Factor Structure of Borderline Personality Disorder Symptomatology in Adolescents

Transcription

1 CanJPsychiatry 2012;57(4): Original Research Factor Structure of Borderline Personality Disorder Symptomatology in Adolescents Mario Speranza, MD, PhD 1 ; Alexandra Pham-Scottez, MD 2 ; Anne Revah-Levy, MD, PhD 3 ; Remy P Barbe, MD, PhD 4 ; Fernando Perez-Diaz, MD 5 ; Boris Birmaher, MD, PhD 6 ; Maurice Corcos, MD, PhD 7 1 Child and Adolescent Psychiatrist, Centre Hospitalier de Versailles, Service de Pédopsychiatrie, Le Chesnay, France and EA4047, University of Versailles Saint Quentin, Versailles, France; Researcher, INSERM U669, University Paris-Sud and University Paris Descartes, UMR-S0669, Paris, France. Correspondence: 177 rue de Versailles, Le Chesnay, France; msperanza@ch-versailles.fr. 2 Psychiatrist, Clinique des Maladies Mentales et de l Encéphale, Hôpital Sainte-Anne, and INSERM U669, University Paris-Sud and University Paris Descartes, UMR-S0669, Paris, France. 3 Psychiatrist, Centre de Soins Psychothérapeutiques de Transition pour Adolescents, Hôpital d Argenteuil, F-95107, Argenteuil, France and INSERM U669, University Paris-Sud and University Paris Descartes, UMR-S0669, Paris, France. 4 Child and Adolescent Psychiatrist, Child and Adolescent Department, University Hospital of Geneva, Department of Psychiatry, University of Geneva, Switzerland; Child and Adolescent Psychiatrist, Department of Psychiatry, University of Pittsburgh, Pennsylvania. 5 Research Engineer, Centre Emotion CNRS UPSR 3246, Hôpital de la Salpétrière, Paris, France. 6 Child and Adolescent Psychiatrist, Western Psychiatric Institute and Clinic, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania. 7 Psychiatrist, Institut Mutualiste Montsouris, Département de Psychiatrie de l Adolescent et du Jeune Adulte and INSERM U669, University Paris-Sud and University Paris Descartes, UMR-S0669, Paris, France. Key Words: borderline personality disorders, adolescents, exploratory factor analysis, confirmatory factor analysis, DSM-IV criteria, Structured Interview for DSM-IV Personality Received August 2011, revised, and accepted October Objective: To examine the factor structure of the Diagnostic and Statistical Manual of Mental Disorders (DSM), Fourth Edition, criteria for borderline personality disorder (BPD) in a sample of adolescents with a borderline symptomatology. Method: The latent structure of borderline criteria, assessed with the Structured Interview for DSM-IV Personality, was explored with a principal factor analysis in a sample of 107 adolescents with a borderline symptomatology drawn from a European research project on BPDs. Results: The principal component analysis revealed 2 homogeneous factors accounting for 66.8% of the variance. The first factor included internally oriented criteria, such as avoidance of abandonment, identity disturbance, chronic feeling of emptiness, and stress-related paranoid ideation. The second factor included externally oriented criteria, such as unstable relationships, impulsivity, suicidal or self-mutilating behaviours, and inappropriate anger. Affective instability was the only criterion loading on both factors. Conclusions: The results of our study suggests that an internal or external dichotomy may be an appropriate way to conceptualize the structure of borderline criteria in adolescents with a borderline symptomatology, with affective instability being a core feature of BPD at this age. W W W Objectif : Examiner la structure factorielle des critères du Manuel diagnostique et statistique des troubles mentaux (DSM), 4e édition, pour le trouble de la personnalité limite (TPL) dans un échantillon d adolescents ayant une symptomatologie de personnalité limite. Méthode : La structure latente des critères de la personnalité limite, évaluée par l entrevue structurée pour le DSM-IV (personnalité), a été explorée par une analyse factorielle principale dans un échantillon de 107 adolescents ayant une symptomatologie de personnalité limite, liés à un projet de recherche européen sur le TPL. Résultats : L analyse factorielle principale a révélé 2 facteurs homogènes responsables de 66,8 % de la variance. Le premier facteur incluait des critères orientés vers l intérieur, comme l évitement ou l abandon, la perturbation de l identité, le sentiment de vide chronique, et l idéation paranoïde liée au stress. Le deuxième facteur comprenait des critères orientés vers l extérieur, comme les relations instables, l impulsivité, les comportements suicidaires ou d automutilation, et la colère exagérée. L instabilité affective était le seul critère qui logeait sur les deux facteurs. Conclusions : Les résultats de notre étude suggèrent qu une dichotomie intérieure ou extérieure puisse être une façon appropriée de conceptualiser la structure des critères de la personnalité limite chez les adolescents avec une symptomatologie de la personnalité limite, l instabilité affective étant une caractéristique fondamentale du TPL à cet âge. 230 W La Revue canadienne de psychiatrie, vol 57, no 4, avril

