Comorbidity of borderline with antisocial and narcissistic personality disorders: a multimethod study Authors: Banafsheh Mohajerin, Parham Mohammadi, Richard Howard Published: 2026 ABSTRACT AbstractBackgroundThe construct of borderline personality disorder (BPD) remains contentious, with some arguing that it embraces too much pathology to be of any real value as a clinical diagnosis. Examining BPD’s comorbidity is key to understanding its heterogeneous vs. homogeneous nature.MethodsThis study aimed to uncover differences in patterns of symptoms and their severity between 3 groups of patients a with a DSM-5 diagnosis of BPD: (i) those with a single borderline diagnosis; (ii) those with a dual diagnosis of BPD and antisocial personality disorder (ASPD); those with a dual diagnosis of BPD and narcissistic personality disorder (NPD). Groups were compared in terms of: (i) borderline symptom severity; (ii) self-report measures of emotional dysregulation, suicidality, and identity disturbance; (iii) DSM-defined borderline symptom profiles. Network analysis assessed whether a stable and reproducible structure of symptoms could be seen across all three groups and whether groups differed in symptom prevalence.ResultsResults of network analyses showed that the three groups shared the same symptom structure but differed systematically in which symptoms were most strongly endorsed: identity/interpersonal dysfunction in BPD, affective/behavioural dysregulation in BPD+ASPD, and a mixed/intermediate profile in BPD + NPD. BPD+ASPD showed the greatest severity of borderline symptoms, greatest emotional dysregulation, and highest impulsiveness and anger. The BPD group showed the highest identity disturbance and suicidality, BPD + NPD the lowest. In their symptom profiles, BPD were more likely to meet criteria for fear of abandonment, unstable relationships, identity disturbance and suicidality, while BPD+ASPD were more likely to meet criteria for impulsivity, affective instability, and anger control difficulty. The three groups differed on measures of identity and interpersonal relationships, supporting the idea that PD reflects disturbances of relationality.ConclusionResults highlight that BPD can manifest in different ways that depend on its comorbidity. These differences require treatment to be tailored to the individual patient, with a focus on identity/interpersonal dysfunction when BPD presents alone, and on affective/behavioural dysregulation when BPD co-occurs with ASPD.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40479-026-00349-z.Keywords: Borderline personality disorder, Comorbidity, Antisocial personality disorder, Narcissistic personality disorder, Network analysis Background Borderline personality disorder (BPD) is a complex and highly heterogeneous mental disorder whose status as a coherent and valid construct continues to be highly debated, with different clinicians and researchers defining the construct—or emphasizing certain facets of it—in their own ways [18]. It is often associated with psychiatric comorbidities that exacerbate symptoms and complicate management [16]. In the latest (11th) iteration of the International Classification of Disorders (ICD-11: World Health Organization, [64]), BPD is neither a diagnosis nor part of the central classification but has been retained as a ‘pattern specifier’. Some have argued that it is misclassified as a personality disorder and properly belongs under the rubric of emotional disorders [39, 53]. Others have argued that it embraces too much pathology to be of any real value as a clinical diagnosis [42, 63]. Echoing the severity-based classification of ICD-11, Watson and Clark, [63] argued that patients diagnosed with BPD have such serious problems, across such a broad range of psychopathology, that any differences observed when borderline patients are compared with others primarily reflect quantitative differences in severity, not BPD per se. This chimes with a psychodynamic perspective, which argues that something fundamental about personality pathology is captured by its severity [17]. Viewed through a psychodynamic lens, borderline personality organization is a broad structural category based on identity pathology, encompassing all severe personality disorders—not just those meeting BPD criteria [9]. BPD shows “rampant comorbidity” with other personality disorders (PDs), particularly antisocial and narcissistic PDs [63]1. Its highest comorbidity is with antisocial PD, between 5.6% and 27% in community samples, more than 50% in clinical and forensic settings [10, 60], and 60–70% in forensic patients whose PD is deemed “dangerous and severe” [12]. Individuals