Sage Journals: Discover world-class research

Abstract

French

Objective:

To examine trajectories of adolescent borderline personality (BP) features in a normative-risk cohort (n = 566) of Canadian children assessed at ages 13, 14, 15, and 16 and childhood predictors of trajectory group membership assessed at ages 8, 10, 11, and 12.

Method:

Data were drawn from the McMaster Teen Study, an on-going study examining relations among bullying, mental health, and academic achievement. Participants and their parents completed a battery of mental health and peer relations questionnaires at each wave of the study. Academic competence was assessed at age 8 (Grade 3). Latent class growth analysis, analysis of variance, and logistic regression were used to analyze the data.

Results:

Three distinct BP features trajectory groups were identified: elevated or rising, intermediate or stable, and low or stable. Parent- and child-reported mental health symptoms, peer relations risk factors, and intra-individual risk factors were significant predictors of elevated or rising and intermediate or stable trajectory groups. Child-reported attention-deficit hyperactivity disorder (ADHD) and somatization symptoms uniquely predicted elevated or rising trajectory group membership, whereas parent-reported anxiety and child-reported ADHD symptoms uniquely predicted intermediate or stable trajectory group membership. Child-reported somatization symptoms was the only predictor to differentiate the intermediate or stable and elevated or rising trajectory groups (OR 1.15, 95% CI 1.04 to 1.28). Associations between child-reported reactive temperament and elevated BP features trajectory group membership were 10.23 times higher among children who were bullied, supporting a diathesis–stress pathway in the development of BP features for these youth.

Conclusions:

Findings demonstrate the heterogeneous course of BP features in early adolescence and shed light on the potential prodromal course of later borderline personality disorder.

Keywords

attention-deficit hyperactivity disorder borderline personality features diathesis–stress trajectory analysis

Understanding the unfolding of personality disturbance in childhood and adolescence is crucial to illuminating our understanding of adult personality psychopathology and to informing prevention and intervention science. As personality disturbance does not simply manifest de novo in adulthood, research exploring the developmental precipitants and comorbid mental health psychopathology in young people with elevated personality disturbance affords an important opportunity to better understand specific vulnerabilities and prodromal features which may foreshadow the later emergence of clinical disorder.^1
–3

Borderline personality disorder (BPD) is characterized by an enduring “pattern of instability in personal relationships, self-image, and affect, and marked impulsivity.”⁴ BPD is estimated to affect 1.6% to 5.9% of the general population and about 10% of people seen in outpatient settings and 20% of people seen in inpatient settings.⁴ High rates of comorbidity exist between BPD and anxiety and depressive disorders,⁵ and there is substantial overlap between BPD and attention-deficit hyperactivity disorder (ADHD) and other problems on the externalizing spectrum, such as impulsivity and poor self-regulation, regarding their behavioural and neuropsychological impairments.⁶ Increasing evidence indicates that borderline personality (BP)-related characteristics or BP-related features are present in children and adolescents and that they confer risk for the development of BPD in adulthood.^2,7

–11

Empirical research progress has highlighted numerous environmental and biological factors and their interactive influence in the development of BPD in adulthood.^12,13 Broadly, it is thought that both adverse early attachment or interpersonal experiences (for example, abuse or neglect) and the core personality traits of affective dysfunction (for example, emotional reactivity and lability) and disinhibition (for example, impulsivity) are integral for the development of BPD in adulthood.¹⁴ Indeed, most contemporary theoretical accounts of the pathogenesis of BPD in adulthood^15

–18 suggest a diathesis–stress process involving these signature risk factors. Despite this generative research, much remains to be understood about the etiological course of BPD. Perhaps most crucial, is whether early subthreshold markers of the disorder have unique developmental patterns of growth across childhood and adolescence and whether they may be differentially associated with biological and environmental risk factors. This question is particularly important given the clinical characterization of the stable instability of patients with BPD¹⁹ and longitudinal research which indicates considerable instability of BPD diagnoses, with at least 50% to 70% of adolescents shifting to a subclinical level of symptoms at subsequent assessments.^8,20,21 As such, it is reasonable to suspect that the potential prodromal course of this personality disturbance in childhood and adolescence may be characterized by multinomial heterogeneity with important individual differences in both the magnitude and direction of change.²²

Guided in part by extant theoretical models of the pathogenesis of BPD in adulthood, as well as the high degree of covariation between BPD and both internalizing and externalizing psychopathology, we examined associations of childhood intra-individual and interpersonal risk factors from ages 8, and 10 to 12 with person-centred trajectories of BP features for ages 13 to 16 in a community-based sample. To the best of our knowledge, it is the first study to identify and examine the developmental course of BP features in adolescence. Research of this kind is important as little work has examined childhood risk factors for the later development of BPD symptoms or clinical BPD.^6,23 This is especially the case regarding the influence of peer social functioning on the development of BP features.² In addition, given limited empirical tests of the biosocial diathesis–stress model of borderline etiology,^15,16 we also examined whether children who carry a putative biological diathesis of reactive temperament were more likely to be in an elevated BP features trajectory group if they also experience being bullied in childhood.

