Construct Validity of the Adolescent Borderline Personality Disorder: A Review

Abstract

Introduction

Although the term borderline personality disorder (BPD) is used to describe adolescents in clinical settings, there is confusion as to what it comprises. To further elucidate that diagnosis, this article reviews its construct validity.

Method

Relevant publications appearing in PsychInfo (1872 to present) were reviewed for the purposes of this article.

Results

Thirty-six of the approximately sixty-five publications selected for consideration were included in this review.

Conclusion

The construct validity of adolescent BPD is supported by internal consistency (comparable to that of adults), group differences (ie this diagnosis segregates BPD from non-BPD adolescents), convergent validity (ie multiple measures of this disorder measure the same pathology) and concurrent validity, whereby these youth manifest functional impairment and distress. By contrast, the adolescent BPD criteria manifest less construct validity than the adult diagnosis in that its criteria did not uniformly predict the overall diagnosis, and showed more criterion overlap with other personality disorders and a broader pattern of axis II comorbidity. Further diminishing its construct validity, factor analysis suggested that adolescent BPD was not a single entity, and its low predictive validity was demonstrated by little diagnostic stability through adolescence into adulthood.

INTRODUCTION

Adolescent psychopathology in general can involve affective lability, impulsivity, identity confusion, and a tendency to idealize peers while devaluing others (Block et al, 1991). Those same features also characterise patients specifically suffering from borderline personality disorder (BPD) (Kernberg, 1975). Despite these similarities, research has shown borderline adolescents to be distinct from their peers (Ludolph et al, 1990; Faulkner et al, 1999; Westen et al, 2003). Block et al (1991) concluded that while all adolescents may struggle with similar developmental crises, the BPD adolescent “has neither the ego strengths nor the facilitated environment to launch her successfully into adulthood”.

Criteria for the diagnosis of BPD in adolescence reflect those of the adult disorder, for which several sets of criteria will be described. According to the DSM IV (American Psychiatric Association, 1994), the diagnosis of BPD in adulthood requires that the individual demonstrate “a pervasive pattern of instability of interpersonal relationships, self-image, and affects, and marked impulsivity beginning by early adulthood and present in a variety of contexts.” Such individuals are further characterised by identity disturbance, concerns about abandonment, recurrent suicidal behaviour, chronic feelings of emptiness and transient, stress-related paranoid ideation or severe dissociative symptoms.

Kernberg (1975) further defined borderline personality organization patients as occupying a borderline area between psychosis and neurosis, in a stable not fluctuating pattern. They are characterised by identity disturbance, primitive defense mechanisms (ie splitting, idealisation, devaluation and projective identification) and predominantly intact reality testing, despite regression and micro-psychotic episodes.

Finally, Gunderson & Singer (1975) identified six features that characterized borderline patients, including the presence of intense affect, a history of impulsive behaviour, a certain level of social adaptiveness, brief psychotic experiences, loose thinking in unstructured situations, and relationships that vacillate between transient superficiality and intense dependency. Despite these well-established criteria for the BPD construct in the adult population, there is uncertainty as to whether there is a corresponding construct for the adolescent population.

Initial conceptualisations of adolescent BPD emphasised its developmental history. Masterson (1978) viewed BPD as a culmination of childhood events initiated by intense feelings of abandonment, a resulting clinging to the maternal figure, and a failure to progress through normal developmental stages of separation- individuation to autonomy. Clinical and empirical findings suggest that BPD derives from traumatic childhood experiences including abuse, maternal neglect and rejection, grossly inappropriate parental behaviour, disrupted attachments, multiple parent surrogates and familial psychopathology in general (Grilo et al., 1999; Ludolph et al, 1990; Goldman et al, 1993).

Although these patient characteristics would appear to apply to adolescent patients we currently diagnose as suffering from BPD, many aspects of this diagnostic construct in adolescents need to be clarified. For example, there has been relatively little investigation of the prevalence, manifestations and long-term ramifications of this diagnosis (Meijer & Treffers, 1991). The following review will consider the empirical findings concerning the construct validity of adolescent BPD, using predominantly the DSM and Gunderson criteria for the adult diagnosis as a reference point.

