Onset in adolescence

The emergence of clinically significant levels of impulsivity and suicidality is related to developmental timing. These symptoms usually become apparent in adolescence (Zanarini, Frankenburg, Khera, & Bleichmar, 2001), a time when neurocircuitry is reorganized (Cicchetti & Rogosch, 1996), and when one sees an increase in many impulsive behaviors, particularly addictions (Chambers, Taylor, & Potenza, 2003).

For this reason, the clinical onset of BPD usually takes place in adolescence. Zanarini et al. (2001) found that most patients have their first clinical presentation at a mean age of 18, with a standard deviation of 6. The same study also noted that many cases had experienced symptoms for several years prior to presentation, supporting the clinical observation that many of these patients describe significant difficulties prior to the adolescent years (Paris, 2003).

Unfortunately, we lack solid information about the childhood precursors of BPD. Although we have suggestive evidence about prepubertal risks and behavioral patterns in this population, we need longitudinal data. Research would benefit from a prospective follow-up of a large cohort of symptomatic children considered to be at long-term risk, paralleling the classic study by Robins (1966) on ASPD. Such a study has not yet been conducted, possibly for the reason that whereas conduct disorder (CD) is usually identifiable in clinical settings, the precursors of BPD during childhood may be subclinical.

Relationship to other impulse spectrum disorders

The central role of impulsivity in BPD suggests that its childhood origins may be similar to other impulse spectrum disorders. ASPD, with its wide range of impulsive behavior (American Psychiatric Association, 2000), is a particularly relevant point of comparison. In spite of obvious differences (impulsive actions directed against others vs. actions directed against the self), there are a number of striking similarities between these two disorders. Long-term outcome and course are similar (Paris, 1997), with both disorders presenting peak symptomatology in young adulthood, followed by a gradual “burnout.” In both disorders, biological measures are correlated with impulsive aggression (Coccaro et al., 1989). It is possible that both disorders emerge from a common matrix of trait dimensions, but have different behavioral expressions (Looper & Paris, 2000).

These observations suggest that ASPD and BPD could have common childhood precursors. In fact, their psychosocial risk factors are similar, with both diagnoses being associated with childhood adversity and family pathology (Paris, 1997). However, although CD is an established precursor of ASPD, with which its symptoms have an obvious continuity (Caspi, 2000; Moffitt, 1993; Robins, 1966, 1978), the relationship of behavioral disorders to BPD is unclear, possibly because it is confounded by gender. One prospective study (Rey, Singh, Morris–Yates, & Andrews, 1997) as well as a retrospective study (Goodman, Hull, Clarkin, & Yeomans, 1999) have observed that girls with conduct symptoms in adolescence often go on to develop a range of personality disorders (including BPD) in young adulthood. This relationship might even be stronger if it were not for the fact that the definition of CD gives more weight to typical male symptoms than to symptoms more common in females (Zoccolillo, Tremblay, & Vitaro, 1996). Gender differences also affect the timing of conduct symptoms, in that aggressive behavior starts earlier in boys than in girls (Crick & Zahn–Waxler, 2003). This sequence, in which impulsive symptoms emerge after puberty, parallels the adolescent onset of impulsive disorders such as substance abuse (Schuckit & Smith, 1996) and eating disorders (Garner & Garfinkel, 1980).

Externalizing and internalizing symptoms

The suicidality associated with BPD in adulthood lacks clearcut parallels in childhood. Although children can think about or threaten suicide (Pfeffer, 2000), attempts are uncommon prior to puberty (Brent, 2001).

