Trauma and Outcomes of Mentalization-Based Therapy for Individuals With Borderline Personality Disorder

The impact of complex trauma on attachment and mentalizing has been demonstrated. In normative development, the attachment system is activated in response to threat (24, 25), leading the individual to seek proximity to responsive attachment figures, which leads in turn to downregulation of distress and discomfort. At the neurobiological level, this response is mediated by a mesocorticolimbic dopaminergic reward system that plays an essential role in downregulating the stress system and is responsible for feeling supported, validated, and understood by close others (26, 27). Trauma typically disrupts the virtuous cycle associated with proximity seeking in the face of stress and adversity, as the individual begins to rely excessively either on attachment hyperactivating or deactivating strategies, or a combination of both. These strategies are considered to be adaptations to an environment characterized by inconsistency in, unresponsiveness of, and/or abuse by attachment figures (28–30). From this perspective, attachment deactivating strategies develop as a means to deal with the (perceived) unresponsiveness of attachment figures. Attachment hyperactivating strategies, in turn, tend to develop as an attempt to elicit care and support from inconsistently responsive attachment figures. Disorganized attachment, characterized by the oscillation between attachment hyperactivating and deactivating strategies, is thought to primarily develop in individuals with a history of complex trauma. These individuals typically find themselves in a catch-22 situation because the same caregivers who are expected to be supportive and comforting, thereby downregulating distress, are also the source of conflict, abuse, and/or neglect (31). A continuous approach-avoidance conflict in relationships emerges, which may help to explain the marked oscillation typical of these individuals’ attachment relationships, leading to severe impairments in the capacity to seek help from others, including mental health professionals (32, 33). Consistent with these assumptions, studies have reported high rates of both preoccupied (reflecting attachment hyperactivating strategies) and disorganized (reflecting a combination of attachment hyperactivating and deactivating strategies) attachment among individuals with borderline personality disorder (34, 35).

The pernicious impact of trauma on mentalizing has also been amply demonstrated (4). Growing up in an environment characterized by abuse and/or neglect is hypothesized to undermine the capacity for mentalizing, particularly in the case of attachment trauma, because the individual is deprived of a responsive caregiver, which is essential to acquiring the capacity to regulate stress and arousal. Feelings of isolation and loneliness typically characterize individuals with borderline personality disorder and a history of trauma, as does their tendency to act out these unmentalized experiences of the self and others as being bad, evil, neglected, neglectful, or unworthy, resulting in reenactment that leads to revictimization (36, 37). Theoretically, revictimization is linked to the familiar yet highly maladaptive pattern where the abuser acts as both the source of anxiety, anger, and conflict, and the source of care, love, and support (33), thereby repeating the unsolvable approach-avoidance conflict typical of the disorganized or disoriented attachment pattern characteristic of many individuals with borderline personality disorder. Moreover, many individuals with a history of trauma perpetrate abuse on their children, partner, and/or others in their environment (e.g., friends, coworkers, employees) because specific features in others (e.g., talent or submissiveness) may trigger their own past experiences of abuse and/or neglect. When the individual is faced with the inability to reflect on these unmentalized experiences, nonmentalizing modes of experiencing both the self and others tend to reemerge, further impairing opportunities to recalibrate one’s mind.

Evidence, although inconclusive, is also emerging that trauma, and particularly attachment trauma, may undermine epistemic trust (i.e., the willingness and openness to consider new knowledge acquired by means of social communications as trustworthy, potentially personally relevant, and of generalizable significance to integrate into one’s life) (38). As a result, the traumatized individual may become completely cut off from social learning and salutogenesis (1, 39). Epistemic vigilance among traumatized individuals may also be accompanied by initial excessive epistemic credulity, often driven by strong wishes to be able to rely on a trusted other, rendering such individuals vulnerable to exploitation and abuse, which then further increases their epistemic vigilance (38).

This study was approved by the Medical Ethical Committee of Erasmus Medical Center, Rotterdam, the Netherlands (NL38571.078.12). Written informed consent was obtained, and the study was registered at the Netherlands Trial Register (NTR2292). Inclusion and exclusion criteria, patient characteristics, and randomization procedures (including study enrollment and allocation) have been described previously (48). Of the 114 randomly assigned patients from the original trial, only 83 patients (MBT-IOP, N=34; MBT-DH, N=49) had data available on childhood trauma, because the trauma assessment was included after inclusion had already started. Patients were assessed at baseline, prior to random assignment, and were then assessed on a range of outcome measures every 6 months, from start of treatment until 36 months afterward.

The primary outcome measure was symptom severity as assessed by the Global Severity Index of the Brief Symptom Inventory (50, 51). Secondary outcomes were features of borderline personality disorder, as measured by the Personality Assessment Inventory–Borderline Features Scale (52), and interpersonal functioning, as assessed by the Inventory of Interpersonal Problems (53, 54).