with both ASPD and BPD demonstrate a significantly greater risk of violent crime than those with ASPD alone [52]. Thus, when comorbid with ASPD, BPD represents a particularly severe and criminogenic constellation of features, representing a severe challenge to treatment efforts [41]. BPD also commonly co-occurs with narcissistic personality disorder (NPD), characterized by a pervasive pattern of grandiosity, feeling privileged, entitled, and expecting preferential treatment, as well as displaying an exaggerated sense of self-importance, and showing arrogant and haughty behaviors or attitudes [1]. Comorbidity of BPD and NPD has been reported as between 17% and 80% in the adult population [59, 62]. It is increasingly recognized that the severity of NPD can range from mild (personality difficulty) to severe [21, 48]. The present study The present study extends previous studies that compared patients with a single (NPD or BPD) diagnosis with those having a dual (NPD + BPD) diagnosis [23, 51]. The former study reported that their NPD + BPD group were much more likely to show co-occurring antisocial PD and any Cluster B PD, suggesting greater PD severity in the comorbid group. Hörz-Sagstetter et al., [23] compared patients with a single BPD diagnosis with those having a dual (BPD + NPD) diagnosis. Comorbidity was associated with more criteria of borderline, histrionic, paranoid, and schizotypal PDs, again suggesting a link between comorbidity and PD severity. Building on these studies, the present study compared three groups with a roughly equal distribution of males and females: (i) a group with a single diagnosis of BPD; (ii) a group with BPD co-occurring with narcissistic personality disorder (BPD + NPD); (iii) a group with BPD co-occurring with antisocial PD (BPD+ASPD). Network analysis Previous researchers have emphasized the importance of embracing novel methodological and theoretical frameworks in the field of comorbidity research (e.g. Köhne and Isvoranu, [31]. Network analysis, a relatively recent technique that visually represents and quantifies connections between symptoms, has been used to identify core symptoms of BPD in adult and adolescent samples [31, 46, 50]. Results have supported conceptualizations of BPD that place either affective instability or identity disturbance as central to the disorder [46, 56]. We employed two network estimation method (see Methods below) to investigate the conditional dependency structure among the BPD criteria across BPD, BPD+ASPD, and BPD + NPD and the relative prevalence of symptoms across groups. Self-report measures of borderline symptoms We used measures of borderline symptom severity, emotion regulation, anger regulation, impulsivity, identity disturbance, and self-harm (see Methods below). Given the very criminogenic nature of BPD+ASPD [15, 26], we expected that this comorbidity, compared with BPD + NPD, would be associated with a greater severity of borderline symptoms and greater emotion dysregulation. Symptom profiles We examined DSM-defined symptom profiles to determine the extent to which they differed between the three groups. Further, and in keeping with a recent reframing of personality disorders as “relational disorders”, involving a fundamental disturbance of ‘self’ and ‘other’ relationships [24, 66], we examined DSM-defined symptom profiles to see whether different blends of self/other dysfunction were associated with BPD when it occurred alone compared with when it co-occurred with antisocial or narcissistic PD. Method Participants Participants gave their informed consent before participating in the study, which was approved by the first author’s university ethics committee. Potential participants were recruited by referral from private and public psychology and psychiatry clinics and through invitations posted on various social media platforms, including Instagram and Telegram. Participants who were eligible to participate then completed a face-to-face interview to determine whether they met at least 5 of 9 criteria required for a BPD diagnosis in DSM-5 [1] and were between 18 and 65 years old. The following exclusions were applied: Participants were excluded who met the criteria for a bipolar disorder or schizophrenia spectrum disorder assessed by SCID-5 (SCID-5, [14]) or a substance use disorder diagnosis [1], assessed by urine drug screening (UDS) or alcohol use disorder identification test (AUDIT, [55]). Participants were excluded who met the criteria for a bipolar disorder or schizophrenia spectrum disorder assessed by SCID-5 (SCID-5, [14]) or a substance use disorder diagnosis [1], assessed by urine drug screening (UDS) or alcohol use