Sustainability of Mentalization‐Based Treatment Programs for Poorly Functioning Patients With Borderline Personality Disorder. Can We Really Keep It Up?

ABSTRACT

Though positive effects of mentalization‐based treatment (MBT) for patients with severe borderline personality disorder (BPD) are increasingly documented, less is known about the sustainability of specialized treatment standards and the maintenance of positive outcomes over time. This study aimed to investigate the organizational and clinical sustainability of an outpatient MBT program across two successive treatment periods. The study compares outpatients referred to MBT in a tertiary‐level, specialist mental health service in 2009–2011 (Period I: n = 96) versus 2011–2015 (Period II: n = 89). Organizational quality was based on the MBT quality manual (2019). By structured clinical interviews and repeated self‐reports, comparisons included baseline characteristics of eligible patients, clinical outcomes during treatment, therapeutic alliance, therapist countertransferences, and treatment adherence. The MBT team, organization, and systems for quality assurance held satisfactory standards and stability across the two time periods largely compatible with manual recommendations. Patient selection to MBT was in accordance with the targeted patient group, admitting poorly functioning young adults with BPD in both periods. Period II included patients with more severe BPD and self‐harming behaviors. Clinical improvement rendered effect sizes in the large range irrespective of time period, and overall 70% remission rates of self‐harming. Patient‐reported alliance to therapists and therapist‐reported countertransference responses were stable with satisfactory levels in both periods. Both periods had low rates of drop‐out, though higher in Period II. Results support positive effects and sustainability of MBT for poorly functioning BPD patients treated in a tertiary outpatient MBT‐unit within a supportive and stable organizational environment.

1. Introduction

Borderline personality disorder (BPD) is associated with considerable burden of disease and societal cost (Laurenssen et al. 2016; Hastrup et al. 2019). Its overall impact is a strong argument for health service investment in specialized treatment. Mentalization‐based treatment (MBT) is one of several effective approaches designed for BPD (Stoffers‐Winterling et al. 2022). Positive effects are recorded for up to 8‐year follow‐up (Bateman et al. 2021; Bateman and Fonagy 2009). Improvements include reduction of BPD features, self‐harm, emergency and hospital services (Hajek Gross et al. 2024; Vogt and Norman 2019), the latter contributing to considerable health cost savings (Bateman and Fonagy 2003). Beneficial outcomes and low drop‐out rates are also replicated in observational studies reflecting “real‐life” conditions (Tong et al. 2022; Kvarstein et al. 2015; Bales et al. 2015). Nonetheless, sustainability of MBT, how quality is maintained over time, is not well‐documented.

At its heart, MBT aims to improve patients' capacity for mentalizing (Bateman et al. 2018). It is a structured psychotherapy approach originally developed from psychodynamic traditions, incorporating pedagogical elements from cognitive therapy. Its format provides interpersonal therapeutic arenas, dyads, and groups. The style of intervention is empathic, explorative, focused on emotionally driven situations (Bateman and Fonagy 2004). Epistemic trust is considered an underlying vulnerability factor, impacting social learning processes and therapeutic change (Fonagy and Allison 2014). Emphasis is therefore on treatment consistency, coherence, and continuity (Bateman et al. 2018).

The importance of organizational factors when implementing specialized BPD treatment has been emphasized across approaches (Flynn et al. 2021; Bales, Verheul and Hutsebaut 2017). Two MBT studies reported how patient outcomes and team cohesion were vulnerable to organizational disruptions (Hutsebaut et al. 2012; Bales, Timman, et al. 2017). The quality manual for MBT recommends standards for sufficient and safe treatment application (Bateman et al. 2014), thus accounting for potential risks involved when taking on treatment for poorly functioning patients with BPD.

Ample psychotherapy research associates a bonding and cooperative working alliance with clinical improvement (Fluckiger et al. 2018). In a study of MBT, the mutual understanding of therapeutic work (tasks in therapy) was particularly notable (Folmo et al. 2020). In a recent study of PD treatments, therapists reported less negative countertransferences during therapies with good outcomes (Øvstebø et al. 2024). The establishment of a satisfactory therapeutic relationship is thus likely to be an essential quality aspect.

