1 N EW R E S E A R C H 59

2 60
3 61
4
5 Borderline Symptoms at Age 12 Signal Risk for Poor 62
63
6 64
7 Outcomes During the Transition to Adulthood: Findings 65
8 66
9
10
Q1Q2
From a Genetically Sensitive Longitudinal Cohort Study 67
68
11 Q3 Jasmin Wertz, PhD, Avshalom Caspi, PhD, Antony Ambler, MSc, Louise Arseneault, PhD, 69
12 70
13 Daniel W. Belsky, PhD, Andrea Danese, MD, PhD, Helen L. Fisher, PhD, Timothy Matthews, PhD, 71
14 Leah Richmond-Rakerd, PhD, Terrie E. Moffitt, PhD 72
15 73
16 74
17 75
18 Objective: Borderline personality disorder in adolescence remains a controversial construct. We addressed concerns about the prognostic significance 76
19 of adolescent borderline pathology by testing whether borderline symptoms at age 12 years predict functioning during the transition to adulthood, at age 77
20 18 years, in areas critical to life-course development. 78
21 Method: We studied members of the Environmental Risk (E-Risk) Longitudinal Twin Study, which tracks the development of a birth cohort of 79
22 2,232 British twin children. At age 12, borderline symptoms of study members were measured using mothers’ reports. At age 18, study members’ 80
23 personality, psychopathology, functional outcomes, and experiences of victimization were measured using self-reports, coinformant reports, and official 81
24 records. 82
25 83
26 Results: At age 18, study members who had more borderline symptoms at age 12 were more likely to have difficult personalities, to struggle with poor
84
27 mental health, to experience poor functional outcomes, and to have become victims of violence. Reports of poor outcomes were corroborated by
85
28 coinformants and official records. Borderline symptoms in study members at 12 years old predicted poor outcomes over and above other behavioral and
86
29 emotional problems during adolescence. Twin analyses showed that borderline symptoms in 12-year-olds were influenced by familial risk, particularly
87
30 genetic risk, which accounted for associations with most poor outcomes at age 18.
88
31 Conclusion: Borderline symptoms in 12-year-olds signal risk for pervasive poor functioning during the transition to adulthood. This association is 89
32 driven by genetic influences, suggesting that borderline symptoms and poor outcomes are manifestations of shared genetic risk. 90
33 Key words: adolescence, borderline, longitudinal, personality, twin 91
34 92
35 J Am Acad Child Adolesc Psychiatry 2019;-(-):-–-.
93
36 94
37 95
38 96
39 97
orderline personality disorder is characterized by adolescence.4 In this study, we addressed concerns about the
40
41
42
43
B pervasive instability in a person’s mood, sense of
self, impulse control, and interpersonal relation-
ships. In adults, borderline personality disorder is consid-
validity of adolescent borderline pathology by testing in
12-year-old adolescents the prognostic significance of
borderline symptoms for psychosocial adjustment during
98
99
100
101
44 ered a valid diagnosis by most clinicians.1 In adolescents, the transition to adulthood, at age 18 years. 102
45 the diagnosis is more controversial.2,3 Although diagnostic In recent years, research has made great strides toward 103
46 104
47
classification systems allow for a diagnosis of borderline establishing the validity of the borderline personality dis- 105
48 personality disorder in adolescence, clinicians are reluctant order diagnosis in adolescents. This research shows that 106
49 to assess and treat borderline symptoms before adult- borderline symptoms can be observed and reliably measured 107
50 hood.4,5 Among the reasons cited for this reluctance are in adolescents, that symptoms are as prevalent in adoles- 108
51 concerns that adolescents’ borderline symptoms may be cents as they are in adults, that symptoms are relatively 109
52 110
transient; that a diagnosis could be stigmatizing; that per- stable across time, and that symptoms predict a diagnosis of
53 111
54 sonality development is still in flux; and that some borderline personality disorder in adulthood.2,6-8 Studies 112
55 borderline symptoms, such as impulsivity and difficulty in also report significant psychosocial impairment in adoles- 113
56 establishing a sense of identity, are inseparable from what is cents who experience borderline symptoms.9 Another 114
57 thought to be a normative degree of storm and stress during approach to testing the validity of borderline personality 115
58 116

