The Complexity of Borderline Personality Disorder: Network Analysis of Personality Factors and Defense Styles in the Context of Borderline Personality Organization
Article information
Abstract
Objective
Borderline personality disorder (BPD) is known to share characteristics with a variety of personality disorders (PDs) and exhibits diverse patterns of defense mechanisms. To enhance our understanding of BPD, it’s crucial to shift our focus from traditional categorical diagnostics to the dimensional traits shared with other PDs, as the borderline personality organization (BPO) model suggests. This approach illuminates the nuanced spectrum of BPD characteristics, offering deeper insights into its complexity. While studies have investigated the comorbidity of BPD with other PDs, research exploring the relationship between various personality factors and defense mechanisms within BPD itself has been scarce. The present study was undertaken to investigate the complex interrelationships between various personality factors and defense styles in individuals diagnosed with BPD.
Methods
Using a network analysis approach, data from 227 patients diagnosed with BPD were examined using the Defense Style Questionnaire and Personality Disorder Questionnaire-4+ for assessment.
Results
Intricate connections were observed between personality factors and defense styles. Significant associations were identified between various personality factors and defense styles, with immature defense styles, such as maladaptive and image-distorting being particularly prominent in BPD in the centrality analysis. The maladaptive defense style had the highest expected influence centrality. Furthermore, the schizotypal, dependent, and narcissistic personality factors demonstrated relatively high centrality within the network.
Conclusion
Network analysis can effectively delineate the complexity of various PDs and defense styles. These findings are expected to facilitate a deeper understanding of why BPD exhibits various levels of organization and presents with heterogeneous characteristics, consistent with the perspectives proposed by the BPO.
INTRODUCTION
Borderline personality disorder (BPD) is a complex disease characterized by distorted self-image, high levels of impulsivity, and notably unstable interpersonal relationships and emotional states [1]. The onset of BPD is influenced by multiple factors, among which genetic predisposition and stress exposure during childhood are recognized as significant contributors [2]. Additionally, BPD represents a considerable societal burden which can be attributed to an array of factors, including elevated suicide rates, severe functional impairments, and a high prevalence of comorbid conditions [3]. However, the diagnosis and treatment of BPD pose substantial challenges owing to the heterogeneous nature of its symptoms, which manifest across a broad spectrum [4]. Owing to the distinct characteristics of BPD, there is increasing advocacy for shifting from the traditional Diagnostic and Statistical Manual of Mental Disorders (DSM)-based categorical diagnosis to a more dimensional approach that considers the unique factors of personality disorders (PDs) [5]. From a dimensional perspective, the diverse clinical manifestations of BPD often exhibit characteristics that are shared with other PDs. Notably, one study highlighted that approximately 80% of individuals diagnosed with BPD also presented with at least one additional co-occurring PD [6]. The most frequently co-occurring PDs in individuals with BPD include paranoid, passive-aggressive, avoidant, and dependent PDs. Additionally, sex-based differences have been observed; women with BPD are more likely to have comorbid-dependent PD, whereas men are more likely to have comorbid antisocial PD [7]. Individuals with BPD who also have co-occurring PDs, such as avoidant, dependent, narcissistic, and histrionic PDs, have a relatively poor prognosis and exhibit heightened severity of clinical symptoms [8,9]. Consequently, the integration of dimensional characteristics, including the identification of factors related to other PDs, offers valuable insights in BPD. This approach not only aids in the assessment of symptom severity but also plays a crucial role in formulating effective treatment strategies.
