The Effectiveness of Dialectical Behavior Therapy Compared to Schema Therapy for Borderline Personality Disorder: A Randomized Clinical Trial

Psychotherapy and Psychosomatics Standard Research Article Psychother Psychosom 2024;93:249–263 DOI: 10.1159/000538404 Received: July 27, 2023 Accepted: March 9, 2024 Published online: July 10, 2024 The Effectiveness of Dialectical Behavior Therapy Compared to Schema Therapy for Borderline Personality Disorder: A Randomized Clinical Trial Nele Assmann a, b Anja Schaich a, b Arnoud Arntz c Till Wagner a Philipp Herzog a, d, e Daniel Alvarez-Fischer a, f Valerija Sipos a Kamila Jauch-Chara a, b Jan Philipp Klein a Michael Hüppe g Ulrich Schweiger a Eva Fassbinder a, b a Department of Psychiatry, Psychosomatics and Psychotherapy, University of Lübeck, Lübeck, Germany; b Department of Psychiatry and Psychotherapy, Christian-Albrechts-Universität zu Kiel, Kiel, Germany; c Department of Clinical Psychology, University of Amsterdam, Amsterdam, The Netherlands; d Department of Psychology, Harvard University, Cambridge, MA, USA; e Department of Psychology, University of Kaiserslautern–Landau, Landau, Germany; f Institute of Neurogenetics, University of Lübeck, Lübeck, Germany; g Department of Anaesthesiology, University of Lübeck, Lübeck, Germany Keywords Borderline personality disorder ·Randomized clinical trials · Effectiveness ·Psychotherapy research ·Treatment outcome Abstract Introduction: In the treatment of borderline personality disorder (BPD), there is empirical support for both dialectical behavior therapy (DBT) and schema therapy (ST); these treatments have never been compared directly. This study examines whether either of them is more effective than the other in treating patients with BPD. Methods: In this randomized, parallel-group, rater-blind clinical trial, outpatients aged between 18 and 65 years with a primary diagnosis of BPD were recruited in a tertiary outpatient treatment center (Lübeck, Germany). Participants were randomized to DBT or ST with one individual and one group session per week over 1.5 years. The primary outcome was the BPD symptom severity assessed with the mean score of the Borderline Personality Disorder Severity Index at 1-year naturalistic follow-up. Results: Between November 26, 2014, and December 14, 2018, we enrolled 164 patients (mean age = 33.7 [SD = 10.61] years). Of these, 81 (49.4%) were treated with ST and 83 (50.6%) with DBT, overall, 130 (79.3%) were female. Intention-to-treat analysis with generalized linear mixed models did not show a significant difference at 1-year naturalistic follow-up between DBT and ST for the BPDSI total score (mean difference 3.32 [95% CI: –0.58–7.22], p= 0.094, d=−24 [–0.69; 0.20]) with lower scores for DBT. Pre-tofollow-up effect sizes were large in both groups (DBT: d= 2.45 [1.88–3.02], ST: d= 1.78 [1.26–2.29]). Conclusion: Patients in both treatment groups showed substantial improvements indicating that even severely affected patients with BPD and various comorbid disorders can be treated successfully with DBT and ST. An additional non-inferiority trial is needed to show if both treatments are equally Nele Assmann and Anja Schaich share first authorship. [email protected] www.karger.com/pps © 2024 The Author(s). Published by S. Karger AG, Basel Correspondence to: Nele Assmann, nele.assmann @ uksh.de This article is licensed under the Creative Commons AttributionNonCommercial4.0InternationalLicense (CCBY-NC)(http://www. karger.com/Services/OpenAccessLicense).Usage and distributionfor commercial purposes requires written permission. Erratum August 30, 2024. DOI: 10.1159/000540975 Downloaded from http://karger.com/pps/article-pdf/93/4/249/4272968/000538404.pdf by Universitätsbibliothek Kiel user on 09 September 2024 effective. The trial was retrospectively registered on the German Clinical Trials Register, DRKS00011534 without protocol changes. © 2024 The Author(s). Published by S. Karger AG, Basel Introduction Several psychotherapy methods are available for treating borderline personality disorder (BPD) with good empirical evidence [1–3], and international guidelines highly recommend disorder-specificpsychotherapy for patients suffering from BPD [4–6]. Two of the most established treatments include dialectical behavior therapy (DBT) [7, 8] and schema therapy (ST) [9, 10]. DBT is the most frequently studied treatment for BPD, and its effectiveness has been