Based on a well‐established MBT‐unit, the present study aims to compare two different time periods considering quality aspects of organization, patient selection, treatment process, and outcomes.

2. Material and Methods

2.1. Design

The study has an observational, longitudinal design.

2.2. Context of Data Collection

The study‐setting was a multidisciplinary, outpatient MBT‐unit on a tertiary specialist mental health service level, the first to implement MBT in Norway as described in a former publication (Kvarstein et al. 2015). The MBT‐unit was part of a Norwegian cross‐regional quality and research collaboration, the Network for Personality Disorders (the Network) (Pedersen et al. 2023). The collaboration entailed standard assessment procedures by therapist interviews and patient self‐reports, feedback systems for patient evaluation, and regular meeting points, training courses, and conferences. Referrals to the MBT‐unit came from specialist (secondary) mental health services within a larger urban district. Poorly functioning adults (18–30 years) with BPD were the prioritized target group. Data from interviews and self‐reports were transferred anonymously to a quality register.

2.3. Sample

The total count included 185 outpatients. Period I included patients referred 2009–June 2011 (n = 96); Period II from July 2011–2015 (n = 89).

2.4. Quality of MBT Organization

A validated scoring system for organizational fidelity was not available. Thirty‐three “quality standards” were therefore developed, based on the MBT Quality Manual (Bateman et al. 2014). Each fulfilled item was represented by score = 1, adding up to a sum‐score (Table 1). Information was collected retrospectively from leaders of the MBT‐unit. MBT quality standards were developed and assessed before outcome analyses were completed. Details on MBT organizational quality scores are presented in Data S2.

TABLE 1.

Quality of MBT organization

		Period I	Period II
	Multidiciplinary
1	Multidisciplinary team	1	1
2	Individual therapist is principal therapist	1	1
3	Therapeutic teams round each patient	1	1
4	Psychiatrist evaluates medication	1	1
5	Secretary receives basic MBT/BPD courses, administers invitations, phone calls, etc.	1	1
SUM		5	5
	MBT‐team
6	Team manager: MBT qualified	1	1
7	Most therapists have primary work in the team, doing MBT.	1	1
8	Team size < 10, but > four therapists*	0	0
9	Caseload per therapist < 18 patients	1	1
SUM		3	3
	Competencies
10	All active therapists have MBT training	0	0
11	All active therapists have manuals	0	1
12	System for quality assessment (therapist fidelity)	0	1
SUM		1	2
	Regular supervision
13	MBT supervision groups < five therapist/therapist pairs	1	1
14	Team—intervision	0	0
15	Team—case discussion/knowledge/process integration	1	1
SUM		2	2
	Supervisors
16	All are qualified MBT supervisors	0	1
17	All skilled in MBT group and individual	1	1
18	All associated with an MBT team practicing MBT	1	1
19	All are not associated with the supervised MBT team	0	0
20	Regular reports on the functioning of the supervision	0	0
21	Quality monitoring system MBT fidelity available	0	1
SUM		2	4
	Treatment
22	MBT psychoeducation group	1	1
23	Combination individual and group formats	1	1
24	Group therapy 1–1,5 h, two therapists	1	1
25	Crises: Therapist availability systems	0	1
26	Treatment duration 18 months**	1	1
27	Monitoring system alliance and outcome	1	1
SUM		5	6
	Target population, referrals, assessment, agreements, and reviews
28	Defined target group: BPD, age‐range, other	1	1
29	Defined exclusion	1	1
30	Clear pathways—referral, management, collaborations other services	1	1
31	Pretreatment assessment (diagnoses/functioning/suicidal risk/medication)	1	1
32	Treatment agreement/plan/dynamic case formulation	1	1
33	Regular progress reviews (patient and therapists)	1	1
SUM		6	6
Sum total		24	28

^*The MBT team had > 10 active therapists in both periods.

^**Duration > 18 months.