Defense styles are indicative of ego functioning and offer insights into an individual’s psychological well-being and personality characteristics. Extensive research has indicated that individuals with BPD are more inclined to employ immature defense styles, such as maladaptive action and image distortion, than those with other PDs [10-12]. Projective identification and splitting have traditionally been cited as the hallmark defense mechanisms in BPD. However, Zanarini et al. [10] have identified alternative primary defense mechanism in BPD, characterized by impulsivity and demanding traits, namely, acting out, emotional hypochondriasis, and undoing. Furthermore, another study revealed a significant correlation between the emotional instability characteristic of BPD and the use of defense mechanisms such as undoing, acting out, passive aggression, projection, and splitting [13-15]. Additionally, understanding defense types can aid in comprehending the characteristics of BPD, making it useful in psychotherapeutic approaches for patients [16]. Kernberg [17] suggested that BPD, exhibiting symptoms of diverse psychiatric conditions, including depression, anxiety disorders, psychotic episodes, bipolar disorder, and substance abuse, can be termed polysymptomatic neurosis. Kernberg [17] also noted that BPD encompasses a range of personality traits that span the spectrum between neurosis and psychosis. In this context, Kernberg [17] proposed the borderline personality organization (BPO) model, which classifies PDs based on the level of severity and integration of the personality structure, rather than distinct categorical disorders [18].
Although various studies, including the BPO model, have reported an association between BPD, various PDs, and defense styles, studies analyzing the relationships between these factors are limited [19]. Although research has explored the comorbidity of BPD with other PDs, studies specifically investigating how BPD’s characteristic features coexist with various other personality factors within the disorder itself remain scarce. Furthermore, there is a lack of investigation into the nature of these personality factors, their relationship with defense styles, and the organization levels they exhibit within the context of BPD. Due to the complicated interactions between personality factors and defense styles, conventional analytical methods may be inadequate to fully understand these relationships [20]. The network model, a comparatively recent development in analytical approaches, visually presents interconnections among various elements in a network format. This aids in elucidating the complex interrelations among multiple factors [21]; therefore, we aimed to use the network model to comprehensively investigate the complex relationships between various personality factors and defense styles in BPD, which were difficult to explain using previous methods. We aimed to explore the existence of various personality factors within patients with BPD and examine their association with defense styles. Through this examination, our goal was to uncover the existence of various levels of organization within BPD, in line with what is proposed by the BPO model.
METHODS
Study design and participants
This study is part of a research project aimed at understanding the relationships and roles of defense mechanisms and personality factors across various psychiatric disorders. It utilizes data retrospectively collected from patients diagnosed with PDs among those visiting the outpatient department of psychiatry at a general hospital, including demographic information and psychological scales. Among 3,596 patients whose data were collected between October 2006 and March 2019, we enrolled 227 patients diagnosed with BPD according to the DSM-IV-TR or DSM-5 criteria. Patients diagnosed with other PDs, intellectual disabilities, autism spectrum disorders, psychotic disorders, psychiatric conditions secondary to organic causes, or substance use, were excluded from the study. Due to the high comorbidity associated with BPD, participants with psychiatric diseases present as comorbidities, excluding those outlined in the exclusion criteria, were included in the study. The initial diagnoses were established through clinical evaluations performed by a board-certified psychiatrist with extensive experience. These initial assessments were subsequently reexamined through a comprehensive review of the medical records in a research meeting involving five psychiatrists. Final diagnosis and patient inclusion were confirmed during this collaborative session. Patients who voluntarily underwent the requisite psychological assessments were selected from the Department of Psychiatry at a university hospital in a metropolitan area. Among the diverse psychological assessments collected retrospectively, including the Minnesota Multiphasic Personality Inventory-2 and the Symptom Checklist-90-Revised were included. Specifically, for the purposes of this research, we utilized the Personality Disorder Questionnaire-4+ (PDQ-4+) and the Defense Style Questionnaire (DSQ) in our analysis.
Ethics
All studies involving participants were approved by the Institutional Review Board (IRB) of the Yeungnam University Hospital (IRB No. 2015-12-020) were conducted in accordance with local legislation and institutional requirements. This study was conducted retrospectively, and informed consent was not obtained from the participants.