Methods

Participants and Procedure

Data were drawn from the McMaster Teen Study, an on-going study examining the relations between bullying, mental health, and academic achievement.²⁴ The study began in the spring of 2008 (Grade 5, age 10; time 1 [T1]). Participants were initially recruited from a random sample of 51 schools within a large Southern Ontario Public School Board. From the recruitment process, 875 participants agreed to be part of the longitudinal arm of the study. Parental consent and child assent were obtained at each time point. The study has maintained consistent yearly approval status from the pertinent university research ethics boards (McMaster University and the University of Ottawa). Data collected annually from T1 (Grade 5; age 10) to T7 (Grade 11; age 16) were included in the present study, as were data from participants’ official school records from Grade 3 (age 8).

Data for our study were selected on the basis of 566 (252 [44.5%] boys, 314 [55.5%] girls) participants who had valid reports of BP features for at least 1-time point across T4 (Grade 8; age 13) to T7 (Grade 11; age 16), thus permitting inclusion in trajectory analyses. Among the analytic sample, 89.8% (n = 508) had complete BP features data at T4, 85.5% (n = 484) had complete data at T5, 79.2% (n = 448) had complete data at T6, and 73.7% (n = 417) had complete data at T7. Data at all 4 time points were available for 62.4% (n = 353) of the sample.

Measures

Indicators of Borderline Personality Features Trajectories

The Borderline Personality Features Scale for Children (BPFS-C) was used to examine child-reported features of BP from T4 (age 13) to T7 (age 16).² The BPFS-C consists of 24 items rated on a 5-point Likert-type scale, with the response set including 0 = not at all true, 1 = hardly ever true, 2 = sometimes true, 3 = often true, and 4 = always true. The BPFS-C is a psychometrically sound measure of BP features.^2,25 Criterion validity with reference to clinical diagnoses of BPD in youth has also been established.²⁶ A BP features composite was calculated as an average of all items. Internal consistencies for the BPFS-C in the present sample were 0.90 at T4 (age 13), 0.91 at T5 (age 14), 0.91 at T6 (age 15), and 0.91 at T7 (age 16).

Predictors of Borderline Personality Features Trajectories

An adaptation of the Olweus Bully/Victim Questionnaire^27,28 was used to measure peer victimization. Internal consistency for the peer victimization items was good at each time point (T1 Cronbach α = 0.82; T2 Cronbach α = 0.80; T3 Cronbach α = 0.82). Scores were averaged across T1 to T3 (ages 10 to 12) to form a composite measure of peer victimization (composite Cronbach α = 0.74).

A shortened version of the Aggressive Behaviour Scale²⁹ was used to assess T1 to T3 (ages 10 to 12) child-reported physical and relational aggression. Children responded on a 4-point scale (0 = not at all true to 3 = completely true). Internal consistencies for both forms of aggression over time were good (T1 Cronbach α = 0.87, T2 Cronbach α = 0.85, T3 Cronbach α = 0.84 for physical aggression and T1 Cronbach α = 0.81, T2 Cronbach α = 0.80, T3 Cronbach α = 0.80 for relational aggression). Physical and relational aggression scores were averaged across T1 to T3 (ages 10 to 12) to provide composite measures of childhood physical and relational aggression (composite Cronbach αs = 0.84 and 0.76 for physical and relational aggression, respectively).

Parent-reported child anxiety, depression, ADHD, and somatization symptoms were assessed from T1 to T3 using the Brief Child and Family Phone Interview Version 3 (BCFPI-3)³⁰ and the parent report somatization subscale of the Behavior Assessment System for Children, Second Edition (BASC-2).³¹ The BCFPI-3 is a 30-minute structured phone interview used to screen for emotional and behavioural issues in children ages 3 to 18 years. The BASC-2 is a multi-method, multidimensional measure of behaviour and self-perceptions of people aged between 2 and 25. Items were rated on a 3-point Likert scale (0 = never to 2 = often). Responses were summed within each scale (items reversed where necessary) to yield total scores. Internal consistency (Cronbach α) of the BCFPI-3 subscales across T1 to T3 (age 10 to 12) ranged from 0.81 to 0.82 (anxiety), 0.81 to 0.82 (depression), and 0.84 to 0.81 (symptoms of ADHD). Scales were averaged across T1 to T3 to provide composite measures of parent-reported symptoms of anxiety, depression, and ADHD symptoms. Parent-report data on the BASC-2 somatization subscale were collected at T3. Internal consistency of this subscale at T3 was 0.75.

Child-reported symptoms of anxiety, depression, and hyperactivity symptoms were assessed from T1 to T3 (ages 10 to 12), and child-reported attention problems and somatization symptoms at T3, using the Self-Report of Personality (SRP) form of the BASC-2.³¹ Attention problem and somatization symptoms were assessed at T3 as these scales are only included in the adolescent version of the BASC-2. Internal consistency of the BASC-2 subscales at relevant time points ranged from 0.85 to 0.88 (anxiety), 0.87 to 0.90 (depression), 0.78 to 0.80 (hyperactivity), 0.83 (attention problems), and 0.68 (somatization problems). Child-reported symptoms of psychopathology were averaged across T1 to T3 to provide composite measures of anxiety, depression, and hyperactivity symptoms. Additionally, the hyperactivity composite (T1 to T3) and the attention problems score at T3 (r = 0.62) were averaged to yield a child report ADHD symptom composite.

