Family therapy and EMDR after child abuse and neglect: moderating effects of child attachment style and PTSD symptoms on treatment outcome Family therapy and EMDR after child abuse and neglect: moderating effects of child attachment style and PTSD symptoms on treatment outcome

Çocuk istismarı ve ihmalinden sonra aile terapisi ve EMDR: Çocuk bağlanma stili ve TSSB semptomlarının tedavi sonucu üzerindeki hafifletici etkileri

Family therapy and EMDR after child abuse and neglect: moderating effects of child attachment style and PTSD symptoms on treatment outcome

 

Authors: Mara L. van der Hoeven, Samantha Bouwmeester, Nathalie E. F. Schlattmann, Ramón J.L. Lindauer & Irma M. Hein

 

Translated by: Specialist Psychologist Gizem Pozam

 

ABSTRACT

Background: Effective and appropriate care and treatment for children to reduce psychosocial problems following exposure to child abuse and neglect (CAN) are vital, given the severity of the symptomatology that may develop.

 

Objective: The aim of this study was to examine whether attachment style and core-cluster posttraumatic stress disorder symptoms act as moderators of treatment outcomes for a new integrative treatment model for trauma and attachment. In this treatment model, family therapy is combined with EMDR, and barriers to trauma processing are addressed first.

 

Method: We included children aged 6–12 years with a history of CAN who had not responded to evidence-based trauma treatment. The targeted treatment outcomes were attachment, posttraumatic stress symptoms, behavioral problems, and emotion regulation problems. We conducted a multiple-baseline ABC Single-Case Experimental Design (SCED). We divided the 12 participants into four groups based on attachment style and core-cluster PTSD symptoms: (1) non-disorganized and re-experiencing; (2) non-disorganized and avoidance/hyperarousal; (3) disorganized and re-experiencing; and (4) disorganized and avoidance/hyperarousal. We compared the four groups with one another and over time, and examined the interaction between groups and the effect over time. We performed nonparametric permutation tests and estimated q-values to control the false discovery rate.

 

Results: Children with a disorganized attachment style generally had more severe symptomatology, with the exception of posttraumatic stress symptoms. The treatment appears to be more effective in targeting and successfully treating children with a non-disorganized attachment style, particularly children with a non-disorganized attachment style and core-cluster PTSD symptoms.

 

Conclusion: Our study underscores the complexity of treating children who develop a complex combination of symptoms following CAN and calls for the continued development of innovative interventions.

 

 

KEYWORDS

Child abuse and neglect; trauma treatment; family therapy; EMDR; posttraumatic stress symptoms; attachment style; moderator analyses; single-case experimental design

 

 

 

1. Introduction

Child abuse and neglect (CAN) in young children involves early relational traumatic events that can lead to a wide range of (developmental) problems across various domains, including physical health problems, posttraumatic stress symptoms (PTSS), behavioral problems, attachment problems, anxiety problems, and depression (Carr et al., 2020). These problems can hinder future development in terms of social, educational, and societal functioning and threaten the continuity of the child's caregiving environment and, in foster care, placement stability (Konijn et al., 2019; Rostill Brookes et al., 2011). Effective and appropriate care and treatment to reduce psychosocial problems following experiences of CAN are therefore vital. Effective trauma-focused, attachment-focused, or family-focused treatments have been developed for different types of problematic behavior. However, these treatments may be too narrow in scope given the complex variety of problems that characterize this particular group of children. Extreme avoidance, behavioral and emotional dysregulation, and interpersonal reactivity in the child can create challenges that are difficult to overcome in the treatment context (Struik et al., 2017; Wesselmann et al., 2017; Wesselmann and Shapiro, 2013).

 

An integrated treatment model comprising different techniques has been developed for this vulnerable group of children to better meet their treatment needs (van der Hoeven et al., 2023). The Integrative Attachment Trauma Protocol for Children (IATP-C; Wesselmann et al., 2014) and its Dutch adaptation (Integratieve Gehechtheidsbevorderende Traumabehandeling voor Kinderen; IGT-K; Schlattmann et al., 2023) combine EMDR (Eye Movement Desensitization and Reprocessing; Shapiro, 2017) and family therapy. The treatment aims to reduce barriers to trauma treatment. These barriers may include poor affect tolerance, disruptive or avoidant child behavior, and difficulties in the parent–child relationship. The original developers of IATP-C conducted case-series research, and the results showed reductions in attachment problems, behavioral problems, and traumatic stress symptoms (Wesselmann et al., 2018). Recently, the first experimental study of IGT-K was conducted. The results showed that this integrative treatment model may successfully reduce problems in this population, for whom treatment options are limited because of difficulties engaging in evidence-based treatments and a lack of other available treatments (van der Hoeven et al., 2023).