2 Factor Structure of Borderline Personality Disorder Symptomatology in Adolescents The diagnosis of BPD in adolescents has been a topic of debate in recent years, with controversial reports concerning its validity and stability over time. 1,2 Clinicians overall are reluctant to diagnose personality disorders during this age period, as adolescents are undergoing fastchanging developmental processes. 3 However, according to the DSM-IV, personality disorder categories can be applied to children and adolescents in those relatively unusual instances in which the person s particular maladaptive personality traits appear to be pervasive, persistent, and unlikely to be limited to a particular developmental stage. 4 Moreover, an emerging body of data is beginning to suggest that BPD can be reliably diagnosed in adolescents, 5 with prevalence rates in the community similar to those found in adults. 6 8 However, more controversial is the notion of stability of BPD features from adolescence to adulthood. Several studies have shown that the categorical stability of BPD in adolescents is relatively low However, this is not unique to adolescents, as demonstrated in a recent 10-year prospective follow-up study that showed that up to 50% of adults with a BPD will eventually recover. 13 In contrast with the categorical diagnoses, it appears that a greater stability can be observed using a dimensional approach with selected symptom criteria, such as mood lability and anger dyscontrol, which emerge as significant predictors of BPD diagnosis in both adolescent and adult samples. 11,12,14,15 Thus conceptualizing BPD from a dimensional, rather than a categorical, approach could be particularly pertinent in adolescents, as a dimensional approach may better account for the developmental variability and heterogeneity observed during this age period. 2,11 Factor analysis has been widely employed as a way to explore clinical heterogeneity and to identify meaningful components within the structure of BPD. Overall, factor analytic studies in adult BPD have yielded conflicting results, with evidence supporting one or more latent personality constructs Differences in methods (the type of interviews) and sampling factors may partially explain the heterogeneity of these results. If the unidimensional structure of BPD has been the most frequently replicated finding, 18,20,22,23 the 3-factor model identified by Sanislow et al, 24,25 which includes disturbed relatedness, behavioural dysregulation, and affective dysregulation, is currently considered as a reference model for it seems to coherently Abbreviations BPD DSM GFI ICD borderline personality disorder Diagnostic and Statistical Manual of Mental Disorders Goodness of Fit Index International Classification of Diseases K-SADS Schedule for Affective Disorders and Schizophrenia for School-Age Children Present and Lifetime Version PCA principal component analysis RMSEA root mean square error of approximation SIDP-IV Structured Interview for DSM-IV Personality Clinical Implications Factor analysis is a useful approach to identify meaningful components of BPD symptomatology in adolescents. The latent structure of borderline symptomatology in adolescence differs from what is known in adult patients. Emotional dysregulation should deserve special attention in adolescents with a BPD symptomatology. Limitations The cross-sectional design of our study does not allow to demonstrate the stability of the identified dimensions. The limited size of the sample may have reduced the statistical power of the analyses and underestimated the number of true latent dimensions. Female patients were overrepresented in the sample thus reducing the generalizability of the results. capture the clinical complexity of BPD patients. According to Sanislow et al s analysis, the heterogeneity of the structure of BPD is reflected in the criteria, which are an admixture of personality traits, symptomatic behaviours, and symptoms, which respectively characterize each of these dimensions. 24 The factor analysis approach has rarely been applied to samples of adolescents with BPD, notwithstanding its potential usefulness in identifying meaningful components implicated in the diagnostic specificity and stability of BPD. Becker et al 21 conducted an exploratory factor analysis on DSM-III-R BPD criteria in a sample of 123 inpatient adolescents using the Personality Disorder Examination. They identified 4 components accounting for 67% of the variance: self-negation or depression, affective dysregulation or irritability, interpersonal dysregulation, and impulsiveness. Although Becker et al 21 employed a similar methodology and an identical Axis II assessment procedure, these components differed markedly, in terms of number and content, from those reported in Sanislow et al s study. 24 These results suggest that adolescents could have a different BPD structure compared with adults. 21 In this respect, Leung and Leung, 8 using a confirmatory approach on a large sample of 4100 adolescents, found that Sanislow et al s 24 original 3-factor model failed to provide a detailed conceptualization of symptom composition of the BDP pathology. However, Leung and Leung s study 8 was conducted on a nonclinical sample of college students using a self-administered questionnaire to rate symptom severity. Thus it is difficult to generalize these results to clinical samples of adolescents with BPD who, in addition to having more severe borderline symptomatology, have high rates of comorbid disorders that may impact the factor structure. 21 Our study aimed to further explore this domain by investigating the factorial structure of DSM-IV BPD criteria in a sample of adolescents with a borderline symptomatology. The Canadian Journal of Psychiatry, Vol 57, No 4, April 2012 W 231