Child-reported reactive temperament was measured at T1 and T2 (ages 10 and 11) using the Frustration or Negative Affect subscale of The Early Adolescent Temperament Questionnaire–Revised (EATQ-R).³² Items were measured on a 5-point response scale (0 = very false to 4 = very true). This scale consisted of 7 items that tap negative affect related to interruption of ongoing tasks or goal blocking. Items were averaged to yield a total scale score. Internal consistency of the negative affectivity scale at both T1 (Cronbach α = 0.82) and T2 (Cronbach α = 0.79) was good. Scale scores at T1 and T2 were averaged to form a negative affectivity composite.

The Education Quality and Accountability Office (EQAO)³³ is an Ontario province-wide standardized assessment of reading, writing, and mathematics administered to all students attending a publicly funded school at the end of grades 3 and 6. The Grade 3 (age 8) EQAO assessment was used in our study. The EQAO provides an objective evaluation of students’ cumulative reading, writing, and mathematics ability throughout the elementary division. For each category, students answer numerous multiple choice, short, and long answer questions, that are then converted into a single measure of achievement. Correlations among the Grade 3 reading, writing, and mathematics assessment scores were strong (mean r = 0.67) and were averaged to form an index of general academic skills.

Semi-parametric group-based methods were used to identify the number and shape of distinct trajectories of BP features using 4 cycles of data (T4 [Grade 8; age 13] –T7 [Grade 11; age 16]). Models were estimated in Mplus 7.11³⁴ through latent class growth analysis. Given the lack of any prior theoretical or empirical work suggesting distinct trajectories of BP features in adolescence, up to a 5-class solution was examined. Evaluation of the best fitting models was based on consideration of the following criteria: 1) the Bayesian information criterion (BIC); 2) the Lo-Mendell-Rubin adjusted likelihood ratio test (LMR-LRT); 3) entropy; 4) a conceptually clear model; and 5) a model with a sufficient number of members in each group to allow for examination of group differences.

Results

Empirically Derived Borderline Personality Features Trajectory Groups

Evaluation of model fit statistics (Figure 1) suggested that a 3-class solution was both empirically and conceptually justified. The 3-group solution is presented in Figure 1. In the 3-group trajectory model, most adolescents demonstrated an intermediate or stable level of BP features (42% of the sample; n = 235; 107 boys, 128 girls) with the second largest group following a low or stable pattern of BP features (34%; n = 195; 109 boys, 86 girls). Finally, there were a group of youth who followed an elevated or rising pattern of BP features (24%; n = 136; 36 boys, 100 girls). Examination of posterior probabilities indicated that children were well matched to their group (0.85 for the intermediate or stable, 0.92 for the low or stable, and 0.90 for the elevated or rising increasing groups, respectively).

Figure 1.

Three trajectory model of borderline personality features. Fit indices for 1 trajectory model (Bayesian Information Criteria [BIC] = 3541.67; Lo-Mendell-Rubin likelihood ratio test [LMR-LRT], not applicable; entropy, not applicable), 2 trajectory (BIC = 2837.52; LMR-LRT, P < 0.001; entropy = 0.78), 3 trajectory (BIC = 2633.15; LMR-LRT, P < .001; entropy = 0.75), 4 trajectory (BIC = 2602.85; LMR-LRT, P = 0.49; entropy = 0.78); 5 trajectory (BIC = 2589.00; LMR-LRT, P = 0.15; entropy = 0.72).

There were differences in the proportion of boys and girls in the BP features trajectory groups (χ² = 28.26, df = 2, P < 0.001). Examination of adjusted standardized residuals in the differences in sex proportions across the 3 trajectory groups indicated that there were significantly more girls in the elevated or rising group (31.8% girls, compared with 14.3% boys) and significantly more boys in the low or stable group (43.3% boys, compared with 27.4% girls).

Covariates Associated with Trajectory Group Membership

Analyses of variance procedures were used to explore whether mental health, peer relations, and intra-individual risk factors differentiated among the BP features trajectory groups. The Benjamini-Hochberg (BH)³⁵ correction for multiple testing was used to protect against type 1 error. For these analyses, estimates of effect size (Cohen d) are provided.³⁶ In follow-up multinomial and logistic regression analyses, the predictive significance of each covariate for membership in the empirically-derived BP features trajectory groups was also examined in conjunction with child sex given evidence that BPD prevalence rates may be higher among women than men.^4,37,38 Because relatively few such sex interactions emerged, we note in Table 2 when effects were significantly larger for boys or girls and describe below relevant interaction effects. Finally, we also tested if some of the covariates were better predictors of BP features trajectory group than others by simultaneously entering all predictors, using the low or stable group as the reference group in a multinomial logistic regression analyses and the intermediate or stable group as the reference group in a binary logistic regression analyses comparing the intermediate or stable and elevated or rising groups.

Means and standard deviations of risk factors by BP features trajectory group are presented in Table 1. As shown in Table 2, all of the indicators with the exception of the general academic competence indicator distinguished the intermediate or stable and the elevated or rising trajectory groups from the low stable trajectory group (BH corrected). Follow-up multinominal regressions revealed that 1 of 26 effects were larger for girls than boys, while 2 of 26 effects were larger for boys than girls. Regarding contrasts between the elevated or rising group and the intermediate or stable group, only child-report measures of mental health symptoms, peer relations, and temperament were significant discriminators of membership in the elevated or rising group compared with the intermediate or stable group (BH corrected). There were no interactions with child sex in follow-up logistic regressions.

