Predictors and moderators of treatment outcomes in phase-based treatment and trauma-focused treatments in patients with childhood abuse-related post-traumatic stress disorder

ABSTRACT

Background: Knowledge of treatment predictors and moderators is important for improving the effectiveness of treatment for PTSD due to childhood abuse.

Objective: The first aim of this study was to test the potential predictive value of variables commonly associated with PTSD resulting from a history of repeated childhood abuse, in relation to treatment outcomes. The second aim was to examine if complex PTSD symptoms act as potential moderators between treatment conditions and outcomes.

Method: Data were obtained from a randomized controlled trial comparing a phase-based treatment (Skills Training in Affect and Interpersonal Regulation [STAIR] followed by Eye Movement Desensitization and Reprocessing [EMDR] therapy; n = 57) with a direct trauma-focused treatment (EMDR therapy only; n = 64) in people with PTSD due to childhood abuse. The possible predictive effects of the presence of borderline personality disorder, dissociative symptoms, and suicidal and self-injurious behaviours were examined. In addition, it was determined whether symptoms of emotion regulation difficulties, self-esteem, and interpersonal problems moderated the relation between the treatment condition and PTSD post-treatment, corrected for pre-treatment PTSD severity.

Results: Pre-treatment PTSD severity proved to be a significant predictor of less profitable PTSD treatment outcomes. The same was true for the severity of dissociative symptoms, but only post-treatment, and not when corrected for false positives. Complex PTSD symptoms did not moderate the relationship between the treatment conditions and PTSD treatment outcomes.

Conclusions: The current findings suggest that regardless of the common comorbid symptoms studied, immediate trauma-focused treatment is a safe and effective option for individuals with childhood-related PTSD. However, individuals experiencing severe symptoms of PTSD may benefit from additional treatment sessions or the addition of other evidence-based PTSD treatment approaches. The predictive influence of dissociative sequelae needs further research.

The study design was registered in The Dutch trial register (https://www.trialregister.nl/trialreg/admin/rctview.asp?TC = 5991) NTR5991 and was approved by the medical ethics committee of Twente NL 56641.044.16 CCMO.

HIGHLIGHTS

• One of first studies that aimed to identify multiple potential predictors and moderators in patients with PTSD related to childhood abuse.

• Only severe PTSD predicted worse treatment outcomes.

• The predictive influence of dissociative sequelae needs further research.

Antecedentes: El conocimiento de los predictores y moderadores del tratamiento es importante para mejorar la eficacia del tratamiento del TEPT debido a abuso en la infancia.

Objetivo: El primer objetivo de este estudio fue probar el valor predictivo potencial de las variables comúnmente asociadas con el TEPT resultante de una historia de abuso infantil repetido, en relación con los resultados del tratamiento. El segundo objetivo fue examinar si los síntomas complejos del TEPT actúan como moderadores potenciales entre las condiciones de tratamiento y los resultados.

Método: Se obtuvieron datos de un ensayo controlado aleatorizado que comparaba un tratamiento basado en fases (Entrenamiento en Destrezas para la Regulación Emocional e Interpersonal [STAIR] seguido de terapia de Desensibilización y Reprocesamiento a través de los Movimientos Oculares [EMDR]; n = 57) con un tratamiento directo centrado en el trauma (sólo terapia EMDR; n = 64) en personas con TEPT debido a abuso en la infancia. Se examinaron los posibles efectos predictivos de la presencia de trastorno límite de la personalidad, síntomas disociativos y conductas suicidas y autolesivas. Además, se determinó si los síntomas de dificultades en la regulación de las emociones, la autoestima y los problemas interpersonales moderaban la relación entre la condición de tratamiento y el TEPT post-tratamiento, corregida por la gravedad del TEPT pre-tratamiento.

Resultados: La gravedad del TEPT pre-tratamiento resultó ser un predictor significativo de resultados menos provechosos del tratamiento del TEPT. Lo mismo ocurrió con la gravedad de los síntomas disociativos, pero sólo después del tratamiento, y no cuando se corrigió por falsos positivos. Los síntomas de TEPT complejo no moderaron la relación entre las condiciones de tratamiento y los resultados del tratamiento del TEPT.

Conclusiones: Los hallazgos actuales sugieren que, independientemente de los síntomas comórbidos comunes estudiados, el tratamiento inmediato centrado en el trauma es una opción segura y eficaz para las personas con TEPT relacionado con la infancia. Sin embargo, los individuos que experimentan síntomas graves de TEPT pueden beneficiarse de sesiones de tratamiento adicionales o de la adición de otros enfoques de tratamiento para el TEPT basados en la evidencia. Es necesario seguir investigando la influencia predictiva de las secuelas disociativas.