demonstrated in several randomized controlled trials (RCTs) [2, 3]. Even though ST is less often studied so far, there are promising results from two large [11, 12] and one small randomized trial [13] as well as studies in routine care [14–16]. DBT and ST have never been compared in a randomized trial so far. Overall, there is a need for comparative clinical trials. Comparing the effectiveness of BPD treatments may improve patients’informed choices and guide decision-making about which treatments to prioritize in health care systems [17]. The programs for BPD (PRO*BPD) study is the first randomized trial to compare the effectiveness of ST and DBT in patients with BPD [18]. It was implemented into a routine tertiary care setting of an outpatient clinic affiliated with inpatient services. This outpatient service was explicitly set up to treat severely affected patients with BPD who are repeatedly seen in emergency rooms or would otherwise be treated in inpatient settings. Research in this population is fundamental considering that BPD is associated with high societal costs [19–22], primarily caused by inpatient treatment. The primary hypothesis was that BPD associated symptoms differ between DBT and ST at the 1-year naturalistic follow-up. Since information from literature does not allow predictions on the directionality of the difference, a two-sided hypothesis was chosen. Secondary outcomes included treatment retention, general symptom severity, depression, dissociation, quality of life, psychosocial functioning, and participation. In addition, based on former research and theoretical considerations regarding DBT and ST, secondary hypotheses were formulated regarding differential effects in the secondary outcomes: We hypothesized that DBT reduces suicidality, self-harm, and dissociation better and faster than ST. At the same time, ST would be better in improving the quality of life and reducing general symptom severity and depressive symptomatology (see study protocol for a detailed explanation [18]). Materials and Methods Study Design The PRO*BPD study design is a single-center randomized clinical trial. The recruitment of participants took place in a tertiary outpatient treatment center of a university clinic affiliated with inpatient treatment in Lübeck, Germany, mainly treating severely affected patients who cannot receive treatment elsewhere or cannot sufficiently be treated by practitioners in private practices. The PRO*BPD study was designed in adherence to the Consolidated Standards of Reporting Trials (CONSORT) guidelines and methodology [23] and follows methodological recommendations for trials of psychological interventions [24] (summarized in online suppl. Table S1; for all online suppl. material, see https://doi.org/10.1159/000538404). The planned sample size was N= 160 based on a power analysis (power 80%; medium effect size, α= 0.05, n= 128) and taking attrition (20%, n= 32) into account [25]. Details about the study design are published in the study protocol [18]. Participants Patients were eligible when they (1) were between 18 and 65 years of age, (2) had a primary diagnosis of BPD (diagnosed with the Structural Clinical Interview for DSM-IV for Axis II; SCID-II-Interview) [26], (3) had a BPD severity score >20 points on the Borderline Personality Disorder Severity Index (BPDSI), Version 4 [27], (4) gave informed consent, and (5) were able and willing to participate reliably in therapy and assessment procedures. Exclusion criteria were (1) a lifetime diagnosis of a psychotic disorder, (2) intellectual deficits (IQ <85), (3) poor German language skills, and (4) acute substance use disorder that required detoxification treatment. Participation was possible after completing detoxification treatment and 4 weeks of abstinence. Participants with cannabis dependence could participate if they committed to working on abstinence during treatment. Randomization and Masking Randomization was conducted by Prof. Michael Hüppe, who was not involved in the data collection and therapy using the program BiaS (11.02) [28]. After completing the baseline assessment, participants were randomized to the two conditions (ST or DBT). Randomization was stratified by sex to avoid biases due to unbalanced sex distribution. The allocation sequence was concealed from both participants and researchers. Raters were blind to the assignment; participants were informed about their condition