METHOD

A search of PsycInfo 1872 to present was carried out, with the use of the following keywords “personality disorder”, “borderline personality disorder”, “borderline”, as well as the terms “adolescent” and “adolescence”. Non-English publications were excluded, as were publications concerning children and young adult populations. Inclusion criteria consisted of an adolescent study population, and the use of validated borderline personality disorder measures. These measures include the Diagnostic Interview for Borderline patients (DIB, Gunderson et al, 1981), and its revised version (DIB-R, Zanarini et al, 1989), the Personality Disorder Examination (PDE, Loranger et al, 1988), the Personality Diagnostic Questionnaire (PDQ, Hyler et al, 1988), and the Millon Adolescent Clinical Inventory (MACI, Millon et al, 1993). Publications assessing BPD using DSM criteria (DSM III, III-R, and IV) were also included. Based on the above exclusion and inclusion criteria, thirty-six publications of the original sixty-five articles selected were ultimately included in the review.

RESULTS

The first consideration with respect to construct validity concerns group differences, and more specifically, whether or not adult BPD criteria differentiate BPD from non-BPD adolescents. Bernstein et al. (1993) found that in a community sample of adolescents, 15% of males and 17.2% of females (aged 11- 17 years), met the criteria for BPD, and Chabrol et al. (2001) estimated its overall frequency to be 14% in a random sample of French high school students. It was also determined that adolescent inpatients with BPD are clinically distinct from their peers and that adult BPD criteria are applicable to adolescents (Block et al, 1991; Ludolph et al, 1990; Faulkner et al, 1999; Westen et al, 2003).

Becker et al (2002) found no significant difference in the baseline rate of BPD when comparing hospitalised adolescents (53 %) with adults (47 %). However, when the diagnostic efficiency of the BPD criteria was examined in these two groups, the positive predictive power of the adult criteria (ie the ability of each criterion to predict the presence of the disorder) was more uniform in the adult than the adolescent group. By contrast, the negative predictive power (ie absence of the criterion renders the diagnosis less likely) of the criteria was comparable between the two groups. Hence, while it is possible to apply the adult criteria to the diagnosis of adolescent BPD, some diagnostic criteria are more predictive of the adolescent disorder than others.

Becker et al (1999) examined the homogeneity, another indicator of construct validity, of the BPD diagnostic construct in adolescent and adult inpatients. To do so, they explored the cohesiveness (internal consistency and mean inter-criterion correlations) within the BPD construct, as well as inter-category associations between BPD and other personality disorder criteria. They found the cohesiveness within the BPD category to be only moderate, but comparable, for both the adolescent and adult groups.

Criterion overlap (ie the extent to which criteria of different personality disorders overlap) was then assessed. In adolescents and adults, the borderline criteria were found to be consistently more related to one another than to other personality disorders criteria. However, the degree of correlation between criteria of BPD and those of other personality disorders was almost always greater in the adolescent group, suggesting a lesser degree of discriminant validity of the adolescent BPD diagnosis.

Convergent validity, a third element of construct validity, was assessed in two ways. First, correlations were sought between the BPD diagnosis and other measures and variables with which it shares an overlap of constructs. Westen et al (2003) examined the relationship between diagnoses derived from each of two measures (ie the DSM-IV axis II criteria and the Shedler- Westen Assessment Procedure-200 for Adolescents (SWAP-200A)) and instruments of adaptive functioning and childhood behaviour profiles. Strong positive associations were found between both diagnostic measures and suicide attempts, psychiatric hospitalizations, and aggressive and externalizing behaviours. These similar findings from two separate measures further support the integrity of the BPD adolescent construct.

