Abstract
Background: Borderline personality disorder (BPD) is characterised by significant clinical heterogeneity. Classifying subtypes of BPD may offer deeper insights into the disorder's complexity and inform more tailored therapeutic strategies. The exploration of data-driven subtyping using cluster-analytic approaches represents a promising avenue for capturing variability in symptomatology and comorbidity profiles. Aim: This systematic review aims to synthesise and critically evaluate the empirical studies that have applied cluster-analytic methods to identify subtypes of BPD in adult populations. It further assesses the consistency of findings and their alignment with theoretical models of the disorder. Methods: A comprehensive search of PubMed, Scopus, and PsycNet was conducted in accordance with the PRISMA guidelines. Eligible studies employed either traditional or probabilistic clustering techniques to classify adult individuals diagnosed with BPD based on the DSM criteria. A total of 29 studies, encompassing 24,345 participants, met the inclusion criteria. The study quality and risk of bias were assessed using the AXIS tool. Results: Most studies identified clinically meaningful BPD subtypes based on dimensions such as affective regulation, effortful control, interpersonal style, and impulsivity or aggression. Several findings supported the existence of internalizing and externalizing profiles, which converge with long-standing theoretical conceptualisations of BPD. However, substantial heterogeneity was observed in subtyping bases, sample characteristics, and analytic procedures. Discussion: Although this review highlights the recurring subtype patterns, the methodological inconsistencies and a lack of longitudinal and treatment-outcome data limit the strength of the conclusions. The future research should prioritise standardised subtyping frameworks and explore the prognostic and therapeutic utility of BPD subtypes in clinical settings.