Again, the influence of gender may make the developmental precursors of BPD difficult to identify. Community studies have consistently found that boys have a higher frequency of externalizing symptoms whereas girls have a higher frequency of internalizing symptoms (Cohen, Cohen, Kasen, & Velez, 1993; Achenbach & McConaughy, 1997). In childhood, boys are more likely to act out aggression (Crick & Zahn–Waxler, 2003), which explains why they are more likely to present in clinical settings. These relationships have a parallel in adults with personality disorders: ASPD patients are about 80% male (Robins & Regier, 1991), whereas about 80% of BPD patients are female (Samuels et al., 2002; Swartz, Blazer, George, & Winfield, 1990; Torgersen, Kringlen, & Cramer, 2001). Similar relationships can be observed in related Axis I diagnoses: mood disorders are more common in adult females (Weissman & Klerman, 1985), whereas substance abuse and criminality are more common in adult males (Robins & Regier, 1991).

The influence of gender could explain why it has been difficult to identify developmental precursors of BPD. Girls at risk for the disorder may have less severe externalizing symptoms, as well as internalizing symptoms that may go unnoticed. However, the concept that children who later develop BPD show a mixture of externalizing and internalizing symptoms has been supported by several lines of empirical evidence. Early-onset depression and dysthymia are frequent markers for the onset of personality disorder in adolescence (Lewinsohn, Rohde, Seeley, Klein, & Gotlib, 2000; Pepper et al., 1995). A longitudinal community study (Crawford, Cohen, & Brook, 2001a, 2001b) found that among adolescents with Cluster B personality disorder symptoms (i.e., in the impulsive spectrum), externalizing symptoms predicted continuing psychopathology in males, whereas a combination of externalizing and internalizing symptoms predicted continuing Cluster B symptoms in females.

Working backward, one would therefore expect to see both behavioral disturbances and mood disturbances before puberty. Adolescents with personality disorders tend to report childhood histories of conduct problems, depression, and anxiety (Bernstein, Cohen, Skodol, Bezirganian, & Book, 1993). There is some evidence that the precursors of these problems can be measurable early in childhood (Caspi et al., 1996). But if these features do not bring children to clinical attention, this would explain why one only begins to observe the onset of a diagnosable personality disorder when impulsive symptoms become marked after puberty.

Borderline pathology of childhood

Another approach to understanding the developmental precursors of impulsivity and suicidality in BPD is to study children who have symptoms that resemble the adult disorder. There are indeed such cases. A clinical literature on “borderline children” (Kernberg et al., 2000) described this population as having a mixture of impulsive and affective symptoms (behavioral problems, suicidal threats, and mood instability) as well as cognitive phenomena similar to those seen in adults with BPD (micropsychotic phenomena such as hallucinations and paranoid trends). Thus, these children have psychopathology in all three of the trait dimensions affected by adult BPD, impulsivity, affective instability (internalizing symptoms), and cognitive symptoms. This syndrome is also common among children whose mothers have a diagnosis of BPD (Weiss et al., 1996).

The concept of borderline pathology in childhood remains controversial, mainly due to a lack of clear boundaries with other disorders (Bemporad, Smith, Hanson, & Cicchetti, 1982; Petti & Vela, 1990). Moreover, most of these patients are males. It is most telling that children with borderline pathology do not generally evolve into adults with BPD. Borderline pathology in childhood might therefore be a general precursor for adult personality disorders in these cohorts. This supposition is supported by a study by Lofgren, Bemporad, King, Lindem, and O'Driscoll (1991), which followed 19 children who had met a set of criteria for borderline pathology developed by Bemporad et al. (1982). By age 18, the cohort had developed a wide range of personality disorders in all the DSM clusters (but not Axis I diagnoses such as schizophrenia or bipolar mood disorder). However, this study did not use a precise definition of borderline pathology at baseline (i.e., criteria giving greater prominence to impulsive symptoms), which may have reduced the likelihood impulsive personality disorders in adulthood.

In our own study (Guzder, Paris, Zelkowitz, & Feldman, 1999; Guzder, Paris, Zelkowitz, & Marchessault, 1996), we used an instrument specifically developed to assess the clinical picture of borderline pathology in childhood, modeled on structured interviews for diagnosis developed for adult BPD. The Child Diagnostic Interview for Borderlines (Greenman, Gunderson, Cane, & Saltzman, 1986) has subscales that assess multiple dimensions of pathology, impulsivity, depression and suicidality, as well as micropsychotic symptoms. However, because most of our cohort were boys, in our follow-up study in adolescence (Zelkowitz, Guzder, Paris, Feldman, & Roy, 2004) we found that only 20% met criteria for a diagnosis of BPD.