Psychological assessment
DSQ
The DSQ is a self-report questionnaire developed by Bond and Vaillant [22] which is used to identify subjects’ defense styles. The questionnaire comprises 88 items, each evaluated on a scale ranging from 1 (not at all) to 9 (completely agree). The DSQ measures defense types through individual elements associated with defense styles, including neurotic denial, passive aggression, non-paranoid projection, acting out, splitting, projective identification, omnipotence, undoing, intimacy, somatization, and health worries. Additionally, the DSQ classifies individual defense styles into four maturity-based categories: maladaptive, image-distorting, self-sacrificing, and adaptive. It also allows for the measurement of scores in each category. The DSQ has been utilized as a valid and reliable tool in research, as evidenced in studies by Bond and Vaillant [22] and Chung et al. [23] In this study, we employed only the overarching defense styles from the DSQ, intentionally excluding specific defense mechanism items. This approach was chosen for two primary reasons: firstly, to avoid the potential reduction in network stability that could result from incorporating an excessive number of nodes, and secondly, because the variability in findings across existing studies concerning individual defense mechanisms—specific to PDs or psychiatric diagnoses—made it challenging to establish consistency. Furthermore, the objective of this research was not to delineate individual defense mechanisms, but rather to investigate the relevance of the organization level of PDs, in a manner similar to the BPO model.
PDQ-4+
The PDQ-4+ is a self-report assessment tool developed by Hyler et al. [24] which employs the DSM-IV diagnostic criteria to evaluate and identify 10 different factors of PDs. In this study, we utilized a modified and standardized version of the PDQ-4+ adapted by Kim et al. [25] This version is capable of assessing 12 PDs, including passive-aggressive and depressive PDs, which are additional diagnoses included in the DSM-IV Appendix. The PDQ-4+ comprises 99 items, each contributing one point to the total score. A score of 30 or higher suggests the potential presence of PD, whereas most individuals in the general population typically score less than 20 points. A higher total score indicated an increased likelihood of PD. For specific PD types, a score ranging from a minimum of three to five points or more is indicative of a specific PD.
Network forming and statistical analysis
Prior to conducting network analysis, the four maturity defense styles of the DSQ and the 12 PD scales of the PDQ-4+ were normalized to a Gaussian normal distribution using the “huge.npn” function of the “huge” package in R [26]. To establish optimized parameters in the network formation, the network model was generated utilizing the “ggmMODSelect” function from the “qgraph” package [27] The “ggmMODSelect” function constructs 100 unregularized graphical lasso network models, subsequently selecting the model that yields the highest Extended Bayesian Information Criterion value. In the developed network, each node represents a single variable representing a psychological trait. The edges linking these nodes represent the partial correlation values conveyed through their thickness and clarity, thereby illustrating the key interrelationships among the variables. As such, the more pronounced and clearer the edges in the visualized network, the stronger the implied association between them. Furthermore, in the visualized network, the blue edges denote positive relationships, whereas those in red represent negative relationships. Given that the network’s nodes embody heterogeneous traits, the expected influence was employed to assess node centrality. This measure is less influenced by the anticipated polarity of edges [28]. The stability of the network was assessed using the “bootnet” function of the “bootnet” package in R software [29]. To assess the stability of network edge weights and Expected Influence centrality, the “bootnet” function employed bootstrapping (n=1,000) techniques to compute the correlation stability coefficient. Correlation stability coefficient reflects the network’s robustness by indicating the correlation between edge values of the network generated after dropping a subset of samples from the total sample and those of the original network. Typically, it is advised not to interpret networks with a correlation stability coefficient lower than 0.25, due to network instability.
