)	0.46	.644
Binary secondary outcomes
Treatment response
Condition	−0.70 (0.62)	−1.90, 0.51d	.259
PTSD remission
Condition	0.17 (0.45)	−0.72, 1.05d	.713
ImRs noncompletion
Condition	−0.04 (0.37)	−0.76, 0.68d	.917

Notes: CAPS-5 = Clinician-Administered PTSD Scale for DSM-5; PCL-5 = PTSD Checklist – DSM-5; Structured Clinical Interview for DSM-5 Personality Disorders; CPD = cluster C personality disorder; OQ-45 = Outcome Questionnaire – 45; WHODAS 2.0 = World Health Organisation Disability Assessment Schedule – 2.0; EQ VAS = EuroQol Visual Analog Scale.

^aEstimates are in their original scale and based on final models containing condition, age, sex, condition and, where appropriate, time, baseline outcome score and the interaction of baseline outcome score with time. See Table S2 for the test statistics of all variables in the models.

^bT0 is reference.

^cIMRS + ST is reference.

^d95% Wald confidence intervals.

Figure 2.

Changes over time per condition in primary and key secondary outcomes. Changes in clinician-rated CAPS-5 total scores are displayed on the left. Self-reported PCL-5 total scores and SCID-5-PD CPD symptom scores over time are displayed on the right. Estimated marginal means were calculated from the pooled estimates after multiple imputation. Error bars show 95%CIs calculated from the estimated marginal means. Abbreviation: CAPS-5 = Clinician-administered PTSD Scale for DSM-5; ImRs = imagery rescripting; ImRs + GST = imagery rescripting with group schema therapy and psychomotor therapy; PCL-5 = PTSD checklist – DSM-5; SCID-5-PD = Structured Clinical Interview for DSM-5 Personality Disorders; CPD = cluster C personality disorder.

Second, there was a large and significant reduction in CPD symptoms in both conditions (d_ImRs= 2.90, 95% CI = 2.41–3.39; d_{ImRs + GST}= 3.04, 95% CI = 2.53–3.54; p < .0001). The condition*time interaction was not significant (p = .600). Because of the low reliability of the CPD scale, we conducted a post-hoc analysis of avoidant and obsessive-compulsive PDs separately (but not dependent, as only several cases of this PD were included). The results of these analyses did not diverge from the composite CPD scale analysis. Other secondary outcome analyses yielded similar results. Main effects of time were significant (ps < .0001) and their respective effect sizes were medium (WHODAS, EQ-VAS) to large (OQ-45, PCL-5). No significant condition*time interactions were found for treatment noncompletion at T2, treatment response and PTSD remission (for both the CAPS-5 total score <12 in the imputed and the DSM-5 threshold definitions of PTSD remission) at T4, and PCL-5, OQ-45, WHODAS and EQ-VAS at any of the time points. Two exceptions were the OQ-45 at T1 (p = .002) and PCL-5 scores at T1 (p = .0001), which were lower in the ImRs condition. Note that T1 is the approximate time point at which ImRs is introduced in the ImRs + GST condition. The condition*time interaction term for PCL-5 became nonsignificant at T2 for both PCL-5 (p = .058) and OQ-45 (p = .228). Analyses of the secondary outcomes without controlling for covariates did not alter the findings.

PTSD remission (CAPS-5 total score <12) at the primary endpoint (T4) was achieved for n (%) = 29 (42.6%) in the nonimputed data, and for n (%) = 36 (27.7%) in the imputed data. When the DSM-5 criteria for PTSD were applied to the nonimputed data, PTSD remission was achieved for n (%) = 18 (26.1%) at T4. Most patients (n [%] = 53 [77.9%] for the nonimputed sample, n [%] = 112 [86%] for the imputed sample) were classified as treatment responders at T4. Two patients showed significant symptom worsening. Furthermore, ImRs completion was not associated with the condition (p = .993). Finally, CAPS-5 total scores at any time point were not associated with treatment completion in any condition.

There were three serious adverse events (SAE) in the study period. SAE were documented and reported by the principal investigator (KT) following procedures in accordance with the Dutch Medical Research Involving Human Subjects Act. One patient in the ImRs + GST condition was hospitalized due to medical problems not related to the study (myocardial infarction). One patient in the ImRs condition was admitted to inpatient treatment for 4 weeks due to increased suicidality. ImRs treatment resumed after discharge. Lastly, one patient in the ImRs + GST condition was briefly hospitalized after ingesting gamma-hydroxybutyric acid during the pre-treatment schema therapy phase, before starting ImRs.