 

Despite these promising results, questions remain about the influence of various child characteristics on treatment outcomes in this population, as not all children showed improvement on all outcome measures (van der Hoeven et al., 2023). Variation was found across cases and outcome measures. Two children showed clinical improvement on all four outcomes, two on three outcomes, three on one outcome (although the outcomes differed for each case), and one on none of the outcomes. This raised further questions about which mechanisms drive treatment outcomes and for whom this treatment works best (Kraemer, 2016). This led us to wonder whether pretreatment variables might also function as moderators of treatment outcomes. We developed a particular interest in the potential role of attachment style and PTSD as moderators, as these constructs may manifest differently in each individual.

 

Within the extensive literature on youth mental health services, we found little research on the contribution of attachment representations to differences in treatment outcomes (Hornstra et al., 2022; Stefini et al., 2013). Attachment can be conceptualized as a child's tendency to seek proximity and contact with an attachment figure (usually a parent), particularly in anxiety-provoking and stressful situations (Ainsworth and Wittig, 1969; Bowlby, 1988). This concept includes the individual's confidence in the availability of an attachment figure to function as a secure base and a safe haven. Three organized attachment styles (secure, anxious-avoidant, and anxious-ambivalent) and one disorganized style have been identified (Ainsworth et al., 1978; Main and Solomon, 1990). Securely attached children trust in the physical and emotional availability of their attachment figures. They develop organized thoughts, behaviors, and feelings that enable them to express their attachment needs adequately and accurately and to expect a positive response from their caregivers (Schore, 2001). Insecure attachment refers to a lack of confidence in the caregiver's availability when the child needs comfort or protection. Anxious-avoidant and anxious-ambivalent patterns emerge when caregivers are not emotionally available or responsive to the child's needs. To maintain the relationship with the attachment figure, children may minimize their attachment needs because they consistently experience low responsiveness or rejection from the attachment figure (anxious-avoidant), or maximize their attachment needs because they experience inconsistent responsiveness from the attachment figure (anxious-ambivalent) (Ainsworth et al., 1978). Children who cannot be classified into one of the three organized attachment categories perceive their attachment figures as sources of both comfort and fear. This creates an insoluble paradox and prevents the child from developing an organized strategy for obtaining help from the attachment figure. It leads to a combination of different types of insecure behavior along with fearful responses toward the attachment figure (Main and Hesse, 1990). Ambivalent and avoidant attachment styles are considered risk factors for maladjustment, while the disorganized attachment style is thought to be most strongly associated with psychopathology (Main, 1996; Zeanah and Gleason, 2015). This blueprint of what to expect in interactions and relationships with another person may also influence interactions and relationships with people other than caregivers (Bowlby, 1980). Attachment styles may therefore be thought to carry over into the therapeutic context and into the way the child positions themselves within therapy or interacts with the therapist. Since the disorganized attachment style is considered to be most strongly associated with psychopathology and most resistant to treatment (de Wolff & Wildeman, 2020), it may be hypothesized that children with the highest scores on the disorganized style would show less improvement on outcome measures than children with the highest scores on a non-disorganized attachment style (ambivalent or avoidant).

 