2.5. Baseline Assessment

Items designed for the Network included patient self‐report on gender, age, occupational status, and substance use, and therapist‐report on aggressive behaviors. Mental health disorders were assessed by the M.I.N.I. International Neuropsychiatric Interview (MINI, (Sheehan et al. 1994)) and the Structured Clinical Interview for Personality Disorders (First 1994). Assessments were performed by trained therapists at the unit (Pedersen et al. 2013). The Severity Indices of Personality Problems (SIPP) is a 118‐item (16 facets, 5 domains) patient self‐report assessing maladaptive personality functioning, rated on a 1–4 scale. Summarizing domains, we report norm‐referenced standardized deviation scores (T‐scores) (Pedersen et al. 2019). The EuroQol EQ‐5D‐3L (van Asselt et al. 2009) is a patient self‐report including a visual analogue scale (VAS) ranging health state from worst to best possible (scores 0–100). Mean VAS‐scores in general population studies range 80–89 and in a PD population, 56 (Saarni et al. 2006; Soeteman et al. 2008).

2.6. Repeated Assessment (Every Third Month First Year, Then Half‐Yearly, and on Termination)

The Working Alliance Inventory‐Short (WAI‐S) was first applied for patient‐rated alliance to the individual therapist. During the study period, it was replaced by WAI‐SR (Working Alliance Inventory‐Short Revised). Both include three 4‐item subscales, Goals, Bonds, and Tasks (7‐point rating scale, scores > 4 signify satisfactory alliance). This study presents Goals and Tasks in a combined sum‐score in accordance with other research (van Benthem et al. 2024). Data S1 presents WAI versions across periods.

The Feeling Word Checklist‐Brief version (FWC‐BV) (Breivik et al. 2020) is a 10‐item therapist‐reported measure of countertransference feelings, arranged in three dimensions (Confident, Inadequate, and Idealized, 0–4 rating‐scale, higher score indicating stronger reaction).

Brief Symptom Inventory (BSI‐18) (Derogatis 2000) is an 18‐item (0–4 scale) patient self‐report including a global mean score (BSI). Severe distress is indicated by high scores.

Circumplex of Interpersonal Problems (CIP) (Pedersen 2002) is a 48‐item (0–4 scale) Norwegian version of the patient self‐report Inventory of Interpersonal Problems—Circumplex version (Alden et al. 1990). We report total mean CIP score. High scores indicate severe problems.

The Work and Social Adjustment Scale (WSAS) is a 5‐item patient self‐report (0–8 scale, sum‐score ranging 0–40) (Pedersen et al. 2017). Severe impairment is indicated by high scores.

The Global Assessment of Functioning (GAF: APA, 1994) is an observer‐based evaluation of functioning and symptom severity (1–100 scale). Clinicians in the MBT team were trained to evaluate GAF. Reliability was acceptable in a former study from the Network (Pedersen et al. 2007). GAF‐scores < 60 indicate clinically relevant impairment.

2.7. Other Assessment

Therapist‐reported treatment duration, regularity, and reasons for termination were recorded at the end of treatment, and patient‐reported self‐harm and suicide attempts, both lifetime and in the recent year, were assessed at the start and end of treatment (items designed for the Network).

2.8. Ethics

All patients gave written consent to use anonymous clinical data for research. Procedures for collection, transferral, and data‐storage in a quality register were approved (Oslo University Hospital, data protection officer). Approval from the Regional Committee for Medical Research and Ethics was not required.

2.9. Statistics

Statistics were based on IBM SPSS Statistics, version 29.0.2.0 (Fluckiger et al. 2018). Mixed models (MM) were chosen for longitudinal analyses (Singer and Willett 2003) and included five linear change models investigating process variables and four investigating outcome variables. The predictor, “time‐period” was investigated in each model. Tables 3 and 5 present MM estimates and indices of model fit (Akaikes Information Criterion, AIC). 0–24‐month Cohen's d effect sizes are based on MM predicted values (small effect size: d = 0.2, medium d = 0.5, large d = 0.8). Modelling procedures are detailed in Data S3.

TABLE 3.

Alliance and countertransference.