Table 1.

Means and standard deviations (SD) of antecedent symptoms of psychopathology, peer relations, and intra-individual risk variables by empirically derived borderline personality features trajectory groups.

Outcome	Borderline personality features trajectory group
	Low or stable		Intermediate or stable		Elevated or rising
	Mean	SD	Mean	SD	Mean	SD
Mental health parent report
Anxiety	3.62	2.35	4.34	2.34	4.43	2.46
Depression	0.68	1.20	1.15	1.44	1.40	1.58
ADHD	3.22	2.43	4.30	2.82	4.49	2.76
Somatization	3.73	2.86	4.48	3.46	5.04	3.25
Mental health child report
Anxiety	6.68	4.45	9.55	5.04	12.43	5.18
Depression	2.80	3.08	4.64	4.42	7.08	4.86
ADHD	3.87	2.71	6.12	3.44	7.66	4.04
Somatization	1.50	2.32	2.35	2.76	3.94	3.37
Peer Relations child report
Peer victimization	0.48	0.43	0.74	0.62	0.92	0.61
Physical aggression	0.25	0.29	0.38	0.39	0.49	0.45
Relational aggression	0.28	0.23	0.40	0.29	0.51	0.34
Intra-individual risk
Child temperament	1.98	0.68	2.20	0.74	2.51	0.57
General academic skills (EQAO Composite); grade 3	3.21	0.59	3.25	0.57	3.21	0.50

ADHD = attention-deficit hyperactivity disorder; EQAO = The Education Quality and Accountability Office.

Table 2.

Planned comparisons on risk factors with borderline personality features trajectory groups.

Outcome	Trajectory group contrasts
	Intermediate or stable (n = 235; 41.5%, compared with low or stable (n = 195; 34.5%)			Elevated or rising (n = 136; 24.0%), compared with low or stable (n = 195; 34.5%)			Elevated or rising (n = 136; 24.0%), compared with intermediate or stable (n = 235; 41.5%)
	B (SE)	P	d	B (SE)	P	d	B (SE)	P	d
Mental health parent report
Anxiety	0.72 (0.23)	0.002^a	0.31	0.80 (0.27)	0.003^a	0.34	0.08 (0.26)	0.75	0.04
Depression	0.48 (0.14)	0.001^a	0.36	0.73 (0.16)	<0.001^a	0.53	0.25 (0.15)	0.10	0.17
ADHD	1.09 (0.26)	<0.001^a	0.41	1.27 (0.30)	<0.001^a	0.50	0.19 (0.29)	0.52	0.07
Somatization	0.76 (0.33)	0.02^a	0.24	1.32 (0.38)	0.001^a	0.44	0.56 (0.36)	0.13	0.17
Mental health child report
Anxiety	2.87 (0.47)^b	<0.001^a	0.60	5.75 (0.55)	<0.001^a	1.21	2.88 (0.53)	<0.001^a	0.57
Depression	1.84 (0.40)	<0.001^a	0.48	4.28 (0.46)	<0.001^a	1.10	2.44 (0.45)	<0.001^a	0.53
ADHD	2.25 (0.33)	<0.001^a	0.72	3.79 (0.38)	<0.001^a	1.15	1.54 (0.36)	<0.001^a	0.42
Somatization	0.85 (0.28)	0.003^a	0.33	2.44 (0.33)	<0.001^a	0.88	1.59 (0.32)	<0.001^a	0.53
Peer relations (child report)
Peer victimization	0.26 (0.05)	<0.001^a	0.48	0.44 (0.06)	<0.001^a	0.87	0.18 (0.06)	0.003^a	0.30
Physical aggression	0.13 (0.04)	<0.001^a	0.38	0.24 (0.04)^c	<0.001^a	0.67	0.11 (0.04)	0.006^a	0.27
Relational aggression	0.12 (0.03)	<0.001^a	0.47	0.23 (0.03)	<0.001^a	0.85	0.11 (0.03)	<0.001^a	0.37
Intra-individual risk
Child temperament	0.23 (0.07)^b	0.001^a	0.32	0.53 (0.08)	<0.001^a	0.84	0.30 (0.07)	<0.001^a	0.45
General academic skills (EQAO composite); Grade 3	0.04 (0.06)	0.51	0.07	–0.00 (0.07)	0.98	0.00	–0.04 (0.07)	0.54	0.08

^aSignificant at the Benjamini–Hochberg adjusted P value; ^bEffect significantly larger for boys; ^cEffect significantly larger for girls

ADHD = attention-deficit hyperactivity disorder; EQAO = The Education Quality and Accountability Office

Analysis of variance F statistics were significant for each measure except General Academic Skills (EQAO Composite).

Estimates of our multinomial model in which all predictors were entered simultaneously are presented in Table 3. Only parent-reported anxiety and child-reported ADHD symptoms were associated with the intermediate or stable trajectory group relative to the low or stable group. Membership in the elevated or rising trajectory group (compared with the low or stable group) was associated with child-reported ADHD and somatization symptoms and relational aggression. Lastly, we conducted a binomial logistic regression model comparing the elevated or rising group to the intermediate or stable group, again entering all predictors simultaneously. The only indicator to significantly distinguish these groups was child-reported somatization symptoms (OR 1.15, 95% CI 1.04 to 1.28).

Table 3.

