Borderline personality disorder: a comprehensive review of diagnosis and clinical presentation, etiology, treatment, and current controversies

The term “borderline” was introduced in the psychiatric literature by Stern¹ and Knight², to identify a patient group showing a level of functioning situated between neuroses and schizophrenic disorders. This patient group was not well defined. An important progress occurred with Kernberg's introduction of the concept of borderline personality organization^{3, 4}, marked by the use of primitive defense mechanisms such as splitting or projective identification, identity diffusion (shifting between all-good and all-bad), and severely disturbed object relationships³. Reality testing was largely intact, differentiating individuals with borderline personality organization from psychotic patients³. Another early contribution was provided by Grinker et al⁵, who empirically identified four features of the “borderline syndrome”: anger, impaired close relationships, identity problems, and depressive loneliness.

In 1980, borderline personality disorder (BPD) was introduced in the DSM-III⁶, based on a study by Spitzer et al⁷, who drew both on research by Gunderson and colleagues^{8, 9} and on Kernberg's concept of borderline personality organization³, by including specific problems of identity and interpersonal relationships characterized by sudden shifts from one extreme to another (e.g., from all-good to all-bad or vice versa). This early research showed that BPD could be discriminated with sufficient accuracy from both schizophrenia and (neurotic) depression, as well as from other personality disorders^{10, 11}.

In the following more than four decades, a plethora of research has been carried out on BPD, much more than on any other personality disorder. This research has focused on the diagnosis of BPD, its etiology (including genetics, neurobiology, and interactions between genetics/neurobiology and adverse childhood experiences), epidemiology, course and prognosis, cognition, and the effectiveness of pharmacotherapies and psychotherapies^12-18.

BPD remains a challenging disorder, from both research and clinical perspectives. At present, for example, there is still controversy concerning its conceptualization as either a specific personality disorder or a level of general impairment in personality functioning^19-21. The treatment of BPD remains challenging as well. As to pharmacotherapy, there is no consistent evidence showing that any psychoactive medication is efficacious for the core features of the disorder¹⁶. Indeed, no medications have been approved by regulatory agencies for treating BPD^{16, 22}. According to the UK National Institute for Health and Care Excellence (NICE), pharmacotherapy should only be used to treat discrete and severe comorbid anxiety or depressive symptoms or psychotic-like features, or to manage acute crises, and should be administered for the shortest time possible²². Psychotherapy is the treatment of choice for BPD, with various approaches having proved to be efficacious in randomized controlled trials (RCTs)^{14, 17, 22}. However, almost 50% of BPD patients do not respond sufficiently to psychotherapy²³, so that further research in this area is clearly warranted. Whether specialized methods of psychotherapy or more generalist approaches are required for the treatment of BPD is a debated issue^24-26.

This paper provides a comprehensive review of BPD diagnosis and clinical characterization, course, epidemiology, risk factors, neurobiology, social cognition and neurocognition, and management. Current controversies (e.g., categorical vs. dimensional approaches to diagnosis; specific vs. generalist psychotherapy interventions) are also discussed, and major areas in which further research is warranted are highlighted.

DIAGNOSIS AND CLINICAL CHARACTERIZATION

The DSM-5 characterizes BPD as a pervasive pattern of instability of interpersonal relationships, self-image and affects, and marked impulsivity, emerging by early adulthood and present in a variety of contexts, as indicated by five or more of a set of nine criteria²⁷ (see Table 1).

The DSM-5 alternative dimensional model requires for BPD the presence of moderate or greater impairment in personality functioning, manifested by difficulties in at least two of the following areas: an unstable self-image (identity); unstable goals and values (self-direction); compromised ability to recognize the feelings and needs of others (empathy); and intense, unstable and conflicted close relationships (intimacy). In addition, four or more of the seven following personality traits are required (at least one of which must be impulsivity, risk taking or hostility): emotional lability, anxiousness, separation insecurity, depressivity, impulsivity, risk taking, and hostility. Impairments in personality functioning and pathological personality traits are required to be relatively pervasive and stable²⁷ (see Table 2).