RESULTS
Demographic data and psychological assessment
Demographic analysis of the 227 participants revealed a mean age of 31.72 years (standard deviation [SD]=12.12 years), with a sex breakdown of 87 males and 140 females. For the DSQ defense maturity scales, the average scores were as follows: 4.78 (SD=1.72) for maladaptive, 4.27 (SD=1.17) for image-distorting, 3.92 (SD=1.45) for self-sacrificing, and 4.18 (SD=1.27) for adaptive. The mean scores on the PDQ-4+ for the 12 PD scales were measured as follows: paranoid PD at 4.08 (SD=1.72), schizoid PD at 3.29 (SD=1.71), schizotypal PD at 4.05 (SD=2.24), antisocial PD at 2.40 (SD=1.89), BPD at 4.88 (SD=1.80), histrionic PD at 3.89 (SD=1.76), narcissistic PD at 3.60 (SD=2.10), avoidant PD at 4.10 (SD=2.03), dependent PD at 4.23 (SD=2.24), obsessive-compulsive PD at 3.84 (SD=1.66), depressive PD at 4.11 (SD=1.85), and negativistic PD at 3.69 (SD=1.91). Detailed demographic data and psychological assessment results are presented in Table 1.
Demographic data and psychological assessment of borderline personality group
Network result
Among the 120 potential edges that could be established between the 16 nodes derived from the personality factors and the maturity levels of defense styles, 35 significant edges were identified. The average weight of the developed edges was calculated as 0.055. The network between the defense styles and personality factors of the 35 significant edges is presented in Figure 1. Among the identified significant edges, 31 had positive weight values with an average weight of 0.24. Conversely, four of these significant edges possess negative weight values, with an average of -0.20. The most strongly associated edge among the nodes was found between the self-sacrificing and adaptive defense styles with a weight value of 0.41. Furthermore, the edge linking maladaptive and image-distorting defense styles, with a weight of 0.40, and the edge between the avoidant and dependen t PD scales, weighing 0.38, exhibited higher weight values than those of the other edges.
Network of personality factors and defense styles in BPD. Each circle (node) represents a personality factor or defense style, while lines (edges) indicate relationships between nodes. These relationships are represented by weight values within the network structure, based on partial correlations. Thicker and darker edges signify stronger relationships. Blue edges denote positive relationships, and red edges represent negative ones. BPD, borderline personality disorder; PD, personality disorder.
The relative values (z-scores) of the expected influence centrality within the network are illustrated in Figure 2. The maladaptive defense style (z-score=1.34) exhibited the highest strength centrality. Furthermore, schizotypal PD (z-score=1.28), dependent PD (z-score=1.13), narcissistic PD (z-score=1.08), and image-distorting defense style (z-score=1.00) demonstrated a higher expected influence centrality than the other scales. The stability of correlation coefficients pertaining to the network edge weights was observed at 0.28, as illustrated in Figure 3A, while the calculated value for expected influence centrality was determined to be 0.52, as depicted in Figure 3B.
Expected Influence centrality of personality factors, defense styles, and network in BPD. BPD, borderline personality disorder; A1, paranoid PD; A2, schizoid PD; A3, schizotypal PD; B1, antisocial PD; B2, BPD; B3, histrionic PD; B4, narcissistic PD; C1, avoidant PD; C2, dependent PD; C3, obsessive-compulsive PD; D1, maladaptive style; D2, image-distorting style; D3, self-sacrifice style; D4, adaptive style; N1, depressive PD; N2, negativistic PD.
Correlation stability of defense styles and personality network in BPD. A: Edge correlation stability of network in BPD. B: Expected Influence centrality correlation stability of the network in BPD. The red line indicates the correlation value from bootstrapping, and the surrounding opaque area represents its 95% confidence interval. Correlation stability reflects the network’s robustness by indicating the correlation between edge values of the network generated after dropping a subset of samples from the total sample and those of the original network. BPD, borderline personality disorder.
DISCUSSION
This study utilized network analysis to show that various personality factors and defense styles are interconnected in a network-like structure. Although numerous studies have established that personality factors are not independent but interrelated. However, most previous research has either categorized similar PDs into clusters for analysis or focused on examining the relationships between individual PD traits [30]. Consequently, the findings of this study are expected to facilitate a more intuitive understanding of the intricate interplay between various PDs and defense styles within one BPD. These characteristics of BPD appear to reflect the disorder’s sharing of various personality factors.