during the first therapy session. Procedure Participants completed assessments at baseline, 0.5 years after the start of treatment, one year after the start of treatment, posttreatment (1.5 years), and two naturalistic follow-up 250 Psychother Psychosom 2024;93:249–263 DOI: 10.1159/000538404 Assmann et al. Downloaded from http://karger.com/pps/article-pdf/93/4/249/4272968/000538404.pdf by Universitätsbibliothek Kiel user on 09 September 2024 assessments (0.5 years and one year). In the following, the 1-year follow-up assessment will be referred to as follow-up (primary outcome). Assessments included self-report measures and interviews (conducted by blind and independent raters trained in the interviews). Due to a wide range of measures, the baseline assessment consisted of three to four appointments with the patients and was scheduled within 3 months before the start of treatment. In case, there were more than 3 months between assessment of the primary outcome and start of treatment (due to availability of treatment slots and missed appointments), the interview of the primary outcome was repeated. Outcomes The primary outcome was the severity of BPD assessed by the total score of the BPDSI-IV at 1-year naturalistic follow-up. The BPDSI is a semi-structured interview including 70 items rating the frequency and severity of the nine BPD traits described in the DSM-IV over the prior 3 months. The BPDSI is a reliable instrument with good psychometric qualities [29, 30]. In the present study, the assessed interrater reliability was excellent (all intra-class correlation coefficients >0.997 [ICC], online suppl. material). In addition, suicidality and the number of suicide attempts were identified using BPDSI data. An increase of 11.70 or more in the BPDSI total score compared to the baseline assessment was defined as deterioration [11]. Secondary outcome measures included a self-rating questionnaire of the burden experienced by BPD manifestations (BPD checklist) [31], the severity of depression (Quick Inventory of Depressive Symptoms, QIDS-SR) [32], global symptom severity (Brief Symptom Inventory, BSI) [33], dissociation (Dissociation Tension Scale, DSS) [34], psychosocial functioning and participation (World Health Organization Disability Assessment Schedule 2.0, WHODAS 2.0 [35] and Work and Social Adjustment Scale, WSAS [36]) and quality of life (World Health Organization Quality of Life questionnaire, WHOQOL) [37]. Comorbid mental disorders were assessed using the German version of the SCID-I and IIInterview [38, 39]. For further information on secondary outcomes, see online supplementary material. Psychopharmacological treatment and demographic parameters were also assessed. Treatment retention was also analyzed as a secondary outcome. To register also unwanted iatrogenic effects of psychotherapy, we recorded adverse events (inpatient treatment, suicide attempts, cases of death) and analyzed deteriorationintheBPDSI. Treatment Patients were randomized to DBT or ST; both included one individual session (60 min) and one group session (120 min) per week and followed a written protocol [7, 8, 10, 40]. Groups consisted of up to 10 BPD patients and two therapists and were offered in a semi-open format. The start of the individual therapy was planned four to 10 weeks before the beginning of group therapy. Eligible patients who could not attend group sessions due to family, professional, or educational duties were provided weekly individual therapy only. The treatment duration was 1.5 years. Patients couldstopthetreatmentprogram prematurely and were considered as an “early success”if their BPDSI score was less than 15 and the therapist team agreed that the patient should stop the program due to remission. Patients could also be “pushed out”of the treatment program based on a decision of the therapist team and the local supervisor [18]. We decided not to define explicit rules for a patient to be “pushed out”of both conditions. It should be noted that this is a difference from the standard DBT protocol [22]. In the PRO*BPD trial, a “push out”wasalwaysdecided by the therapist team and the local supervisors and was consistent with the respective treatment manual except for the deviation mentioned above. After the end of study treatment, no further treatment was recommended. As it was a naturalistic follow-up, further treatment was not prohibited but registered at follow-up