An important aspect omitted in the DSM-5 criteria for BPD is regression proneness (i.e., showing emotions or behaviors not adequate to age) in unstructured situations, one of the reasons for many of the treatment problems occurring with the disorder²⁸. Regression proneness has been empirically demonstrated by use of unstructured psychological tests such as the Rorschach or the Thematic Apperception Test (TAT)^29-32. In these tests, patients with BPD tend to show bizarre-idiosyncratic primary process thinking, usually associated with the activation of low-level defense mechanisms and object relations^31-33.

In the ICD-11, the categorical system of personality disorders has been replaced by a dimensional approach similar to the DSM-5 alternative model³⁴. Of the DSM-5 personality disorders, only BPD remains distinct and unique, by use of the “borderline pattern specifier”. In the ICD-11, a diagnostician's task is to rate the severity level of personality dysfunction as “mild”, “moderate” or “severe”. In addition, the patient may be described on five domains (negative affectivity, detachment, dissociality, disinhibition, and anankastia). While in the clinical setting most patients with BPD can be expected to be classified as having a severe personality disorder, the ICD-11 allows to rate BPD patients in whom some areas of personality functioning are relatively less affected as suffering from a moderate personality disorder³⁵.

The ICD-11 borderline pattern specifier may be applied in the presence of at least five of the following requirements: a) frantic efforts to avoid real or imagined abandonment; b) unstable and intense interpersonal relationships, which may be characterized by vacillations between idealization and devaluation; c) identity disturbance, manifested in unstable self-image; d) a tendency to act rashly in states of high negative affect, leading to potentially self-damaging behaviors; e) recurrent episodes of self-harm; f) emotional instability due to marked reactivity of mood; g) chronic feelings of emptiness; h) inappropriate intense anger or difficulty controlling anger; and i) transient dissociative symptoms or psychotic-like features. Further manifestations which may be present include a view of the self as inadequate; an experience of the self as profoundly different and isolated from other people; and proneness to rejection hypersensitivity (see Table 3).

Proposals to describe BPD by the five-factor model of personality³⁶ characterize it by high levels of both neuroticism (anxiousness, angry hostility, depressiveness, impulsiveness, vulnerability) and openness (high openness to feelings and actions), and by low levels of both agreeableness (low compliance) and conscientiousness (low deliberation)^{37, 38}. Another approach to define and conceptualize BPD focuses on major dimensions of psychopathology: most researchers agree that the dimensions which capture the essence of the disorder are emotional dysregulation, impulsivity and behavioural dysregulation, and interpersonal hypersensitivity³⁸.

With nine DSM-5 criteria and a threshold for diagnosis of five positive criteria, there are 256 theoretically possible ways to meet the criteria for BPD³⁹. Thus, despite conceptual coherence⁴⁰, BPD appears to be a heterogeneous diagnostic category which may include patient subtypes⁴¹. A cluster analysis, for example, found three clusters: a large one with “core” BPD symptoms; an extravert/externalizing one characterized by high levels of histrionic, narcissistic and antisocial features; and a small one of patients with marked schizotypal and paranoid features⁴².

Although still utilized with caution, the diagnosis of BPD in adolescents is no longer controversial. Early detection of BPD (or subthreshold features of the disorder) facilitates a timely treatment of these young patients, reducing individual suffering and societal costs⁴³. In the past, several arguments were used against BPD diagnosis prior to the age of 18, including the not uncommon occurrence of affective instability and irritation regarding self-image in adolescents, and the potential harm due to stigmatization. Today, there is a consensus regarding the potential appropriateness and usefulness of BPD diagnosis in the youth. This is also reflected by the latest developments in the ICD-11 and DSM-5^{27, 34}, where the age threshold for the diagnosis has been omitted. The diagnosis of BPD can be regarded as being as reliable and valid in adolescence as in adulthood^{44, 45}. A community-based study conducted in the US found a point prevalence for adolescents at around 1% and a cumulative prevalence of 3% up to the age of 22⁴⁶. As in adults, prevalence rates in outpatient and inpatient psychiatric settings are considerably higher^{47, 48}.

In older patients with BPD, symptoms shift to more depression, emptiness and somatic complaints^{49, 50}. Emotional dysregulation, unstable interpersonal relationships, anger and attachment insecurity persist, whereas impulsivity and identity disturbances decrease^{49, 50}. Self-harm may take other forms, such as non-adherence to medical regimes or misuse of medication⁵⁰.