We also found little research on the moderating role of PTSD-related variables in treatment outcomes in young people. Studies have shown that children with more severe PTSD at baseline achieve a smaller response to direct trauma-focused treatments (de Roos et al., 2021; Lindebø Knutsen et al., 2020; Wamser-Nanney et al., 2016). The potential moderating role of dysfunctional posttraumatic cognitions in treatment outcome has been examined, with studies finding no evidence for such a moderating role (de Roos et al., 2021 Lindebø Knutsen et al., 2020). Trauma type has also been investigated as a moderator, with mixed results. While some studies found no evidence (Danzi and La Greca, 2021; Goldbeck et al., 2016), others found larger effect sizes for treatment outcomes in PTSD, internalizing problems, and functioning among children exposed to sexual abuse (Kane et al., 2016; Silverman et al., 2008). Children with higher levels of emotional dysregulation showed smaller reductions in symptoms during treatment (Sharma-Patel and Brown, 2016). Children enrolled in IGT-K are characterized by relatively high levels of emotional dysregulation and/or avoidance. These symptoms can complicate the treatment process (Struik et al., 2017; Wesselmann et al., 2017; Wesselmann and Shapiro, 2013). Accordingly, children with the highest scores in the avoidance and/or hyperarousal clusters might show less improvement on outcome measures than children with the highest scores in the re-experiencing cluster. In addition, different combinations of attachment style and core PTSD cluster type may together lead to different treatment outcomes across the population. Since a disorganized attachment style and avoidance or hyperarousal as core-cluster PTSD symptoms have been found to complicate the treatment process, children with this combination of symptoms are hypothesized to show less improvement on all outcome measures than those with other combinations. Children with a non-disorganized attachment style and re-experiencing as their core PTSD symptoms are hypothesized to show greater improvement than those with other combinations of attachment style and core PTSD symptoms.

 

The aim of this study was to identify potential moderators of children's treatment outcomes in the integrative treatment model for trauma and attachment. We were particularly interested in whether attachment style (disorganized versus non-disorganized) and core-cluster PTSD symptoms (avoidance or hyperarousal versus re-experiencing) act as moderators of treatment outcomes in terms of levels of attachment problems, PTSD, behavioral problems, and emotion regulation problems.

 

1. Methods

2.1. Study design

This article was part of the first study on the effectiveness of the integrative treatment model for trauma and attachment in children, described extensively elsewhere (van der Hoeven et al., 2023). The Medical Ethics Committee of Amsterdam UMC, AMC, approved this project under project number W17_169.

 

We conducted a multiple-baseline ABC Single-Case Experimental Design (SCED). Participants were randomized to baseline lengths of three, four, or five weeks (Kratochwill et al., 2010). The intervention consisted of 33 weeks and was divided into two phases: a phase involving preparatory techniques and a phase involving trauma-processing interventions.

 

2.2. Setting and participants

We conducted the study at two academic centers for child and adolescent psychiatry in the Netherlands, with recruitment taking place between May 2018 and December 2020. Written informed consent was obtained from guardians, and verbal informed consent was obtained from the children. The following eligibility criteria were established: (1) children aged four to 12 years with a stable living situation (with biological, foster, adoptive, or other resource parents with whom it should be possible to establish an attachment relationship)—if the parent participating in treatment was responsible for past abuse, the parent was required to have received treatment and the abuse to have ended; (2) referral to specialized youth mental health services for assessment and treatment of PTSD, attachment problems, behavioral problems, and/or self-regulation problems, with scores at the clinical cutoff for at least three of these four outcome measures; (3) a history of child maltreatment within primary attachment relationships; and (4) an unsuccessful attempt at direct evidence-based trauma-focused treatment or a clinical judgment that the child could not engage in direct trauma-focused treatment because of strong avoidance and dysregulated behavior. Exclusion criteria were fetal alcohol syndrome, autism spectrum disorder, a child IQ score of 70 or below, or a clinical judgment that the current caregiver's psychiatric problems would interfere with the caregiver's (emotional) availability to the child during treatment.

 

Fifteen children were found eligible to participate in the study. One participant withdrew consent shortly before treatment began because of unexpected changes in the child's caregiving system. One participant dropped out after six treatment sessions because the caregivers could not make the time commitment required by the treatment. One participant completed only seven measurements and therefore could not be included in the analyses, although this participant completed treatment. Four participants completed most measurements but dropped out during the final treatment phase. Their reasons for dropping out were not directly related to the treatment or their problems, but rather to unexpected court decisions concerning changes in placement or visitation arrangements with biological parents. We therefore decided to include these four participants, along with the eight participants who completed both the study and treatment in full, in the analyses of this study. Seven of the 12 participants included in this study were boys. The mean age of these 12 children at inclusion was 9.72 years (SD = 1.85; range = 6–12 years). Participants' background variables and pretreatment scores are shown in Table 1.

 

2.3. Procedure

After consent was obtained, the child and caregivers were assigned to a pair of therapists, and the length of the baseline phase was randomly determined. Throughout the baseline and intervention phases, caregivers completed four questionnaires (CRIES, AISI, SDQ, and the Emotional Control scale of the BRIEF) weekly by telephone on a predetermined date.

