Who Improves in Therapy for Borderline Personality Disorder? A Landmark Review of 311 Studies Offers Surprising Answers

For more than a decade, clinicians treating borderline personality disorder have operated with a striking gap in their knowledge: despite the existence of several effective psychotherapies, remarkably little was known about which patients respond best, which factors predict improvement, and which psychological changes actually drive recovery. The last major systematic review of treatment response predictors in this population was published back in 2012. Now, a team of researchers led by Anne Sonley and Shelley McMain at the Centre for Addiction and Mental Health in Toronto, working with colleagues at the University of Toronto, the University of Winnipeg, The New School, NYU Grossman School of Medicine and Imperial College London, has filled that gap with one of the most comprehensive syntheses ever attempted in the field. Their systematic review, published in BMC Psychiatry, sifted through nearly nineteen thousand studies to identify the patient-related factors that shape psychosocial treatment outcomes in borderline personality disorder, and the results challenge several long-standing clinical assumptions.

The scale of the undertaking is itself remarkable. Following PRISMA guidelines, the researchers systematically searched five major biomedical and psychological databases: MEDLINE, EMBASE, PsycINFO, Web of Science and CINAHL. The initial sweep returned 18,973 unique articles, a mountain of evidence that the team then screened with rigorous inclusion criteria. Studies were eligible if they examined the effects of psychosocial treatments for individuals diagnosed with borderline personality disorder across treatment modalities, and if they reported on predictors, moderators or mediators of outcomes. Pharmacological-only trials were excluded, since the review focused specifically on talk therapies and structured psychological interventions. In the end, 311 studies met the criteria for inclusion, an extraordinary evidence base that reflects both the growth of psychotherapy research in this field and the intensity of scientific interest in a condition that affects a significant proportion of psychiatric patients worldwide.

Because the volume of included research was so large, the authors made a pragmatic decision to split the reporting into two parts. The current paper, the first in the series, concentrates exclusively on patient-related factors: the characteristics that individuals bring with them into the therapy room. Treatment-related predictors, such as features of the therapies themselves and the clinicians delivering them, will be reported in a subsequent publication. This division matters scientifically, because it allows the authors to disentangle questions about who benefits from questions about what kind of treatment works best for whom. The review was pre-registered in March 2020 in the PROSPERO International Prospective Register of Systematic Reviews, a step that lends transparency to the methods and guards against retrospective bias in how the evidence was selected and synthesized.

Among the most consequential findings are the variables that appear not to matter. Demographic characteristics, including age, gender, education and employment status, showed no apparent association with either change in borderline personality disorder symptoms or with dropping out of treatment. Equally striking, common co-morbid mental health conditions, including substance use disorders, major depressive disorder and anxiety disorders, also appeared unrelated to how much patients improved or whether they stayed in therapy. This is a message with real clinical weight. Patients with borderline personality disorder frequently carry additional diagnoses, and they are sometimes steered away from intensive psychotherapy on the assumption that depression, anxiety or active substance use will undermine their progress. The review suggests that such exclusions may not be justified by the evidence, and that these patients should not be presumed poor candidates for psychological treatment simply because of their demographic profile or co-occurring conditions.

The picture becomes more nuanced when it comes to baseline symptom severity. The question of whether patients who start treatment with more severe borderline personality disorder symptoms improve less, more, or at the same rate as those with milder presentations has long divided clinicians. Here the review found mixed findings: studies examining baseline BPD symptoms as a predictor of symptom change at follow-up did not converge on a consistent answer. Some investigations suggested one direction of effect, others the opposite, and still others no relationship at all. This inconsistency is itself informative, highlighting that severity of presenting symptoms cannot yet be used as a reliable guide to prognosis across psychosocial treatments. For a field in which triage decisions often hinge on perceived severity, the absence of a dependable signal is a finding that deserves close attention from service planners and clinicians alike.

Amid the null and mixed results, two psychological variables emerged as genuinely promising. Mindful non-judging, the capacity to observe inner experience without evaluating it as good or bad, and emotion regulation, the ability to manage and modulate emotional responses, both showed promise as predictors and as mediators of improvement in borderline personality disorder symptoms. The distinction between predictors and mediators is technically important. A predictor is a characteristic measured before or during treatment that forecasts the outcome; a mediator is a change that occurs during treatment and statistically explains how the therapy produces its effect. If mindful non-judging and emotion regulation function as mediators, they may represent the active ingredients through which therapies such as dialectical behaviour therapy and related approaches exert their influence. That, in turn, opens the door to deliberately strengthening these capacities, whether within standard treatment protocols or through targeted adjunctive modules, to amplify clinical gains.

Yet the review also exposes a structural weakness in the scientific literature itself. The authors report that many of the 311 studies examined unique predictor-outcome pairs, meaning that few findings were replicated across independent research teams. When a single study tests a novel combination of predictor and outcome, a positive result cannot be distinguished from chance, and a null result cannot be distinguished from an underpowered test. This fragmentation is precisely why the authors did not conduct a meta-analysis: the heterogeneity of methodologies, outcome measures, predictor definitions and analytic approaches across studies made pooling the results statistically unfeasible. Instead, the review offers a narrative synthesis, mapping the terrain and flagging which regions are well-trodden and which remain unexplored. The conclusion is sobering but constructive: the field needs coordinated replication of a small set of candidate predictors rather than continued proliferation of one-off analyses.

The methodological implications extend beyond borderline personality disorder research. The pattern documented here, in which hundreds of studies each probe idiosyncratic questions with heterogeneous designs, is common across psychotherapy research more broadly. The Toronto-led team argues that further replication and systematic identification of predictors may aid in the refinement of treatment development and implementation for this population. In practical terms, robust predictors could eventually allow clinicians to match patients to the treatments most likely to help them, monitor the psychological mechanisms that signal progress, and intervene early when the expected mediators of change fail to move. Robust moderators, variables that indicate for whom a treatment works better or worse, could similarly personalize care. None of that personalization is currently possible on firm evidence, but this review clarifies exactly which candidate variables have earned continued investigation and which can be set aside.

For patients and families, the headline message is quietly hopeful. The evidence reviewed suggests that recovery in psychosocial treatment is not confined to any demographic group, and that co-occurring depression, anxiety or substance use need not stand in the way of meaningful improvement. The psychological skills that appear most promising, non-judgmental awareness and emotion regulation, are trainable capacities at the heart of several established therapies, which means the mechanisms of change may be within reach of deliberate clinical cultivation. As the second part of this systematic review, focused on treatment-related factors, prepares to follow, the field now possesses something it has lacked since 2012: a current, comprehensive and transparently conducted map of what does and does not predict who gets better. Turning that map into personalized, mechanism-informed care is the task the next decade of research must now take up.

Subject of Research: Patient-related predictors, moderators and mediators of psychosocial treatment outcomes in borderline personality disorder

Article Title: Predictors, moderators and mediators of psychosocial treatment outcomes in borderline personality disorder: a systematic review (part I – patient related factors)

Article References: Predictors, moderators and mediators of psychosocial treatment outcomes in borderline personality disorder: a systematic review (part I – patient related factors). (n.d.). https://doi.org/10.1186/s12888-026-08761-z

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08761-z

Keywords: borderline personality disorder, psychotherapy, systematic review, treatment outcomes, predictors, mediators, moderators, emotion regulation, mindfulness, co-morbidity, BMC Psychiatry, mental health