KEYWORDS:

• PTSD
• childhood abuse
• predictors
• moderators
• treatment

PALABRAS CLAVE:

• TEPT
• abuso infantil
• predictores
• moderadores
• tratamiento

1. Introduction

For long time, there has been a, sometimes heated, debate about the optimal treatment for individuals diagnosed with post-traumatic stress disorder (PTSD) stemming from childhood physical and sexual abuse (De Jongh et al., 2016). Childhood sexual and physical abuse are considered to be types of traumatic experiences associated with the development of Complex PTSD (Cloitre et al., 2012), a term that was initially launched by Herman (1992). She pleaded for a separate diagnosis with symptoms such as changes in affect regulation, consciousness, identity, perception of the perpetrator, and systems of meaning. These symptoms would form the basis for the diagnosis of DES-NOS (Disorders of Extreme Stress Not Otherwise Specified; Van der Kolk et al., 2005), the former Complex PTSD diagnosis. Recently Complex PTSD has been acknowledged as an official diagnosis according to the International Classification of Diseases 11th revision ICD-11 (ICD-11; World Health Organization, 2018), but involves fewer and different symptoms than the proposed DES-NOS diagnosis. Complex PTSD involves symptoms of ‘Disturbances in Self-Organization’ (DSO; i.e. problems with affect regulation, negative self-concept, and interpersonal problems; World Health Organization, 2018), in addition to the full diagnostic criteria of PTSD. With the inclusion of Complex PTSD as a distinct classification within the ICD-11 discussions involving the diagnosis and treatment of patients with PTSD resulting from childhood abuse have become even more relevant. To this end, some studies have identified a clear distinction between diagnoses of PTSD and Complex PTSD (e.g. Rink & Lipinska, 2020), whereas others have not found much greater variability in clinical presentations among adult patients with PTSD (Achterhof et al., 2019). Nevertheless, there are indications that patients with a history of childhood abuse may exhibit more severe PTSD symptoms or display symptoms of Complex PTSD (Rink & Lipinska, 2020), which further fuels the debate on whether patients with PTSD related to childhood abuse can benefit from evidence-based PTSD treatments to the same extent as individuals with other trauma histories.

Existing evidence-based trauma-focused therapies may lead to less favourable outcomes in patients with a history of early and repeated trauma, such as childhood abuse (Cloitre et al., 2002; Cloitre et al., 2012; Karatzias et al., 2019). One suggestion to improve the treatment outcomes in this patient group is to offer a phase-based treatment consisting of three phases (ISTSS; Cloitre et al., 2012). Phase one attempts to teach the patient skills for handling interpersonal problems and emotion regulation difficulties that could interfere with tolerating trauma-focused treatments in phase two (Cloitre et al., 2002). Phase three integrates the learned skills in daily life. Regarding the application of a phased approach to patients with a history of childhood abuse, the most studied protocolized first-phase programme is STAIR (Skills Training in Affect and Interpersonal Regulation; Cloitre et al., 2002; Cloitre et al., 2010).

Recently, two randomized controlled studies have examined the effects of adding STAIR to a trauma-focused treatment compared to a trauma-focused treatment without the addition of STAIR in patients with PTSD related to childhood abuse (Raabe et al., 2022; Van Vliet et al., 2021). Van Vliet et al. (2021) found that among patients with PTSD related to repeated sexual and/or physical abuse before the age of 18 years (N = 121), treatment with EMDR-only therapy and EMDR preceded by STAIR were equally effective with STAIR with regard to the primary outcome variable being PTSD severity (between conditions ES d = .12). Additionally, for the secondary outcome measures (i.e. complex PTSD symptoms, dissociative symptoms and general psychopathology) no significant differences in change were found between the treatment conditions from pretreatment to posttreatment. Similarly, Raabe et al. (2022) found that for individuals with PTSD related to repeated sexual or physical abuse before the age of 15 years (N = 61), Imagery Rescripting (ImRs) as a standalone treatment and ImRs preceded by STAIR were equally effective regarding both primary (i.e. PTSD symptoms; d = .18) and secondary outcomes (i.e. feelings of shame, guilt and anger, dissociative symptoms, depression, emotion regulation and interpersonal problems). These studies provide support for the effectiveness of trauma-focused treatments for patients with PTSD related to childhood abuse and, from a cost-effectiveness perspective, for the use of trauma-focused treatment as astandalone approach without the addition of a stabilizing treatment track. However, it remains unclear whether specific patient characteristics are associated with less favourable treatment outcomes, and whether moderation by patient characteristics differs between treatment conditions. Investigating patient characteristics that either predict or moderate treatment outcomes is important because it can aid the development of personalized treatment approaches for patients suffering from symptoms characteristic of Complex PTSD (ISTSS, 2018; Kraemer, 2013).