 

2.4. Treatment and treatment fidelity

The treatment examined in the study was 'Integratieve Gehechtheidsbevorderende Traumabehandeling voor Kinderen' (IGT-K), the Dutch adaptation of the Integrative Attachment Trauma Protocol for Children (Wesselmann et al., 2014; Wesselmann et al., 2018). IGT-K combined family therapy with EMDR therapy. The protocol targeted barriers such as extreme avoidance and dysregulated behavior that made it difficult for the child to benefit from trauma-focused treatments. Treatment began with a 17-week preparatory phase, with the initial sessions held with caregivers alone for psychoeducation. This was followed by ten caregiver–child sessions involving work to strengthen attachment and improve self-regulation (phase B). Subsequently, treatment continued with a 16-week EMDR trauma-processing phase. For a detailed description of the treatment protocol, we refer to previous publications and the therapist manual (Schlattmann et al., 2023; van der Hoeven et al., 2023; Wesselmann et al., 2014; Wesselmann et al., 2018). A description of how therapists were trained and how treatment fidelity was ensured can also be found there.

 

2.5. Measures

Measurements were taken before and after treatment and weekly during treatment to assess attachment problems, PTSD, emotion regulation difficulties, and behavioral problems. All questionnaires were completed by the primary caregiver. We used the following measures.

 

The Attachment Insecurity Screening Inventory (AISI; Polderman & Kellaert-Knol, 2012) was used to assess attachment problems. This questionnaire assesses primary caregivers' perceptions of insecure-avoidant, insecure-ambivalent/resistant, and insecure-disorganized attachment relationships with their children. The questionnaire contains 20 items rated on a 6-point Likert scale: never, sometimes, regularly, often, very often, and always. Together, the three subscales yield a total insecure attachment score ranging from 20 to 120. A score of 50 or higher indicates insecure attachment. The Dutch version of the AISI has acceptable psychometric properties, with Cronbach's α ranging from .65 to .80 (Spruit et al., 2018).

 

 

 

To measure PTSD, we used the caregiver-report version of the Children's Revised Impact of Event Scale, which consists of 13 items (CRIES-13; Perrin et al., 2005; Verlinden et al., 2014). These items can be scored as never (0), rarely (1), sometimes (3), or often (5). The items can be summed into three scales: avoidance, re-experiencing, and arousal. The total score can range from zero to 65. A score of 30 or higher indicates a higher risk of developing PTSD. The Dutch version of the CRIES-13 has good psychometric properties, with a Cronbach's α of .87 (Verlinden et al., 2014).

 

We measured emotional and behavioral difficulties using the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997; van Widenfelt et al., 2003). This is a 25-item questionnaire describing positive and negative attributes of children and adolescents. The questionnaire consists of five subscales: emotional symptoms, conduct problems, hyperactivity-inattention symptoms, peer problems, and prosocial behavior. Together, the first four scales yield a total difficulties score, with higher scores reflecting greater difficulties. The total score can range from zero to 40. We used the total difficulties score in our study. The Dutch version of the SDQ has good psychometric properties, with a Cronbach's α of .70 (Muris et al., 2003).

 

To measure self-regulation problems, we used the Emotional Control subscale of the Behavior Rating Inventory of Executive Function (BRIEF; Gioia et al., 2000a; Huizinga and Smidts, 2009). Scores on this scale range from 10 to 30. Higher scores indicate greater difficulties with emotional control. Overall, the BRIEF has demonstrated good psychometric properties, with Cronbach's α values between .78 and .90 (Gioia et al., 2000b; Huizinga and Smidts, 2009, 2010).

 

2.6. Data analysis

We focused on the potential moderating effects of attachment style (disorganized versus non-disorganized) and core-cluster PTSD symptoms (avoidance/hyperarousal versus re-experiencing) on treatment response. Treatment outcome measures were levels of attachment problems, PTSD, behavioral problems, and emotion regulation problems. Regarding attachment scores, we examined which attachment style (anxious-avoidant, anxious-ambivalent, or disorganized) had the highest AISI score for each participant. We then combined participants in the anxious-avoidant and anxious-ambivalent groups into a single 'non-disorganized' group. Regarding PTSD symptoms, we examined which of the three clusters (re-experiencing, avoidance, or hyperarousal) had the highest CRIES score for each individual. We then combined participants in the avoidance and hyperarousal groups into a single 'avoidance/hyperarousal' group. We divided participants into four groups for the attachment and core-cluster PTSD symptom variables, based on their relatively highest-scoring AISI or CRIES subscale measured before treatment. The four groups were as follows: (1) non-disorganized and re-experiencing; (2) non-disorganized and avoidance/hyperarousal; (3) disorganized and re-experiencing; and (4) disorganized and avoidance/hyperarousal.