The prevalence of trauma in childhood was measured with the Dutch version of the Childhood Trauma Questionnaire (CTQ) (55). The CTQ is a retrospective self-report questionnaire that measures five categories of childhood trauma, including emotional, physical, and sexual abuse and emotional and physical neglect. Each subscale is measured with five items rated on a 5-point Likert scale: never true, rarely true, sometimes true, often true, and very often true. Each subscale score ranges from 5, no history of abuse or neglect, to 25, extreme history of abuse and neglect. In this study, participants were classified as having a substantial history of childhood trauma in any of five specific categories by using the following cutoff scores from the CTQ manual: 13 or higher for emotional abuse, 10 or higher for physical abuse, 8 or higher for sexual abuse, 15 or higher for emotional neglect, and 10 or higher for physical neglect. Research has shown good psychometric properties for both the original CTQ (55, 56) and the Dutch translation (57).

A detailed description of MBT-DH and MBT-IOP has been provided elsewhere (48). Briefly, MBT-DH consisted of a 5-day per week hospitalization program, and MBT-IOP consisted of a 2-day per week outpatient treatment program. Both MBT-DH and MBT-IOP included weekly individual sessions, but the intensity of group therapy differed markedly between the programs. Treatment adherence to the MBT model during the intensive treatment phase was rated as adequate by three independent raters and did not differ between the treatment groups.

Analyses were performed on all available data by using SPSS Statistics, version 25.0, for Windows. The prevalence of trauma was examined by using both dimensional (means) and categorical scores on the CTQ at baseline, with two-tailed independent sample t tests and chi-square tests, as appropriate.

For predictor and moderator analyses, multilevel random-effect modeling was used to best accommodate the missing data that are inevitable in longitudinal follow-up and to deal with the dependency of repeated measures within subjects over time. Time points were coded –6, –5, –4, –3, –2, –1, and 0, implying that regression coefficients involving time measured the rate of change from baseline to the 36-month follow-up and regression intercepts referenced group differences at the last time point. On the basis of previous analyses, random slopes of change were assumed. Models with quadratic time functions or quadratic (interaction) terms did not show a better fit on the basis of Akaike’s information criterion (58) or Schwarz’s Bayesian information criterion (59). For parsimony, we therefore report linear models. Treatment groups were coded 0=MBT-IOP and 1=MBT-DH. Consequently, differences in slope refer to differences in the slope of MBT-DH compared with MBT-IOP. For the mixed models, the main effect of trauma is reported, along with the interaction effects with treatment group and the rate of change from baseline to 36 months across all levels of trauma (for both treatment groups combined). The critical coefficients for each trauma predictor were, first, the two-way trauma (present or absent) predictor × time interaction, indicating the predictive value of the level of trauma on the rate of change, and then the three-way predictor × time × group interaction, indicating the moderating role of trauma on the rate of change. To investigate the robustness of the models, we also performed mixed model analyses for the two-way interaction (predictor × time) separately. Because these analyses yielded similar results, only estimates from the three-way interaction model are reported. Results of the two-way interaction models are available by request from the first author (M.L.S.).

There were no significant pretreatment differences between patients randomly assigned to MBT-IOP or to MBT-DH in terms of demographic or clinical characteristics or baseline prevalence of outcome measures (Table 1). There were also no differences in the prevalence of trauma types and overall trauma between patients in the two groups. In total, 86% of patients had scores above the cutoff point for the presence of substantial trauma in at least one of the trauma categories (Table 1). Emotional neglect was most common (66%), followed by emotional abuse (57%) and sexual abuse (40%). Physical neglect (37%) and physical abuse (22%) were somewhat less prevalent, but rates were substantially higher than among community samples. Approximately 60% of patients reported experiencing two or more types of trauma; less than 15% did not meet criteria for substantial trauma.

Results of the multilevel analyses are summarized in Table 2. Multilevel estimates and model parameters from the three-way interaction models of all trauma predictors for each outcome measure are available in the online supplement.

For symptom severity, in contrast to our predictions, patients with and without trauma showed similar rates of change in MBT-DH, and patients with a history of emotional or physical neglect tended to show a faster rate of change in MBT-DH compared with those without such a history (Figure 1). Yet, patients scoring above the cutoff score (indicating substantial emotional neglect, emotional abuse, sexual abuse, or physical neglect) showed a slower rate of change in MBT-IOP compared with patients without trauma. For physical abuse, a similar trend was observed, but this trend did not reach significance (p=0.062). Thus, contrary to our expectations, trauma had a differential impact on treatment outcome in MBT-IOP and MBT-DH in terms of improvement in general distress.

Trauma was less predictive of treatment outcome in terms of improvement in borderline symptoms and interpersonal problems. For borderline symptoms, only emotional neglect showed a significant moderating effect (p=0.028). As Figure 2 shows, patients in MBT-IOP with substantial emotional neglect showed a slower rate of improvement in borderline personality disorder features compared with those who had not experienced substantial levels of emotional neglect, whereas in MBT-DH, patients with a history of emotional neglect showed a slightly greater rate of improvement compared with those without such a history. None of the other trauma types predicted changes in borderline symptoms in MBT-DH or MBT-IOP.

A trend was observed for history of physical abuse to moderate changes in interpersonal problems in MBT-IOP versus MBT-DH. Patients with substantial previous physical abuse tended to show a slower rate of change in MBT-IOP compared with those without such a history, whereas in MBT-DH, there was a trend for patients with high levels of physical abuse to show greater improvement in interpersonal problems compared with patients without a history of physical abuse (Figure 3). None of the other types of trauma predicted changes in interpersonal problems.