A second approach to the determination of convergent validity is to evaluate if the adolescent BPD construct overlaps with other theoretically-related constructs, such as cluster B personality disorders. Becker et al (2000) found that the BPD group of patients (adolescents and adults) showed an overrepresentation of all kinds of personality disorders, in contradistinction to a non-borderline group with personality disorders, thus showing mild support for the adolescent construct.

Further, the adolescent BPD group had a relatively broad distribution of clusters A, B and C axis II co-morbidities, where-as the pattern of adult BPD comorbidity was more restricted to cluster B. The authors suggest that the broad pattern of comorbid overlap found in the adolescent sample suggests a more diffuse range of psychopathology in adolescents, and thus challenges the notion of such a construct in that population.

A fourth means by which construct validity can be measured is through factor analysis. Westen et al (2003) used factor analysis to develop an empirically- and clinically-based classification of adolescent personality disorders. Using the SWAP-200A, and allowing for both dimensional and categorical diagnoses of personality disorders, they discerned five statistically independent, non-overlapping personality disorders: antisocial-psychopathic, emotionally-dysregulated, avoidant-constricted, narcissistic and histrionic. They further examined correlations between these factors and other variables, including axis II diagnoses and measures of adaptive functioning. The axis II-defined diagnosis of adolescent BPD was strongly correlated with two separate factors (ie the emotionally-dysregulated and histrionic factors), suggesting that the unitary adult-based axis II BPD criteria may not be applicable to the adolescent population.

The emotionally-dysregulated factor then positively correlated with suicide attempts, and psychiatric hospitalizations, to name a few, while the histrionic factor positively correlated with suicide attempts, amongst other variables. The authors suggested that while those two personality configurations share many of the features of the current borderline diagnosis, maintaining them as two separate entities may be a more accurate reflection of personality pathology in adolescence. The combination of these two factors into one diagnosis may represent an “artifact of overlapping diagnostic categories”. Their findings support the diagnosis of personality disorders in adolescence, but raises questions specifically as to the validity of a BPD construct during that developmental stage.

To examine the concurrent validity of adolescent BPD, and thus its clinical relevance, investigators have sought correlations between it and various measures of distress and dysfunction. Bernstein et al (1993) originally validated BPD diagnoses against indicators of functional impairment and distress, measures of problems at school and work and measures of psychopathology, and found BPD to be associated (ie increased odd ratio) with eleven out of twelve measures. Levy et al (1999) found adolescent personality disorders in general to be associated with significantly greater levels of impairment and distress. Finally, Westen et al (2003) correlated BPD diagnoses (derived using three different diagnostic measures) with conceptually relevant measures of impairment and distress, finding similar patterns of dysfunction and distress in association with the BPD construct, irrespective of the diagnostic tool employed. The above findings demonstrate that the concurrent validity of adolescent BPD has been consistently supported when assessed in terms of associated distress and dysfunction.

Again in the realm of concurrent validity, female adolescent inpatients with BPD, in contrast to those without, were found to have more frequent time-outs, behave more aggressively, have been subjected to more drug and/or alcohol screens and have poorer self-concepts (Faulkner et al, 1999; Pinto et al, 1996). In a similar fashion, Westen et al (2003) found that BPD was the personality disorder most highly correlated with suicide attempts in a sample of adolescent patients.

Conversely, adolescent inpatients who had attempted suicide, compared to “never-suicidal” inpatients, were more likely to have a personality disorder, and a greater prevalence and severity of borderline traits and disorder (Wade, 1987; Brent et al, 1993). Although not directly relevant to concurrent validity, increased suicidality (defined on a spectrum ranging from suicide ideation to completion) in BPD adolescents has been associated with the number of stressful life situations, familial alienation, impaired functioning, and co-existing major depressive disorder (Runeson & Beskow, 1991; Young & Gunderson, 1995; Friedman et al, 1983).

As further markers of concurrent validity, and in this context of stress, adolescent BPD has also been associated with concurrent axis I disorders, such as depression (McManus et al, 1984), bipolar disorder (Kutcher et al, 1990) and substance abuse (Grilo et al, 1996), and Marton et al (1989) found that BPD was the most common personality disorder (ie a rate of 30%) diagnosed in their sample of depressed adolescents.