assessments. If needed, patients had the opportunity to see their former therapist monthly or less frequently. We had some deviations from our original protocol due to the COVID-19 pandemic and therefore conducted a sensitivity analysis excluding all participants affected by the pandemic (online suppl. material). Therapists Therapists were advanced DBT and ST therapists and therapists who were new to ST or DBT but were experienced in CBT, with training before administering the treatment and learning the method under close supervision. Local and external certified specialists trained therapists for the specific method in several workshops. They participated in weekly supervision sessions under the direction of the locally approved supervisors (DBT: VS, US, ST: EF) and team meetings. Treatment Integrity All individual and group sessions were videotaped, and adherence and competence were rated in a random selection of session tapes from different treatment stages (each treatment stage 6 months) by trained raters. The videos of the sessions were rated altogether after the end of the study treatment. Individual sessions were rated using an adapted version (May 2014) of the Therapy Adherence and Competence Scale for therapy of BPD by Young et al. [41] which was also used in Giesen-Bloo et al. [11] and the Dialectical Behavior Therapy Adherence Checklist –Individual Therapy (DBT AC-I) observer-rated version [42]. The latter was chosen because ratings with the originally planned DBT adherence manual [43] were only possible by DBT-specialists certified for this and were not covered by the funding of our study. Adherence to ST group sessions was rated using a scale combining the Schema Therapist Competency Scale for individual therapy sessions (STCS-I-1) [44] and the Group Schema Therapy Rating Scale (GSTRS) and its revised version (GSTRS-R) [45]. Each item of the scales was code 0/1 for adherence and 1–6 for competence. For the DBT group sessions, there was no freely available rating scale. Ratings of DBT group sessions would have been available from certified raters who did not participate in this study. Unfortunately, funding was not available for these ratings. Ten videos of DBT group sessions were rated using the ST scale to ensure discrimination. Details on the rater training can be found in the online supplementary material. Statistical Analysis Originally, the hypothesis that there is a difference between patients treated with DBT and ST regarding the reduction of BPD symptoms was operationalized as a difference between the slopes over time (baseline to 1-year naturalistic follow-up) [18]. DBT versus Schema Therapy in Borderline Personality Disorder Psychother Psychosom 2024;93:249–263 DOI: 10.1159/000538404 251 Downloaded from http://karger.com/pps/article-pdf/93/4/249/4272968/000538404.pdf by Universitätsbibliothek Kiel user on 09 September 2024 Due to unexpected baseline differences between DBT and ST (especially self-rated BPD symptoms and depression, see online suppl. Table S2), we had to adapt our analysis plan and had to control for the baseline scores (see below and online suppl. material). Moreover, controlling for baseline scores is recommended independently from baseline differences [46]. As a consequence, we had to change the operationalization of the primary hypothesis and tested the difference at 1-year naturalistic follow-up. The statistical analyses were performed with SPSS, version 28. Data analyses followed the intention-to-treat (ITT) principle and used all available data. Patients were included in the ITT sample if they were randomized and attended at least one individual session where they were informed about treatment condition. Missing values were not imputed as generalized linear mixed models (GLMMs) can validly estimate effects under the same assumptions as multiple imputation. A step-by-step report of the statistical analyses can be found in the online supplementary material. Given skewed distributions, continuous outcome variables were analyzed with GLMM gamma regression. As gamma regression cannot handle zeros, a small value (0.01) was added to all scores, if zeros existed. Where estimation allowed, we included a random effect of the treatment cohort to account for the fact that cohorts of patients participated in group therapy together. Time was modeled according to the best model fit, e.g., linear, logarithmic, or segmented development of scores