Multinomial logistic regression of empirically derived borderline personality features trajectory groups with symptoms of psychopathology, peer relations, and intra-individual risk factors: combined model.

Outcome	Trajectory group contrasts
	Intermediate, compared with low or stable				Elevated or rising, compared with low or stable
	B (SE)	e^b	P	95% CI	B (SE)	e^b	P	95% CI
Mental health parent report
Anxiety	0.15 (0.06)	1.17	0.01	(1.04 to 1.31)	12 (0.08)	1.13	0.11	(0.97 to 1.31)
Depression	0.06 (0.13)	1.07	0.63	(0.82 to 1.38)	0.09 (0.15)	1.10	0.55	(0.81 to 1.48)
ADHD	–0.03 (0.06)	0.98	0.69	(0.86 to 1.11)	–0.04 (0.08)	0.96	0.59	(0.82 to 1.12)
Somatization	–0.02 (0.05)	0.98	0.66	(0.89 to 1.08)	–0.03 (0.06)	0.97	0.56	(0.86 to 1.08)
Mental health child report
Anxiety	0.07 (0.04)	1.07	0.08	(0.99 to 1.15)	0.07 (0.05)	1.07	0.13	(0.98 to 1.18)
Depression	–0.01 (0.06)	0.99	0.81	(0.88 to 1.10)	0.05 (0.06)	1.05	0.47	(0.93 to 1.18)
ADHD	0.23 (0.06)	1.26	<0.001	(1.12 to 1.42)	0.28 (0.07)	1.33	<0.001	(1.16 to 1.52)
Somatization	0.09 (0.06)	1.09	0.15	(0.97 to 1.24)	0.24 (0.07)	1.27	0.001	(1.11 to 1.46)
Peer relations (child report)
Peer victimization	0.41 (0.32)	1.50	0.20	(0.81 to 2.78)	0.23 (0.37)	1.25	0.54	(0.61 to 2.57)
Physical aggression	–0.36 (0.53)	0.72	0.53	(0.25 to 2.02)	–0.07 (0.58)	0.93	0.90	(0.30 to 2.92)
Relational aggression	0.78 (0.64)	2.17	0.23	(0.62 to 7.63)	1.74 (0.73)	5.67	0.02	(1.36 to 23.58)
Intra-individual risk
Child temperament	–0.02 (0.22)	0.98	0.93	(0.64 to 1.50)	0.20 (0.29)	1.22	0.50	(0.69 to 2.15)
General academic skills (EQAO composite); Grade 3	0.32 (0.25)	1.37	0.21	(0.84 to 2.24)	0.23 (0.30)	1.26	0.44	(0.70 to 2.29)

ADHD = attention-deficit hyperactivity disorder; EQAO = The Education Quality and Accountability Office.

Test of Diathesis–Stress for Borderline Personality Features Trajectory Group Membership

In a final analysis we tested whether early adolescent child-reported reactive temperament interacted with peer victimization to predict BP features trajectory group. In this model child-reported reactive temperament was considered a proxy reflecting biological susceptibility to the development of BP when triggered by childhood bullying (that is, harsh interpersonal experiences; a diathesis–stress model). As in prior work,⁷ we defined interaction as the epidemiological construct of biological synergy³⁹ in which the effect of 2 risk factors present together is greater than the sum of their independent effects. We also used the 2 × 4 table method⁴⁰ to present relative risks among children with different combinations of reactive temperament and being bullied. Because the 2 × 4 method is designed for binomial outcome variables and requires dichotomization of continuous treatment measures we collapsed the intermediate or stable and elevated or rising trajectory groups into an elevated BP features group and dichotomized at the 80th percentile both the focal predictor (child-reported reactive temperament, scale score = 2.79) and the moderator (peer victimization, scale score = 1.13). Biological interaction was evaluated by calculating 3 measures including the relative excess risk due to interaction (RERI) relative to the risk without exposure, the attributable portion (AP) due to interaction, which is interpreted as the AP of membership in the elevated BP features trajectory group due to interaction among children possessing both risk factors, and the synergy (S) index, which can be interpreted as the excess risk from both risk factors when there is an interaction relative to the excess risk from both risk factors without interaction. If there is no biological interaction, RERI and AP will equal 0 and S will equal 1.⁴¹

The point estimates of all 3 interaction metrics suggested evidence for biological interaction between child-reported reactive temperament and peer victimization in the prediction of elevated BP features trajectory group membership (Table 4). Children possessing both risk factors of child-reported reactive temperament (top 20%) and being bullied (top 20%) were 10.23 (95% CI 2.40 to 43.58) times more likely to be in the elevated borderline features trajectory group than were children with neither risk factor. Evidence for biological interaction was also obtained (although attenuated in magnitude) when using the original continuous metric of the predictor variables in follow-up analyses using a method developed by Knol et al⁴² (results available on request).

Table 4.

Two × 4 table of diathesis–stress interaction between child-reported childhood reactive temperament and peer victimization in predicting membership in elevated borderline personality features trajectory groups.