Individuals with BPD are likely to have co-occurring lifetime mood disorders (83%), anxiety disorders (85%), substance use disorders (78%), and other personality disorders (53%)^51-53. BPD and bipolar I or II disorder co-occur in about 10-20% of patients with either disorder^{54, 55}. Although BPD is often comorbid with major depressive disorder or bipolar disorder, the additional diagnosis of BPD should not be made in an episode of those disorders if there is no evidence that the typical BPD symptomatological pattern persists over time.

Among people with attention-deficit/hyperactivity disorder, the lifetime rate of BPD was found to be 37.7%⁵⁶. Eating disorders are also common among individuals with BPD, with median rates of 6% for anorexia nervosa, 10% for bulimia nervosa and 22% for eating disorders not otherwise specified⁵³. Of individuals with BPD, 30% were diagnosed with post-traumatic stress disorder (PTSD), and 24% of individuals with this latter disorder were diagnosed with BPD⁵⁷.

Although there is a considerable overlap between BPD and the construct of complex PTSD (CPTSD) introduced in the ICD-11 – both disorders include problems in affect regulation, self-concept and interpersonal relationships – there is evidence that they can be empirically differentiated^{58, 59}. In particular, difficulties in affect regulation in CPTSD are ego-dystonic, stressor-specific and variable over time, whereas in BPD they tend to be ego-syntonic and persistent. Moreover, in contrast with the unstable self-concept in BPD, individuals with CPTSD have a consistently negative sense of self. Finally, the high rates of impulsivity and suicidal and self-injurious behaviors of BPD are not observed in CPTSD⁵⁹.

The above high levels of comorbidity may be an artefact of the categorical approach to psychiatric disorders, as also evidenced by the considerable overlap between BPD and the general psychopathology or p factor^60-63. It has been argued that this overlap may represent a more parsimonious way not only to explain the high “comorbidity” associated with BPD, but also its large negative impact on functioning⁶⁴.

BPD can be reliably diagnosed by semi-structured interviews. Several reliable and validated interview methods exist^65-69. In addition, self-report questionnaires and projective techniques such as the Rorschach or the TAT have proved to be helpful, especially with regard to assessing the level of personality functioning^{28, 29, 31, 32, 54} (see Table 4). Sensitive diagnostic instruments for BPD in the elderly, however, need to be developed⁵⁰.

COURSE

BPD seems to be less stable over time than traditionally believed⁵⁴. Considerable rates of recovery and relatively low rates of relapse have been reported in both short-term and long-term naturalistic follow-up studies^{54, 82}. In a 10-year prospective follow-up study, 50% of patients with BPD achieved recovery (i.e., symptomatic remission and good social and vocational functioning during the past two years), while 93% of them showed symptomatic remission lasting two years, and 86% remission lasting four years⁸². Thirty-four percent of patients lost their recovery and 30% their remission status after a two-year long remission⁸². Of note, most individuals received pharmacotherapy or psychotherapy, so that the above remission rates may not reflect the natural history of untreated BPD⁸³.

A meta-analysis of studies on the long-term course (≥5 years) of BPD reported a mean remission rate of 60%, associated with high heterogeneity between studies (I²=80.9%)⁸⁴. Excellent recovery (i.e., remission of symptoms and good social and full-time vocational functioning) was achieved in 39% of BPD patients compared with 73% in other personality disorders⁸⁵.

Patients with BPD show poorer social functioning than those with other mental disorders, including major depressive disorder and other personality disorders^{86, 87}. Only approximately 16% of people with BPD were reported to be married or living with a partner⁸⁸. Social functioning was found to be unstable and highly associated with the symptomatic status^{83, 88, 89}. Those patients who experienced change in personality pathology showed some improvements in functioning^{83, 88-91}. There is evidence that changes in personality traits (defined by the five-factor model) are followed by changes in BPD psychopathology, but not vice versa⁹². Traits were found to be more unstable in BPD than in patients with other personality disorders, indicating a “stable instability”⁹³.

BPD features tend to decline over time, and this process seems to be in part influenced by temperament⁹⁴. However, diagnostic instruments may not be sensitive enough to tap the shift in symptoms in older populations to more depression, emptiness and somatic complaints^{49, 50}.