Key factors to consider when studying individuals with PTSD related to childhood abuse, to develop more personalized treatments, include the severity of dissociative symptoms, the presence of comorbid borderline personality disorder (BPD) and suicidal and self-injurious behaviour. Dissociative symptoms are commonly found in PTSD diagnoses (Lyssenko et al., 2018) with the highest dissociation scores being observed in patients with a history of childhood abuse (Vonderlin et al., 2018). Clinicians are concerned about possible symptom exacerbation during trauma-focused treatments (Van Minnen et al., 2012), and that dissociation could hinder fear activation, which is considered a key mechanism of successful PTSD treatment (Cooper et al., 2017). Although there is evidence to suggest that dissociation is not predictive of the effectiveness of PTSD psychotherapy in patients with various trauma backgrounds (Hoeboer et al., 2020), questions about its impact on treatment outcomes, specifically in patients with PTSD related to a history of childhood abuse remain unanswered. The presence of borderline personality disorder (BPD) with comorbid suicidal and self-injurious behaviours is frequently observed in patients with PTSD (Harned et al., 2010; Zlotnick et al., 2003) and has been found to be closely associated with the presence of Complex PTSD (Resick et al., 2012). A meta-analysis indicated that the effectiveness of psychotherapies for PTSD caused by diverse traumas is unlikely to be influenced by the presence of BPD (Slotema et al., 2020). However, in a particular patient group with a history of childhood abuse, BPD may partly explain the less profitable PTSD treatment outcomes. In addition, most studies among patients with PTSD related to childhood abuse excluded individuals with comorbid BPD and/or suicidal and self-injurious behaviour or did not report treatment outcomes for this subgroup (Bohus et al., 2013; Van Minnen et al., 2012). This knowledge gap makes it unclear whether these symptoms affect PTSD treatment outcomes. Separate DSO symptoms (i.e. emotion regulation difficulties, interpersonal problems, and negative self-esteem) are potentially relevant moderators between PTSD treatment outcomes, and are considered to particularly benefit from treatments such as STAIR. From this point of view, it has been reasoned that patients would benefit more from trauma-focused treatment preceded by STAIR than from a standalone trauma-focused treatment (Cloitre et al., 2002; Cloitre et al., 2010). Conversely, Hoeboer et al. (2021) found that the DSO symptoms cluster did not moderate between three forms of prolonged exposure (PE); that is, prolonged exposure as standalone treatment with weekly sessions (16 sessions), an intensified form of prolonged exposure with three sessions per week (14 sessions), and prolonged exposure (8 sessions) preceded by STAIR (8 sessions). However, they did not study whether or not STAIR, as an actual addition to trauma-focused treatment, would be associated with better treatment outcomes than trauma-focused treatment alone, which is a core question related to the discussion on the treatment of patients with PTSD related to childhood abuse (ISTSS, 2018).

Therefore, the first aim of the present study was to assess potential predictors of PTSD treatment outcomes in patients with PTSD stemming from repeated sexual and/or physical abuse before the age of 18, using EMDR therapy as a trauma-focused treatment. The second aim was to examine possible moderators influencing the relationship between treatment conditions and PTSD treatment outcomes, with direct trauma-focused treatment (EMDR therapy) and phase-based treatment (EMDR therapy preceded by STAIR) as treatment conditions. More specifically, we hypothesized that the presence of a borderline personality disorder or the severity of suicidality, self-injury, or dissociative experiences at pre-treatment would predict less favourable PTSD treatment effects. Furthermore, we hypothesized that disturbances in self-organization would moderate the relationship between treatment conditions and treatment outcomes and that, the presence of more DSO symptoms would be related to worse outcomes in the direct trauma-focused (EMDR only) condition than in the phase-based condition, as STAIR aims to address these disturbances (Cloitre et al., 2002).

2. Method

2.1. Design

The present study used data from the TOPRON study, a randomized controlled trial designed to determine whether a phase-based approach (STAIR followed by EMDR; n = 57) would be more effective than an immediate trauma-focused approach (EMDR only; n = 64) in people with childhood trauma-related PTSD (Van Vliet et al., 2021). The design paper of this study (Van Vliet et al., 2018) is available at https://doi.org/10.1186/s13063-018-2508-8. The study design was registered in NTR5991 and approved by the medical ethics committee Twente NL 56641.044.16 CCMO.

2.2. Participants

From September 2016 until December 2019, participants were recruited in two outpatient mental health organizations in the Netherlands. After patients received information about the study and had time to read and overthink the information, patients were asked to sign an inform consent form when they were willing to participate. Patients were included when aged between 18 and 65 years and diagnosed with PTSD as verified with the CAPS-5 (Clinician-Administered PTSD scale for DSM-5 [Weathers et al., 2013]). Furthermore, they reported being victims of repeated sexual and/or physical abuse before the age of 18 by a caretaker or a person in a position of authority, as identified by the LEC-5 (Weathers et al., 2013). Patients were excluded if they had insufficient mastery of the Dutch language and in case of an acute risk of suicidality, for which direct crisis intervention was needed. In addition, they were excluded if they had received treatment for PTSD in the past year with at least eight sessions (of any evidence-based programme), reported being a victim of ongoing physical and/or sexual abuse, reported alcohol or drug dependency according to DSM-5 criteria (American Psychiatric Association, 2013) or had intellectual disability.

2.3. Assessment schedule

Patient characteristics and PTSD diagnosis and severity were assessed during the baseline assessment, at post-treatment (T4), and at three- and six-month follow-up. In addition to PTSD symptom change, which was the main outcome of this study, we investigated the severity of suicidality and self-injury, presence of a borderline personality disorder, and severity of dissociative experiences as potential predictors at baseline. Furthermore, we investigated DSO symptoms at baseline (i.e. interpersonal problems, emotion regulation problems, and problems in self-esteem) as possible moderators influencing the relationship between treatment conditions and PTSD treatment outcomes. For a detailed description of the instruments that were used to assess these symptoms, see paragraph 2.5 Measurements.