Another key concern of any construct is its predictive validity: its stability, and how it evolves over time. Korenblum et al (1990) assessed personality disorder clusters A, B (including BPD), and C longitudinally in children at the ages of 13, 16 and 18. Although no subjects were diagnosed within cluster B at the age of 13, 40 % of the sample showed personality dysfunction compatible with cluster B personality disorders by the age of 18. When the two year stability of BPD was examined in adolescents, persistence rates were relatively low (ie ranging from 23% to 33%), and even within these ‘stable’ BPD patients, symptoms varied over time (Bernstein et al, 1993; Garnet et al, 1994; Mattanah et al, 1995).

Others assessed adolescent BPD predictive validity by evaluating impairment and distress at future time periods. Levy et al (1999) found adolescents with overall personality disorders showed increased levels of drug use and greater frequencies of inpatient treatment during the ensuing two years. However, they found no differences between those with and without personality disorders in terms of current employment, alcohol abuse, legal difficulties, psychiatric symptoms, or social and family relationships.

Several authors have also investigated the diagnostic trajectory of personality-disordered adolescents. Johnson et al (1999) found cluster B personality disorder diagnoses to be associated with increased risk in early adulthood of disruptive, mood, substance use and personality disorders, but not anxiety disorders or suicidality. Further, for those with cluster B disorders in adolescence, each additional co-morbidity almost doubled the odds of having a personality disorder as an adult (Kasen et al 1999). Lastly, Lewinsohn et al (2000) found that, for those adults prone to a mood disorder, BPD symptoms in adolescence predicted a more complex form of that mood disorder in adulthood.

Conversely, adolescent axis I pathology has been retrospectively noted among young adults later diagnosed with personality disorders, and Kasen et al (1999) found disruptive and depressive disorders in adolescence to be associated with cluster B personality disorders as adults. Similarly, the presence and number of axis I diagnoses (Lewinsohn et al, 1997) in adolescence were found to be associated with the general category of personality disorders in adulthood, and specifically adolescent anxiety disorders and attention deficit hyperactivity disorder (Rey et al, 1995) with the development of BPD later on (Lewinsohn et al, 1997). No association was found between adolescent BPD and violence and criminal behaviours in adulthood (Johnson et al. 2000).

CONCLUSION

From the perspective that adult criteria for the BPD construct have traditionally been used as an anchor from which to diagnose BPD in adolescence, two conclusions can be drawn from the above review. First, a number of factors contribute to the notion that adolescent BPD is a valid construct. For example, it has a comparable degree of internal consistency and group differences to the adult construct, and it shows a high degree of concurrent validity, as adolescents diagnosed with BPD also show functional impairment and distress. As well, on a test of its convergent validity, there was a high correlation between two separate BPD diagnostic measures.

Second, and by contrast, several factors detract from the notion that it is a valid construct. For example, BPD criteria among adolescents, when compared to adults, were less uniform with respect to diagnostic efficiency, showed a broader pattern of axis II co-morbidity and more criterion overlap with other personality disorders, suggesting that BPD is more diffuse in nature in adolescents. Additionally, factor analysis suggested that BPD was not a single entity in adolescence. Lastly, the construct showed a low degree of predictive validity, with little diagnostic stability, including associations with both axis I and axis II disorders, but not necessarily BPD, in adulthood.

This mixed picture concerning the validity of the adolescent BPD construct raises some questions. First, could these inconsistencies suggest non-homogeneity among adolescents currently diagnosed with BPD? Second, could the current diagnostic construct comprise both a state and/or a trait? Finally, do these discrepancies in measures of validity, particularly predictive, reflect multiple adolescent developmental pathways, normal and abnormal, following three potential suggested trajectories: diagnostically stable, diagnostically transient, and diagnostically variable (Mattanah et al, 1995)? Further research is warranted to elucidate the answers to these questions, and to thus clarify the nature of adolescent BPD.