 

To assess the value of the intervention for each outcome measure within each group, we calculated effect sizes in terms of Cohen's d. We compared the four groups, the three phases (baseline, preparation, and treatment), and the interaction between groups and the effect over time. First, for each of the four groups, we aggregated all observations within a phase for each outcome measure and calculated the mean. For the main effect of group, we assessed whether the groups differed in their means aggregated across all observations in the three phases. For the main effect of time, we aggregated observations from the different groups within a phase and compared the means of the three phases. For the interaction effect, we assessed whether the differences between phases differed across the four groups.

 

2.6.1. Randomization tests

For the randomization tests, we followed Ernst (2004) and performed nonparametric permutation tests using the Shiny App Single Case Designs, which runs on R version 4.3.2 and uses, among other things, the 'shiny' package (Bouwmeester, 2021). For group comparisons, observed mean differences were first calculated by aggregating scores within a group for each of the three phase comparisons. The scores from the two phases being compared were then shuffled and randomly assigned to one of the phases. This was done for each of the four groups and repeated 1,000 times. We then calculated the squared mean differences for these randomly generated phases and summed them. The p-value was then calculated as the proportion of randomly generated t-values that were more extreme than the observed t-values. When the overall p-value was significant, we performed post-hoc analyses to check which of the four groups differed. We estimated q-values to control the false discovery rate (Storey, 2002, 2003). As many statistical tests were performed, q-values will be reported in addition to p-values. The q-value is an analogue of the p-value that incorporates correction for multiple testing. The q-value is defined as the minimum false discovery rate at which an observed score is deemed significant.

 

 

1. Results

Table 2 shows the allocation of the 12 participants to the four groups. See Table 3 for descriptive values. Table 4 presents effect sizes (Cohen's d) for phase comparisons for each outcome measure within each of the four groups. Table 5 shows p- and q-values for the main effect of group, the main effect of time (three phases: baseline, preparation, and intervention), and the interaction effect (time × group) for the four outcome measures. Tables 6–8 then show the results of post-hoc tests for group, time, and time × group. Figures 1–4 show the symptom trajectories of each of the four groups for each outcome measure.

 

3.1. Attachment problems

A small effect size was found for the non-disorganized × avoidance/hyperarousal group. There was a main effect of group, p < .001. Post-hoc tests showed that, averaged across all phases, all groups differed from one another except the non-disorganized × re-experiencing group and the non-disorganized × avoidance/hyperarousal group, p = .910. The disorganized × re-experiencing group had the highest mean, followed by the disorganized × avoidance/hyperarousal group. The two groups of children with a non-disorganized attachment style had the lowest means. No significant effect of time was found, p = .926, indicating that the baseline, preparation, and treatment means did not differ when all groups were combined. The results showed a significant time × group interaction effect, p < .001. However, post-hoc tests showed that none of the pairwise comparisons had a p-value below .05.

 

 

 

3.2. PTSS

A medium effect size was found for the non-disorganized x re-experiencing group. There was a main effect of group, p < .001. Post-hoc tests showed that, averaged across all phases, the groups differed from one another except for the non-disorganized x re-experiencing group and the disorganized × avoidance/hyperarousal group, p = .074, and the disorganized x re-experiencing group and the disorganized × avoidance/hyperarousal group, p = .078. The non-disorganized × avoidance/hyperarousal group had the highest mean, followed by the non-disorganized × re-experiencing group, the disorganized × avoidance/hyperarousal group, and the disorganized × re-experiencing group. No significant effect of time was found, p = .458, indicating that the baseline, preparation, and treatment means did not differ when all groups were combined. The results showed a significant time × group interaction effect, p = .001. Post-hoc tests revealed four significant comparisons. The decrease in scores from baseline to the treatment phase in the non-disorganized × re-experiencing group differed significantly from the increase in scores in the disorganized × re-experiencing group, p = .043, and the disorganized × avoidance/hyperarousal group, p < .001. The same result was found for the difference between the preparation and treatment phases: the decrease in scores in the non-disorganized × re-experiencing group differed significantly from the increase in scores in the disorganized × re-experiencing group, p = .029, and the disorganized × avoidance/hyperarousal group, p = .001.