disorder identification test (AUDIT, [55]). Two participants were excluded who, within a 3-month period prior to the study, had received psychotropic medications (risperidone, quetiapine, and clozapine). Since these medications typically target symptoms such as anger, aggression, and impulsiveness, they could have impacted scores of the Impulsive Behavior Scale and the State-Trait Anger Expression Inventory–2 (UPPS-P and STAXI-2 respectively, see details below). Two participants were excluded who, within a 3-month period prior to the study, had received psychotropic medications (risperidone, quetiapine, and clozapine). Since these medications typically target symptoms such as anger, aggression, and impulsiveness, they could have impacted scores of the Impulsive Behavior Scale and the State-Trait Anger Expression Inventory–2 (UPPS-P and STAXI-2 respectively, see details below). To preserve relatively “pure” symptom profiles of participants, we excluded those who, within the previous year, had received any type of psychotherapy that lasted more than 2 months. The rationale for this is that some types of short-term psychotherapy lasting more than two months have shown effectiveness in reducing BPD symptoms [27, 32, 54]. To preserve relatively “pure” symptom profiles of participants, we excluded those who, within the previous year, had received any type of psychotherapy that lasted more than 2 months. The rationale for this is that some types of short-term psychotherapy lasting more than two months have shown effectiveness in reducing BPD symptoms [27, 32, 54]. The assessment was carried out by four clinical psychologists and one psychiatrist. After a BPD diagnosis was confirmed, a doctoral-level psychologist, who was not blind to the study aims, checked that the inclusion and exclusion criteria were fulfilled. All participants completed a battery of self-report measures (see below for details) in a single session. A clinical psychologist with a master’s degree who was blind to the study aims checked that all questions had been answered. Clinical diagnosis Structured Clinical Interview for DSM-5 Personality Disorders (SCID-5-PD; [13]) is a widely used semi-structured interview to assess personality disorders DSM-5. It has acceptable psychometric properties [6]. In the current study, we used the Persian version, which has good reliability and validity [19], and its kappa score for Cluster B PDs ranged from 0.75 to 0.82. The Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders (SCID-5; [14]) is a semi-structured interview used to assess DSM-5 diagnoses. It has good psychometric properties [45]. We used its Persian version [57], which has acceptable validity and reliability (0.71 < kappa < 0.88). Self-report measures Alcohol Use Disorder Identification Test (AUDIT, [55]), a 10-item self-report questionnaire that is widely used to screen for excessive alcohol consumption and drinking problems. We used the Persian version [49], which has acceptable psychometric properties (α = 0.75). Borderline Symptom List (BSL-23; [7]) is a self-report measure used for assessing BPD symptom severity within the past week. It consists of 23 items to which participants responded using a five-point Likert scale. It has strong psychometric properties [7]. We used its Persian version which has good psychometric properties [43] and in this study showed excellent internal consistency (α = 0.92). Following the methodology used by Kleindienst et al., [30], mean BSL-23 scores were subdivided into the following six severity categories: i) none / low: 1.52 (2 SD below the mean); b) mild: 1.93 (1 SD below the mean); c) moderate, equal to mean source (x̄ =2.35); d) high: 1.93 (1 SD above the mean); very high 1.52 (2 SDs above the mean); extremely high (=/>1.53). The Difficulties in Emotion Regulation Scale (DERS, [20]) is a 36-item self-report questionnaire for assessing emotion regulation difficulties on a 5-point rating scale. In the current study, we used its Persian version [36] whose internal consistency was good = 0.85). State-Trait Anger Expression Inventory–2 (STAXI-2, [58]) is a 57-item self-report measure widely used to assess state anger, trait anger, and anger expression using a four-point Likert scale. In this study, we used its Persian version [2] whose internal consistency is acceptable (α = 0.78). For assessing anger, we used four subscales that indicate anger expression-in (AX-In, α = 0.79) and anger expression-out (AX-Out, α = 0.91), anger control-in (AC-I, α = 0.92) and anger control-out (AC-O, α = 0.85), UPPS-P Impulsive Behavior Scale (UPPS-P, [34]) is a widely used 59-item self-report questionnaire that comprises five subscales: Negative Urgency, Positive Urgency, (Lack of) Perseverance, (Lack of) Premeditation, and Sensation Seeking. We used its Persian version [3] that has excellent internal consistency (α = 0.85). Self-concept and Identity Measure (SCIM, [28]) is a widely used self-report scale comprising 27 items, answered using a seven-point Likert scale, that is used to assess identity disturbance. We used the Persian version [61] that has excellent internal consistency (α = 0.92). Sample characteristics Of 388 individuals with a confirmed BPD diagnosis, 128 additionally met DSM-5 criteria for ASPD and 130 met DSM-5 criteria for NPD [1]. Participants were divided into three groups: BPD (N = 128, 32.8%), BPD + NPD (N = 130, 33.9%) and BPD+ASPD (N = 128, 33%). Nearly 15% of the BPD+ASPD group met three DSM-5 criteria of ASPD, 60% met four criteria, 23% met five criteria, and nearly 2% met six criteria. 75% of the NPD + BPD group met five criteria of NPD, 17% met six criteria, 7% met seven criteria, and nearly1% met eight criteria. Demographic and clinical data are detailed in Table S1 in the Supplementary Information. To summarize: roughly half the sample were female, and groups did not differ significantly on sex (X2 =5.00, p = 0.08) or age (F (2,385) = 1.38, p = 0.25). Most participants (46%) had high school diplomas and most (38%) were single. Statistical plan By using G-power and indicating alpha 0.05, β = 0.20, and power (1-β) = 0.75, at least eighty-two participants were needed in each group. All analyses were performed in SPSS 23. Analyses indicated no violation of normality, homogeneity of variance, and linearity assumptions, and the significance level was set at alpha = 0.05 (two-tailed) in all statistical analyses. We used independent t-tests and χ2 to compare groups on demographic and severity measures. In the network analysis, each of the nine DSM-5 BPD criteria is represented as a node, and a line (edge) is drawn between two criteria whenever they remain statistically associated after controlling for every other criterion — that is, whenever they show a conditional dependency. A network was estimated separately for each of the three groups (BPD, BPD+ASPD, BPD + NPD), using two complementary methods that are described more fully in the Appendix to the Supplementary Information. Because the nine criteria are scored as binary (present/absent) items, we first fitted a Mixed Graphical Model (MGM) with the Ising estimator, which is designed for binary data. The MGM produced a fully connected network in every group (all 36 possible edges present), meaning that each criterion was statistically associated with every other; at this level, the network structure alone could not distinguish the groups. To obtain a sparser and more readable picture of which symptom relationships stood out, we then applied Fused Graphical Lasso (FGL). FGL prunes weak edges and, through a fusion penalty, encourages similar structure across groups, so that the resulting networks highlight both the shared backbone of BPD symptomatology and the points at which the three groups diverge. FGL assumes Gaussian-distributed data; we treated the binary criteria as numeric for this step and interpret its results with that caveat in mind. Results Network analysis MGM analysis Results of MGM analysis using the Ising model revealed fully connected networks for all three subtypes, with all 36 possible edges detected. This complete connectivity pattern indicates that every BPD criterion shows conditional dependence on every other criterion, even after controlling for all remaining criteria. All 36 edges were shared across the three groups, resulting in identical network topologies, as seen in Fig. 1. This finding provides strong evidence for a unified, highly interconnected symptom structure in BPD, with no subtype-specific conditional dependencies at the chosen regularization strength (λ = 0.1). MGM/Ising network of the nine DSM-5 BPD criteria. All 36 possible edges were present, and network topology was identical across BPD, BPD+ASPD, and BPD + NPD; therefore, a single network is shown. Node abbreviations: ABA = fear of abandonment; REL = unstable interpersonal relationships; IDE = identity disturbance; IMP = impulsivity; SUI = suicidal behaviour; AFF = affective instability; EMP = chronic emptiness; ANG = intense anger; DIS = dissociation/paranoid ideation Despite identical edge structures, criterion prevalence differed markedly across the three groups. As seen in Fig. 2, fear of abandonment (ABA) had the highest prevalence (77.7%) in the BPD group, while in the BPD+ASPD group impulsive behaviors had the highest prevalence (83.6%). There was no clear dominance