3 Original Research Methods Subjects Our study sample was drawn from a European longitudinal research project investigating the diagnostic stability of BPDs from adolescence to young adulthood (European Research Network on BPD). 26 The research network was composed of 5 academic psychiatric centres in France, Belgium, and Switzerland. To increase the validity of the diagnosis of BPD, the recruitment followed a 2-stage sampling procedure. From January to December 2007 all consecutive in- and outpatient adolescents (aged 15 to 19 years) were clinically screened by their consulting psychiatrists for the diagnosis of BPD according to DSM-IV criteria. All adolescents fulfilling a clinical diagnosis of BPD were then referred to the research team for a further assessment (see methods described below) and confirmation of the diagnosis of BPD using the SIDP-IV. 27 Adolescents with a diagnosis of schizophrenia, with any chronic and (or) serious medical illness involving vital prognosis, and adolescents with a mental retardation, were excluded from the sample. Adolescents (n = 107) with a clinical diagnosis of BPD were referred to the research team. A full DSM-IV diagnosis of BPD according to the SIDP-IV interview was confirmed only for 85 adolescents (78% of the referred girls and 92% of the referred boys). For the purpose of our study (which was to examine the factor structure of DSM-IV BPD criteria in adolescents with a borderline symptomatology), all referred adolescents with a clinical diagnosis of BPD, not limited to those with a full diagnosis of BPD on the SIDP-IV, were included in the factor analysis. We used a broader study group encompassing a wider range of borderline pathology severity to encompass the problem highlighted by Nunnally and Bernstein, 28 who demonstrated that when participants are sampled to be as homogeneous as possible on a given characteristic (as a high severity cut-off), internal consistency estimates are low and may be influenced by other sources of variation that are not controlled by the study design. The study was approved by the French Ethics Committee (authorization number , Hôtel Dieu). Results were collected in an anonymous database according to the requirements of the French National Committee for Private Freedoms. After a complete description of the study was given to the subjects, written informed consent was obtained from adolescents and their parents. Procedure All the participants completed a research protocol, which consisted of a diagnostic evaluation of Axis I and II disorders and a self-report questionnaire eliciting sociodemographic data. Axis II disorders were ascertained through the SIDP-IV, a semi-structured interview assessing each of the 10 DSM-IV personality disorders, including BPD. 27 Reliability and validity of the SIDP-IV have been established in adolescents and young adults Axis I disorders were assessed using the K-SADS-PL. 33 Diagnostic interviews were conducted 232 W La Revue canadienne de psychiatrie, vol 57, no 4, avril 2012 by a research team of 5 doctoral- or master s-level clinicians (psychologists or psychiatrists) familiar with DSM-IV Axis I and II disorders and experienced in assessment and (or) treatment of psychiatric adolescents. To reach high levels of reliability, the research evaluation team participated in several training sessions, including commented scoring of videotaped interviews and a training session conducted by the developers of the K-SADS-PL. 33 Final research diagnoses were established by the best-estimate method for the interviews and any additional relevant data from the clinical record according to the longitudinal, expert, alldata (commonly referred to as LEAD) standard. 