 

 

 

 

 

 

 

 

 

3.3. Behavioral problems

A large effect size was found for the non-disorganized × re-experiencing group. There was a main effect of group, p < .001. Post-hoc tests showed that, averaged across all phases, all groups differed from one another except the non-disorganized × re-experiencing group and the disorganized × re-experiencing group, p = .514. The disorganized × avoidance/hyperarousal group had the highest mean, followed by the non-disorganized × re-experiencing group and the disorganized × re-experiencing group. The non-disorganized × avoidance/hyperarousal group had the lowest mean. There was a significant effect of time, p = .011. The post-hoc test showed that baseline scores were higher than treatment scores, p = .005, and preparation scores were higher than treatment scores, p = .004. The results showed a significant time × group interaction effect, p = .002. Post-hoc tests revealed three significant comparisons. The decrease in scores from baseline to treatment in the non-disorganized × re-experiencing group was greater than the decrease in scores in the disorganized × avoidance/hyperarousal group, p = .027. The same result was found for the difference between the preparation and treatment phases: the decrease in scores in the non-disorganized × re-experiencing group was greater than the decrease in scores in the disorganized × avoidance/hyperarousal group, p = .016. There was also a significant difference in the decrease in scores from the preparation phase to the treatment phase between the non-disorganized × re-experiencing group and the disorganized × re-experiencing group, p = .016.

 

3.4. Emotion regulation problems

A large effect size was found for both non-disorganized groups and a medium effect size for the disorganized × avoidance/hyperarousal group. There was a main effect of group, p < .001. Post-hoc tests showed that, averaged across all phases, the mean of the non-disorganized x avoidance/hyperarousal group was lower than those of the other three groups, which did not differ from one another. There was a significant effect of time, p = .001. The post-hoc test showed that baseline scores were higher than treatment scores, p = <.001, and preparation scores were higher than treatment scores, p = .011. However, the q-value for the latter was not significant (q = .083). The results showed a significant time × group interaction effect, p < .001. Post-hoc tests revealed two significant comparisons. The decrease in scores from baseline to treatment in the non-disorganized × re-experiencing group was greater than the decrease in scores in the disorganized × avoidance/hyperarousal group, p = .022 (although the q-value was not significant: q = .314), and the disorganized × re-experiencing group, p = .035.

 

1. Discussion

In this study, we aimed to identify potential moderators of treatment response in children participating in a new integrative treatment model for trauma and attachment. We examined attachment style (disorganized vs. non-disorganized) and core-cluster PTSD symptoms (avoidance or hyperarousal vs. re-experiencing) as moderators of treatment outcomes in terms of levels of attachment problems, PTSD, behavioral problems, and emotion regulation problems.

 

Overall, children with a disorganized attachment style were found to experience generally more severe symptom levels. These findings are consistent with previous research showing that children with a disorganized attachment style have higher levels of externalizing problem behavior than children with other insecure styles (Badovinac et al., 2021; Fearon et al., 2010; Groh et al., 2017; O'Connor et al., 2011). Disorganized attachment is thought to be linked to emotion regulation processes: children who experience their caregivers as sources of both safety and fear may perceive the world as unpredictable and frightening and make less use of problem-solving coping strategies because they have missed opportunities to develop appropriate emotion regulation strategies (Brumariu, 2015; Brumariu et al., 2012; Lyons-Ruth and Jacobvitz, 2008). These vulnerabilities in attachment and emotion regulation may lead to greater difficulties with behavioral or emotional control, as disorganized attachment has been found to be associated with both internalizing and externalizing behaviors (Walsh et al., 2008). Children with a non-disorganized attachment style and re-experiencing experienced more behavioral and emotion regulation problems than children with a non-disorganized attachment style and avoidance/hyperarousal. Children relatively more affected by re-experiencing symptoms may become more distressed by intrusions, which may lead to emotional dysregulation and behavioral problems (Levendosky et al., 2002; Miller Graff et al., 2016).