EPIDEMIOLOGY

The age of onset of BPD varies, but symptoms are usually manifest in early adulthood²⁷. In the adult general population, rates for BPD range between 0.7 and 2.7%^{95, 96}. In primary care, psychiatric outpatients and psychiatric inpatients, prevalence rates of 6%, 11-12% and 22%, respectively, have been found^{96, 97}. In a US community sample, 2.7% of individuals had been diagnosed with BPD in their lifetime, with only slightly higher rates for women compared to men (3% vs. 2.4%)⁵². In a psychiatric outpatient setting, however, considerably higher rates of BPD were found in women compared to men (72% vs. 28%)⁹⁷. There are gender differences in comorbidity: men with BPD display more frequently substance abuse and antisocial personality disorder, while women more frequently present with mood, anxiety and eating disorders, and PTSD⁹⁸.

The rate of death by suicide is higher among individuals with BPD than in patients with other personality disorders (5.9% vs. 1.4%)⁹⁹. These results are consistent with those of a recent meta-analysis which reported suicide rates of 2 to 5% (mean 4%) over follow-up periods of 5 to 14 years among people with BPD⁸⁴. Suicide attempts occurred in more than 75% of BPD individuals¹⁰⁰.

In addition, BPD patients have a higher prevalence of somatic comorbidities – such as endocrine, metabolic, respiratory, cardiovascular and infectious (e.g., human immunodeficiency virus infection, HIV; hepatitis) diseases – than persons without BPD^{101, 102}. Mortality by non-suicide causes is clearly increased, with 14% of BPD patients and 5.5% of those with non-BPD personality disorders dying over a 24-year follow-up⁹⁹. Compared with patients without BPD who had other mental disorders or medical conditions, BPD was associated with a 2.3-fold increase in mortality rate during a 2-year follow-up¹⁰¹.

Patients with BPD die on average 14-32 years earlier than subjects in the general population⁹⁹, while some studies report lower lifetime loss (6-7 years)¹⁰¹. Loss of lifetime years is more pronounced in men¹⁰¹. Compared to individuals without BPD, men with BPD had a poorer lifetime expectancy than women with BPD, with an odds ratio (OR) of 2.40 (95% CI: 1.93-2.54) vs. 2.21 (95% CI: 2.08-2.77)¹⁰¹.

These data suggest recommending BPD patients to engage in regular medical check-ups¹⁰³. Increased health problems and associated higher mortality may reflect both unhealthy lifestyle and more direct neurobiological dysregulation of the stress and immune system associated with high levels of early adversity in BPD. Indeed, chronic physical diseases are strongly associated with “immature” personality¹⁰⁴, for which BPD may serve as a prominent example.

BPD is associated with intensive treatment utilization, and with societal costs exceeding those of anxiety and depressive disorders, diabetes, epilepsy and Parkinson's disease^{54, 87, 101, 105}. Thus, BPD constitutes a significant public health concern.

RISK FACTORS

It is currently hypothesized that, in BPD, genetic factors and adverse childhood experiences interact to influence brain development via hormones and neuropeptides^{54, 106}. Adverse childhood experiences are thought to modulate gene expression and lead to stable personality traits that may predispose to BPD⁵⁴.

There is familial aggregation of BPD^{54, 107}, with recent data from a Swedish population-based study estimating heritability at 46%¹⁰⁸. The risk of receiving a BPD diagnosis was increased 4.7-fold for full siblings¹⁰⁸. The hazard ratio in identical twins was 11.5 (95% CI: 1.6-83.3). However, no single nucleotide polymorphisms associated with BPD have been identified^{38, 109}, and some evidence of a genetic overlap of BPD with bipolar disorder, major depression and schizophrenia has emerged¹⁰⁹. Results of epigenetic studies yielded inconsistent results and are often limited by small sample size^{38, 110}. Further large scale studies that are sufficiently powered to detect effects of genes on BPD phenotype are required³⁸. In addition, more reliable measures of this phenotype are needed.

Adverse childhood experiences – including physical, sexual and emotional abuse, and neglect – are significantly associated with BPD^{111, 112}. Consistent with these findings, BPD has been associated with high levels of disorganized and unresolved patterns of attachment¹¹³. Borderline personality traits were associated with prior significant negative experiences in 12-year-old children¹⁰⁷. This effect was more pronounced when families had psychiatric histories. While multiple psychosocial factors, including maltreatment, are associated with an increased risk for BPD, these findings do not seem to be disorder-specific¹¹¹.