2.4. Treatment

For a complete description of the two treatment conditions, see Van Vliet et al. (2018; 2021). Both STAIR and EMDR therapy were delivered twice a week in sessions lasting 90 min each. STAIR was conducted according to the treatment programme described by Cloitre and her colleagues (Cloitre et al., 2002). The EMDR therapy was conducted according to the Dutch translation of the EMDR standard protocol (De Jongh & Ten Broeke, 2013). In both treatment arms, patients first received one 90 min session before the start of the actual treatment, consisting of psychoeducation and formulation of a hierarchy of relevant traumatic experiences with the most disturbing memory at the top. After the first session, patients started with STAIR sessions or EMDR therapy sessions, depending on the condition they were randomly assigned to. After treatment (eight sessions of STAIR followed by 16 sessions of EMDR therapy, or only 16 sessions of EMDR therapy), patients were not allowed to receive any psychological therapy during the six-month follow-up period.

2.5. Measurements

2.5.1. Outcome measurements

The presence and severity of PTSD were assessed using the Clinician-Administered PTSD Scale for DSM-5 (CAPS-5; Weathers et al., 2013) at pre-treatment, post-treatment, and both follow-up measurement points. The pre-treatment score was used as a covariate for the post-treatment scores to determine whether the severity of pre-treatment scores would directly affect the post-treatment scores. Since the CAPS is the gold standard for classifying PTSD and its severity, we included only the total scores in the calculations. All the more since the internal consistency in validation of the Dutch CAPS-5 has turned out to be high (α = .90; Boeschoten et al., 2018). To ensure that our trained independent professionals scored the CAPS correctly we calculated the inter-rater reliability using the intra-class correlation coefficient (ICC) after the independent rating of the same patient by four trained professionals. With an ICC of .999, the inter-rater reliability was excellent for this outcome measure in the present study.

Although the study protocol described the PSS-SR (Foa et al., 1993) as the primary outcome variable (Van Vliet et al., 2018), the CAPS-5 total scores were used in the present study because they contained fewer missing values in the dataset.

2.5.2. Predictors

The presence of a borderline personality disorder was determined using the Structured Clinical Interview for DSM-IV Axis II personality disorders (SCID-II interview; First et al., 1997), with items 97 and 98 determining the severity of self-injury (score 1 for absent, 2 for doubtful, and 3 for present; for statistical analyses, the scores were recoded as 0 for absent, 1 for doubtful, and 2 for present). The Dutch translation and adaptation have been evaluated well (mean Interrater reliability .84 [Lobbestael et al., 2011]; test-retest reliability .63 [Weertman et al., 2003]). The dichotomous outcomes of the SCID-II were used, (i.e. whether the patient fulfilled the diagnostic criteria for a personality disorder) therefore, we were unable to calculate the Cronbach’s alpha for this measure.

The severity of suicidality was determined using Item 9 of the Beck Depression Inventory (BDI; Beck et al., 1996), a self-report questionnaire with a scale from 0 to 3 (with a score of 0 indicating the absence of suicidal thoughts, a score of 1 indicating suicidal thoughts, but no intention to carry them out, and a score of 2 indicating a clear intention to commit suicide). If a score of 3 indicated that a patient was acutely suicidal and in need of direct crisis interventions, the patient was permanently excluded for participation and received the needed crisis interventions. Since this measure was categorical, we were unable to calculate Cronbach’s alpha.

Severity of dissociative symptoms was measured using the Dutch translation of Dissociative Experiences Scale (DES-II; Carlson & Putnam, 1993; Van IJzendoorn & Schuengel, 1996; Cronbach's α = .93 in the present study at baseline). It is a 28-item self-report questionnaire, with scores ranging from 0 to 100.

2.5.3. Moderators

The severity of interpersonal difficulties was indexed by the Dutch translation of the Inventory of Interpersonal Problems (IIP; Horowitz et al., 2000; Cronbach's α = .85 at baseline of the current study). Each of the 32 IIP items was scored on a 5-point scale from 0 (not at all) to 4 (very strongly). The psychometric properties of the IIP are satisfactory (Barkham et al., 1996).

The severity of difficulties in emotion regulation was indexed using the Dutch version of the Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004). The reliability of this study at baseline was high (Cronbach's α = .92). It consists of 36 items, each rated on a 5-point scale and the instrument has been validated with good results (Gratz & Roemer, 2004).

To assess the severity of DSO symptom self-esteem the subscale for negative cognitions about self from the Dutch version of the Posttraumatic Cognitions Inventory (PTCI; Foa et al., 1999) was used existing of 21 items, which has high reliability at baseline in this study (Cronbach's α = .94). Patients have to score on a Likert scale from 1 (‘I totally disagree’) to 7 (‘I totally agree’). The Dutch PTCI had good psychometric properties (Van Emmerik et al., 2007).