 

However, we found the opposite results for the PTSD outcome variable. Both groups of children with a disorganized attachment style had lower PTSD levels at the beginning of treatment than the two groups of children with a non-disorganized attachment style. Furthermore, PTSD levels in the two groups of children with a disorganized attachment style increased over the course of treatment. This finding contradicts previous research indicating that children with a disorganized attachment style in infancy showed higher levels of PTSD at age 8.5 than children with a non-disorganized attachment style (MacDonald et al., 2008). However, the sample in that study differed from ours, as it consisted of children with intrauterine cocaine exposure. It is conceivable that PTSD is harder to recognize in children with a disorganized attachment style because it is less distinct and overshadowed by behavioral problems.

 

In addition, in this exploratory study, we found several moderating effects on treatment outcomes in three of the four problem domains. Overall, these results may suggest that the treatment is more effective for children with a non-disorganized attachment style. To our knowledge, there is relatively little literature on the role of attachment style and type of core-cluster PTSD symptoms in the effectiveness of treatment for CAN-related problems, making it difficult to interpret our findings conclusively. Some studies have found that changes in posttraumatic cognitions play a mediating role in the different treatment outcomes between evidence-based trauma-focused treatment and control groups (Jensen et al., 2018; Pfeiffer et al., 2017). Our study appears to be among the first to investigate the role of attachment style and core-cluster PTSD symptoms in treating a complex population. Accordingly, we propose initial possible hypotheses that might clarify our findings, and these should therefore be considered speculative.

 

Looking at our findings, the results indicate that much of the difference in symptom trajectories during treatment between children with a non-disorganized attachment style and those with a disorganized attachment style emerged during the trauma-processing phase. This leads us to hypothesize that more caregiver–child sessions than currently specified in the preparatory phase of the protocol may be needed to help children with a disorganized attachment style transform their disorganized attachment patterns into non-disorganized patterns. Furthermore, given the number of traumatic events characterizing our study population, it is hypothesized that rapid and continuous processing of traumatic events may be overwhelming for these children. It may therefore be preferable to integrate sessions into the trauma-processing phase in which no actual trauma processing takes place and attention is focused on attachment and self-regulation, allowing the child some respite. Finally, a disorganized attachment style has been found to be associated with dissociation (Joubert et al., 2012; MacDonald et al., 2008; Ogawa et al., 1997). It is therefore hypothesized that children may gain greater awareness of their bodies and feelings during the attachment and regulation phase. As a result, their symptoms may be perceived more intensely and consequently show an increase in reports. In addition, working with what happened in the past may evoke feelings of sadness, anger, and grief. The treatment phase may therefore need to last even longer. Another possibility that occurred to us is that, following the attachment and regulation phase, some children may need a trauma-focused approach other than EMDR, such as Trauma-Focused Cognitive Behavioral Therapy. The adult literature has found that, at an individual level, patients may differ substantially in their treatment response to the two treatments (Deisenhofer et al., 2018). This may also be true for children. However, all these lines of reasoning remain speculative, as this topic has been little studied. There are indications that a disorganized attachment style has a stronger association with psychopathology (Borelli et al., 2010; Gazzillo et al., 2020; Zeanah et al., 2016). A disorganized attachment style may therefore be more difficult to treat than ambivalent or avoidant attachment styles.

 

As the treatment appeared to be more effective in terms of PTSD for the group of children with a non-disorganized attachment style and re-experiencing, it is possible that, during treatment, a child stops avoiding or resisting trauma-related stimuli and therefore temporarily experiences more PTSD symptoms in the form of hyperarousal and re-experiencing, leading to temporarily higher PTSD levels (Larsen et al., 2016). Children with re-experiencing as their core PTSD symptoms may therefore benefit more quickly from trauma processing. However, this is only speculation.

 

In summary, our findings contribute to the currently limited research on attachment style and PTSD as potential moderators of treatment outcomes. The results indicate differences in symptom trajectories across the four groups comprising our study sample, governed by attachment style and core-cluster PTSD symptoms. However, the preliminary hypotheses derived from this study's results require further research to examine and elaborate on these findings. According to our results, the current treatment protocol appears clinically more suitable for children with a non-disorganized attachment style, particularly children with a non-disorganized attachment style and core-cluster PTSD symptoms. Despite the positive results for this group, outcomes remain unsatisfactory for the group of children with a disorganized attachment style. Although causality is not yet clear, this is concerning, given that a disorganized attachment style is known to be associated in some way with severe forms of psychopathology later in life (Gazzilli et al., 2020; Lecompte and Moss, 2014). Insufficient effectiveness among interventions for attachment problems is a well-known issue, as other existing interventions targeting attachment disorganization do not significantly alter or prevent disorganized attachment (Bakermans-Kranenburg et al., 2005). This highlights how far the scientific field still has to go in developing interventions for trauma- and attachment-related symptoms.