34 Interrater reliability for SIDP-IV was calculated from independent ratings of 10 videotaped interviews. The kappa coefficient for the presence or absence of a BPD was very good (0.84). The intraclass correlation coefficient for borderline SIDP-IV score was excellent (0.95). Statistical Analysis The internal consistency of the SIDP-IV was assessed using the Cronbach alpha coefficient. Intercorrelations between items were calculated from polychoric correlations for ordinal variables. The factorial structure of the sample was explored in 2 ways. In the first step, we performed a PCA with a promax oblique rotation. One of the more critical decisions in a PCA is to determine the correct number of factors to retain and rotate. To avoid the under- or overestimation of the number of true latent dimensions, each model was evaluated according to several criteria 35 : inspection of the eigenvalues (greater than one), analysis of the screeplot (completed by a graphical simulation of screeplots calculated with random normal data), 36 Glorfeld s 37 extension of parallel analysis, and interpretability. We retained only factor loadings of 0.35 or greater. We used a promax oblique rotation because this technique yields more interpretable factors with a simpler structure than those obtained with an orthogonal rotation, especially in situations in which factors are probably correlated. The factorability of the correlational matrix was explored with the Bartlett sphericity test (P < 0.05). The Kaiser-Meyer-Olkin test of sampling adequacy was used to verify that the data satisfied the thresholds for supporting an interpretable factor analysis (0.6 or greater). As a second step, we conducted a confirmatory factor analysis using the 3-factor structure of the Sanislow et al model. 16 Model fit was evaluated using the GFI, the RMSEA, and the ratio between the chi-square and the number of degrees of freedom. 38 Finally, to explore the association between the factor structure of BPD criteria and Axis I comorbidity, we performed a series of stepwise linear regression analyses, with age, sex, and current Axis I disorders as independent variables and the components derived from the principal component analysis as dependent variables. Axis I diagnoses were included as groups of related disorders to have a stronger power of analysis. The criterion for inclusion of a variable in the equation was set at a P value of Statistical analyses were performed using SAS,

4 Factor Structure of Borderline Personality Disorder Symptomatology in Adolescents Table 1 Frequency of current Axis I disorders in the research sample (n = 107) Disorder n % Mood disorders Major depression Dysthymia 10 9 Bipolar disorder 9 8 Depressive disorder NOS 2 2 Posttraumatic stress disorders Eating disorders Anorexia Bulimia Disruptive behaviour disorders Attention-deficit hyperactivity disorder 9 8 Oppositional defiant disorder 9 8 Conduct disorder Substance-related disorders Alcohol related disorders Drug related disorders At least 1 disorder NOS = not otherwise specified version 9.1 (SAS Institute Inc, Cary, NC). The graphical representation of the simulated screeplots and the parallel analysis for component retention were calculated using the r program. Results There were no significant differences between the recruitment centres in terms of subject age and educational level, numbers of BPD criteria and in- or outpatient ratios. The research sample was characterized by a severe borderline symptomatology. The mean score on the SIDP-IV for the entire sample (n = 107) was 15.4 (SD 5.8) (with a minimum required score for a BPD diagnosis of 15 and a maximum score of 27). The mean age was 16.6 (SD 1.4) years. Eighty-nine per cent (n = 95) were girls and 63% (n = 62) were recruited from inpatient units. There was a high frequency of current comorbid disorders. Most adolescents met the criteria for at least 1 Axis I disorder (n = 90, 84%). Mood disorders were the most frequently observed comorbidity (n = 48, 45%) followed by eating disorders (n = 38, 37%), disruptive behaviour disorders (n = 22, 21%), and substance use disorders (n = 17, 16%) (Table 1). Table 2 shows the mean scores and correlations between SIDP-IV criteria. Affective instability, inappropriate anger, suicidal or self-mutilating behaviour, and impulsivity were the most highly endorsed criteria, and the stress-related paranoid ideation criterion was the least endorsed criterion. The 9 BPD criteria were moderately intercorrelated. Internal consistency was in the medium range, with a Cronbach alpha of Affective instability and inappropriate anger were highly correlated with the other BPD criteria. The highest correlations were between inappropriate anger and impulsivity (0.63) and between inappropriate anger and affective instability (0.65). Concerning the PCA for the BPD criteria, the Kaiser- Meyer-Olkin measures (0.80) and the Bartlett sphericity test (P < 0.001) showed that the sample met the criteria for factor analysis. The inspection of the eigenvalues with loadings of greater than 1 indicated that only 2 factors could be extracted from the data. The simulated screeplots calculated with random normal data and the results of the Horn Parallel Analysis confirmed that 2 components had to be retained. The 2-factor solution with a promax oblique rotation accounted for 66.8% of the variance, with each factor explaining almost one-half of the total variance (factor 1: 33.9% of the variance and factor 2: 32.9% of the variance). Correlation between factors was With the threshold of 0.35, the first factor included exclusive loadings for avoidance of abandonment, identity disturbance, chronic feeling of emptiness, and stress-related