over time. We used a piece-wise regression model if the primary inspection of the data indicated that change during treatment had a different slope than change during the naturalistic follow-up period, and a comparison of model fit confirmed this impression. For piece-wise coded regression, we defined all assessments of the treatment phase as one section and the follow-up assessments separately. As predefined in the study protocol, we included the use of psychotropic medication as a running covariate coding the use for every assessment point. Since this can be discussed controversially, we conducted a sensitivity analyses without the medication covariate. The covariance structure for the repeated part and time was modeled according to the best model fit. Additional covariates were included as we observed systematic baseline differences between ST and DBT in almost all outcome variables with higher scores for DBT patients (online suppl. Table S2). We included a centered baseline global severity score based on principal component analysis, including all outcome variables at baseline and the number of comorbidities on axis I and II as well as the centered baseline variable as a covariate and excluded the baseline assessment in the dependent time variable for all outcomes [46]. Therefore, the main effect of treatment after 1-year naturalistic follow-up was the primary test of treatment differences. Thus, the fixed effects in the GLMM included time, medication, treatment, and their interaction, as well as global severity, and the particular baseline score was also included in the fixed part. A further fixed effect of time piece-wise and its interaction with treatment and medication were included if a piecewise regression model was used. The exact configuration of the GLMM for each outcome variable is specified in online supplementary Table S3 and a definition of all variables included in the model can be found in online suppl. Table S4. For a detailed stepby-step description of the statistical analyses, refer to online supplementary material. The number of suicide attempts during the last 3 months was analyzed using generalized estimating equations with a Tweedie distribution (because of many zero counts) with a log link and first-order autoregression for the repeated part. Treatment retention was analyzed using GLMM survival analysis. Effect sizes were expressed as r for the fixed effects in the GLMM analyses and Cohen’sdas conventional withinand between-groups effect size. For a description of effect sizes, see online supplementary material. Additionally, we conducted a completer analysis of the primary outcome, including all patients that completed treatment after 1.5 years or left the program earlier after remission. Further, we conducted a sensitivity analysis including only those patients who received the combination of group and individual therapy. We also conducted a sensitivity analysis following the original analytic plan including all assessment points and not controlling for baseline differences. Lastly, we conducted a sensitivity analysis excluding the medication covariate. Results Patient Flow and Sample Characteristics Between November 26, 2014, and December 14, 2018, 255 patients were screened for eligibility, and 164 were included in the PRO*BPD trial’s ITT sample of the PRO*BPD trial. The last follow-up assessment was conducted on January 20, 2022. The CONSORT flow diagram of participant recruitment is presented in Figure 1. Table 1 displays the demographic and clinical characteristics of the ITT sample. The mean number of sessions before the start of group therapy was M= 4.3 (DBT and ST), and the average number of individual sessions was M= 44.48 (DBT) and M= 46.7 (ST) for the intent-to-treat sample and M= 52.4 (DBT) and M= 52.1 (ST) for the completer sample. Treatment Integrity Overall, 500 videos out of 7,475 individual sessions and 43 videos out of 436 group sessions were rated by trained raters. The interrater reliability was assessed for 20 ratings of individual sessions (ten ST and ten DBT) and 19 ratings of group sessions (ten ST and nine DBT). Interrater reliability showed to be very good both for the ratings of the individual sessions (ICC between 0.88 and 0.961) and for the ST ratings of the group sessions (ICCs between 0.991 and 0.992). There was significant discrimination between ST and DBT at the ST, respectively, DBT adherence and competence scales (all p<0.009). ST group adherence and competence ratings also indicated a good differentiation between the conditions (all p<0.001). 