Other groups of researchers who have studied this group of children have used different terminology to describe the same population (without necessarily acknowledging the similarity). One of these constructs is multiple complex developmental disorder (Cohen, Paul, & Volkmar, 1987; Lincoln, Bloom, Katz, & Boksenbaym, 1998), a diagnostic term designed to reflect the multiple pathological dimensions seen in these children.

Another term, proposed by a research group studying the precursors of adult psychosis, is multidimensionally impaired disorder (Kumra et al., 1998), which describes children with micropsychotic symptoms, daily periods of emotional lability, impaired interpersonal skills, and cognitive deficits in information processing. Kumra et al. (1998) followed 26 of these patients to a mean age of 15, and although most showed remission of psychotic symptoms, many developed features of a chronic mood disorder (Nicolson et al., 2001). Because chronic depression in adolescence is associated with impulsive personality disorders (Pepper et al., 1995), these findings can readily be interpreted as supporting the concept that children who show problems with mood, impulsivity, and cognitive defects are at risk for adult personality pathology.

If children with borderline pathology are at risk for developing personality disorders, and if their symptoms fall in the impulsive spectrum, the risk factors associated with this syndrome may provide clues to the origins of adult BPD. Another useful aspect of studying children with borderline pathology is that, in contrast to retrospective reports from adult patients, one can make direct observations about risk factors that avoid recall bias, and the data can be validated by informants. In fact, the psychosocial risk factors associated with borderline pathology of childhood are strikingly parallel those described by adults with BPD (Paris, 1994; Zanarini, 2000), supporting earlier clinical observations that these children tend to come from dysfunctional families characterized by trauma, neglect, and separation (Bemporad et al., 1982; Kestenbaum, 1983). Sexual and physical abuse, as well as neglect, are common among children and adolescents with this clinical picture (Goldman, D'Angelo, & DeMaso, 1993; Goldman, D'Angelo, DeMaso, & Mezzacappa, 1992; Guzder et al., 1996, 1999). As is the case for adult BPD, children with borderline pathology also have first-degree relatives with disorders in the impulsive spectrum (Feldman et al., 1995; Goldman et al., 1992, 1993; Guzder et al., 1996, 1999; Weiss et al., 1996). Borderline pathology of childhood is associated with many of the neuropsychological abnormalities seen in adult BPD (O'Leary, 2000), learning disabilities, attention deficit disorder, and abnormal EEG patterns (Petti & Vela, 1990; Lincoln et al., 1998; Paris, Zelkowitz, Guzder, Joseph, & Feldman, 1999), reflecting defects in “executive functioning” that tend to be associated with impulsive personality traits (Stein, Hollander, Cohen, & Frenkel, 1993).

Retrospective studies of childhood adversity in adults with BPD

In the literature, the most common method of determining the childhood precursors of BPD has been to collect retrospective data from adult patients about their early experiences. Given the well-known limitations of retrospective recall, as well as the well-known tendency for adult borderline patients to have distorted perceptions of life events, recall bias is a serious issue in this population (Paris, 1995). Nonetheless, a large body of data using this approach has had some degree of consistency.

In adult patients with BPD, the most robust findings are that a majority of patients report some form of childhood adversities or trauma, with close to half reporting severe problems, sexual and physical abuse, parental neglect, and highly dysfunctional families (Zanarini, 2000). These reports led some clinicians (e.g., Herman, 1992) to propose a posttraumatic mechanism for borderline pathology. However, the links between these risk factors and adult personality pathology are highly complex.