paranoid ideation. Identity disturbance and stress-related paranoid ideation had the highest loadings. The second factor included exclusive loadings for unstable relationships, impulsivity, and high loadings for suicidal or self-mutilative behaviour and inappropriate anger, although these 2 criteria also loaded on the first factor. Impulsivity and inappropriate anger had the highest loadings. Affective instability was the only criterion loading equally on both factors (Table 3). Several stepwise linear regressions were performed to explore the association between the factor structure of BPD criteria derived from the PCA and Axis I comorbidity, The best model predicting the first factor accounted for 14% of the variance (Adjusted R² = 14.7; F = 9.50, df = 2, 97, P = 0.001) and included mood disorders (OR 0.35; 95% CI 0.32 to 1.01, P = 0.001) and posttraumatic stress disorder (OR 0.18; 95% CI to 0.97, P = not significant). The second factor explained 21% of the variance (Adjusted R² = 20.8; F = 9.68, df = 3, 96, P = 0.001) and included mood disorders (OR 0.33; 95% CI 0.29 to 0.97, P = 0.001), disruptive disorders (OR 0.22; 95% CI 0.08 to 0.94, P = 0.02), and substancerelated disorders (OR 0.19; 95% CI 0.01 to 0.93, P = 0.02). Table 4 presents the confirmatory factor analysis of the sample according to Sanislow et al s model. 25 The uncorrelated 3-factor model had satisfactory global fit indices (GFI = 0.93; RMSEA = 0.059; chi-square and degrees of freedom ratio = 1.37; Probability > chi-square = 0.105). The reliability of the 3 factors was adequate although slightly lower for the behavioural dysregulation factor (including impulsivity and suicidal or self-mutilating behaviours). Discussion Our study aimed to explore the factorial structure of BPD criteria in a sample of adolescents with a borderline symptomatology using a dual exploratory and confirmatory strategy. Overall, results showed that a 2-factor solution The Canadian Journal of Psychiatry, Vol 57, No 4, April 2012 W 233

5 Original Research Table 2 Correlations among the DSM-IV criteria for BPD in the research sample (n = 107) BPD criterion Mean (SD) Avoidance of abandonment 1.3 (1.1) Unstable relationships 1.5 (1.1) 0.26 Identity disturbance 1.4 (1.1) 0.29 a 0.23 Impulsivity 2.2 (1.0) a 0.15 Suicidal or self-mutilating 2.1 (1.1) 0.38 b 0.28 a 0.25 a 0.32 b behaviour Affective instability 2.1 (1.1) 0.41 b 0.24 a 0.46 b 0.39 b 0.39 b Chronic feeling of emptiness 1.5 (2.2) 0.27 a b b 0.43 b Inappropriate anger 2.2 (1.0) b 0.37 b 0.63 b 0.49 b 0.65 b 0.38 b Stress-related paranoid ideation 1.2 (1.0) 0.44 b 0.35 b 0.54 b b 0.39 b 0.34 b 0.34 b BPD criteria assessed with the SIDP-IV: a score of 0 = not present, 1 = subthreshold, 2 = present, 3 = strongly present. Correlational matrix: polychoric correlations between ordinal variables. All tests are 2-tailed. Normalized Cronbach α = 0.77 a P < 0.05; b P < 0.01 Table 3 Promax solution with 2 factors in the research sample (n = 107) Factor loading BPD criterion Factor 1 Factor 2 Stress-related paranoid ideation 0.82 a 0.25 Identity disturbance 0.77 a 0.27 Chronic feeling of emptiness 0.64 a 0.32 Avoidance of abandonment 0.54 a 0.32 Affective instability 0.63 a 0.66 b Inappropriate anger b Impulsivity b Suicidal or self-mutilating behaviour b Unstable relationships b Promax oblique rotation. Total per cent of variance = 66.8%. a Cells representing the criteria that correspond to Factor 1 (from stress-related to affective instability) b Cells representing the criteria that correspond to Factor 2 (from affective instability to unstable relationships) Factor 1: Per cent of variance = 33.9% Factor 2: Per cent of variance = 32.9% seems the most appropriate to conceptualize the structure of BPD criteria in a well-characterized sample of in- and outpatient adolescents. The inspection of the criteria loading on the dimensions derived from the PCA highlights 2 components that could be coherently understood according to their predominant internal or external content. The first dimension included only internally oriented criteria, such as avoidance of abandonment, identity disturbance, chronic feeling of emptiness, and stress-related paranoid ideation. These criteria highlight the lack of cohesive self-experience (chronic feeling of emptiness and avoidance of abandonment) and the instability of identity (identity disturbance and stress-related paranoid ideation) of adolescents with a BPD 234 W La Revue canadienne de psychiatrie, vol 57, no 4, avril 2012 symptomatology. The second dimension contained mainly externally oriented criteria, such as inappropriate anger, impulsivity, suicidal or self-mutilating behaviour, and unstable relationships. This association is consistent with the view that self-mutilation is used by adolescents with BPD in an attempt to control destabilizing emotions, such as inappropriate anger. 