4. Discussion

To our knowledge, the present study is the first RCT directly comparing the addition of PD treatment (GST) to trauma-focused treatment (ImRs) for patients with PTSD and comorbid CPD. Strikingly, no differences between conditions were found for any of the primary and secondary outcome measures at the primary endpoint. Large reductions in PTSD severity, general psychiatric symptom severity, CPD symptoms and moderate improvements in health status and general functioning were observed in both conditions. In addition, some evidence in this RCT suggests that the PTSD symptom reduction is specific to ImRs, because self-reported PTSD symptoms and general psychiatric symptom severity tended to decrease only after the introduction of ImRs in the ImRs + GST condition. What makes these findings particularly noteworthy is that patients in the ImRs + GST condition received a significantly higher dosage of psychotherapy compared to those in the ImRs only condition. Moreover, the higher dosage of psychotherapy was specifically designed to address the pervasive and persistent avoidance, passivity, and control mechanisms associated with CPD. Although our study was originally designed and powered to demonstrate the superiority of one treatment over the other, the fact that there were no significant differences in any of the outcomes and that the absolute size of the differences was close to zero, despite the substantial difference in psychotherapy dosage, raises the suggestion that standalone ImRs might be equivalent to ImRs + GST.

Although no causal conclusion pertaining to ImRs can be drawn from the current study due to the lack of a control group for ImRs, the wide-ranging reduction in symptoms may be understood from the hypothesized working mechanism of ImRs. It has been proposed that ImRs works through revaluation of memory representations of early adverse experiences (Arntz, 2012; Brockman & Calvert, 2017). Changing the meaning of these memory representations, which are often laden with feelings of self-blame, shame, fear, anger and/or disgust, can be expected to have profound effects on the sequelae of these early adverse experiences, including PTSD and personality disorder symptoms. In fact, CPD remission was reached for most personality disorders in both conditions. This result is consistent with the hypothesis that ImRs would be effective in treating PD (Arntz, 2011), and findings that comorbid PD status significantly changes in response to other trauma-focused treatments, such as prolonged exposure and eye movement desensitization and reprocessing (De Jongh et al., 2020; Kolthof et al., 2022; Markowitz et al., 2015). Whether the present results are replicated in patients with PTSD and comorbid borderline PD is currently being tested in a parallel RCT (Snoek et al., 2020).

Treatment noncompletion in the present study was approximately twice as high as reported in a meta-analysis on dropout rates in trauma-focused treatments (Lewis et al., 2020), although estimates of treatment noncompletion in the literature are highly heterogeneous (Schottenbauer et al., 2008). Low dropout rates for ImRs were found in recent trials (Boterhoven de Haan et al., 2020; Raabe et al., 2022), but this finding does not seem to generalize to patients with PTSD and comorbid CPD. Addressing the pervasive and persistent avoidance, passivity and control behavior present in CPD by adding GST to ImRs did not reduce dropout rates. An alternative explanation of high dropout rates may be that they may be associated with the effects of the COVID-19 pandemic. Unfortunately, the relative impact of the COVID-19 pandemic on dropout rates could not be assessed.

The present study possesses notable strengths. First, the study was adequately powered to detect medium differences between conditions. Second, the present study is the first study on the treatment of PTSD and CPD comorbidity and the first study to directly compare any TFT and PD-focused treatments. Third, treatments were provided by experienced therapists who received adequate training and supervision in the relevant treatments. Third, therapists were adequately trained and supervised, and therapist competence and adherence ratings were generally high and satisfactory. Note that the presence of ImRs and GST elements above a certain level does not imply higher treatment adherence. Various items on the rating scale are not applicable in every session (e.g. item 6 of the GSTRS ‘therapist self-discloses in an appropriate manner that serves the therapy process’) and were thus rated as absent. However, this does not imply lower quality adherence. Finally, the study was performed with a sample representative of patients encountered in a naturalistic setting in terms of high symptom and comorbidity levels and extensive trauma histories.