Inherited and environmental risk factors are thought to contribute independently and interactively to the etiology of BPD. Recent findings on familial clustering and heritability of clinically diagnosed BPD, which revealed a 54% contribution from unshared, individually unique environmental factors, point in this direction¹⁰⁸.

There is increasing evidence that BPD is associated with both early and later adversity, leading to vicious interpersonal cycles. This is, for instance, evidenced by high levels of revictimization in romantic relationships and bully-victim relationship with peers, leading to increasing levels of distrust in others and social isolation^114-118. Moreover, there is growing evidence that social deprivation and societal inequality may increase the risk for BPD, which may be related to high levels of distrust and sensitivity to social rejection and injustice in individuals with BPD^119-121. These results point to the need of considering vulnerability to BPD from a broad, socio-ecological and transactional perspective^{113, 115}.

NEUROBIOLOGY

A large number of studies have been conducted on the neurobiological underpinnings of BPD. Although several brain areas and neurotransmitters have been identified as potentially involved, only few findings have been confirmed by meta-analyses.

At the neuroendocrinological level, dysfunctions of the hypothalamic-pituitary-adrenal (HPA) axis, with altered levels of cortisol, have been suggested to underlie the impaired stress responses characteristic of BPD. One meta-analysis found significantly lower mean basal cortisol levels in individuals with BPD compared to non-psychiatric controls, with a small effect size of g=–0.32 (95% CI: –0.56 to –0.06, N=546, n=12, I²=53%)¹²². Yet, a more comprehensive meta-analysis found no significant differences in singular cortisol assessments between individuals with BPD and healthy controls or individuals with other mental disorders, although heterogeneity between studies was high and moderate, respectively¹²³. In a sub-analysis of five studies investigating continuous cortisol output, BPD patients’ cortisol response to psychosocial challenges was blunted relative to healthy controls as well as to individuals with other personality disorders¹²³. It is unclear whether disturbed HPA axis functioning is specifically associated with BPD or may rather be understood as a consequence of trauma exposure common in many psychiatric disorders¹²⁴. However, research evidence is consistent with the allostatic load hypothesis, suggesting that the blunted cortisol response in BPD reflects a compensatory down-regulation consequent to adversity and stress.

Oxytocin has been also implicated in BPD, with particular relevance for interpersonal functioning, given its purported role in attachment behavior and social cognition¹²⁵. A recent meta-analysis found decreased oxytocin levels among women with BPD (standardized mean difference, SMD=–0.46, 95% CI: –0.90 to –0.02; N=131, n=4, I²=64%)¹²⁶. However, the number of studies included was small, heterogeneity was moderate, and there were no significant differences with other personality disorders¹²⁶. Furthermore, the administration of exogenous oxytocin in BPD patients has yielded inconsistent and paradoxical effects¹²⁷. Further research is required to determine the role of oxytocin in BPD, in particular whether the observed impairments in the oxytocinergic system reflect a transdiagnostic vulnerability factor associated with early adversity and disturbed parent-infant attachment¹²⁵, or psychopathology in general¹²⁶.

In terms of neural systems, the most widely held hypothesis suggests a fronto-limbic imbalance in BPD, in which emotion dysregulation is mediated by hyperactivity of limbic structures (e.g., amygdala, hippocampus and anterior cingulate cortex) and abnormal functioning of prefrontal structures¹²⁸. However, only tentative conclusions can be drawn on the neurobiology of BPD, as most neuroimaging studies are severely underpowered¹²⁹.

The most robust meta-analytic result of neuroimaging studies in BPD is hyperactivity of the amygdala and hippocampal area during emotional processing experiments^130-132, which seems to be accompanied by impairments in habituation of the amygdala to repeated negative stimuli^133-138. While earlier meta-analyses found a reduction in hippocampal and amygdala volume in BPD^{139, 140}, a more recent and comprehensive meta-analysis reported no gray matter alterations¹⁴¹. Although the amygdala is assumed to be involved in emotional evaluation and recognition of subjectively dangerous situations, its exclusive role in processing negative emotions has recently been challenged, as studies have shown that amygdala activation is only marginally involved in the prediction of subjective fear ratings¹⁴², correlates with the experiencing of positive emotions¹⁴³, and might rather indicate saliency for faces than threats¹⁴⁴. Furthermore, despite the common conceptualization of the amygdala as the brain's “fear center”, inconsistent meta-analytic evidence has been found for its involvement in processing threats^{145, 146}. Hence, negative emotional experiencing cannot be confidently inferred from amygdala hyperactivity in BPD¹⁴⁷.