39 The high levels of impulsiveness associated with this factor may probably facilitate the use of these externalizing strategies. Affective instability is the only criterion loading equally on both factors, suggesting that in adolescence emotional dysregulation could be a core feature of BPD, with other symptoms being strategies to cope with the emotional instability. Unexpectedly, the unstable relationships criterion loaded with the external factor in our sample. The typical fluctuations between idealization and devaluation, which are frequently associated with identity disturbances and fears of abandonment in adult BPD, seem in adolescence more related to affective instability and impulse control. Even if the literature is sparse, the internal and external dichotomy found in our study can be traced to other factor analysis studies using DSM-III-R 40 or DSM-IV 17 criteria. In addition, this dichotomy is in line with the proposals of Krueger, 41 who has suggested the use of using an internalizing or externalizing framework to characterize personality disorders in general, and BPDs in particular. 41,42 As an example, in a recent study using a hierarchical model strategy, James and Taylor 43 showed that BPD appears as a multidimensional indicator of both internalizing (anxious misery) and externalizing psychopathology, albeit with a stronger internalizing component. Our factor analysis is also coherent with ICD-10 conceptualization of the emotionally unstable personality disorders, which includes, on a severity gradient, an impulsive subtype, which almost totally overlaps with our external factor, and a borderline subtype, which is close

6 Factor Structure of Borderline Personality Disorder Symptomatology in Adolescents Table 4 Confirmatory factor analysis according to the 3-factor model of Sanislow et al 24 Factor Standardized loadings Indicator reliability Error variance Reliability Variance t Disturbed relatedness Unstable relationships Identity disturbance Chronic feeling of emptiness Stress-related paranoid ideation Behavioural dysregulation Impulsivity Suicidal or self-mutilating behaviour Affective dysregulation Affective instability Inappropriate anger Avoidance of abandonment n = 107; GFI = 0.932; RMSEA = 0.059; χ 2 = 32.9; df = 24; Probability > χ 2 = to our internal factor. 44 It is useful to mention that, in our sample, the criteria corresponding to the ICD-10 impulsive subtype were also the most highly endorsed criteria, thus suggesting that this externalizing subtype could be more frequent in adolescence and more related to age-specific symptoms and behaviours, such as affective instability, impulsivity, and inappropriate anger. This is indirectly confirmed by the regression analyses, which show that, while both factors were positively associated with mood disorders, the external factor was also positively associated with disruptive disorders and substance-related disorders. A similar result can be found in Becker et al s 21 study in which the affective dysregulation and impulsiveness factors, which contain several criteria loading on our external factor, were solely correlated with typical adolescent externalizing Axis I disorders, such as oppositional defiant disorder and substance use disorders. This association may represent a specific feature of adolescents with BPD and may question, at this age, the independence of impulsivity and suicidal or self-mutilating behaviours from anger dyscontrol and affective instability criteria, which in Sanislow et al s model 24 are separated between the behavioural and the affective dysregulation factors. Moreover, although the results of the confirmatory factor analysis revealed that Sanislow et al s 3-factor model 24 could fit with the data of our sample, the theoretical coherence of the structure, the external validation given by the comorbidity, and the principle of parsimony, seem to better support the pertinence, in our sample, of a 2-factor latent structure of BPD criteria. Some limitations of our study must be considered when interpreting the findings. The limited size of the sample may have reduced the statistical power of the analyses and underestimated the number of true latent dimensions. However, compared with other studies, our sample is one of the largest found in the adolescent BPD literature, with several statistical methods converging toward the identification of a 2-factor solution. The limited number of indicators by which DSM-IV delineates BPD may have imposed a restraint on the number of recovered factors. 45 However, the use of DSM-IV criteria as the primary method of conceptualizing the underlying diagnostic structure has been largely employed in psychiatric literature to be considered as valid. 16 As expected when studying BPD, females were overrepresented. This may represent a selection bias and explain the differences in the comorbidity profile of our sample (with higher eating disorders and lower substance use disorders) when compared with other studies. 