However, some limitations apply. First, no control group was included to compare imagery rescripting with. As a result, the symptom reductions cannot be causally attributed to ImRs. However, the large within-condition effect sizes are comparable to recent trials on ImRs for PTSD, adding to the growing evidence base showing ImRs can be as effective as other trauma-focused treatments (Boterhoven de Haan et al., 2020; Raabe et al., 2022). Second, treatment and measurement noncompletion may have impacted the results, although ImRs completion rates were equal for the two conditions and analyses suggest that noncompletion did not bias the results. Lower baseline scores on self-reported PTSD symptoms and general psychiatric symptom severity were associated with treatment noncompletion, which may indicate that patients with lower baseline scores need fewer treatment sessions. Third, we specifically tested a group schema therapy format. Thus, our findings may not be generalized to individual schema therapy. We chose for a group format to facilitate the surfacing of interpersonal problems that are believed to be central to (cluster-C) personality disorders, although we readily admit that there is no definitive empirical evidence supporting this choice. Conversely, group therapy may have introduced significant additional stress and anxiety in this patient population, compared to an individual ST format. Indeed, there may be a differential effect for ST format. For example, Arntz and colleagues recently found that predominantly GST was less effective than combined individual and group ST in treatment of borderline PD (Arntz et al., 2022), although direct comparisons between GST and individual ST formats for CPD are lacking. Moreover, meta-analytical evidence on group versus individual psychotherapy formats indicates that therapy format does not predict outcome in general (Burlingame et al., 2016). Relatedly, GST group sizes were relatively small (i.e. < 6 members) at certain stages of the study due to treatment dropout and group member absence, which may have negatively impacted the presumed therapeutic effects of group dynamics, although, again, therapy format has not been found to be a significant predictor of outcome in general. Future studies may test the differential effects of therapy format in CPD patients. Fourth, PD change after treatment was measured by conducting a structured interview developed to assess the presence of symptoms since at least early adulthood. Future studies should incorporate dimensional measures of personality functioning, for example by using the Avoidant Personality Disorder Severity Index, which was only recently validated (Baljé et al., 2023). Fifth, the treatment duration, dosage and, consequently, timing of measurements, differed between conditions. Ideally (at least, in terms of internal validity) duration and dosage of treatment and the timing of measurements are balanced between conditions. However, we were specifically interested in the added effect of PD treatment. In fact, trauma-focused treatment can be considered treatment as usual in the present design. We believed that measuring as close to the end of trauma-focused treatment as possible was the best way to assess treatment effects in terms of external validity. Finally, while we included patients from different age groups, and varying socio-economic and cultural backgrounds, the sample consisted predominantly of females, potentially limiting the generalizability of our findings to males and other genders.

5. Conclusions and implications

In conclusion, PTSD severity, general psychiatric symptoms and comorbid PD symptoms were greatly reduced at 12 months after starting ImRs in a sample of severely traumatized patients with PTSD and comorbid CPD. Moderate effects were found for health status and general functioning. Strikingly, the addition of GST did not yield superior treatment effects on any of the predefined outcomes. These findings carry several important clinical implications. Firstly, they suggest that offering concurrent treatment programs for both PTSD and CPD as the initial approach may not be ideal. It may be worthwhile for future research to explore the advantages of using a stepped care model, where trauma-focused treatment serves as the first step in addressing comorbid PTSD and personality disorder. Alternatively, future research may test the superiority of other PD treatments or treatment formats (e.g. individual) concurrent to trauma-focused treatment.

Secondly, the finding that CPD symptoms are highly changeable after a relatively short treatment like ImRs has the potential to impact the treatment of personality disorders, which are typically lengthy and intensive. While ImRs, when integrated into schema therapy, has shown success in treating personality disorders, standalone ImRs may be an effective PD treatment. Further studies with adequate control groups and patients with a primary diagnosis of CPD and other personality disorders are needed to test this hypothesis.

Funding Statement

This work was supported by Stichting Steunfonds Joodse Geestelijke Gezondheidszorg. The funder did not have any role in the study design and reporting of the results.

Author statement contributors

AE drafted the manuscript, performed the data curation and formal analyses, and was involved at all stages of the study. AB, JD, IA, AS, MB, CV, OH, and KT were involved in conceptualization and reviewing of the manuscript. AE was supervised by KT, JD and AB. KT, IA, AS were involved in investigating and project administration. MB, KT, AB, and JD were involved in developing the methodology. KT was responsible for funding acquisition.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Clinical trial registration

The present study was registered under NCT03833531 on ClinicalTrials.gov and was approved by the regional Medical Ethics Committee.

Patient consent

All patients consented to participate in the study and gave permission to use the data for scientific research by signing an informed consent form.

Data availability statement

Access to deidentified, individual participant data in an online data repository might be granted upon reasonable request, with a signed data access agreement. Data requests should be submitted to the corresponding author.