pattern in the BPD + NPD group. Coded by which group showed the highest prevalence for each criterion, Fig. 2 reveals a clear two-cluster pattern. The BPD group had the highest prevalence of criteria associated with identity and interpersonal functioning — fear of abandonment (ABA), unstable and intense interpersonal relationships (REL), identity disturbance (IDE), suicidal behaviour (SUI) and chronic emptiness (EMP) — whereas the BPD+ASPD group had the highest prevalence of criteria associated with affective and behavioural dysregulation — impulsivity (IMP), affective instability (AFF), intense anger (ANG) and dissociation/paranoid ideation (DIS). The BPD + NPD group occupied an intermediate position, with no criterion for which it was clearly the dominant group. This means that although all three groups shared the same underlying symptom architecture (see Fig. 1), they differed markedly in which parts of that architecture were most endorsed: an identity/interpersonal profile in the BPD group and an affective/behavioural profile in the BPD+ASPD group. MGM Combined network. Type1 = BPD group, Type 2 = BPD+ASPD, Type 3 = BPD + NPD, ABA= Fear of abandonment, REL=unstable and intense interpersonal relationships, IDE=Identity disturbance: IMP=Impulsivity, SUI=Suicidal behavior, AFF= Affective instability, EMP= Emptiness, ANG=Anger, DIS= Dissociation FGL analysis Results of FGL analysis, shown in Fig. 3, revealed relatively sparse networks with substantial shared structure across the three BPD subtypes. The BPD group exhibited 21 edges, the BPD+ASPD exhibited 20 edges, and the BPD + NPD exhibited 23 edges. Notably, 20 edges (87–100% of each network) were shared across all groups, indicating a strong common network architecture underlying BPD symptomatology regardless of comorbidity condition. Four edges showed partial sharing (present in two of the three groups), specifically: fear of abandonment–suicidality, interpersonal relationships-affective instability, suicidal behavior-affective instability, and dissociation-interpersonal relationships. FGL Analysis among three groups. Type1 = BPD group, Type 2 = BPD+ASPD, Type 3 = BPD + NPD, ABA= Fear of abandonment, REL=unstable and intense interpersonal relationships, IDE=Identity disturbance: IMP=Impulsivity, SUI=Suicidal behavior, AFF= Affective instability, EMP= Emptiness, ANG=Anger, DIS= Dissociation The sparser FGL networks shown in Fig. 3 presents a picture that converges with the prevalence patterns shown in Fig. 2. Although 20 edges were shared across all three groups — a stable common backbone of BPD symptomatology — the four edges that were not uniformly present all involved criteria that separate the two clusters identified above: fear of abandonment, suicidality, interpersonal relationships and dissociation on the identity/interpersonal side, and affective instability on the affective/behavioural side. The small number of group-specific edges therefore lie along the identity/interpersonal versus affective/behavioural axis, rather than cutting across it. Taken together, the network analyses support the view that all three groups shared the same symptom structure but differed systematically in which symptoms were most strongly endorsed: BPD was characterised by identity/interpersonal dysfunction, BPD+ASPD by affective/behavioural dysregulation, and BPD + NPD by a mixed profile that fell between the other two. Self-report measures Means and standard deviations for all self-report measures are shown in Table S2 in Supplementary Information. MANOVA results (see Table 1) showed that groups differed significantly in borderline symptom severity (BSL-23), emotion dysregulation (DERS), self-harm (DSHI), identity disturbance (SCIM), and anger-in and anger-out (STAXI). Table 1MANOVA results of BPD vs. BPD + NPD vs. BPD+ASPDMeasuresFηp2BPD vs. BPD + NPDBPD vs. BPD+ASPDBPD + NPD vs. BPD+ASPD(M differences)95% CI for differences(M differences)95% CI for differences(M differences)95% CI for differencesBSL-23340.38**0.63-0.84(-2.61, 0.92)-17.09**(-18.87, -15.32)-16.25**(-18.02, -14.47)DERS173.38**0.47-5.02**(-6.28, -3.76)-9.80**(-11.07, -8.53)-4.87**(-6.04, -3.51)DSHI346.65**0.644.95**(4.49, 5.40)1.98**(1.52, 2.44)-2.97**(-3.42, -2.51)SCIM212.84**0.5218.99**(16.73, 21.25)6.04**(3.77, 8.31)-12.95**(-15.22, -10.68)STAXI-2AX-In259.37**0.572.76**(2.26, 3.23)4.57**(4.08, 5.05)1.82**(1.33, 2.31)AX-Out105.07**0.35-0.56*(-1.11, -0.02)-3.08**(-3.62, -2.54)-2.51**(-3.06, -1.97)AC-In77.07**0.28-0.20(-0.63, 0.21)1.78**(1.36, 2.21)1.99**(1.57, 2.41)AC-Out49.80**0.200.32(-0.05,0.69)-1.16**(-1.53, -0.78)-1.48**(-1.86, -1.11)UPPS-PNU167.94**0.46-2.51**(-3.03, -1.99)-3.89**(-4.40, -3.37)-1.37**(-1.89, -0.85)LPR527.15**0.73-1.51**(-1.95, -1.08)-5.69**(-6.13, -5.26)-4.18**(-4.61, -3.74)BSL-23 = Borderline Symptoms List 23; DERS=Difficulties in Emotion Regulation Scale; DSHI=Deliberate Self-Harm Inventory; SCIM=Self-concept and Identity Measure; STAXI-2 = State-Trait Anger Expression Inventory–2; AXI= Anger expression-In; AXO= Anger expression-Out; ACI= Anger control-In; ACO= Anger control- out; UPPS-P= Impulsive Behavior Scale; NU=Negative Urgency; LPR=lack of Premeditation*<0.01, **<0.001 MANOVA results of BPD vs. BPD + NPD vs. BPD+ASPD BSL-23 = Borderline Symptoms List 23; DERS=Difficulties in Emotion Regulation Scale; DSHI=Deliberate Self-Harm Inventory; SCIM=Self-concept and Identity Measure; STAXI-2 = State-Trait Anger Expression Inventory–2; AXI= Anger expression-In; AXO= Anger expression-Out; ACI= Anger control-In; ACO= Anger control- out; UPPS-P= Impulsive Behavior Scale; NU=Negative Urgency; LPR=lack of Premeditation *<0.01, **<0.001 Borderline symptom severity. The groups differed significantly in severity of BPD symptoms, with the BPD+ASPD group having a significantly (p < 0.001) higher score than both the BPD and BPD + NPD groups. The latter groups did not differ significantly from one another. Scores on the 6 categories of borderline symptom severity (see under Methods) are shown in Fig. 4, where it may be seen that the majority of BPD and BPD + NPD group members had a moderate level of severity, while most of the BPD+ASPD group reported a high level. Proportion of each group showing different levels of borderline symptom severity FIGURE 4 HERE. Emotion dysregulation Groups differed significantly on emotion dysregulation (DERS), with the BPD+ASPD group reporting the greatest difficulty with emotion regulation, the BPD + NPD reporting the least difficulty, and the BPD group reporting an intermediate level of difficulty. Self-harm and identity disturbance Groups also differed significantly on self-harm measured by DSHI. The BPD group scored highest on this measure and significantly higher than both comorbid groups, with BPD + NPD scoring lowest. This pattern of results was also seen in the case of identity disturbance measured by SCIM, with the BPD group scoring highest and the BPD + NPD group showing the least disturbance. Anger dysregulation and impulsivity The BPD group had the highest scores on anger expression-in and the BPD+ASPD had the lowest scores in comparison to other groups. The BPD + NPD group was intermediate, with greater anger expression-in compared with the BPD+ASPD (P < 0.001) and lower anger expression-in than the BPD group (p < 0.001). This pattern was reversed in the case of anger expression-out, with the BPD+ASPD group scoring highest (p < 0.001), and the BPD group scoring lowest. BPD and BPD + NPD groups did not differ significantly on either anger control-in (p = 0.71) or anger control-out (p = 0.11). The BPD+ASPD group scored highest, and the BPD group lowest, on both impulsivity measures (negative urgency and lack of premeditation). The scores of the BPD + NPD group were intermediate between the other two groups. Diagnostic criteria (DSM-5-TR) Borderline symptom profiles The symptom profiles of the three groups, shown in Fig. 5, overlapped considerably, but some notable between-group differences emerged. BPD scored highest on fear of abandonment, unstable relationships, identity disturbance and suicidal behaviour, while BPD+ASPD showed the highest frequency of impulsivity, affective instability and difficulty controlling anger (p < 0.001). Notably, the BPD group showed significantly (p < 0.01) higher frequencies for abandonment fears, unstable relationships, identity disturbance, suicidal behaviour and chronic emptiness. The BPD+ASPD group had significantly (p < 0.001) higher frequencies of impulsivity, affective instability and anger control difficulties. Compared with BPD + NPD, BPD had higher frequencies of abandonment fears (p < 0.01), identity disturbance and suicidal behaviour (p < 0.001), but significantly lower frequencies of impulsivity, affective instability, anger control difficulties and dissociation/paranoid ideation (p < 0.001). Borderline symptom profiles compared between the three groups Discussion This study used different methodological approaches to address the question of whether BPD should be viewed as a homogeneous or heterogeneous construct. Results of network analysis indicated both homogeneity and heterogeneity. Homogeneity was supported by results suggesting a largely uniform network structure across the three groups, with complete conditional dependence among all BPD criteria. This is consistent with previous research that suggested