		Fixed effects				AIC
		Intercept estimate	p	Slope estimate	p
Dependent variable	Predictor	Mean (SE)		Mean (SE)
WAI
Goals&Tasks		10.0(0.30)	< 0.001	0.03 (0.02)	0.05	1352
Goals&Tasks	Δ (II–I)	−1.3 (0.59)	0.03	0.07 (0.03)	0.05	1351
Bond		5.2 (0.16)	< 0.001	0.01 (0.009)	> 0.1	996
Bond	Δ (II–I)	−0.23 (0.33)	> 0.1	0.013 (0.02)	> 0.1	999
FWC‐BV
Idealized		1.2 (0.1)	< 0.001	0.001 (0.005)	> 0.1	738
Idealized	Δ (II–I)	0.26 (0.2)	> 0.1	−0.02 (0.011)	> 0.1	738
Inadequate		0.59 (0.07)	< 0.001	0.00 (0.004)	> 0.1	535
Inadequate	Δ (II–I)	0.13 (0.15)	> 0.1	−0.004 (0.008)	> 0.1	538
Confident		2.36 (0.10)	< 0.001	0.006 (0.006)	> 0.1	760
Confident	Δ (II–I)	−0.13 (0.21)	> 0.1	0.001 (0.011)	> 0.1	762

Tab. 6 demonstrates MM estimates for aspects of alliance (WAI) and countertransference (FWC‐BV) for each outcome model. Row one demonstrates estimates for the linear model, and row two estimates for the model with predictor. Differences between time periods are indicated by Δ (II–I). Positive estimates indicate that the estimate for Period II > Period I. Indices of model fit are given by Akaike's information criterion (AIC).

TABLE 5.

Clinical outcome.

Dependent variable	Predictor	Fixed effects				AIC	d
		Intercept estimate	p	Slope estimate	p
		Mean (SE)		Mean (SE)
BSI		2.08 (0.1)	< 0.001	−0.03 (0.003)	< 0.001	1590	1.5
BSI	Δ (II–I)	0.14 (0.1)	> 0.1	0.002 (0.01)	> 0.1	1592
CIP		1.83 (0.04)	< 0.001	−0.02 (0.001)	< 0.001	889	1.3
CIP	Δ (II–I)	0.12 (0.1)	> 0.1	−0.004 (0.004)	> 0.1	890
WSAS		25.3 (0.6)	< 0.001	−0.34 (0.04)	< 0.001	3977	2.2
WSAS	Δ (II–I)	2.15 (1.1)	> 0.1	−0.07 (0.08)	> 0.1	3978
GAF		48.8 (0.4)	< 0.001	−0.34 (0.03)	< 0.001	3755	3.8
GAF	Δ (II–I)	−0.44 (0.8)	> 0.1	−0.02 (0.06)	> 0.1	3759

Table 5 demonstrates MM estimates. For each outcome model, row one demonstrates estimates for the linear model and row two estimates for the model with predictor. Differences between the two time periods are indicated by Δ (II–I). Indices of model fit are given by Akaike's information criterion (AIC). d indicates Cohen's d 0–24‐month effect sizes based on MM predicted values.

Patients had different numbers of assessments (range included in MM analyses: 1–7). For all models, mean assessment number was > 3. A variable indicating assessment number per individual was investigated as a predictor to evaluate possible bias of missing assessments (Hedeker and Gibbons 1997). Number of assessments was not associated with deviating intercept level nor deviating change rate (for all models p > 0.05) (See also Data S3).

Possible impacts of different WAI versions were investigated in WAI models. Different WAI versions were not associated with deviating intercept levels nor changes (for all WAI subscale models p > 0.05). WAI investigations also included final models controlling for different WAI versions.

Other descriptive statistics included multivariate analyses of variance for continuous data. We report exact p‐values for values less than p = 0.1 and greater than p = 0.01. For comparisons of change during treatment for categorical data (proportions), McNemar's test was used for paired samples (start–end) and Wald's test for independent samples (comparison of end point data). For comparison of other categorical data, p‐values based on chi square tests are presented. Considering the multiple comparisons of baseline data, Bonferroni correction would imply an adjusted significance level to p < 0.001.

3. Results

3.1. Organization

MBT organization scores indicated that most quality standards were met and stable across time periods (Table 1). The highest score was in Period II (28 of total 33 in Period II, versus 24 of 33 in Period I). Scores were in both periods below standards for team size and supervision. In Period II, a greater number of therapists had completed MBT training, MBT manuals and MBT fidelity scoring were available, and systems for crisis intervention were implemented. A relative increase of psychologists within the multidisciplinary team was found in Period II. In Period II, the number of MBT groups for BPD was reduced from nine to eight, due to the introduction of a pilot for patients with avoidant PD. MBT organization with details on therapists and the MBT team is presented in Data S4.