252 Psychother Psychosom 2024;93:249–263 DOI: 10.1159/000538404 Assmann et al. Downloaded from http://karger.com/pps/article-pdf/93/4/249/4272968/000538404.pdf by Universitätsbibliothek Kiel user on 09 September 2024 Mean adherence to the respective manuals was 93.33% in ST and 99.18% in DBT individual therapy. Mean therapeutic competence was 5.27 for ST and 4.99 for DBT. Mean adherence in ST group therapy was 98.10%, and mean therapeutic competence was 5.41. Fig. 1. CONSORT diagram of participant flow. DBT versus Schema Therapy in Borderline Personality Disorder Psychother Psychosom 2024;93:249–263 DOI: 10.1159/000538404 253 Downloaded from http://karger.com/pps/article-pdf/93/4/249/4272968/000538404.pdf by Universitätsbibliothek Kiel user on 09 September 2024 Table 1. Descriptive statistics at baseline Treatment condition all (N= 164) DBT (n= 83) ST (n= 81) Characteristic Age, years, M(SD) 33.71 (10.61) 34.54 (11.15) 32.85 (10.01) Gender, n(%) Male 33 (20.1) 17 (20.5) 16 (19.8) Female 130 (79.3) 66 (79.5) 64 (79.0) Non-binary 1 (0.6) –1 (1.2) Relationship status, n(%) Partner 82 (50.0) 41 (49.4) 41 (50.6) No partner 82 (50.0) 42 (50.6) 40 (49.4) Education level, n(%) No/primary education 9 (5.5) 5 (6.0) 4 (4.9) Lower secondary education 92 (56.1) 46 (55.4) 46 (56.8) Upper secondary education 49 (29.9) 26 (31.3) 23 (28.4) Tertiary education 14 (8.5) 6 (7.2) 8 (9.9) Ethnic background, n(%) German 151 (92.1) 78 (94.0) 74 (91.4) Different 12 (7.3) 5 (6.0) 7 (8.6) Work status, n(%) Working 20 (12.2) 7 (8.4) 13 (16.0) Studying 17 (10.4) 7 (8.4) 10 (12.3) Homemaker 12 (7.3) 6 (7.2) 6 (7.4) Disability pension 41 (25.0) 22 (26.5) 19 (23.5) Unable to work due to sick leave 44 (26.8) 25 (30.1) 19 (23.5) Unemployed 15 (9.1) 8 (9.6) 7 (8.6) Retirement pension 3 (1.8) 3 (3.6) – Other 12 (7.3) 5 (6.0) 7 (8.6) BPD severity BPDSI at baseline, M(SD) 32.80 (8.71) 33.19 (9.38) 32.39 (8.01) Number of BPD criteria (SKID-II) 7.26 (1.26) 7.23 (1.32) 7.30 (1.21) Comorbid Disorders Number of comorbid SCID-I diagnoses, M(SD) 3.99 (1.95) 4.12 (1.88) 3.85 (2.03) Number of comorbid SCID-II diagnoses (excl. BPD), M(SD) 1.35 (1.12) 1.37 (1.06) 1.32 (1.18) Comorbid personality disorders, n(%) Paranoid personality disorder 37 (22.6) 17 (20.5) 20 (24.7) Schizoid personality disorder 1 (0.6) –1 (1.2) Schizotypal personality disorder 6 (3.7) 3 (3.6) 3 (3.7) Antisocial personality disorder 3 (1.8) 1 (1.2) 2 (2.5) Histrionic personality disorder 12 (7.3) 9 (10.8) 3 (3.7) Narcissistic personality disorder 13 (7.9) 4 (4.8) 9 (11.1) Avoidant personality disorder 69 (42.1) 39 (47.0) 30 (37.0) Dependent personality disorder 21 (12.8) 14 (16.9) 7 (8.6) Obsessive-compulsive personality disorder 59 (36.0) 27 (32.5) 32 (39.5) Comorbid disorders axis I, n(%) Any comorbid axis I diagnosis 161 (98.2) 82 (98.8) 79 (97.5) Affective disorders 108 (65.9) 59 (71.1) 49 (60.5) Anxiety disorders 144 (87.8) 74 (89.2) 70 (86.4) Somatic symptom disorder 21 (12.8) 11 (13.3) 10 (12.3) Substance use disorder 51 (31.1) 24 (28.9) 27 (33.3) Eating disorders 76 (46.3) 41 (49.4) 35 (43.2) 254 Psychother Psychosom 2024;93:249–263 DOI: 10.1159/000538404 Assmann et al. Downloaded from http://karger.com/pps/article-pdf/93/4/249/4272968/000538404.pdf by Universitätsbibliothek Kiel user on 09 September 2024 Main Outcome The GLMM showed a significant reduction of BPD severity measured by the BPDSI-IV total score over time (0.5 years to 1-year naturalistic follow-up) with large effect sizes (pre-post and pre-follow-up) for both conditions (Table 2; Fig. 2a). However, treatment had no significant effect at 1-year naturalistic follow-up (t 131 =1.69,p=0.94, d=−0.24 [−0.69; 0.20]). See online supplementary Table S5 for the effects of GLMM covariates on all outcomes. The completer and sensitivity analyses did not show different results (see online suppl. material). Secondary Outcomes BPDSI-Based Secondary Outcomes All BPDSI subscales except for (para)suicide showed asignificant reduction over time (0.5 years to 1-year naturalistic follow-up) (online suppl. Table S6). For this subscale, the within-group pre-follow-up effect sizes were also large in both conditions indicating a substantial improvement between baseline and 0.5 years which was not included in the GLMM. The within-group pre-follow-up effect