a largely replicable BPD network structure across samples that differed in age and levels of BPD pathology [46]. However, heterogeneity of the borderline construct is supported by differences between groups in terms of symptom endorsement rates. Specifically, the BPD group was characterized by identity/interpersonal dysfunction, manifested by fear of abandonment, unstable and intense interpersonal relationships, identity disturbance, suicidal behaviors, and emptiness. In contrast, the BPD+ASPD group was characterized by affective/ behavioral dysregulation, manifested by impulsiveness, affective instability, anger, and dissociation. The BPD + NPD group showed a mixed presentation with moderate symptom endorsement, suggesting that narcissistic personality traits may be a protective factor for individuals with borderline personality disorder [11]. Heterogeneity of the borderline construct was further supported by the self-report data. The greatest severity of borderline symptoms was seen in the BPD+ASPD group, consistent with BPD and ASPD exerting mutually synergistic effects on borderline symptomatology [4, 25]. Consistent with ASPD comorbidity representing a distinct aggravating factor in the link between BPD and aggression and violence (Gaudin et al., 2025), the BPD+ASPD group showed the highest level of emotion dysregulation and outwardly directed anger, and the highest level of impulsivity, particularly emotionally driven impulsivity (negative urgency). Notably, emotional impulsivity has been found to be a key nexus linking borderline symptom dimensions with aggression in borderline patients [35]. Our BPD+ASPD group resembles the high-severe class of BPD patients described by Black et al., [5]. This high-severe class showed high rates of comorbid personality disorders (including ASPD) and high rates of disturbed relationships, abandonment fears, and intense anger [5]. Membership of this class was predicted by high impulsiveness. Our BPD group, by contrast, showed the highest level of identity disturbance, self-harm, and inwardly directed anger. Our BPD group most closely resembles the empty/dissociation/identity disturbance group identified by Black et al., [5], who showed relatively high rates of emptiness, identity disturbance, and stress-related paranoia/dissociation. Consistent with the suggestion that narcissistic personality traits may be a protective factor for individuals with borderline personality disorder [11], the BPD + NPD group scored lower on emotion dysregulation, self-harm, and identity disturbance compared with the other two groups. Self-report data were corroborated by the borderline symptom profiles shown by the three groups, with the BPD group showing high scores on a triad of borderline symptoms, namely identity disturbance, frantic efforts to avoid abandonment, and emptiness. The BPD group were also most likely to have unstable relationships and to engage in suicide behaviour, consistent with this triad of borderline symptoms – identity disturbance, chronic feelings of emptiness, and frantic efforts to avoid abandonment – being specifically linked to a high risk of suicide [65]. If PDs are viewed as essentially relational disorders [24, 66], different blends of self/other dysfunction would be expected to be seen when different PDs co-occur. Evidence supporting disturbed relationality is contained in data comparing our three groups on symptoms of identity disturbance and interpersonal relationships. The BPD group showed the highest degree of identity disturbance measured by the SCIM while the BPD + NPD group showed the least disturbance. This was corroborated by the borderline symptom profiles shown in Fig. 5, where the BPD group showed the greatest identity disturbance and the BPD + NPD group the least. From Fig. 5, it may be seen that disturbed interpersonal relationships also differed between groups, being lowest in the BPD+ASPD group and higher in both the other groups. However, this evidence for variation in disturbed relationality being associated with different patterns of comorbidity is suggestive only, and further evidence will be required to confirm it. Strengths and limitations To our knowledge, this is the first study to use network analysis to investigate BPD symptoms in a large clinical sample of individuals with BPD segregated into sub-groups with and without NPD and ASPD comorbidity. A notable strength is the inclusion of sufficient numbers of male participants in all three borderline sub-groups, addressing deficiencies in prior research that has focused predominantly on female borderline patients. We acknowledge several limitations to our study. First, rather than