21 More generally, factor analyses are very sensitive to sampling bias and comorbidity features. Thus, although this 2-factor model may appear as conceptually appealing in adolescence, it should be generalized with caution to all adolescents with BPD as it may reflect the specificity of the clinical sample. Conclusions In summary, our study showed that a 2-factor solution separates BPD criteria according to 2 main dimensions, internally and externally oriented. Our study also indicates that affective instability appears as a main feature of BPD during adolescence, suggesting that emotional dysregulation could be a core feature of BPD at this age, with other symptoms being strategies to cope with the emotional instability. The identification of specific clinical dimensions involved in BPD functioning, such as emotional dysregulation, may offer the advantage of better developing targeted interventions for adolescents with BPD. This dimensional approach is appealing and in line with the recommendations issued by the working group on personality disorders for the future DSM-5. Although the advantages and inconveniences of a traitbased system, compared with a more clinically meaningful The Canadian Journal of Psychiatry, Vol 57, No 4, April 2012 W 235

7 Original Research prototype approach for classifying personality disorders are under inquiry, 46 it is worth noting dimensional and categorical conceptualizations of personality pathology are not competing. 47 On the contrary, clinical prototypes should be constructed from combinations of well-defined and empirically validated dimensions identified from crosssectional and longitudinal data on clinical and community samples. Acknowledgements This research was supported by a grant from the Wyeth Foundation for Child and Adolescent Health (2005) and by a grant from the Eli Lilly Foundation (2006). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. This work was conducted in a European collaborative research project on BPD (European Research Network on Borderline Personality Disorders [EURNET-BPD]). All members of the EURNET-BPD should be acknowledged for this work (authors excluded): Corinne Dugré-Lebigre, Véronique Delvenne, Lionel Caihol, Gwenolé Loas, Philippe Stéphan, David Cohen, Jean Marc Baleyte, Sylvie Nezelof, Julien Daniel Guelfi, and Bruno Falissard. References 1. Bondurant H, Greenfield B, Tse SM. Construct validity of the adolescent borderline personality disorder: a review. Can Child Adolesc Psychiatr Rev. 2004;13(3): Miller AL, Muehlenkamp JJ, Jacobson CM. Fact or fiction: diagnosing borderline personality disorder in adolescents. Clin Psychol Rev. 2008;28(6): Bleiberg E. Borderline disorders in children and adolescents: the concept, the diagnosis, and the controversies. Bull Menninger Clin. 1994;58(2): American Psychiatric Association (APA). Diagnostic and statistical anual of mental disorders. 4th ed. Text revision. Washington (DC): APA; Becker DF, Grilo CM, Morey LC, et al. Applicability of personality disorder criteria to hospitalized adolescents: evaluation of internal consistency and criterion overlap. J Am Acad Child Adolesc Psychiatry. 1999;38(2): Zanarini MC, Frankenburg FR, Hennen J, et al. The longitudinal course of borderline psychopathology: 6-year prospective followup of the phenomenology of borderline personality disorder. Am J Psychiatry. 2003;160(2): Chabrol H, Montovany A, Chouicha K, et al. Frequency of borderline personality disorder in a sample of French high school students. Can J Psychiatry. 2001;46(9): Leung SW, Leung F. Construct validity and prevalence rate of borderline personality disorder among Chinese adolescents. J Pers Disord. 2009;23(5): Mattanah JJ, Becker DF, Levy KN, et al. Diagnostic stability in adolescents followed up 2 years after hospitalization. Am J Psychiatry. 1995;152(6): Bondurant H, Greenfield B, Tse SM. Construct validity of the adolescent borderline personality disorder: a review. Can Child Adolesc Psychiatr Rev. 2004;13(3): Crawford TN, Cohen P, Brook JS. Dramatic-erratic personality disorder symptoms: II. Developmental pathways from early adolescence to adulthood. J Pers Disord. 2001;15(4): Crawford TN, Cohen P, Brook JS. Dramatic-erratic personality disorder symptoms: I. Continuity from early adolescence into adulthood. J Pers Disord. 2001;15(4): W La Revue canadienne de psychiatrie, vol 57, no 4, avril Zanarini MC, Frankenburg FR, Reich DB, et al. Time to attainment of recovery from borderline personality disorder and stability of recovery: a 10-year prospective follow-up study. Am J Psychiatry. 2010;167(6): McGlashan TH, Grilo CM, Sanislow CA, et al. Two-year prevalence and stability of individual DSM-IV criteria for schizotypal, borderline, avoidant, and obsessive-compulsive personality disorders: toward a hybrid model of axis II disorders. Am J Psychiatry. 2005;162(5): Zanarini MC, Frankenburg FR, Ridolfi ME, et al. Reported childhood onset of self-mutilation among borderline patients. J Pers Disord. 2006;20(1): Sanislow CA, Grilo CM, McGlashan TH. Factor analysis of the DSM-III-R borderline personality disorder criteria in psychiatric inpatients. Am J Psychiatry. 2000;157(10): Clarkin JF, Hull JW, Hurt SW. Factor structure of borderline personality disorder criteria. J Pers Disord. 