sizes for all other subscales were large; only in the subscale anger the effect size in the ST condition was medium. This subscale was the only one with a significant treatment effect at 1-year naturalistic follow-up: patients in the DBT condition had significantly lower scores at this time point. The suicidality score showed no significant effects, and the within-group pre-post effect sizes were medium in both conditions (online suppl. Table S6). For the number of suicide attempts in the last 3 months, there were also no significant effects. The within-group pre-follow-up effect sizes were small to medium. Other Secondary Outcomes All GLMM of secondary outcomes showed a significant time effect (0.5 years to 1-year naturalistic followup), indicating improvement, but no effect of treatment at 1-year naturalistic follow-up (Table 2). The within-group pre-follow-up effect sizes were large for the BPD Checklist, BSI, QIDS-SR, WSAS, and WHODAS and small to large for the DSS (Table 2) and the WHOQOL subscales (online suppl. Table S7). Treatment Retention Figure 2b shows treatment retention during the 1.5 years treatment period. Deviation contrasts of the GLMM survival analysis showed that in the second quarter, significantly more patients dropped out from DBT, t 870 =−2.79, p= 0.005 (ST: 0% vs. DBT: 7.2%), whereas in the fifth quarter, significantly more patients dropped out from ST t 870 =−2.21, p= 0.028 (ST: 4.9% vs. DBT 0%). The overall dropout rate was 22.9% for DBT and 23.5% for ST also including few “push-outs”(3.6% DBT; 1.2% ST). Details on reasons for drop out can be found in online supplementary Table S8. Deterioration and Adverse Events Deterioration was found in <1% of all assessments (0.74% DBT, 0.76% ST). For an overview of adverse events during the study period, see online supplementary material. Effects of Psychotropic Medication The proportion of patients taking psychotropic medication is descripted in online supplementary Table S9 separately for each assessment point and different medication categories. The main effects and interactions involving psychotropic medication in the GLMMs are reported in online supplementary Table S5 and interpreted in the online supplementary results. Treatment during the Naturalistic Follow-Up Phase An overview of the treatments received during the naturalistic follow-up period can be found in online supplementary Table S8. Briefly, 3.0% (0.5 years) and Table 1 (continued) Treatment condition all (N= 164) DBT (n= 83) ST (n= 81) Psychiatric medication 1 ,n(%) 126 (76.8) 64 (77.1) 62 (76.5) Previous treatment, n(%) 154 (93.9) 82 (98.8) 72 (88.9) Previous psychotherapeutic treatment, n(%) 151 (92.1) 80 (96.4) 71 (87.7) 1 Including medication irregularly taken as needed. DBT versus Schema Therapy in Borderline Personality Disorder Psychother Psychosom 2024;93:249–263 DOI: 10.1159/000538404 255 Downloaded from http://karger.com/pps/article-pdf/93/4/249/4272968/000538404.pdf by Universitätsbibliothek Kiel user on 09 September 2024 Table 2. Estimated means, 95% CIs, and effect sizes for the primary and secondary outcomes and effects (time, time piece-wise, and treatment) of the GLMM Estimated means (95% CI), effect size Cohen’sdEffects of the GLMM a DBT ST between-group effect size d b [95% CI] Outcome and time point M[95% CI] within-group effect size d c [95% CI] M[95% CI] within-group effect size d c [95% CI] observed estimated tdf pr BPDSI total Baseline M= 32.80 e –Time −10.12 244 <0.001 0.54 0.5 years 25.20 [23.53; 26.99] 1.00 [0.54; 1.45] 27.76 [25.98; 29.66] 0.64 [0.19; 1.08] −0.24 [−0.58; 0.10] −0.27 [−0.61; 0.07] Time piece-wise 4.98 260 <0.001 0.29 1 year 21.21 [19.64; 22.91] 1.70 [1.17; 2.17] 23.43 [21.71; 25.28] 1.29 [0.81; 1.77] −0.04 [−0.39; 0.31] −0.21 [−0.56; 0.14] Treatment at 1-year FU 1.69 131 0.094 0.15 1.5 years (post) 17.85 [16.02; 19.89] 2.33 [1.77; 2.89] 19.77 [17.76; 22.01] 1.94 [1.41; 2.47] −0.01 [−0.40; 0.37] −0.17 [−0.56; 0.22] Time × Treatment 0.08 245 0.936 <0.01 0.5 years FU 17.58 [15.78; 19.59] 2.39 [1.83; 2.95] 20.20 [18.11; 22.53] 1.86 [1.34; 2.38] −0.17 [−0.59; 0.26] −0.19 [−0.61; 0.24] Time piece-wise × Treatment 0.49 260 0.628 0.03 1 year FU 17.31 [15.03; 19.94] 2.45 [1.88; 3.02] 20.63 [17.79; 23.93] 1.78 [1.26; 2.29] −0.15 [−0.60; 0.30] −0.24 [−0.69; 0.20] BPD checklist d Baseline M= 84.67 e –Time −7.06 291 <0.001 0.38 0.5 years 61.92 [55.31; 69.33] 0.94 [0.49; 1.39] 72.08 [64.39; 80.70] 0.48 [0.04; 