Research on abnormal prefrontal functioning lacks spatial specificity in BPD^{147, 148}, and meta-analyses have yielded conflicting results, with an earlier one finding abnormal functioning in prefrontal areas¹³¹, while the most recent and comprehensive one reported no significant differences to healthy controls¹³², although again the marked heterogeneity of BPD may be an important factor explaining inconsistent findings.

Connectivity analyses could test assumptions of reduced prefrontal top-down regulation on limbic areas such as the amygdala. However, only very few studies have investigated connectivity during emotion regulation tasks in BPD¹⁴⁹. A considerable number of studies have investigated resting-state connectivity in BPD, yielding conflicting results with respect to the fronto-limbic imbalance hypothesis^150-152.

Taken together, to date there is only weak evidence that a fronto-limbic imbalance underlies emotion dysregulation in BPD¹⁴⁷. Moreover, most neuroimaging findings lack specificity to BPD and might rather relate to transdiagnostic factors of psychopathology^{131, 153} or to childhood maltreatment^{134, 147, 154-157}. Recent research efforts point to the possible role in BPD of impairments in the temporoparietal junction¹⁵⁸, which is thought to play a crucial role in distinguishing self from other, so that its impairments might underlie the typical self-other distinction problems (i.e., identity diffusion) observed in BPD patients. However, meta-analyses are not yet available and the small number of studies preclude drawing strong conclusions.

In summary, although brain areas and neurotransmitters have been identified as potentially involved in BPD, an integrated and empirically supported neurobiological model of the disorder does presently not exist. Research on the neurobiology of BPD is complicated by several factors, including the high prevalence of comorbidities, the heterogeneity of the condition, the use of medication, as well as substantial differences in experimental designs.

SOCIAL COGNITION AND NEUROCOGNITION

Over the past decade, empirical studies on social cognition have advanced our understanding of interpersonal and emotional dysfunction in BPD. The disorder appears to be characterized by relatively severe impairments in mentalizing, i.e., the capacity to understand the self and others in terms of intentional mental states, as a result of largely affect-driven, externally-cued processing of social information. Results are not always consistent, which may be due to the type of tasks used (e.g., some social cognition tasks show ceiling effects or primarily rely on “cold” social cognition, whilst mentalizing impairments mainly tend to emerge in high-arousal contexts in BPD patients) and the influence of factors involved in the etiology of the condition (e.g., severity of trauma or attachment style).

A recent systematic review¹⁵⁹ of experimental studies on social cognition in BPD based on the Systems for Social Processes approach of the Research Domain Criteria included four meta-analyses, concerning more basic (i.e., emotion recognition accuracy and reaction time) and more complex (i.e., understanding of mental states and ostracism) features of mentalizing with regard to others. Individuals with BPD showed reduced accuracy for recognizing facial emotional expression in others compared to healthy controls, with a significant moderate effect size of g=–0.41 (95% CI: –0.57 to –0.25; n=18, I²=21%). There was no evidence for differences with respect to reaction time in detecting facial emotions (g=0.27, 95% CI: –0.04 to 0.59, n=8, I²=27%). As to the widely held hypothesis of an anger bias in BPD, the evidence of the systematic review was inconsistent, although the number of included studies was very small (n=4). Another meta-analysis found evidence for an attentional bias to negative and personally relevant negative words rather than an attentional bias towards facial stimuli¹⁶⁰.

Strong rejection sensitivity (ostracism) was found in BPD. Following perceived social exclusion, individuals with BPD experienced substantially more negative emotions and reported a greater threat to needs relative to healthy controls, with a large effect size (g=1.13, 95% CI: 0.67-1.59, n=10)¹⁵⁹. Although there was significant heterogeneity and evidence for publication bias, people with BPD showed greater levels of ostracism compared to individuals with other mental disorders (e.g., social anxiety disorder, major depressive disorder), with a medium effect size (g=0.67, 95% CI: 0.16-1.18). These findings from experimental studies are consistent with those of other meta-analyses, reporting strong expectancy of social rejection assessed by self-report in BPD compared to normal controls^{120, 161, 162}. However, heterogeneity between studies was again large, and there was evidence for publication bias.

