Leveraging a well-characterized longitudinal female sample with and without carefully diagnosed childhood ADHD, we examined theory-informed predictors of young adult BPD—considered both categorically and dimensionally—from childhood and late-adolescent timespans. Although we emphasize caution regarding interpretation of findings due to our small sample size, this investigation extends research from our laboratory on developmental predictors of self-harm behaviors [87] to include borderline personality disorder as a criterion measure. We note that individuals with BPD—a condition characterized by intense and pervasive dysregulation of emotion, behavior, cognition, and relationships—may or may not engage in self-harm.

First, regarding our categorical measure of BPD, using binary logistic regressions with correction for small sample size, we found that—as hypothesized—a cumulative history of childhood adversity, as measured by the ACE score, predicted BPD diagnosis. Low EF in childhood was also a significant predictor, even after adjusting for ACE scores and demographic covariates. Regarding late-adolescent predictors, hyperactivity/impulsivity, inattention, internalizing, and externalizing symptoms each independently predicted young adult BPD diagnosis after adjusting for covariates, but maternal depression did not. In stringent analyses accounting for all independently significant late adolescent predictors, only symptoms of inattention were independently (albeit marginally) related to young adult BPD diagnosis.

Second, with respect to our dimensional measure of BPD features, we found—consistent with hypotheses—that both childhood hyperactivity/impulsivity and a cumulative history of childhood adversity from the ACE score predicted young adult BPD features, with adjustment for covariates. Furthermore, childhood inattention, externalizing symptoms, internalizing symptoms, overt aggression, relational aggression, negative peer nominations, and maternal parenting stress due to dysfunctional interactions also independently predicted young adult BPD features after adjusting for covariates. Yet in the final model, including all childhood predictors with a medium (or larger) effect size that had survived covariates, only childhood hyperactivity/impulsivity and the ACE score maintained significance. As for late-adolescent predictors of the dimensional outcome, hyperactivity/impulsivity, inattention, externalizing symptoms, internalizing symptoms, maternal psychopathology, and maternal parenting stress due to dysfunctional interactions maintained significance after adjusting for covariates, but low executive functioning did not. In the final analysis, adding all surviving predictors in the same model, only late-adolescent externalizing and internalizing symptoms maintained significance. Finally, as for exploratory moderator analyses, we found an interaction between low childhood executive functioning and low socioeconomic status at baseline was significant, suggesting that socioeconomic disadvantage may compound the predictive effects of low executive functioning with respect to later BPD dimensional scores.

Overall, the child and adolescent predictors of later BPD are largely consistent with those from previous investigations [8, 10, 18, 25, 26, 43], emerging here from a carefully controlled prospective investigation. Regarding ADHD symptoms, almost 75% of women who met criteria for BPD in young adulthood had diagnoses of childhood ADHD, most often characterized by high levels of impulsivity (ADHD-C). This finding is consistent with both cross-sectional and longitudinal research, as well as theoretical models of the developmental course of individuals with high levels of early impulsivity, related to BPD as an end-point [11].

For ADHD dimensions, our findings add to the limited number of studies examining hyperactive, impulsive, and inattentive symptoms and their severity across development, especially beginning in childhood [8, 25,26,27]. That hyperactivity/impulsivity in childhood and adolescence did not significantly predict later categorical BPD diagnosis was unexpected and may relate to our small sample size. Yet regarding our dimensional measure, when adjusting for covariates and other predictors, hyperactivity/impulsivity in childhood did significantly predict later BPD features. This finding is consistent with the only two other known studies to our knowledge that have examined prospective associations between childhood impulsivity and later BPD [26, 27]. Two studies have found prospective prediction from both adolescent hyperactivity/impulsivity and inattention to later BPD [8, 25]—along this line, note our marginally significant prediction of categorical BPD from late-adolescent inattentive symptoms: In both Carlson et al. [25] and a recent machine learning study of 128 variables related to risk for BPD, inattention in adolescence emerged as an important predictor [8]. Each dimension of ADHD appears to play an important role in the development of BPD symptoms.

The finding linking low EF in childhood to young adult BPD diagnosis is also consistent with previous (yet limited) research. In the only known prospective longitudinal study to date examining childhood EF as predictive of later BPD, a composite measure of EF at age 5 predicted BPD symptoms at age 12 [26]. When we examined low childhood EF and BPD dimensional features, our findings were not significant. Still, moderator analyses revealed that girls with both low EF and low socioeconomic status were at especially risk for high levels of BPD features. Perhaps difficulties in low EF are related to high or clinically significant BPD in the context of socioeconomic disadvantage.

Regarding internalizing and externalizing symptoms, we found that high levels of each were related to later BPD features after adjusting for other important predictors—but not when we measured BPD categorically. Furthermore, multiple forms of aggression in childhood including both overt and relational aggression, plus negative peer nominations, predicted young adult BPD features, but these findings did not maintain significance in the presence of other important predictors. Thus, symptoms of aggression and peer preference are important in childhood as risk factors for later BPD, yet other factors—childhood hyperactivity/impulsivity and trauma-remain statistically superior. Indeed, there was substantial overlap in our measures of aggression, negative peer nomination, hyperactivity/impulsivity, and broadband externalizing symptoms. Overall, our key findings replicate those from Stepp et al. [43], who showed that adolescent internalizing and externalizing symptoms predict adult BPD (for additional research, see Belsky et al., [26]; and Geselowitz et al., [10]). Adolescence appears to be a particularly sensitive period during which vulnerability for the development of severe and pervasive dysregulation across the lifespan may be realized [44].