0.93] −0.03 [−0.39; 0.33] −0.32 [−0.68; 0.04] Treatment at 1-year FU 1.65 86 0.102 0.18 1 year 54.61 [49.14; 60.70] 1.32 [0.84; 1.79] 64.51 [57.90; 71.87] 0.82 [0.36; 1.27] −0.09 [−0.47; 0.28] −0.26 [−0.63; 0.12] Time × treatment 0.436 291 0.663 0.03 1.5 years (post) 48.17 [42.89; 54.10] 1.69 [1.19; 2.19] 57.73 [51.22; 65.08] 1.15 [0.68; 1.62] −0.02 [−0.46; 0.42] −0.20 [−0.64; 0.24] 0.5 years FU 42.48 [36.87; 48.95] 2.07 [1.54; 2.60] 51.67 [44.66; 59.78] 1.48 [0.99; 1.98] −0.19 [−0.66; 0.27] −0.25 [−0.72; 0.22] 1 year FU 37.47 [31.40; 44.70] 2.45 [1.88; 3.02] 46.24 [38.60; 55.40] 1.82 [1.30; 2.33] −0.12 [−0.63; 0.38] −0.25 [−0.76; 0.25] DSS Baseline M= 2.90 e −Time −2.76 179 0.006 0.20 0.5 years 2.12 [1.73; 2.58] 0.45 [0.01; 0.89] 2.40 [1.99; 2.90] 0.27 [−0.17; 0.71] −0.07 [−0.43; 0.28] −0.15 [−0.50; 0.21] Treatment at 1-year FU 1.31 246 0.193 0.08 1 year 1.82 [1.54; 2.16] 0.67 [0.22; 1.11] 2.06 [1.74; 2.43] 0.49 [0.05; 0.93] 0.13 [−0.24; 0.50] −0.15 [−0.52; 0.23] Time × treatment 0.48 178 0.633 0.04 1.5 years (post) 1.57 [1.27; 1.93] 0.88 [0.43; 1.33] 1.76 [1.42; 2.18] 0.71 [0.26; 1.16] −0.03 [−0.47; 0.42] −0.11 [−0.56; 0.33] 0.5 years FU 1.57 [1.30; 1.89] 0.88 [0.43; 1.33] 1.90 [1.57; 2.30] 0.60 [0.16; 1.05] −0.09 [−0.57; 0.40] −0.20 [−0.68; 0.29] 1 year FU 1.57 [1.23; 1.99] 0.88 [0.43; 1.33] 2.05 [1.57; 2.69] 0.49 [0.05; 0.94] 0.12 [−0.40; 0.63] −0.26 [−0.78; 0.26] 256 Psychother Psychosom 2024;93:249–263 DOI: 10.1159/000538404 Assmann et al. Downloaded from http://karger.com/pps/article-pdf/93/4/249/4272968/000538404.pdf by Universitätsbibliothek Kiel user on 09 September 2024 Table 2 (continued) Estimated means (95% CI), effect size Cohen’sdEffects of the GLMM a DBT ST between-group effect size d b [95% CI] Outcome and time point M[95% CI] within-group effect size d c [95% CI] M[95% CI] within-group effect size d c [95% CI] observed estimated tdf pr BSI Baseline M= 1.94 e –Time −5.20 246 <0.001 0.31 0.5 years 1.64 [1.50; 1.79] 0.44 [0.01; 0.88] 1.77 [1.63; 1.92] 0.24 [−0.19; 0.68] 0.10 [−0.25; 0.46] −0.16 [−0.52; 0.19] Treatment at 1-year FU 1.45 147 0.150 0.12 1 year 1.49 [1.38; 1.61] 0.69 [0.25; 1.14] 1.64 [1.52; 1.77] 0.44 [−0.00; 0.88] 0.08 [−0.29; 0.45] −0.19 [−0.56; 0.19] Time × treatment 0.71 245 0.481 0.05 1.5 years (post) 1.35 [1.23; 1.48] 0.95 [0.49; 1.40] 1.53 [1.39; 1.67] 0.63 [0.18; 1.07] 0.10 [−0.34; 0.55] −0.15 [−0.60; 0.30] 0.5 years FU 1.23 [1.09; 1.39] 1.20 [0.73; 1.66] 1.42 [1.25; 1.61] 0.82 [0.37; 1.27] −0.07 [−0.56; 0.41] −0.20 [−0.68; 0.29] 1 year FU 1.11 [0.95; 1.31] 1.45 [0.97; 1.93] 1.32 [1.12; 1.55] 1.01 [0.55; 1.47] 0.05 [−0.47; 0.57] −0.21 [−0.73; 0.31] QIDS−SR Baseline M= 16.26 e –Time −5.03 202 <0.001 0.33 0.5 years 14.27 [13.26; 15.35] 0.42 [0.08; 0.76] 14.30 [13.33; 15.33] 0.42 [0.08; 0.75] 0.20 [−0.16; 0.55] −0.01 [−0.36; 0.35] Treatment at 1-year FU 1.35 147 0.179 0.11 1 year 13.23. [12.33; 14.64] 0.67 [0.31; 1.02] 13.67 [12.77; 14.64] 0.56 [0.21; 0.90] 0.16 [−0.22; 0.53] −0.08 [−0.46; 0.29] Time × treatment 1.31 203 0.193 0.09 1.5 years (post) 12.26 [11.30; 13.30] 0.91 [0.52; 1.30] 13.08 [12.04; 14.20] 0.70 [0.30; 1.10] 0.11 [−0.34; 0.55] −0.08 [−0.53; 0.36] 0.5 years FU 11.36 [10.26; 12.59] 1.16 [0.73; 1.57] 12.51 [11.25; 13.90] 0.85 [0.41; 1.27] −0.07 [−0.55; 0.41] −0.12 [−0.61; 0.36] 1 year FU 10.53 [9.27; 11.97] 1.40 [0.96; 1.83] 11.96 [10.46; 13.68] 0.99 [0.51; 1.46] 0.08 [−0.44; 0.60] −0.18 [−0.7; 0.34] WSAS Baseline M= 23.59 e –Time −2.39 411 0.018 0.12 0.5 years 18.93 [16.69; 21.48] 0.67 [0.23; 1.11] 21.23 [18.75; 24.03] 0.32 [−0.12; 0.76] 0.02 [−0.33; 0.38] −0.19 [−0.55; 0.16] Treatment at 1-year FU −0.10 273 0.924 0.01 1 year 18.45 [16.65; 20.11] 0.75 [0.30; 1.19] 20.04 [18.08; 22.22] 0.50 [0.05; 0.94] 0.10 [−0.27; 0.48] −0.12 [−0.50; 0.26] Time × treatment −0.89 412 0.374 0.04 1.5 years (post) 17.97 [16.28; 19.83] 0.83 [0.38; 1.28] 18.93 [17.07; 20.99] 0.67 [0.22; 1.12] 0.19 [−0.26; 0.64] −0.05 [−0.50; 0.40 0.5 years FU 17.51 [15.59; 19.66] 0.91 [0.46; 1.36] 17.88 [15.76; 20.25] 0.85 [0.22; 1.12] 0.12 [−0.36; 0.60] −0.02 [−0.50; 0.45] 1 year FU 17.06 [14.72; 19.77] 0.99 [0.53; 1.44] 16.88 [14.39; 19.80] 1.02 [0.56; 1.48] 0.24 [−0.28; 0.75] 0.01 [−0.50; 0.53] DBT versus Schema Therapy in Borderline Personality Disorder Psychother Psychosom 2024;93:249–263 DOI: 10.1159/000538404 257 Downloaded from http://karger.com/pps/article-pdf/93/4/249/4272968/000538404.pdf by Universitätsbibliothek Kiel user on 09 September 2024