Notably, one meta-analysis found a larger difference in negative affectivity following social inclusion (d=1.00, 95% CI: 0.76-1.25, I²=78%) than social rejection (d=0.68, 95% CI: 0.57-0.80, I²=68%) in individuals with BPD compared to non-BPD groups¹²⁰. However, heterogeneity was high and significant. Although these findings await confirmation, disturbed perceptions of both social exclusion and inclusion might be one explanation for the marked instability in close relationships in BPD. Further evidence for this comes from a meta-analysis of 26 studies on romantic attachment in BPD patients¹⁶³. The disorder was significantly correlated with attachment anxiety (r=0.48, I²=77%), but also with attachment avoidance (r=0.30, I²=74%)¹⁶³. Heterogeneity was significant. Hence, a combination of both forms of attachment difficulties might underlie BPD, which is consistent with the assumption that the disorder, and its severe cases in particular, is related to a disorganization of the attachment system characterized by strong push-pull cycles in close interpersonal relationships^{164, 165}.

The above-mentioned meta-analysis of experimental studies¹⁵⁹ also found, in BPD patients compared to healthy controls, a significantly poorer understanding of mental states in others, as assessed with Theory of Mind (ToM) tasks¹⁶⁶, with a medium effect size (g=–0.45, 95% CI: –0.75 to –0.16, n=24). However, there was high heterogeneity between studies (I²=85%). Individuals with BPD also showed greater deficits in inferring others’ mental states in comparison to people with other mental disorders, with a medium effect size (g=–0.53, 95% CI: –1.03 to –0.03). Heterogeneity was high (I²=64%). These findings are largely consistent with those of other meta-analyses of studies using ToM tasks^{167, 168}.

Moreover, in a meta-analytic evaluation¹⁶⁹, significant impairments were found in studies of mentalizing involving ToM tasks in BPD compared to healthy controls (d=0.36, 95% CI: 0.24-0.48, n=31, N=2,737, I²=50%). Deficits in mentalizing assessed by self-report were more pronounced (d=1.84, 95% CI: 1.64-2.04, n=4, N=595, I²=0%). These findings are consistent with a meta-analysis finding a strong correlation between deficits in mentalizing with regard to the self, assessed in terms of emotional awareness or alexithymia, in BPD compared to healthy controls (r=0.52, 95% CI: 0.41-0.61, n=15)¹⁷⁰.

Yet, one recent meta-analysis found evidence for a role of excessive mentalizing or hypermentalizing in BPD (r=0.26, 95% CI: 0.12-0.39, n=10), which was, however, comparable to other mental disorders¹⁷¹. Although hypermentalizing may be related to psychopathology in general rather than BPD in particular, these findings suggest that BPD is not simply associated with general deficits in mentalizing, but with a specific imbalance which can be expressed in hypomentalizing as well as hypermentalizing. This interpretation is consistent with research findings suggesting that BPD is associated with a predominance of automatic, affect-driven and largely externally-based mentalizing, with little possibility for more controlled, cognitive and internally-based mentalizing, specifically in high-arousal contexts¹⁷². However, more longitudinal research is needed, as there is evidence that mentalizing problems and BPD features reciprocally interact over time, and meta-analytic evidence for a specific mentalizing profile in BPD patients is currently lacking.

A meta-analysis of 3,543 participants¹⁷³ found that BPD symptomatology was associated with less frequent use of adaptive emotion regulation strategies (i.e., cognitive reappraisal, problem solving, and acceptance) and more frequent use of maladaptive ones (i.e., suppression, rumination, and avoidance). The role of rumination as a dysfunctional emotion regulation strategy in BPD was also confirmed by two recent meta-analyses^{174, 175}. Furthermore, a meta-analysis found stronger self-report of experienced shame in comparison to healthy controls, with a large effect size of d=1.44 (n=10, N=3,543)¹⁷⁶. However, there was significant heterogeneity and evidence for publication bias.

Lastly, there is preliminary evidence of negative self-evaluation^{159, 177}, lack of cooperation/trust^{178, 179}, impairments in self-other distinction¹⁸⁰, disturbed interoception¹⁸¹, and splitting¹⁷⁹ in BPD patients, but meta-analytic evaluations have yet to confirm these hypothesized deficits.