Although maternal psychopathology—plus both child and late-adolescent maternal parenting stress due to dysfunctional interactions—were each independently associated with young adult BPD features after adjusting for covariates, their effect sizes were small. Any implications require replication. Findings from other research indicate that parental invalidation and negative parenting practices may well be stronger predictors [53].

Consistent with a large body of research linking a history of childhood adversity/trauma with later BPD [88], we found that a cumulative history of childhood adversity, measured by the ACE scale, was a crucial predictor of both young-adult BPD diagnosis and dimensional features. This measure of cumulative history of childhood adversity is retrospective—and may therefore be better characterized as a subjective experience of childhood trauma rather than objective experiences of childhood trauma. These findings are consistent recent data finding that risk of psychopathology is high among individuals with subjective reports of childhood maltreatment regardless of whether these experiences were validated by objective measures [89]. As well, the ACE measure we used constitutes the gold standard in the field.

Our results support theories that transactions between dispositional and environmental factors over time can lead to a cycle of dysregulation of emotion, behavior, and cognition as well as difficult interpersonal relationships [1, 11]. Indeed, findings support Linehan’s Biosocial Theory plus recent developmental models of females with ADHD [11]. That is, behaviorally expressed impulsivity may be a risk factor for a range of outcomes, including BPD. As development progresses, children with trait impulsivity may experience childhood trauma—linked with, for example, intergenerational trauma, maladaptive parenting practices (especially in relation to the child’s impulsive behavior), and/or parents’ own behavioral impulsivity—which may then transactionally escalate the development of BPD. The original impulsivity may, via heterotypic continuity, come to be expressed as a combination of internalizing and externalizing dimensions, leading to BPD [11, 90]—a suggestion requiring further empirical investigation. Future research should include prospective temporal assessment of these domains, as well as other environmental mediators (e.g., peer relationship influences, substance use) and valid measures of behavioral parental invalidation, to assess multi-factor etiological influences.

Clinical implications

Although our sample size is too small to draw definitive clinical recommendations, we provide several ideas for possible clinical and public health implications, emphasizing caution in interpretation of results related to study limitations (see below for more detail in this regard). First, findings highlight the longstanding effects of early experiences of adversity and trauma. Prevention of these childhood experiences, especially through public health initiatives, cannot be overemphasized. Second, our results reveal the importance including global EF deficits in childhood as indicators of risk for BPD, in addition to the focus on childhood impulsivity. These findings have implications for guiding early clinical assessment and intervention (e.g., through early EF skills training) to prevent later BPD. In short, we highlight the need for interventions before the adolescent period, which appears to be an especially sensitive time of risk [11].

Children with histories of adversity/trauma and/or deficits in EF could receive interventions targeting emotion dysregulation, a mechanism linked to the development of BPD, such as Dialectical Behavior Therapy for Children (DBT-C) [91], Parent–Child Interaction Therapy (PCIT) [92], or creative combinations of these therapies [93]. Widespread assessment of early risk factors to identify individuals at risk remains a challenge. We also recommend that evidence-based treatments for severe emotion dysregulation (i.e., DBT) include remediation of EF deficits.

Limitations and future directions

Our study has several important limitations. First, our sample size is small for the categorical BPD variable, with only 19 females meeting diagnostic criteria for BPD in young adulthood, clearly limiting statistical power. Note that we used Firth’s penalized likelihood method to statistically account for our small sample [94]. Furthermore, only a subset of our sample completed the self-report dimensional BPD measure. We emphasize the need for replication and cautious interpretation of findings. Second, we did not have symptom-level data available for our categorical measure of BPD, preventing us from evaluating clinician-assessed dimensions of BPD symptoms. Future research would benefit from examining dimensional severity of BPD symptoms, as well as specific traits, some of which have recently been linked to increased risk for a suicide attempt [95]. Third, several measures—including those of BPD features, cumulative trauma history, and ADHD (in part) were self-reported—and may thus be subject to bias. Fourth, we were not able to peform mediator analyses and therefore cannot add to the literature on potential “driving” mechanisms between childhood and adulthood (e.g., emotion dysregulation). Fifth, we did not separate predictor symptom domains of hyperactivity vs. impulsivity, as psychometrics are superior when using the full 9-item Hyperactivity/Impulsivity scale. As well, this measure is more consistent with the DSM’s layout of symptoms. Although we support the separation of theses symptoms in future research—see the excellent national analysis by Tiger et al. [27]—we elected to include the full 9-item scale. Sixth, our measure of externalizing symptoms in late adolescence (ASR, ABCL) included measures of aggression, but given our multiple testing, we did not examine aggression per se during this developmental window. Future research would benefit from examining aggressive symptoms across development, given empirical research and theory linking high levels aggression and peer problems with later BPD [11, 18]. Seventh, there is controversy over whether the Rey-Osterrieth Complex Figure Test captures meaningful variance in executive functioning [96]. It could be that this measure is a better index of visual-motor integration and overall neuropsychological functioning than of executive functioning [97, 98]. Future research should investigate different domains of neuropsychological functioning and BPD development. Eighth, there is definitional overlap between ADHD and BPD, given that both are characterized by impulsivity, which could account for some of the present results. Finally, a key limitation is the timing of our BPD measure—we measured BPD only during young adulthood, but some participants may already have met criteria for BPD in adolescence.

Still, key strengths include a carefully diagnosed, ethnically and socioeconomically diverse sample of females; emphasis on multi-domain and multi-informant measures; high sample retention; and a prospective (and ongoing) longitudinal design. Moreover, we included stringent use of covariates and statistical penalization. Finally, we examined multiple domains of risk for BPD simultaneously and included several measures of BPD symptomology.