3. Statistical analysis

Analyses were conducted using SPSS version 25 (2017). The power calculation was based on a repeated-measures ANOVA, with treatment condition (phase-based treatment [EMDR preceded by STAIR] versus direct trauma-focused treatment [EMDR only]) as the between-subjects factor and time (pre-treatment, post-treatment/six month follow-up) as the within-subjects factor (Van Vliet et al., 2018). Inspection of the Mahalanobis distances of all the variables in the original dataset indicated two cases as statistical outliers. Little’s Missing Completely At Random test was used to examine the missing data, which showed an insignificant result (χ²(81) = 76.17, p = .63), indicating that there were no systematically missing data in the current dataset. Missing data were handled by multiple imputations using the mice package in R (2020; Van Buuren, 2012). Continuous variables were imputed using predictive mean matching, and categorical variables were imputed using logistic regression (binary or ordinal, depending on the number of categories). The interactions were not imputed but were calculated from the imputed variables during the imputation process and were not used to impute the other variables (see Van Buuren, 2012, for details about these procedures). As the maximum percentage of missing data for the variables was 33.9%, at least 34 imputations were deemed necessary, and we generated 50 imputed datasets as recommended by Graham (2009, 2012). The imputation methods (predictive mean matching and logistic regression) ensured that all imputed values were within the possible range of the observed values, and only the imputed values of the categorical variables were rounded (default method for logistic regression). As SPSS does not provide pooled outcomes for all statistics and tests, we calculated the pooled outcomes using functions in R for the following results: standard deviations of separate variables, multiple correlations, R-squared values (adjusted and unadjusted) and corresponding F-tests, R-squared change values and corresponding F-tests, standardized regression coefficient beta, and Variance Inflation Factor (VIF) scores.

Before the regression analyses, preliminary analyses were conducted to check for violations of assumptions of normality, linearity, multicollinearity, and homoscedasticity. The results indicated no violation of these assumptions. Analyses performed with and without outliers did not lead to different conclusions; therefore, only the results with outliers are presented herein. Outcomes without outliers are available upon request.

To examine possible predictors of PTSD treatment outcomes and moderators between treatment outcome and condition, two hierarchical regression analyses were performed, with the pretreatment PTSD score as a covariate (to exclude the potential conclusion that DSO symptoms are just expressions of severe PTSD) and the scores at post-treatment and at the six-month follow-up as outcome variables. In the analyses, the severity of suicidality and self-injury, the presence of borderline personality disorder, and the severity of dissociative experiences were included as predictors, and emotion regulation problems, interpersonal problems, and self-esteem problems were included as predictors and moderators. To examine moderators influencing the relation between condition and treatment outcome, the interaction terms between condition and the three separate DSO symptoms (Frazier et al., 2004; Warner, 2013) were performed. Variables were entered into four blocks: 1) PTSD severity at pre-treatment, 2) treatment condition (i.e. phase-based treatment versus direct trauma-focused treatment), 3) various psychopathology symptoms at pre-treatment, and 4) interaction terms between treatment form and DSO variables. The same analyses were performed with the core PTSD symptoms from the CAPS as a covariate (Block 1) instead of the total CAPS PTSD symptom score (i.e. not including the D-criteria scores), to ensure that the results were not mainly influenced by the DSO symptoms included in the CAPS.

4. Results

Possible baseline differences between the two treatment groups were analyzed using chi-squared tests and t-tests. Categorical variables did not differ significantly between the two conditions (presence of Complex PTSD F [1, 14523] = 0.57, p = .452; presence of borderline personality disorder F [1, 7476] = 0.79, p = 0.374; severity of suicidality F [2, 11652] = 0.22, p = .800; severity of self-injury F [2, 1083117] = 0.77, p = .464). Table 1 presents the means, standard deviations, and baseline differences between the two conditions for the continuous predictors and moderators and the PTSD scores at post-treatment and the six-month’s follow-up. No significant differences were found at baseline between the two conditions, except for the severity of dissociative symptoms, with a significantly higher mean score in the EMDR therapy only condition (t [8926] = 2.10, p = .035). As for the DES, a significant difference was found at pre-treatment between the two conditions, analyses with and without outliers on the DES were performed. As the outcomes did not lead to different conclusions, the presented results are with outliers included.

Table 1. The pooled means, standard deviations and pooled t-scores for continuous variables pre-treatment and the PTSD severity at post-treatment and at six month’s follow-up.

Variable	EMDR condition M (Sd)	STAIR-EMDR condition M (Sd)	t	df**	p
PTSD pre-treatment	39.34 (9.61)	37.61 (8.32)	1.05	119	.295
Dissociation pre-treatment	28.12 (19.62)	21.38 (13.40)	2.10	8926	.035
Emotion Regulation Problems pre-treatment	117.05 (27.14)	110.55 (21.86)	1.36	3632	.174
Interpersonal Problems pre-treatment	1.70 (0.61)	1.59 (0.54)	1.00	4991	.316
Self-esteem pre-treatment	88.83 (28.83)	80.97 (27.62)	1.46	4919	.146
PTSD post-treatment	18.26 (14.10)	18.06 (13.34)
PTSD six months’ follow-up	16.27 (12.76)	15.89 (13.64)

Note. *significant value p < .05; **the 50 imputations cause high degrees of freedom.