3.2. Patient Selection

Patients admitted to MBT were in accordance with the target group considering age, psychosocial functioning, and PD‐status. Mean GAF was 48 (SD 5); overall health‐related quality of life and measures of psychosocial functioning indicated considerable burden. Table 2 and Data S5 compare status on referral in the two periods. Frequencies of BPD, number of BPD criteria, and PD criteria were higher in Period II; though BPD dominated both periods. There were no significant differences in other psychosocial factors and mental health disorders. Self‐harming behaviors were frequent, and more frequent in Period II. There were not significant differences in aggressive behaviors and substance abuse. SIPP scores indicated personality problems in both groups, with the poorest scores within domains of identity and self‐control. Self‐control was significantly poorer in time Period II, with the largest differences for the facet “emotional regulation.”

TABLE 2.

Patients admitted to MBT.

	Period I		Period II
	%	Mean (SD)	%	Mean (SD)	P difference
Sociodemografics
Females	83		80		> 0.1
Age		28 (6)		27 (5)	> 0.1
Sum of months in > 50% work/study last year		4.7 (4.7)		4.7 (4.7)	> 0.1
EQ‐5D‐3L VAS		45 (19)		43 (19)	> 0.1
Personality disorder
BPD	67		81		0.05
Sum of BPD criteria		5.2 (2)		5.8 (2)	0.01
Avoidant PD	16		16		> 0.1
Paranoid PD	12		14		> 0.1
Obsessive Compulsive PD	9		12		> 0.1
NOSPD	20		13		> 0.1
Number of PD criteria		14 (6)		15 (6)	0.02
Number of other mental health disorders		2.6 (1.4)		2.5 (1.5)	> 0.1
Self‐harm/suicide attempts
Self‐harming (recent year)	79		92		0.01
Self‐harming (ever)	77		92		0.01
Suicidal thoughts (recent year)	83		92		> 0.1
Suicide attempts (recent year)	30		29		> 0.1
Suicide attempts (ever)	52		65		0.05
Sumvariable: selfharm, suicide thoughts, attempts		3.2 (1.4)		3.5 (1.2)	0.09
Aggression
Violent behaviors, not against people	41		35		> 0.1
Violent behaviors against other people	27		21		> 0.1
Police‐record of violence	10		6		> 0.1
Sumvariable: Aggression		0.8 (1.0)		0.6 (0.9)	> 0.1
Substance abuse last 3 months
Alcohol	74		83		> 0.1
Prescribed medication	20		26		> 0.1
Other substances	29		28		> 0.1
Sumvariable: Substance abuse		2.6 (2.1)		2.9 (1.9)	> 0.1
Patient‐reported personality problems
SIPP T self‐control		26 (14)		22 (13)	0.03
SIPP emotion regulation		2.2 (0.7)		1.8 (0.5)	0.001
SIPP T social concordance		33 (15)		33 (14)	> 0.1
SIPP T identity		21 (11)		18 (11)	> 0.1
SIPP T responsibility		28 (14)		29 (12)	> 0.1
SIPP T relation		33 (10)		32 (10)	> 0.1

3.3. Process

All aspects of WAI and FWC‐BV held a satisfactory level through therapy irrespective of time period (Table 3). A tendency of increasing scores of goals and tasks during treatment was stronger in Period II; though not maintained when controlling for WAI‐S/WAI‐SR differences (Table 3). Time periods rendered minimal improvement of model fit.

Mean treatment duration differed by time period and was shorter in time Period II (Table 4). The proportion with treatment duration shorter than 6 months and the proportion “drop‐outs” were higher in time Period II. There were not significant differences in regularity of attendance (Table 4).

TABLE 4.

Duration and adherence.