The relationship between childhood trauma and adult outcome demonstrates both multifinality, in that trauma is a risk factor for other mental disorders (Paris, 2000a), and equifinality, in that BPD can be seen in patients with and without trauma histories (Paris, 1994). About a third of patients with BPD report no childhood trauma at all, in spite of presenting with virtually identical symptoms to those who have such histories (Paris, 1994). Moreover, about half of these reports concern single traumatic incidents, which are known to have fewer sequelae, as opposed to multiple incidents that have more severe effects (Paris, 1997). Notably, community studies of the long-term sequelae of childhood adversities such as physical abuse (Malinovsky–Rummell & Hansen, 1993) and sexual abuse (Browne & Finkelhor, 1986; Rind & Tromovitch, 1997) have shown that only a minority develop mental disorders of any kind. Finally, there is no specificity between childhood trauma and the type of psychopathology seen in adulthood; this relationship can be better understood in the light of personality trait profiles that mediate vulnerability (Paris, 2003).

Similar conclusions apply to measures of other adversities, such as childhood neglect and parental separation (Paris, 1994). These are general risk factors for psychopathology that have little specificity for type of disorder, and which do not necessarily lead to disorder as well. Although not every patient with BPD comes from a dysfunctional family, about half do (Paris, 1994; Zanarini, 2000). Because most children have a high degree of resilience to all but the most severe and prolonged insults (Rutter, 1989), there must be predisposing factors that explain why some develop personality disorders and others do not.

High-risk studies

Some years ago, Gunderson and Zanarini (1989) recommended that the relationship of early adversity to the development of BPD could be clarified by carrying out follow-up studies on children known to have been exposed to the risk factors identified in retrospective research. For example, one might follow children with a history of early neglect and trauma to determine whether there is a higher risk of BPD. However, this type of study has rarely been carried out, largely due to the difficulties, both practical and ethical, of doing so.

One exception derives from the work of a research group headed by Widom (Horwitz, Widom, McLaughlin, & White, 2001), who identified a sample gathered from records of documented court cases of childhood abuse and neglect, and interviewed the subjects about 20 years later, compared with a matched control group who did not experience abuse or neglect. Horwitz et al. (2001) reported that abused and neglected women had higher rates of dysthymia, ASPD and alcohol problems than did controls. However, these relationships were no longer significant when recent life stressors were considered. Notably, this cohort also showed a high rate of resilience (Widom & Kuhns, 1996), as has been observed in many other studies of children at risk (e.g., Werner & Smith, 1992). As Rutter (1987a) has emphasized, childhood, children exposed to abuse and neglect do not necessarily develop psychopathology, unless there is a cumulative risk arising from multiple adversities, as well as continued exposure to adversities in adolescence and young adulthood.

Prospective longitudinal studies

To identify the developmental precursors of BPD, it would be useful to measure adult outcomes in a community sample in which precursors of psychopathology have been identified prior to puberty. This method was used in the Albany–Saratoga study, a large-scale community study of a cohort of children followed into adulthood (Kasen, Cohen, Skodol, Johnson, & Brook, 1999). The results showed that childhood adversities such as neglect and abuse were associated with personality disorder symptoms (Johnson, Cohen, Brown, Smailes, & Bernstein, 1999). The limitation of this approach is that community studies are better at identifying dimensional traits than disorders, particularly disorders with a relatively low prevalence. The prevalence of BPD is less than 1%; a Norwegian community sample (Torgersen et al., 2001) reported 0.7%, an American community sample (Samuels et al., 2002) found 0.5%, but an American sample of university students (Lenzenweger, Loranger, Korfine, & Neff, 1997) found only 0.3%. Moreover, subjects with these problems might be more likely to suffer attrition in a longitudinal community study. For these reasons, Johnson et al. (1999) were not able to identify full-blown cases of BPD; analyses could only be conducted on the total number of personality disorder symptoms.

Another long-term prospective study of early adversity has made use of a birth cohort from New Zealand (Fergusson, Lynskey, & Horwood, 1996; Fergusson & Mullen, 1999). Although reports of childhood sexual abuse in young women were only obtained in adolescence, reports of such experiences were associated with an increased risk for both depression and suicidality. However, there were no measures of personality disorder in that study.