using the diagnostic criteria from Sect. "Method" of DSM-5 to identify our borderline sample, it might have been preferable to use the criteria contained in Sect. "Statistical Plan" (the Alternative Model of Personality Disorder (AMPD). Although research suggests moderate overlap between criteria contained in Sects. "Method" and "Statistical Plan", selection of patients according to AMPD criteria might not only have captured unique presentations of borderline pathology not captured by Sect. "Method" but may also have better captured their heterogeneity [22]. Second, exclusion of patients who had either received psychotropic medication in the three months prior to the study, or had received psychotherapy within the previous year, may have limited the sample in several ways. For example, by excluding individuals who had sought treatment in the previous year we may have limited our sample to patients with less symptom severity overall. Third, histrionic personality disorder was not included due to insufficient referrals of patients with this diagnosis. Fourth, we did not investigate possible sex differences in borderline symptoms in our three groups, an important omission given that borderline symptoms have been found to be differently expressed in male and female adults and adolescents [40]. Finally, we did not differentiate between grandiose and vulnerable sub-types of narcissism. Our results pertain predominantly to the grandiose sub-type which is mostly captured by the DSM-5 criteria for NPD [8, 37, 47]. Treatment implications Evidence from this study supports the heterogeneity of BPD and underscores the likelihood that differences found between BPD and other groups largely reflect quantitative differences in PD severity rather than BPD per se [63]. Our results confirm previous findings, reviewed above, that when it co-occurs with ASPD, BPD is a particularly severe form of personality disorder whose symptoms, particularly those related to emotion dysregulation, are responsive to treatment only in the short term, showing high remission in the longer term [38]. Treatment approaches need to recognize the heterogeneity of BPD and be tailored to the specific ways in which externalizing and internalizing features are manifested in borderline patients, e.g. intense anger and impulsivity versus suicidal gestures and chronic feelings of emptiness. Treatment also needs to consider the different ways in which disturbances of relationality (self/other disturbances) are manifested in patients with different combinations of internalizing and externalizing features. Such an approach to treatment is consistent with a move away from viewing BPD as a distinct diagnostic category towards viewing it as reflecting broad dimensions of personality function captured by Criterion A of AMPD: [44]). These dimensions include problems of identity, internal models of relationships, self-reflection, emotion awareness, and regulation [29]. Disturbed identity is a prime candidate in this respect since, among measures of AMPD Criterion A, it was identified by Gilbert et al., [18] as the strongest predictor of borderline features. As Kerber et al., [29] suggest, future research should investigate personality function both as a potential treatment indicator and as a treatment outcome. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements None. Author contributions BM conceived the study and was overall responsible for data collection and analysis and for drafting the manuscript. PM carried out the network analysis and wrote up the results. RH was a major contributor to writing the manuscript. Funding This research was not funded by an external funding agency. Data availability Date will be made available upon reasonable request to the first author. Declarations Ethics approval and consent to participate The authors followed the ethical principles of the Helsinki Declaration regarding research involving human participants. All participants gave their informed consent to participate in the research and use of the data for research and teaching purposes. The ethics review board of Shahid Beheshti University of Medical Science, Iran, approved the study procedures. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes “Comorbidity” can mean two quite different things, referring either to: (a) the covariation (or correlation) between two diagnoses within a sample or the population; or (b) the co-occurrence between two diagnoses within an individual [33]. In this paper, we refer to comorbidity in the latter sense. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement Date will be made available upon reasonable request to the first author.