Table 2 presents the results of the hierarchical regression analyses identifying potential predictors and moderators using the CAPS-5 at post-treatment as the dependent variable and the CAPS at pre-treatment as a covariate. The CAPS total score explained 24.1% of the variance on the post-treatment score, which was a significant effect (F [1,111.83] = 33.38, p < .001; beta = .490). In the second block, we entered the condition, in the third block all predictors were entered, and in the fourth block the interactions between the separate DSO symptoms and conditions were entered, all explaining low, non-significant differences of the variance. In this final analysis, pre-treatment PTSD severity proved the best predictor of treatment outcomes (beta = .41, p < .001) and led to a positive predictive effect of dissociative experiences on PTSD treatment outcomes, with pre-treatment PTSD severity as a covariate (beta = .28, p = .009; see Table 2). The other predictors and moderators in the model did not lead to a significant difference. We conducted a Benjamini-Hochberg correction on the outcomes to control for false positives (significance level (k/m)*alpha, with alpha = .05), showing that the lowest p-value (.012 for dissociation) is not lower than the corrected significance level (.0083). This means that the predictive value of dissociative experiences on PTSD treatment outcomes should be interpreted with caution.

Table 2. Pooled outcome of hierarchical regression analysis with CAPS scores at posttreatment as the dependent variable and pretreatment PTSD score as the covariate.

Pooled	B	SE	β	t	p	FMI
Block 1
Constant	−10.534	5.011		−2.102	.036	.085
CAPS pretreatment	.745	.128	.490	5.810	.000	.108
Block 2
Constant	−11.276	5.323		−2.118	.034	.108
CAPS pretreatment	.751	.129	.495	5.798	.000	.113
Condition	1.096	2.388	.040	.459	.646	.154
Block 3
Constant	−5.112	6.702		−.763	.446	.181
CAPS pretreatment	.624	.148	.411	4.222	.000	.126
Condition	2.107	2.451	.077	.860	.390	.212
Suicidality	2.104	2.062	.104	1.020	.308	.229
Self-injury	1.766	1.567	.112	1.127	.260	.242
BPD	−1.150	3.232	−.033	−.356	.722	.197
Dissociation	.225	.086	.283	2.604	.009	.283
Emotion Regulation	−.121	.074	−.221	−1.633	.103	.229
Interpersonal Problems	3.238	2.537	.136	1.276	.202	.236
Self-esteem	−.021	.066	−.044	−.322	.747	.253
Block 3
Constant	−1.455	7.919		−.184	.854	.156
CAPS pretreatment	.629	.149	.414	4.225	.000	.127
Condition	−7.275	11.630	−.267	−626	.532	.185
Suicidality	2.011	2.079	.099	.967	.334	.227
Self-injury	2.024	1.629	.128	1.242	.215	.258
BPD	−1.184	3.249	−.034	−.364	.716	.193
Dissociation	.224	.089	.282	2.518	.012	.305
Emotion Regulation	−.133	.099	−241	−1.343	.180	.223
Interpersonal Problems	2.470	3.314	.104	.745	.456	.244
Self-esteem	−.036	.087	−.075	−.417	.677	.229
Condition × Emotion Regulation	.043	.153	.178	.279	.781	.280
Condition × Interpersonal Problems	1.738	5.377	.111	.323	.747	.271
Condition × self-esteem	.021	.123	.070	.170	.865	.238

Note. CAPS = Clinician-administered PTSD scale, BPD = Borderline Personality Disorder, B = Unstandardized regression coefficient, SE = Standard error, β = standardized regression coefficient, FMI = Fraction of missing information.

The model with the CAPS-5 total scores as a covariate and the CAPS score at six months’ follow-up as the dependent variable explained 19.5% of the variance in the post-treatment score, which is a significant effect (F-change [1, 97.29] = 21.62, p < .001; beta = .441; see Table 3). In the second, third, and fourth blocks, we entered the condition, predictors, and interactions between separate DSO symptoms and conditions, all adding low percentages and a non-significant explanation of the variance. None of the potential predictors and moderators caused a significant difference in the model.

Table 3. Pooled outcome of hierarchical multiple regression with CAPS scores at six month’s follow-up as the dependent variable and pretreatment PTSD score as the covariate.