	Period I		Period II
	%	Mean (SD)	%	Mean (SD)	P difference
Current treatment
Duration (months)		29 (15)		20 (12)	< 0.001
Duration < 6 months	4		15		0.02
Therapist defined drop‐out	8		12		0.07
Regular attendance	60		60		> 0.1

3.4. Outcomes

Improvement of BSI‐, CIP‐, WSAS‐, and GAF‐scores was highly significant; estimated 24‐month effect sizes were in the large range (d > 1). Outcome trends did not differ by time period (Table 5). “Time‐periods” explained 0%–5% intercept variation, no additional slope variation, and minimal improvement of model fit (Table 5).

In the total sample, 85% reported self‐harming within 12 months of starting MBT. Frequencies of self‐harming declined in both groups (start–end change: p < 0.001). In the last year of MBT, 23% (Period I) and 34% (Period II) reported self‐harming. Overall remission rate was 70%. The difference between periods was nonsignificant (Period I: 75%, Period II: 63%, p = 0.212).

In the total sample, fewer patients reported suicide attempts at the end of treatment (start–end change: p < 0.001). Thirty percent reported suicide attempts within 12 months of starting MBT, 9% during the last year of MBT. The overall remission rate was 88%. The difference between periods was non‐significant (Period I: 89%, Period II: 87%, p = 0.801).

4. Discussion

This study demonstrates a treatment unit aiming to deliver MBT and organized accordingly. Across different evidence‐based BPD treatment approaches, further research on implementation within clinical settings is recommended (Flynn et al. 2021). As the first unit in Norway to deliver MBT, the present study adds a long‐term perspective to its clinical utility.

4.1. Quality of Organization

In our study we demonstrate high stability of MBT organization. Across periods the internal framework for therapeutic work was clearly defined in terms of a common therapeutic approach, a set treatment format, regular meetings for team collaboration, case discussions, and supervision. The importance of an internal organizational framework in order to maintain treatment quality is increasingly recognized, and common challenges are addressed across approaches. Studies include MBT, Dialectical Behavioral Therapy, the more pragmatic guideline‐informed treatment (GIT‐PD), and multisystemic treatments addressing younger patients and complex family settings (Bales, Timman, et al. 2017; Bales, Verheul and Hutsebaut 2017; Beattie et al. 2022; Swales et al. 2012; King et al. 2018) (Henggeler and Schaeffer 2016; Hutsebaut et al. 2020). Among reported long‐term challenges are management of staff turnover, maintenance of therapist competence, and financial funding issues. The present study also identified therapist turnover and changes in team composition. The higher proportion of psychologists in Period II may reflect changing health service requirements. Possibly compensating for therapist instability, increased access to MBT manuals and qualification courses in the same period were likely contributors to MBT maintenance and engagement.

Other stabilizing factors present in our study sample may have included the use of common assessment and feedback systems. Such are known to support treatment processes, keep therapists on track, facilitate team collaborations by contributing to shared understanding, and enable identification of poor responders (Gual‐Montolio et al. 2020). In our study, the Network provided such a supportive framework, which was operative in both periods (Pedersen et al. 2023). Systematic assessment and outcome monitoring is also recommended in the MBT quality manual (Bateman et al. 2014).

The current unit pioneered MBT in Norway, but by the end of Period II, approximately half of the patients in the Network received some kind of MBT intervention (Kvarstein et al. 2023); MBT‐therapists met for half‐yearly MBT supervisory seminars. This broader implementation of MBT and contact between units may also have supportive and calibrating qualities, counteracting negative effects of isolation and also contributing to therapist recruitment.

4.2. Patient Selection

Selection of eligible patients is a central aspect of a specialized treatment of importance for treatment success. In the present study, the target sample was unchanged across periods. MBT research has indicated particular utility of MBT programs for patients with clinically severe conditions (Bateman and Fonagy 2013; Kvarstein et al. 2017; Smits et al. 2024). Our study revealed increasing intake of patients with more severe emotional dysregulation and a larger proportion with self‐harming behaviors. This development may reflect the ongoing research and discussions on health service prioritizations. As a tertiary‐level service, it is also natural that more afflicted patients were selected when MBT was increasingly established elsewhere.