Deficits in neurocognition in BPD were demonstrated in a meta-analysis of 207 effect sizes across cognitive domains, reporting a medium overall effect size for impaired neuropsychological functions in BPD compared to healthy controls (d=–0.48, 95% CI: –0.58 to –0.43, N=9,332)¹⁸². However, heterogeneity was significant. The strongest impairments were found for decision making (d=–1.41, 95% CI: –0.91 to –1.91), memory (d=–0.57, 95% CI: –0.64 to –0.58), and executive functioning (d=–0.54, 95% CI: –0.64 to –0.43)¹⁹⁸. These results are in line with other meta-analyses^{183, 184}.

In summary, meta-analyses support a complex pattern of alterations in social cognition and neurocognition in BPD. The most robust findings are impairments in emotion recognition accuracy, an attentional bias towards negative stimuli, marked rejection sensitivity following social exclusion as well as inclusion, imbalances in mentalizing, dysfunctional emotion regulation, and deficits in neurocognition. Limitations are that most meta-analyses showed substantial heterogeneity, and results are often not specific to BPD. Further research is required to develop a more comprehensive understanding of the role of social cognition and neurocognition in BPD.

MANAGEMENT

As a first step of management, BPD patients need to be informed about the diagnosis, expected course, putative risk factors, and treatment options⁵⁴. Psychotherapy should be recommended as the first-line treatment, with pharmacotherapy as a possible adjunctive treatment in specific situations. Clear boundaries should be set, response to provocative behavior should be avoided, and a consistent approach should be agreed upon with all involved clinicians, in order to prevent a situation in which some of them are regarded as “bad” and others as “good”. If present, life-threatening behaviors need to be addressed first.

CONTROVERSIES

CONCLUSIONS

BPD is a common mental disorder, associated with considerable functional impairment, intensive treatment utilization, and high societal costs. The construct of BPD is internally consistent and more homogeneous than often assumed²⁶². However, it is still controversial whether BPD is better represented by a categorical or dimensional approach¹⁹. Future research is required to clarify this issue. This is also true for the elucidation of the risk factors, the neurobiological underpinnings, and the role of social cognition and neurocognition in the disorder.

With regard to treatment of BPD, pharmacotherapy is presently only recommended for severe and discrete comorbid mental disorders and for the short-term treatment of crises. Psychotherapy has proven to be efficacious in BPD¹⁷ and is recommended as first-line treatment²². With regard to the different types of psychotherapy, there is presently no reliable evidence that one method is superior to others^{17, 216}. Some differences in efficacy that were recently reported are based on a few trials²¹⁶. As a limitation, rates of non-response and relapse are relatively large²³. Thus, psychotherapy needs to be further improved.

Future studies of psychotherapy in BPD are recommended to focus on patients at risk of non-response and on improving long-term effects, as well as on reducing self-harm behavior and suicidal ideation²⁶³. Taking the high dropout rate into account²³⁴, an-other focus should be on patients prematurely terminating treatments. By studying dropouts, researchers can learn which aspects of a treatment are experienced by patients as not beneficial or even harmful, and in which way treatments may be improved. Thus, patients who drop out of a treatment can provide important information²⁶⁴.

As another limitation, the quality of psychotherapy studies was found to be modest^{17, 216}. Further high-quality studies are required, in both adults and adolescents. Taking the shift from categorical to dimensional concepts into account²⁰, research on psychotherapy of BPD (and of personality disorders in general) needs to take dimensional outcome measures (e.g., Level of Personality Functioning Scale²⁷), as well as personality traits, into account. Treatment research on dimensionally defined (severe) personality disorders is required²⁶⁵.

In addition, high-quality head-to-head comparisons of the major forms of psychotherapy with a sufficient statistical power, adequate treatment implementation, and control of bias and researcher allegiance are needed. Such trials may also examine presumed mechanisms of change. For these head-to-head comparisons, proponents of each approach need to be included on an equal basis (adversarial collaboration)²⁶⁶. Funding organizations are encouraged to support these comparative trials, since large samples may be required to detect small but clinically significant differences, implying considerable study costs. As the differences in efficacy between the major psychotherapeutic approaches do not seem to be substantial at the group level^{17, 216}, identifying what works for whom seems to be a promising strategy. Individual participant data meta-analysis may be helpful in this regard²¹⁶.