 

4.1. Limitations

We based our results on a small sample of 12 children. Our findings are therefore preliminary and entirely exploratory. They should be interpreted with caution. The limited sample size also prevented us from classifying our sample into more than the current four groups of moderator variables. We therefore could not create separate categories for children experiencing avoidance and hyperarousal as core-cluster PTSD symptoms. Although this treatment focused specifically on children with severe avoidance and dysregulation, it would be interesting to investigate whether these core-cluster PTSD symptoms individually account for different symptom trajectories. In addition, as this article concerns the first experimental study of a new treatment model, the number of treatment sessions was predetermined. However, the therapists participating in the study informed us that, when the study ended and data collection stopped, in some cases they continued treating the child with EMDR because they were able to conduct EMDR with the child and there were still targets to reprocess. We may therefore have ended the study too early and consequently missed the full trajectory of symptom change. This may also explain the difference in PTSD-related outcomes between our study and that of the original developers. In their study, some children continued treatment for up to 24 months (Wesselmann et al., 2018) and may therefore have had more time to effectively address trauma targets. In addition, we used the CRIES-13 to assess PTSD. The CRIES-13 is based on the DSM-IV criteria for PTSD. We therefore could not include DSM-5 criterion D, 'Negative alterations in cognitions and mood associated with the traumatic events.' In future studies, we will use the Dutch version of the Child and Adolescent Trauma Screen 2 (CATS-2), which includes criterion D (Kooij and Lindauer, 2019; Sachser et al., 2017, 2022). Finally, we did not include the duration of the child's past exposure to CAN among the participant variables. This factor may play a role in explaining the different findings.

 

4.2. Recommendations for future research and clinical practice

This study highlights the importance of continuing to develop innovative interventions for children for whom evidence-based interventions are insufficient. Leaving these children and their symptoms untreated is not an option. We recommend that future research further investigate more personalized, innovative treatment models while also focusing on which interventions work for whom. SCED research has proven to be a useful study design for examining complex populations. In recent years, SCED methodology and analyses have advanced (Jamshidi et al., 2022; Kratochwill et al., 2021), enabling us to gain richer insight into the potential developmental trajectories of different types of symptoms. As moderator analyses are also possible within SCED research, we may gain a deeper understanding of why some patients respond to particular treatment modules while others do not. Contextual factors, such as family changes or changes in living conditions, should also be included as potential moderators. In particular, we call for the development of innovative and integrated treatment models for children with a disorganized attachment style. Finally, as our efforts represent only an initial attempt to determine what works for whom in CAN-related treatments, we encourage other researchers to include attachment and core-cluster PTSD moderator analyses in studies of (other) trauma-focused treatments.

 

1. Conclusion

Our overall findings indicate that the integrated treatment model for trauma and attachment, developed specifically for a group of children with severe symptomatology and often a history of unsuccessful treatment attempts, successfully treats children with a non-disorganized attachment style, particularly those with a non-disorganized attachment style and re-experiencing as their core-cluster PTSD symptoms. However, this treatment does not reach and successfully treat every participant: there are still children who do not benefit (sufficiently) from the treatment in its current form. This appears particularly true for children with a disorganized attachment style. This underscores the complexity of treating children suffering from a complex combination of symptoms following CAN and calls for continuous adjustment of existing treatment protocols and the development of innovative interventions for this particular group of children.

 

 

Data availability statement

Due to the nature of this research, the participants in this study did not agree to have all their data shared publicly, so supporting data are not available.

 

Ethics approval statement

The Medical Ethics Committee of Amsterdam UMC, AMC, approved this project under project number W17_169.

 

Acknowledgments

We are grateful to the children, their families, and the therapists who participated in this study.

 

Disclosure statement

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

 

Funding

This work was supported by ZonMw: .

 

ORCID Irma M. Hein http://orcid.org/0000-0002-4664-2784

 

 

The q-value is defined as the minimum false discovery rate at which an observed score is deemed significant.