Background: The major aim of this study is to compare the effects of emotion focused (experiential) and cognitive interventions of schema therapy (ST) on emotion regulation deficits in patients with borderline personality disorder (BPD) according to DSM-V (alternative model) criteria. In a randomized, single-blinded parallel-group design clinical effects as well as effects on neurotransmitter metabolism and connectivity will be compared.

Method: While the 9-weeks treatment protocol of particular interest includes emotion focused interventions (ST-EF, n=60) such as chair dialogs, imagery rescripting or role play, the active control condition (ST-AC, n=60) is restricted to cognitive interventions, e.g. psychoeducation or pro/contra discussions. MEGA-PRESS 1H-MR spectroscopy and resting-state functional MR imaging (rs-fMRI) will be used before/after treatment protocols (T0-T1) and 6 months after the end of therapy (T2) to assess the effects on glutamate (Glx) and GABA metabolism in key regions of the target networks (executive control network, ECN: dorsolateral prefrontal cortex, DLPFC; salience network, SN: anteromedial cingulate cortex, aMCC; default mode network, DMN: pregenual cingulate cortex, pgACC) and to investigate the corresponding altered connectivity in these networks. The biological aberrations at T0 as compared to healthy controls (n=60) and treatment effects (at T1 and T2, n≥40 in each condition) on these aberrations will be linked to clinical effects measured by an extensive test battery with particular interest on emotion regulation, and specified by the Reliable Change Index (RCI). For longitudinal data mixed model analysis will be performed.

The main questions are (1) whether the emotion regulation deficit and the pattern of BPD-specific symptomatology are associated with a specific pattern of Glx and GABA concentrations in the DLPFC, aMCC and pgACC and corresponding deviations of functional connectivity within the ECN, SN and DMN. Hypothesis: Depending on primary and secondary outcome measures at T0, altered RSFC in the DMN, SN and ECN and corresponding altered Glx or GABA concentrations are assumed. (2) whether both treatment conditions have different clinical effects on the ability to regulate emotions and whether the respective clinical effects are associated with the changes in neurobiological aberrations. Hypothesis: It is hypothesized that the ST-EF condition will improve emotion regulation skills more effectively than the control condition. Only in the ST-EF condition are higher response and remission rates expected in the primary and secondary outcome measures, as well as effects on the ECN, SN and DMN with corresponding changes in RSFC and Glx or GABA concentrations.

2.3 Work program and proposed research methods

The work program is divided into the following three work packages:

1. Recruitment of patients and healthy subjects including diagnostic and psychological assessments as well as assessment of clinical outcome.
2. Collecting the MR-related (Glx, GABA, RSFC).
3. Data evaluation and analysis of the results.

Recruitment of patients and healthy subjects Patients will be recruited from routine admissions to the psychiatric day-hospital. According to our power calculation, we aim to include at least 120 patients with BPD according to the general and borderline specific criteria of the alternative DSM-5 Model for Personality Disorders (Section III of DSM 5, p 761ff (APA, 2013), which includes a specific dimensional pattern of impairments in personality functioning and pathological personality traits for diagnosis.

A total of at least 60 healthy volunteers will be recruited by newspaper advertisement. They will be examined using a semi-structured interview and self-ratings (SCL-90-R) to assess current mental status, to exclude personal or first-degree family history of psychiatric disorders (including substance abuse) or history of neurological or major medical conditions that might affect brain function.

Description of therapy conditions with/without emotion focused schema therapy interventions:

Deficits in emotion regulation are widely seen as a core pathology in BPD. Hence, quicker and more intense emotional activation can a priori be expected in every BPD patient. This study makes use of the fact that some interventions in ST evoke intense emotions (child modes), e.g. to foster awareness of child's needs, while others aim to establish a "healthy distance" from overintense emotions to enable adequate parental care (healthy adult mode). With regard to emotions schema therapy techniques have been divided into therapy relationship, experiential, cognitive, and behavioral techniques (Fassbinder et al., 2016). Schem therapy assumes that by using these strategies, the patient's fear of emotions reduces while willingness to overcome experiential avoidance increases (Fassbinder et al., 2016). This approach also clarifies "old" (biographically linked) goals of emotion regulation behavior (now counting as dysfunctional or a strategies of emotional avoidance), now replacing them by "new" (actual) goals that require and enhance new (functional) emotion regulation behavior.