The most strongly associated edge among the nodes was identified between self-sacrificing and adaptive defense styles. Furthermore, the edge linking maladaptive and image-distorting defense styles and the edge between avoidant and dependent PD exhibited higher weight values than other edges. Such strong connections between nodes are indicative of similar shared characteristics. Self-sacrificing and adaptive defense styles are known to be relatively mature and associated with neurotic tendencies, whereas maladaptive and image-distorting defense styles are more primitive and linked to severe tendencies [31]. These trends have also been observed within BPD in our results.
When examining the relationships among the PDs uncovered in this study, we noted that most traits within the same cluster established significant edges, suggesting shared characteristics among the PDs in that cluster. Interestingly, obsessive-compulsive PD did not establish significant edges with other disorders in cluster C, and paranoid PD did not form significant edges with other disorders in cluster A. Conversely, it was also noted that significant edges emerged between PD traits traditionally categorized in different clusters. For example, the paranoid PD in cluster A formed a significant edge with the obsessive-compulsive PD in cluster C. In the case of schizoid PD, significant edges were formed with narcissistic, dependent, and antisocial PD. Given these findings, employing network analysis is anticipated to enhance our understanding of the diverse characteristics of PDs, which may not be readily discernible through traditional cluster-based analyses. Within individuals with BPD, various dimensions of personality factors exist, each displaying correlations that are both similar to and distinct from traditional clusters. This suggests the presence of various characteristics within the disorder.
In the centrality analysis using the expected influence centrality, the maladaptive defense style was identified as the node with the highest centrality. Numerous studies have identified maladaptive defense styles as a characteristic of BPD [10,12]. Notably, it has been recognized that this maladaptive defense style is linked with various aspects of BPD, including mentalization [32]. Several studies have reported a correlation between immature defense mechanisms and impairments in mentalization, with Shahar and colleagues [33] specifically noting that this association strengthens in situations with a high emotional burden. Considering the results of these previous studies, the high centrality of maladaptive defense styles revealed in our study suggests that it plays a significant role in the core pathology of BPD.
Schizotypal PD was identified as having the highest centrality among all the personality factors. Schizotypal personality factors have further been known to correlate with the severity of psychopathology, a relationship documented in both the general population and in groups diagnosed with a range of PDs [34] Through the framework of the BPO, Kernberg [17] distinguished PDs not as discrete categories, but rather based on a continuum of severity and functional level. Within this framework, he classified PDs with psychotic traits, such as schizoid PD, as the most severe form of PD [17]. Considering previous studies suggesting that schizotypal PD traits influence the severity of PD pathology regardless of diagnosis, and the high centrality of schizotypal PD, we discovered that it could act as a hub PD, affecting the severity of various PDs [34]. Furthermore, the high centrality of dependent PD observed in this study may be indicative of its association with one of the vulnerable characteristics of patients with BPD, specifically their sensitivity to the dynamics of interpersonal rejection [35,36]. Furthermore, this result suggests a potential link between the core features of BPD and the tendencies observed in patients with dependent PD. Additionally, the high centrality of narcissistic PD may be related to the dynamically low self-esteem and narcissistic vulnerabilities often exhibited by patients with BPD, suggesting a connection between narcissistic PD and its underlying psychological patterns in BPD [37,38].
Network analysis enabled the identification of various interrelationships between defense styles and personality factors. The adaptive defense style, identified as the most mature defense style, established a positive edge with obsessive-compulsive PD, a member of cluster C. Additionally, the self-sacrificing defense style formed a positive edge with dependent PD, which was also categorized within cluster C. Conversely, the maladaptive defense style, which belongs to the immature defense type, formed positive edges with schizoid PD from cluster A, BPD from cluster B, and avoidant PD from cluster C. The image-distorting defense style formed positive correlations with antisocial PD and narcissistic PD, both of which belonged to cluster B. The mature defense types predominantly established positive edges with the traits of cluster C PDs. This pattern suggests that cluster C PDs, reflecting characteristics of relatively less severe psychopathology than clusters A and B, may also exhibit the neurotic features associated with BPO [39].