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 1

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 WERTZ et al.

pathology is to examine the significance of adolescent age 18 outcomes within genetically identical twin pairs
117 borderline symptoms for adult adjustment. Previous find- growing up in the same family who differed in adolescent 176
118 ings suggest that adolescents who display borderline symp- borderline symptoms when assessed at age 12. Because these 177
119 toms experience adjustment difficulties in adulthood.10 twins share all of their family-wide environment and genes, 178
120 179
121
However, studies investigating the clinical and psychoso- these analyses effectively control for familial risk factors 180
122 cial outcomes of adolescent borderline symptoms are sparse, shared between members of a family. 181
123 and a recent review of the literature concluded that many of 182
124 the studies are limited by problems such as sampling bias, METHOD 183
125 high attrition, and a narrow range of psychosocial out- Participants 184
126 185
comes.11 The aim of our study was to extend previous Participants were members of the E-Risk Longitudinal
127 186
128 research by drawing a comprehensive picture of how ado- Twin Study, which tracks the development of a birth cohort 187
129 lescents with borderline symptoms fare during the transition of 2,232 British children.15 Briefly, the E-Risk sample was 188
130 to adulthood. Seven years ago, we described predictors and constructed in 1999
2000, when 1,116 families (93% of 189
131 correlates of borderline symptoms measured in 12-year-old those eligible) with same-sex 5-year-old twins participated 190
132 study members of the Environmental Risk (E-Risk) Lon- in home-visit assessments. This sample comprised 56% 191
133 192
134
gitudinal Twin Study, a population-representative birth monozygotic (MZ) and 44% dizygotic (DZ) twin pairs; sex 193
135 cohort of twins born in the United Kingdom.12 Study was evenly distributed within zygosity (49% male sex). The 194
136 members have now been followed up to age 18, with high study sample represents the full range of socioeconomic 195
137 retention (93%). At age 18, we assessed study members’ conditions in Great Britain, as reflected in the families’ 196
138 performance on a wide range of outcomes in areas critical to distribution on a neighborhood-level socioeconomic index 197
139 198
positive life-course development: personality functioning, (ACORN [A Classification of Residential Neighborhoods],
140 199
141 mental health, functional outcomes, and experiences of developed by CACI, Inc., Arlington, Virginia, for com- 200
142 victimization. Using these data, we tested the hypothesis mercial use):16,17 25.6% of E-Risk families live in “wealthy Q 4 201
143 that adolescent borderline symptoms predict poor outcomes achiever” neighborhoods compared with 25.3% nation- 202
144 during the transition to adulthood. wide; 5.3% compared with 11.6%, in “urban prosperity” 203
145 In addition to analyzing implications of borderline neighborhoods; 29.6% compared with 26.9%, in 204
146 205
147
symptoms in 12-year-olds for outcomes at age 18, we tested “comfortably off” neighborhoods; 13.4% compared with 206
148 whether symptoms contributed to poor outcomes inde- 13.9%, in “moderate means” neighborhoods; and 26.1% 207
149 pendently of comorbid adolescent psychopathology and compared with 20.7%, in “hard-pressed” neighborhoods. 208
150 familial risk. We tested the role of comorbid psychopa- “Urban prosperity” neighborhoods are underrepresented in 209
151 thology to investigate whether borderline symptoms E-Risk because such households are often childless. 210
152 211
demonstrate incremental validity beyond common disorders Home visits were conducted when the children were 7
153 212
154 that clinicians assess in adolescents who present with years old (98% participation), 10 years old (96%), 12 years 213
155 emotional and behavioral dysregulation, such as conduct old (96%), and 18 years old (93%). At ages 5, 7, 10, and 12 214
156 disorder, depression, and anxiety. Previous studies, years, assessments were carried out with participants as well 215
157 including our own, show that many adolescents who display as their mothers (or primary caretakers); the home visit at 216
158 borderline symptoms also experience symptoms of these age 18 included interviews with participants only. Each 217
159 218
160
other disorders.7,12 We tested the incremental validity of age twin was assessed by a different interviewer. These data are 219
161 12 borderline symptoms by accounting for comorbid supplemented by searches of official records and by ques- 220
162 behavioral and emotional problems when evaluating effects tionnaires that are mailed, as developmentally appropriate, 221
163 of adolescent borderline symptoms on age 18 outcomes. to teachers and coinformants nominated by participants 222
164 We tested the role of familial risk because adolescent themselves. There were no differences between participants 223
165 224
borderline behaviors are strongly influenced by risk factors who did and did not take part at age 18 in terms of so-
166 225
167 originating in families, both environmental and genetic.12,13 cioeconomic status assessed when the cohort was initially 226
168 Familial risk factors implicated in adolescent borderline defined (c2 ¼ 0.86, p ¼ .65), age 5 IQ scores (t ¼ 0.98, 227
169 symptoms, such as harsh parenting, maltreatment, and ge- p ¼ .33), age 5 behavioral and emotional problems (t ¼ 228
170 netic susceptibility, also predict psychosocial adjustment in 0.40, p ¼ .69 and t ¼ 0.41, p ¼ .68, respectively) or age 12 229
171 young adulthood.13,14 These findings raise the possibility borderline symptoms (t ¼ 0.30, p ¼ .76) The Joint South 230
172 231
173
that poor outcomes are not due to adolescent borderline London and Maudsley and the Institute of Psychiatry 232
174 symptoms themselves, but that symptoms index familial risk Research Ethics Committee approved each phase of the 233
175 for poor outcomes. We tested this hypothesis by comparing study. Parents gave informed consent, and twins gave assent 234