MR-based investigation protocol

Scheduled acquisitions (scan time including patient positioning up to 1.5h):

• High-resolution, T1-weighted 3D whole brain scan using the MP-RAGE sequence TR/TE/TI = 2530/2.7/1100 ms; α = 7°; 176 sagittal slices of 1 mm thickness; FOVAP×HF = 256 × 256 mm²; in-plane matrix: 256 × 256 pixels; acquisition time: 6 min.
• Resting state fMRI scan of the entire brain using T2*-weighted EPI; TR/TE = 2520/30 ms; whole brain coverage with 45 transverse slices (thickness 2.5 mm); FoVAP×LR = 220×210 mm²; in-plane matrix: 88×84 pixels; 240 volumes; TA = 10 min.
• 1H-MR spectra with MEGA-PRESS sequence in three voxels placed in the aMCC (SN network) and left and right DLPFC (ECN network).

Manual adjustment of field homogeneity and water suppression (5 min); TR/TE = 2000/68 ms; 256 alternating single acquisitions with subsequent application of frequency-selective RF pulses at 1.9 ppm and 7.5 ppm, respectively; additional series acquisition of twelve water non-suppressed single scans with varying echo times (TR = 20 s, TE's = 30, 40, 50, 60, 75, 90, 110, 150, 175, 200, 250, 300 ms), TA = 17 min per voxel; total MRS duration of 50 - 55 min.

Photoplethysmogram (PPG) and respiration recordings during MRI acquisition with 500 Hz using an MR-compatible polygraph MP150 (BIOPAC Systems Inc., Goleta, CA, USA). Respiratory activity will be monitored by a strain gauge transducer incorporated in a belt tied around the chest, approximately at the level of the xiphoid process. The PPG sensor will be attached to the proximal phalanx of the left index finger.

Analysis of baseline (T0) MRI-data:

Separate analysis for rs-fMRI and 1H-MRS data:

A seed-based correlation analysis will be used to explore the whole-brain RSFC pattern of the regions being central to the study, i.e., the aMCC and DLPFC. Analysis of variance (ANOVA) will be applied to test for RSFC differences between patient groups and controls.

The same analyses will be applied to analyze group differences of the local Glx and GABA concentrations as well as the Glx/GABA balance in aMCC and DLPFC between patients and controls.

Associations between rs-fMRI and 1H-MRS data:

A linear mixed model analysis will be performed to investigate differences of associations between RSFC (SN and ECN) and Glx or GABA concentrations at the aMCC and DLPFC between patients and controls.

Follow-up (T0-T1;T1-T2) analysis of treatment effects on MRI target parameters (RSFC, Glx, GABA)

Separated analysis for rs-fMRI and 1H-MRS data:

Analysis of variance (ANOVA) for repeated measures will be applied to test for effects of group belonging (ST-EF, ST-AC, and healthy subjects; ST-EF vs. ST-AC) on changes of RSFC, Glx and GABA over time, as well as for within-subject changes between T0 and T1, and T1 and T2 scans.

Associations between RSFC and the local Glx and GABA concentrations:

A linear mixed model analysis will be performed to investigate changes after 9 weeks of treatment (T0 vs. T1) or 6month of catamnesis (T1 vs. T2) of either therapeutic condition (ST-EF, ST-AC) compared to unspecific changes over time in healthy controls as well as between treatment groups.

In a first step, this analysis will be performed separately in both treatment groups and healthy controls and for both neurotransmitters (Glx, GABA) and both spectroscopic and functional connectivity target regions (aMCC, DLPFC).