Exposure		n		Effect size
Reactive temperament	Peer victimization	Low or stable	Elevated	RR (95% CI)
+	–	19	52	[RR₀₁] 2.00 (1.14 to 3.52)
–	+	12	65	[RR₁₀] 3.96 (2.07 to 7.58)
+	+	2	28	[RR₁₁] 10.23 (2.40 to 43.58)
–	–	162	226	REF [1.0]
Total		195	371
		Biological interaction measure		Point estimate (95% CI)
		RERI		5.27 (–9.67 to 20.21)
		AP		0.52 (–0.23 to 1.26)
		S		2.33 (0.41 to 13.13)

AP = attributable portion due to interaction; RERI = relative excess risk due to interaction; RR = relative risk; S = synergy index.

RERI is calculated as RR₁₁ – RR₁₀ – RR₀₁ + 1; AP = RERI/RR₁₁; S = [RR₁₁ - 1]/[(RR₁₀ -1) + (RR₀₁ - 1)].

The reference group (REF) are individuals who possess neither exposure factor (that is, not in the top 20% of either reactive temperament or peer victimization).

Discussion

Our study is the first to our knowledge to model person-centred developmental trajectories of BP features over the course of adolescence and examine their associations with non-age-overlapping antecedent risk factors in a community-based sample. Trajectory modelling supported the identification of 3 distinct BP features trajectory groups—low or stable, intermediate or stable, and elevated or rising—and provides further evidence that a distinct group of people with elevated BP features are present and identifiable in childhood and adolescence.^2,7,9 The preponderance of girls in the elevated or rising BP features trajectory group provides preliminary support for a downward extension of sex differences that have been noted in BPD diagnoses in adults^4,37,38 and is consistent with prior work using the BPFS-C which found that girls had elevated total scores relative to boys.² Nevertheless, empirical research regarding sex differences in BPD has been inconsistent^13,43,44 and remains a question of much interest for clinicians and researchers.⁴⁵

Results of planned comparisons are consistent with existing empirical evidence indicating high rates of comorbidity and symptom overlap between BPD and anxiety and depressive disorders, as well as the common behavioural and neuropsychological impairments between BPD and ADHD and other symptoms on the externalizing spectrum.^5,46 These results also converge with more recent work^7,47,48 which has found associations between childhood ADHD, internalizing, and externalizing symptoms, lower IQ, and poor behavioural control, and subsequent adolescent BPD symptoms.

Burke and Stepp⁴⁷ and Stepp et al⁴⁸ have called particular attention to the possibility that childhood ADHD may represent part of an early developmental path to BPD as both disorders share a common impulsivity component. In addition, in a recent prospective follow-up of girls,⁴⁹ it was found that ADHD was associated with self-harm behaviour in young adults. Thus some of the biological states that underlie the development of BPD and other mood disorders may present in the context of ADHD symptoms or more broadly as trait impulsivity in childhood.⁵⁰

To our knowledge, no studies have examined the predictive significance of childhood somatization symptoms for the development of BP features in adolescence. One possibility explaining the robust predictive significance of somatization symptoms is that they may reflect prodromal elements of the histrionic or emotional hypochondriasis element of BPD as they reflect the tendency of the child or adolescent to be overly sensitive and complain about relatively minor physical problems or ailments, and to overreport the occurrence of various physical complaints—all of which by definition are help seeking and involve attaining the attention of caretaking figures.⁵¹ Findings for relational aggression are consistent with previous work^23,52 calling attention to this form of aggression as particularly relevant in the development of BP symptoms. Lastly, the few predictor interaction effects with child sex suggested that they may operate differently for girls and boys. Additional research examining the moderating role of child sex in predicting adolescent BP features is needed.

The results of our diathesis–stress interaction test suggested that youth were more likely to be in an elevated BP features trajectory group if they experienced higher levels of peer victimization in the context of their own self-reported reactive temperament. These findings provide suggestive evidence that the invalidating environment component of the diathesis–stress model of BPD¹⁵ etiology may continue to be more fully developed to include those occurring in the context of peer relations.¹⁶ They also underscore the importance of identifying children at risk for or currently experiencing high levels of peer victimization for potential prevention services or mental health intervention.

Limitations

Several limitations of our study should be considered when interpreting our results. We were unable to control for early family or interpersonal variables (for example, physical maltreatment) as these types of data were not available in the McMaster Teen Study. Biological and genetic markers were also not available in this study. Evidence that adult BPD is highly heritable,⁵³ suggests that genetic factors play an important role in the etiology of BP features. Finally, shared method variance may in part account for unique child report (but not parent report) predictive effects on elevated or rising BP features trajectory group membership. Despite these limitations, a key strength of the study was the temporal precedence of predictor variable sets from BP features data.

Conclusions

Our findings provide important evidence to mental health practitioners to identify ports of entry for prevention and intervention strategies aimed at reducing the development of BP features in adolescence and young adulthood.^23,54 It is important to point out, however, that this study, as with other studies of BP symptoms or features in childhood and adolescence,^{2,5,7,23,46,47,55} did not examine a clinical diagnosis of BPD and that absolute levels of BP features for people in the elevated or rising trajectory group were modest in degree. An important task for future work will be to examine whether elevated BP features in childhood and adolescence are predictive of later clinical BPD diagnoses.

Footnotes

Acknowledgments

We thank Heather Brittain, Amanda Krygsman, and Patricia McDougall for the help with the study.