Pooled	B	SE	β	t	p	FMI
Block 1
Constant	−8.623	5.278		−1.634	.103	.197
CAPS pretreatment	.641	.134	.441	4.777	.000	.208
Block 2
Constant	−9.125	5.649		−1.615	.107	.229
CAPS pretreatment	.645	.136	.444	4.745	.000	.217
Condition	.742	2.428	.028	.305	.760	.203
Block 3
Constant	−3.642	7.127		−.511	.609	.260
CAPS pretreatment	.609	.161	.419	3.792	.000	.246
Condition	1.695	2.483	.064	.683	.495	.215
Suicidality	−.171	2.235	−.009	−.076	.939	.330
Self-injury	2.258	1.654	.149	1.365	.173	.305
BPD	−4.474	3.482	−.135	−1.285	.199	.293
Dissociation	.095	.086	.124	1.102	.271	.268
Emotion Regulation	−.138	.079	−.263	−1.758	.079	.295
Interpersonal Problems	2.172	2.623	.095	−828	.408	.270
Self-esteem	.056	.065	.122	.867	.386	.219
Block 3
Constant	.732	8.397		.087	.931	.234
CAPS pretreatment	.612	.162	.421	3.783	.000	.246
Condition	−9.899	11.533	−.379	−.858	.391	.152
Suicidality	−.303	2.234	−.016	−.136	.892	.317
Self-injury	2.501	1.722	.165	1.453	.147	.322
BPD	−4.527	3.511	−.137	−1.289	.198	.294
Dissociation	.100	.088	.130	1.130	.259	.281
Emotion Regulation	−.183	.103	−.348	−1.778	.076	.270
Interpersonal Problems	1.914	3.297	.084	.580	.562	.219
Self-esteem	.068	.089	.147	.769	.442	.243
Condition × Emotion Regulation	.116	.146	.505	.790	.430	.196
Condition × Interpersonal Problems	.799	5.334	.054	.150	.881	.243
Condition × self-esteem	−.032	.123	−.110	−.264	.792	.217

Note. CAPS = Clinician-administered PTSD scale, BPD = Borderline Personality Disorder, B = Unstandardized regression coefficient, SE = Standard error, β = standardized regression coefficient, FMI = Fraction of missing information.

The same analyses were performed with the core PTSD symptoms from the CAPS as a covariate instead of the total CAPS PTSD symptom score (i.e. not including the D-criteria scores). The results were comparable both at post-treatment and at the 6-month follow-up (see Supplemental Table 1 and 2).

5. Discussion

This study investigated potential predictors and moderators of PTSD treatment outcomes among patients with a history of childhood abuse who were treated within two outpatient mental health settings and compared the results of phase-based treatment with those of direct trauma-focused treatment for this target group. Our first hypothesis, that the presence of a borderline personality disorder or the severity of suicidality, self-injury, or dissociative experiences at pre-treatment would predict less favourable PTSD treatment effects, was partly supported. The results showed that the severity of PTSD symptoms at the start of treatment was the best predictor of less beneficial PTSD treatment outcomes. When corrected for pretreatment PTSD severity, only dissociative symptoms predicted worse PTSD treatment outcomes but only post-treatment and when corrected for false positives these results did not hold true, which means that these should be interpreted with caution. In the long term, at the six-month follow-up, dissociative symptoms were not found to be associated with worse PTSD treatment outcomes. This indicates that none of the patient characteristics measured prior to treatment were predictive of less favourable post-treatment treatment outcomes, except for PTSD severity. This also holds true for the severity of dissociative symptoms which was found to have an albeit disputable significant impact on treatment outcomes, but only a immediately post-treatment and not on the long term. This means that pre-treatment characteristics should not be a reason to withhold patients a PTSD treatment or to deliver another type of (e.g. stabilizing) treatment form. Our second hypothesis, stating that DSO symptoms (i.e. interpersonal problems, problems in affect regulation, and negative self-concept) moderate the relationship between treatment conditions and treatment outcomes, was not supported by the present data. Based on these data, the choice for a trauma-focused treatment or a phase-based treatment should not depend on the severity of interpersonal problems, emotion regulation problems or negative self-concept.

This study is the first to examine self-injury and suicidality as potential predictors of PTSD outcomes. Hence, besides pretreatment PTSD severity, we only found dissociative symptoms to be predictive of PTSD treatment outcomes; however, this predictive effect did not persist in the long term and did not hold true when correcting for false positives. This could be explained by the fact that dissociative symptoms diminish in response to trauma-focused treatment (Van Vliet et al., 2021). The finding that DSO symptoms did not moderate between PTSD treatment outcomes and conditions is consistent with the results of a study that compared 16 sessions of Prolonged Exposure therapy (PE) with eight sessions of PE preceded by eight sessions of STAIR in the same target group (Hoeboer et al., 2021). Our results showed that even when STAIR was added to trauma-focused treatment (instead of replacing a part of the trauma-focused treatment), the severity of the separate DSO symptoms did not moderate between the treatment conditions and PTSD treatment outcomes. In other words, the severity of DSO symptoms does not seem to predict whether a patient will profit more from phase-based treatment than from trauma-focused treatment, or vice versa.

The finding that the presence of a borderline personality disorder was not found to be a predictor of PTSD treatment outcomes is in line with the results of a meta-analysis of patients with regular PTSD, which showed that psychotherapy for PTSD in general is effective in patients with borderline personality disorder (Slotema et al., 2020). This suggests that irrespective of the type of trauma background, the presence of BPS may not influence the results of treatment outcomes for PTSD. In contrast to the results of a meta-analysis showing that dissociative phenomena are not predictive of PTSD treatment outcomes (Hoeboer et al., 2020), we found that dissociation symptoms were predictive of worse posttreatment PTSD outcomes. This difference may be explained by differences in target groups, because childhood trauma has been found to be associated with the presence of more severe dissociative symptoms than other trauma types (Schalinski & Teicher, 2015). Hoeboer et al. (2020) included studies on patients with PTSD from all types of trauma backgrounds. Furthermore, self-injury and suicidality did not predict PTSD treatment outcome.