1993;7: Fossati A, Madeddu F, Maffei C. Borderline personality disorder and childhood sexual abuse: a meta-analytic study. J Pers Disord. 1999;13(3): Sanislow CA, Grilo CM, Morey LC, et al. Confirmatory factor analysis of DSM-IV criteria for borderline personality disorder: findings from the collaborative longitudinal personality disorders study. Am J Psychiatry. 2002;159(2): Johansen M, Karterud S, Pedersen G, et al. An investigation of the prototype validity of the borderline DSM-IV construct. Acta Psychiatr Scand. 2004;109(4): Becker DF, McGlashan TH, Grilo CM. Exploratory factor analysis of borderline personality disorder criteria in hospitalized adolescents. Compr Psychiatry. 2006;47(2): Becker DF, Anez LM, Paris M, et al. Exploratory factor analysis of borderline personality disorder criteria in monolingual Hispanic outpatients with substance use disorders. Psychiatry Res. 2010;178(2): Clifton A, Pilkonis PA. Evidence for a single latent class of diagnostic and statistical manual of mental disorders borderline personality pathology. Compr Psychiatry. 2007;48(1): Sanislow CA, Grilo CM, McGlashan TH. Factor analysis of the DSM-III-R borderline personality disorder criteria in psychiatric inpatients. Am J Psychiatry. 2000;157(10): Sanislow CA, Morey LC, Grilo CM, et al. Confirmatory factor analysis of DSM-IV borderline, schizotypal, avoidant and obsessive-compulsive personality disorders: findings from the Collaborative Longitudinal Personality Disorders Study. Acta Psychiatr Scand. 2002;105(1): Corcos M, Pham-Scottez A, Speranza M. European Research Network on Borderline Personality Disorder (EURNET-BPD). 57th Annual Meeting of the American Academy of Child and Adolescent Psychiatry New research poster session no 312, 2010; New York (NY): October 28 31, Pfohl B, Blum N, Zimmerman M. Structured Interview for DSM-IV Personality. Washington (DC): American Psychiatric Press; Nunnally J, Bernstein I. Psychometric theory. 3rd ed. New York (NY): McGraw Hill; Waldo TG, Merritt RD. Fantasy proneness, dissociation, and DSM-IV axis II symptomatology. J Abnorm Psychol. 2000;109(3): Lobbestael J, Van Vreeswijk MF, Arntz A. An empirical test of schema mode conceptualizations in personality disorders. Behav Res Ther. 2008;46(7): Glenn CR, Klonsky ED. Emotion dysregulation as a core feature of borderline personality disorder. J Pers Disord. 2009;23(1): Chabrol H, Montovany A, Callahan S, et al. Factor analyses of the DIB-R in adolescents. J Pers Disord. 2002;16(4): Kaufman J, Birmaher B, Brent D, et al. Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL): initial reliability and validity data. J Am Acad Child Adolesc Psychiatry. 1997;36(7):

8 Factor Structure of Borderline Personality Disorder Symptomatology in Adolescents 34. Pilkonis P, Heaper C, Ruddy J, et al. Validity in the diagnosis of personality disorders: the use of the LEAD standard. Psychol Assess. 1991;3: Velicer WF, Eaton CA, Fava JL. Construct explication through factor or component analysis: a review and evaluation of alternative procedures for determining the number of factors or components. In: Goffin RD, Helms E, editors. Problems and solutions in human assessment: honoring Douglas N Jackson at seventy. New York (NY): Guilford; p Horn JL. A rationale and test for the number of factors in factor analysis. Psychometrika. 1965;30: Glorfeld LW. An improvement on Horn s parallel analysis methodology for selecting the correct number of factors to retain. Educ Psychol Meas. 1995;55: Marsh HW. Confirmatory factor analysis models of factorial invariance: a multifaceted approach. Struct Equ Modeling. 1994;1: Linehan MM. Skills-training manual for treatment of borderline personality disorder. New York (NY): Guilford; Blais MA, Hilsenroth MJ, Castlebury FD. Content validity of the DSM-IV borderline and narcissistic personality disorder criteria sets. Compr Psychiatry. 1997;38(1): Krueger RF. Continuity of axes I and II: toward a unified model of personality, personality disorders, and clinical disorders. J Pers Disord. 2005;19(3): Eaton NR, Krueger RF, Keyes KM, et al. Borderline personality disorder co-morbidity: relationship to the internalizingexternalizing structure of common mental disorders. Psychol Med. 2011;41(5): James LM, Taylor J. Revisiting the structure of mental disorders: borderline personality disorder and the internalizing/externalizing spectra. Br J Clin Psychol. 2008;47(Pt 4): Whewell P, Ryman A, Bonanno D, et al. Does the ICD-10 classification accurately describe subtypes of borderline personality disorder? Br J Med Psychol. 2000;73(Pt 4): Preacher KJ, MacCallum RC. Exploratory factor analysis in behavior genetics research: factor recovery with small sample sizes. Behav Genet. 2002;32(2): Shedler J, Beck A, Fonagy P, et al. Personality disorders in DSM-5. Am J Psychiatry. 2010;167(9): Eaton NR, Krueger RF, South SC, et al. Contrasting prototypes and dimensions in the classification of personality pathology: evidence that dimensions, but not prototypes, are robust. Psychol Med. 2010; Sep 22:1 13; [Epub ahead of print]. The Canadian Journal of Psychiatry, Vol 57, No 4, April 2012 W 237