Author Note

All authors were involved in all aspects of the study; authors contributed to the design and implementation and interpretation of the results. All authors contributed to the review, writing, and editing of the manuscript. The authors maintained control over the final contents of the manuscript and the decision to publish in The Canadian Journal of Psychiatry. Analyses were conducted by Dr Haltigan.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and (or) publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and (or) publication of this article: This study was supported by the Social Sciences and Humanities Research Council of Canada (833-2004-1019) and the Canadian Institutes for Health Research (201009MOP-232632-CHI-CECA-136591).

References

Cicchetti

Crick

. Precursors and diverse pathways to personality disorder in children and adolescents. Dev Psychopathol. 2009;21:683–685.

Crick

Murray-Close

Woods

. Borderline personality features in childhood: a short-term longitudinal study. Dev Psychopathol. 2005;17:1051–1070.

Lenzenweger

Cicchetti

. Toward a developmental psychopathology approach to borderline personality disorder. Dev Psychopathol. 2005;17:893–898.

American Psychiatric Association (APA). Diagnostic and statistical manual of mental disorders. 5th ed. Washington (DC): APA; 2013. p. 645.

Zanarini

Frankenburg

Dubo

. Axis 1 comorbidity of borderline personality disorder. Am J Psychiatry. 1998;155:1733–1739.

Beauchaine

Klein

Crowell

. Multifinality in the development of personality disorders: a biology × sex × environment interaction model of antisocial and borderline traits. Dev Psychopathol. 2009;21:735–770.

Belsky

Caspi

Arseneault

. Etiological features of borderline personality related characteristics in a birth cohort of 12-year old children. Dev Psychopathol. 2012;24:251–265.

Bernstein

Cohen

Velez

. Prevalence and stability of the DSM-III-R personality-disorders in a community-based sample of adolescents. Am J Psychiatry. 1993;150:1237–1243.

Trull

. Borderline personality features in nonclinical young adults: 1. Identification and validation. Psychol Assess. 1995;7:33–41.

10.

Chabrol

Montovany

Chouicha

. Frequency of borderline personality disorder in a sample of French high school students. Can J Psychiatry. 2001;46:847–849.

11.

Winograd

Cohen

Chen

. Adolescent borderline symptoms in the community: prognosis for functioning over 20 years. J Child Psychol Psychiatry. 2008;49:933–941.

12.

Lieb

Zanarini

Schmahl

. Borderline personality disorder. Lancet. 2004;364:453–461.

13.

Leichsenring

Leibing

Kruse

. Borderline personality disorder. Lancet. 2011;377:74–84.

14.

Gratz

Tull

Reynolds

. Extending extant models of the pathogenesis of borderline personality disorder to childhood borderline personality symptoms: the roles of affective dysfunction, disinhibition, and self-and emotion-regulation deficits. Dev Psychopathol. 2009;21:1263–1291.

15.

Linehan

. Cognitive-behavioral treatment of borderline personality disorders. New York (NY): Guilford Press; 1993.

16.

Crowell

Beauchaine

Linehan

. A biosocial developmental model of borderline personality: elaborating and extending Linehan’s theory. Psychol Bull. 2009;135:495–510.

17.

Paris

. Childhood trauma as an etiological factor in personality disorders. J Pers Disord. 1997;11:34–49.

18.

Gunderson

Lyons-Ruth

. BPD’s interpersonal hypersensitivity phenotype: a gene-environment-developmental model. J Pers Disord. 2008;22:22–41.

19.

Lenzenweger

Castro

. Predicting change in borderline personality: using neurobehavioral systems indicators within an individual growth curve framework. Dev Psychopathol. 2005;17:1207–1237.

20.

Chanen

Jackson

McGorry

. Two-year stability of personality disorder in adolescent outpatients. J Pers Disord. 2004;18:526–541.

21.

Grilo

Becker

Edell

. Stability and change of DSM-III-R personality disorder dimensions in adolescents followed up 2 years after psychiatric hospitalization. Compr Psychiatry. 2001;42:364–368.

22.

Nagin

. Group-based modeling of development. Cambridge (MA): Harvard University Press; 2005.

23.

Underwood

Beron

Rosen

. Joint trajectories for social and physical aggression as predictors of adolescent maladjustment: internalizing symptoms, rule-breaking behaviors, and borderline and narcissistic personality features. Comprehensive Psychiatry. 2011;23:659–678.

24.

Vaillancourt

Brittain

McDougall

. Longitudinal links between childhood peer victimization, internalizing and externalizing problems, and academic functioning: developmental cascades. J Abnorm Child Psychol. 2013;41:1203–1215.

25.

Sharp

Mosko

Chang

. The cross-informant concordance and concurrent validity of the Borderline Personality Features Scale for Children in a community sample of boys. Clin Child Psychol Psychiatry. 2011;16:335–349

26.

Chang

Sharp

. The criterion validity of the Borderline Personality Features Scale for Children in an adolescent inpatient setting. J Pers Disord. 2011;25:492–503.

27.

Olweus

. The revised Olweus Bully/Victim Questionnaire. Bergen (NO): Research Center for Health Promotion (HEMIL Center), University of Bergen; 1996.

28.

Vaillancourt

Trinh

McDougall

. Optimizing population screening of bullying in school-aged children. J Sch Violence. 2010;9:233–250.

29.

Little

Henrich

Jones

. Disentangling the “whys” from the “whats” of aggressive behaviour. Int J Behav Dev. 2003;27(2):122–133.

30.