4.3. Treatment Process

The two sources for process evaluation, patients and therapists, were quite coherent; overall levels were satisfactory and did not differ by period. Levels of working alliance and positive therapist countertransference feelings may seem higher than could have been expected considering the target group of patients (Folmo et al. 2020; Øvstebø et al. 2024). Initial motivation and engagement may relate to positive expectations of a tertiary level, highly specialized treatment. Nonetheless, stability signals a productive therapeutic setting. The therapeutic climate suggests that therapies addressing patients with complex relational and emotional difficulties can be focused in a manageable way for both patients and therapists.

Patients' adherence to treatment is a cornerstone of effective therapy, also reflecting the therapeutic climate, and is known to challenge BPD treatment (Barnicot et al. 2011). In our study, drop‐out rates were generally low, in line with other MBT studies (Remeeus et al. 2024). Nonetheless, rates of drop‐out were higher in Period II. Although differences in severity of condition across periods were small, this may explain the greater challenge of treatment adherence in Period II.

4.4. Clinical Outcomes

An essential finding in this study is the maintenance of highly significant improvements for patients during treatment and large effect sizes capturing global qualities of mental distress, interpersonal problems, and psychosocial functioning. Positive results include not least stability of high remission rates of self‐harming behaviors, despite more severe conditions in the last period. The overall improvement kept on level with reasonable expectations of MBT as an evidence‐based treatment for BPD (Vogt and Norman 2019). It is also noteworthy that changes in the multidisciplinary team composition did not seem to disturb patient outcomes.

A former MBT study, based on Period I, indicated satisfactory MBT fidelity scores (Kvarstein et al. 2015). However, the present study lacked sufficient data to compare how the therapies reflected the MBT method across periods. Key questions for further research are how method and organizational quality interact and relate to process and outcomes. In line with former research (Henggeler and Schaeffer 2016), we would assume significant interactions between therapist fidelity and the sum of organizational factors, together promoting favorable treatment processes and outcomes. However, in line with principles of the GIT‐PD model, we would expect limited sustainability of a specific in‐session therapeutic focus without a surrounding framework updated on BPD, providing team support and supervision. As a single‐center study, the present scope is limited. However, multicenter research designs would allow such further investigation.

5. Strengths and Limitations

The large sample size and longitudinal design are major strengths. The study uses appropriate longitudinal statistics (Singer and Willett 2003) and includes statistical consideration of possible missing data patterns (Hedeker and Gibbons 1997). Nonetheless, missing data in longitudinal series are a limitation and can be expected in naturalistic treatment settings. The comparison is uncontrolled, but MM analyses control for baseline variation. The clinical setting is a tertiary‐level outpatient clinic that represents an expert‐driven context. As such, it may not reflect MBT implementation on lower health service levels. Quality of MBT organization was assessed retrospectively involving a small group of clinicians and team leaders active in the two periods. Compared to present‐time registrations, validity may thus be limited. Outcome measures were largely based on patient self‐report. We therefore included clinician‐performed observations (GAF) that were conducted by the patients' therapists, who were trained in the use of GAF and knew the patient well over time. Being a nonindependent rating, scores may be biased. However, as demonstrated in Table 3, main longitudinal trends of self‐reported versus therapist observational measures were comparable. It is generally a limitation that the study did not include independent patient perspectives on service organization.

6. Conclusions

MBT was maintained as a high quality treatment across time periods. The study thus demonstrates high sustainability. Quality aspects included stability of organization, satisfactory therapeutic processes, low levels of drop‐out, effect sizes in the large range, and 70% remission of self‐harming behaviors. The study highlights the potential importance of an expert‐driven context in achieving and sustaining treatment outcomes.

Author Contributions

All authors have been engaged in discussion of results and manuscript development. In addition, authors have specifically contributed with the following: E.H.K. was responsible for study design, data analyses, and drafting of manuscripts; K.B., E.N.E., and E.H.K. contributed to data collection; K.B., A.E., and Å.L.B. with special competence on MBT quality and implementation; G.P. was responsible for the data collection, organization of data, and psychometrics; and T.W. with special clinical research competence.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: Supporting information.

Kvarstein, E. H. , Bremer K., Baltzersen Å. L., et al. 2025. “Sustainability of Mentalization‐Based Treatment Programs for Poorly Functioning Patients With Borderline Personality Disorder. Can We Really Keep It Up?.” Personality and Mental Health 19, no. 4: e70036. 10.1002/pmh.70036.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.