One of the strengths of the current study is that it is one of the first to identify multiple potential predictors and moderators in the treatment of a patient group with PTSD related to childhood abuse, without excluding patients with a borderline personality disorder and/or self-injurious or suicidal behaviour. Second, in line with the Complex PTSD guidelines (Cloitre et al., 2012), STAIR was used as an add-on to trauma-focused therapy within the phase-based treatment condition rather than as a replacement for (a part of) the trauma-focused sessions. The third strength is the six-month follow-up PTSD measurements, rendering it possible to determine the impact of a variety of predictor and moderator variables on the long-term effects of trauma-focused treatment. Conversely, a limitation of the present study is that the sample size was small, and there were missing data (data were imputed) that decreased statistical power, restricted the number of predictors and moderators that could be included, and limited the ability to examine higher-order interactions. In addition, suicidal and self-injurious behaviours were only measured by single items, which also limited the conclusions. One noteworthy aspect for discussion pertains to the source of these findings, which were derived from a dataset originating from an RCT that has been used for various analyses, including one based on the personalized advantage index, PAI (Bremer et al., 2023). However, an essential distinction lies in the fact that the present study aimed to assess the actual predictive value of a wide variety of variables, previously assumed to have a differential impact on treatment outcomes. This differs from the PAI study, a distinct analytical approach that sought to investigate a broader question of whether certain patients with childhood-related PTSD and potential symptoms of complex PTSD would benefit more from a phased treatment approach, whereas others might benefit more from a direct trauma-focused approach. This was achieved by utilizing all variables measured in the study and employing machine learning techniques to identify the combination of variables with the strongest predictive value. Although it is important to acknowledge that relying on a singular dataset for different analyses carries the inherent risk of over-publication of results (Supak Smolcić, 2013), it is also important to establish that the results of both studies converge on the notion that there are no compelling grounds to favour phase-based treatment over trauma-focused treatment.

The fact that both the severity of PTSD symptoms and the severity of dissociative symptoms were predictive of less favourable PTSD treatment outcomes (the latter only at post-treatment and not at follow-up) suggests that individuals who develop severe PTSD symptoms as a result of exposure to adverse childhood events have a relatively worse long-term treatment prognosis. Conceivably, PTSD treatment results for this specific patient group were not optimal. To this end, future studies should focus on improving the treatment outcomes for this target group to enhance the effects of trauma-focused PTSD treatment. This should not be STAIR, because Complex PTSD symptoms were not found to moderate between PTSD treatment outcome and the EMDR condition preceded by STAIR or the EMDR-only condition. Since no clear moderators were found that influenced the relationship between treatment conditions and PTSD treatment outcomes, future studies should focus on investigating other possible moderators or possible combinations of factors that may influence treatment outcomes on one treatment or the other, rather than separate symptoms.

In conclusion, the present findings provide valuable insights into treatment predictors and moderators for patients with severe PTSD related to childhood abuse, addressing unresolved questions involving ongoing discussions on the treatment of these patients. The results of this study, although with some limitations (e.g. size of the sample and the amount of missing data), suggest that trauma-focused treatments can lead to improved outcomes in patients with PTSD related to childhood abuse. However, individuals with severe PTSD symptoms may experience less favourable outcomes. The same holds true for dissociative symptoms, but this turned out to only pertain tot heir status immediately after treatment, and not in the longer run, and not when corrected for false positives. Further research is necessary to replicate these findings and develop treatments for patients at risk of achieving less profitable outcomes. Nonetheless, these and findings of other recent studies on this topic convert to suggest that even patients with severe forms of PTSD profit from evidence-based trauma-focused therapies.

Ethics statement

The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures were approved by the Medical Ethics Committee Twente (merged with the Institutional Review Board United), reference number P16–03.

Consent statement

All patients received the required information about the study to decide whether to participate. Written informed consent was obtained after the procedure was fully explained.

Author contribution

All authors have made substantial contributions to the design of this study. NvV and SvH were responsible for the implementation of this study. NvV, SvH, and NB were responsible for the inclusion process and the data collection. NvV drafted the paper under supervision of MvD, RH, and AdJ. All authors were involved in critically revising the manuscript and made contribution to analysis and interpretation of the data. NvV, AdJ, MM, and SvH supervised the therapist involved in the study. MH, RH, and NvV were involved in the statistical analyses. All authors read and approved the final manuscript.

Disclosure statement

Ad de Jongh receives income for published books on EMDR therapy and for the training of postdoctoral professionals in this method. Mariel Meewisse and Sietske van Haren receive income for the training of postdoctoral professionals in EMDR and Cognitive Behavioral Therapy. There is no conflict of interest in the present study for any of the other authors.

Data availability

The data that support the findings of this study are available on request from the corresponding author (NvV).

Additional information

Funding

Funding for this study was provided by Stichting tot Steun VCVGZ, Dimence Mental Health Care, and the Dutch EMDR association.