Cunningham

Pettingill

Boyle

. The Brief Child and Family Phone Interview (BCFPI). Hamilton (ON): Canadian Centre for the study of Children at Risk, Hamilton Health Sciences Corporation, McMaster University; 2000.

31.

Reynolds

Kamphaus

. Behavior assessment system for children–2nd edition manual. Minneapolis (MN): Pearson; 2004.

32.

Ellis

Rothbart

. Revision of the Early Adolescent Temperament Questionnaire. Poster presented at the 2001 Biennial Meeting of the Society for Research in Child Development; 2001; Minneapolis (MN).

33.

Education Quality Accountability Office (EQAO). EQAO’s Executive Summary of the Technical Report for the 2008–2009 Assessments of Reading, Writing and Mathematics, Primary Division, (Grades 1–3) and Junior Division (Grades 4–6), Grade 9 Assessment of Mathematics, and Ontario Secondary School Literacy Test; 2010. Toronto (ON): EQAO; 2011.

34.

Muthén

. Mplus user’s guide. 7th ed. Los Angeles (CA): Múthen & Múthen; 1998–2012.

35.

Benjamini

Hochberg

. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. 1995;57:289–300.

36.

Cohen

. A power primer. Psychol Bull. 1992;112:155–159.

37.

Sansone

. Gender patterns in borderline personality disorder. Innov Clin Neurosci. 2011;8:16–20.

38.

Swartz

Blazer

George

. Estimating the prevalence of borderline personality disorder in the community. J Pers Disord. 1990;4:257–272.

39.

Rothman

Greenland

Walker

. Concepts of interaction. Am J Epidemiol. 1980;112:467–470.

40.

Botto

Khoury

. Commentary: Facing the challenge of gene-environment interaction: the two-by-four table and beyond. Am J Epidemiol. 2001;153:1016–1020.

41.

Andersson

Alfredsson

Källberg

. Calculating measures of biological interaction. Eur J Epidemiol. 2005;20:575–579.

42.

Knol

van der Tweel

Grobbee

. Estimating interaction on an additive scale between continuous determinants in a logistic regression model. Int J Epidemiol. 2007;36:1111–1118.

43.

De Moor

Distel

Trull

. Assessment of borderline personality features in population samples: is the Personality Assessment Inventory-Borderline Personality Features Scare measurement invariant across sex and age? Psychol Assess. 2009;21:125–130.

44.

Zanarini

Horwood

Wolke

. Prevalence of DSM-IV borderline personality disorder in two community samples: 6,330 English 11-year-olds and 34,653 American adults. J Pers Disord. 2011;25:607–619.

45.

Sharp

Michonski

Steinberg

. An investigation of differential item functioning across gender of BPD criteria. J Abnorm Psychol. 2014;123:231–236.

46.

Grant

Chou

Goldstein

. Prevalence, correlates, disability, and comorbidity of DSM-IV borderline personality disorder: results from the Wave 2 National Epidemiological Survey on Alcohol and Related Conditions. J Clin Psychiatry. 2008;69:533–545.

47.

Burke

Stepp

. Adolescent disruptive behavior and borderline personality disorder symptoms in young adult men. J Abnorm Child Psychol. 2012;40:35–44.

48.

Stepp

Burke

Hipwell

. Trajectories of attention deficit hyperactivity disorder and oppositional defiant disorder symptoms as precursors of borderline personality symptoms in adolescent girls. J Abnorm Child Psychol. 2012;40:7–20.

49.

Hinshaw

Owens

Zalecki

. Prospective follow-up of girls with attention-deficit/hyperactivity disorder into early adulthood: continuing impairment includes elevated risk for suicide attempts and self-injury. J Consult Clin Psychol. 2012;80:1041–1051.

50.

Peluso

MAM

Hatch

Glahn

. Trait impulsivity in patients with mood disorders. J Affect Disord. 2007;100(1–3):227–231.

51.

Zanarini

Franekburg

. Pathways to the development of borderline personality disorder. J Pers Disord. 1997;11:3–104.

52.

Ostrov

Houston

. The utility of forms and functions of aggression in emerging adulthood: association with personality disorder symptomatology. J Youth Adolesc. 2008;37:1147–1158.

53.

Bornovalova

Hicks

Iacono

. Stability, change, and heritability of borderline personality traits from adolescence to adulthood: a longitudinal twin study. Dev Psychopathol. 2009;21:1335–1353.

54.

Carlson

Egeland

Sroufe

. A prospective investigation of the development of borderline personality symptoms. Dev Psychopathol. 2009;21:1311–1334.

55.

Vaillancourt

Brittain

McDougall

. Predicting borderline personality disorder symptoms in adolescents from childhood physical and relational aggression, depression, and ADHD. Dev Psychopathol. 2014;26:817–830.

Identifying Trajectories of Borderline Personality Features in Adolescence

Abstract

Objective:

Method:

Results:

Conclusions:

Keywords

Methods

Participants and Procedure

Measures

Indicators of Borderline Personality Features Trajectories

Predictors of Borderline Personality Features Trajectories

Results

Empirically Derived Borderline Personality Features Trajectory Groups

Covariates Associated with Trajectory Group Membership

Test of Diathesis–Stress for Borderline Personality Features Trajectory Group Membership

Discussion

Limitations

Conclusions

Footnotes

Acknowledgments

Author Note

Declaration of Conflicting Interests

Funding

References