Network analysis revealed that BPD was characterized by complex interactions between various defense styles and personality factors. In particular, immature defense styles, such as the maladaptive and image-distorting, appear to play a critical role in these interactions. The prominence of immature defense styles in BPD may partly explain the characteristic social and interpersonal difficulties, along with the range of symptoms and impulsive behaviors commonly associated with the disorder. BPD encompasses a spectrum of personality factors, from severe psychotic personality levels to neurotic personality levels. The diverse spectrum of personality factors within BPD supports the classification of borderline characteristics into various organization levels within the BPO framework, such as low BPO, high BPO, and neurotic levels, indicating a nuanced approach to understanding the disorder [17]. These diverse personality dimensions are interlinked within a network, and this interplay may illuminate the complexity and varied symptomatology of BPD. These characteristics suggest that BPD, much like the concept of BPO describes BPD as a condition with a spectrum of various levels, is a disorder with multidimensional features.
The limitations of this study are its exclusive focus on a single group diagnosed with BPD, which limits the ability to conclusively attribute the observed relationships between personality factors and defense characteristics as unique to BPD. Hence, future research employing the same methodological approach should explore these relationships in both a general healthy control group and in groups with different PDs. Due to the characteristics associated with BPD comorbidity, despite excluding conditions outlined in the exclusion criteria, the majority of participants were found to have various Axis I psychiatric disorders, including depressive and anxiety disorders. Such comorbid disorders could potentially affect the outcomes of the PDQ and DSQ assessments and the relationships between them. Additionally, future studies should not only focus on BPD but also include all PDs, aiming to investigate the correlations between organization levels and personality factors as delineated in the BPO theory. This study carries the typical limitations inherent in retrospective research. Consequently, it is recommended that future research should involve prospective studies utilizing structured interviews. Furthermore, while organic brain syndrome, substance use disorders, and schizophrenia were excluded, the presence of various comorbid conditions, such as depression, anxiety, in BPD means that such comorbidities could affect the outcomes. Despite these limitations, the significance of this study lies in its exploration of personality factors and defense styles through a network analysis of a relatively large sample of patients with BPD. This approach sheds light on the complex interplay between various personality factors and defense styles in BPD, affirming theories such as Kernberg’s BPO, which emphasizes the multifaceted nature of BPD. This understanding underscores the intricate dynamics involved in BPD, and offers valuable insights for future research and treatment methodologies.
Notes
Availability of Data and Material
Data, analytic codes, and other underlying research materials are available from the first author by request.
Conflicts of Interest
The authors have no potential conflicts of interest to disclose.
Author Contributions
Conceptualization: Bon-Hoon Koo. Data curation: Seokho Yun, So-Hye Jo, Hye-Jin Jeon, Hye-Geum Kim, Eun-Jin Cheon. Formal analysis: Seokho Yun. Hye-Jin Jeon. Funding acquisition: Bon-Hoon Koo. Investigation: SoHye Jo, Hye-Jin Jeon, Hye-Geum Kim, Eun-Jin Cheon, Bon-Hoon Koo. Methodology: Seokho Yun, Bon-Hoon Koo. Project administration: BonHoon Koo. Resources: Bon-Hoon Koo. Software: Seokho Yun. Validation: Seokho Yun, Bon-Hoon Koo. Visualization: Seokho Yun. Writing—original draft: Seokho Yun, Hye-Jin Jeon. Writing—review & editing: Hye-Jin Jeon, Hye-Geum Kim, Eun-Jin Cheon, Bon-Hoon Koo.
Funding Statement
This work was supported by the 2020 Yeungnam University Research Grant.
Acknowledgements
The authors would like to thank all the participants of this study.