In the second (more important) step, potential treatment condition driven differences in the correlations between metabolic and RSFC parameters will be investigated by applying the same model including the factor treatment group.

Clinical outcome

Evaluation of changes of core symptoms and their clinical significance:

In a follow-up (T0-T1; T1-T2) analysis of treatment effects, the individual change of each patient in measures of the primary (Emotion Regulation Inventory, ERI) and secondary (Borderline Personality Disorder Severity Index - fourth version, BPDSI-IV; Levels of Personality Functioning Scale, SEFP; Personality Inventory for DSM-5 (Criterion B), PID-5; Borderline Symptom List (95 Items) BSL-95; Inventory of Interpersonal Problems, IIP-D; Young Schema Questionnaire (short version), YSQ-S3r, and Schema Mode Inventory (118 item version), SMI) outcome criteria will be estimated by the Reliable Change Index (RCI) (Wise, 2004) in each therapeutic condition.

Evaluation of changes in primary and secondary criteria and their clinical significance: In each therapeutic condition the improvement of each patient in measures of the primary (ERI) and secondary (BPDSI-IV, SEFP, PID-5, BSL-95, IIP-D, YSQ-S3r, and SMI) outcome criteria will be estimated by the Reliable Change Index (RCI), which summarizes changes at the level of an individual in the context of observed changes for the whole sample.

The RCI provides information about the functioning of an individual before/after therapy on an outcome measure, including the normative information about this measure by dividing the difference between pre- and post-treatment scores by the standard error (that includes also the reliability coefficient, not only the standard deviation of the measure). The change will be considered as being unlikely the product of measurement error (i.e., considered reliable), if the RCI is greater than 1.96. When the individual has a change score greater than 1.96 it can be assumed that the individual has improved. More detailed information on RCI scores and their meaning in each outcome measurement is given in the uploaded material.

Interrelation between clinical outcome parameters and neurobiological findings:

A linear mixed model analysis will be used that includes those MRI-data (RSFC, Glx, GABA) that have been significantly affected by therapy (T0 vs. T1) and those clinical scores that reached the highest RCI. The detected significant interactions will be compared between the ST-EF and ST-AC treatment condition.

Inclusion Criteria:

• Patients are eligible if they

  • fulfil diagnostic criteria (1.-4. see below) of BPD according to the Alternative DSM-5 Model (Sec-tion III of DSM 5, p 761ff (APA, 2013),
  • are between 18 and 30 years of age,
  • are female,
  • are right-handed,
  • are free of any treatment with antidepressants, antipsychotic or mood stabilizing medication (sporadic use of tranquillizers to induce sleep will be permitted)
  • have never been in psychotherapy before (in particular DBT, ST, CBT, TBP),
  • are willing to participate in (group) therapy,
  • are willing to participate in the study

Exclusion Criteria:

• Patients will be excluded if they

  • acutely suffer suicidal ideas,
  • suffer manifest affective or anxiety disorder ac-cording to DSM-5 criteria or have a lifetime his-tory of or current psychotic disorder (except brief psychotic disorder according to the DSM-5),
  • present a clinically manifest trauma-related dis-order (in particular PTSD),
  • suffer dissociative identity disorder, substance dependence needing clinical detoxification, ano-rexia nervosa with BMI < 14 or a serious and/or unstable medical illness,
  • have an IQ below 80,
  • are at any risk in terms of MRI investigations

Parallel-group: Both conditions (emotion focussed, ST-EF and active control, ST-AC) are highly standardized and active in terms of their expected clinical effect, which was a precondition for a positive ethics approval.

Randomized: Patients are randomly assigned to the ST-EF or ST-AC condition according to a randomization list established by an independent statistician, and applied to incoming help-seeking patients by an independent admission management (managing nurse, not a member of the research team).

Single-blinded: Researchers are blind as regards group belonging of participants.