2 www.jaacap.org Journal of the American Academy of Child & Adolescent Psychiatry

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 OUTCOMES OF BORDERLINE SYMPTOMS AT AGE 12

between age 5 and 12 years and then informed consent at items were very similar to items on the Achenbach scales
235 age 18. used in E-Risk (eg, the SWAP item “Tends to be angry or 294
236 hostile” was similar to the Achenbach scale item “Angry and 295
237 Assessment of Borderline Symptoms hostile”). In these cases, we used the Achenbach scales item 296
238 297
239
When study members were 12 years old, we collected in- instead of the SWAP item to avoid asking mothers to rate 298
240 formation on their borderline symptoms during interviews the same item twice. All items and their descriptive statistics 299
241 with mothers, using items from the Shedler-Westen are reported in Table 1. Mothers were asked how well each 300
242 Assessment Procedure 200-item Q-Sort for Adolescents item described their child (0, not true; 1, somewhat or 301
243 (SWAP-200-A)5 supplemented with items from the sometimes true; 2, very true or often true). Data were 302
244 303
Achenbach System of Empirically Based Assessment.18 available for 2,141 (99.8%) of participating members at age
245 304
246 Items were selected from the set of SWAP-200-A items 12. Item responses were summarized into two measures that 305
247 most commonly used by a sample of 294 doctoral-level have been previously developed and described.12 First, a 306
248 clinicians to describe adolescent patients meeting DSM-IV dimensional borderline symptoms scale was computed by 307
249 diagnostic criteria for adult borderline personality disorder summing up across items, with an internal consistency 308
250 (Table 1).19 Of the 15 items selected from the SWAP, 5 reliability of a ¼ .86 (mean, 4.24; SD, 4.54; range, 0–26). 309
251 310
252 311
253 312
254 TABLE 1 Mothers’ Ratings of Offspring’s Borderline Symptom Items in Adolescence, at Age 12 Years 313
255 Statement Is Very True or Often True of Child 314
256 315
257 Statement Full Sample Boys Girls 316
258 Easily jealous 9.0% 11.1% 6.9% 317
259 Falls for new friends intensely, 5.7%a 5.5% 5.6% 318
260 expects too much too 319
261 quickly 320
262 Changes friends constantly, 4.3% 4.7% 3.8% 321
263 loves them one day and 322
264 hates the next 323
265 324
Fears they will be rejected or 3.0% 3.2% 2.8%
266 325
abandoned
267 326
268 Feels others are out to get 2.1% 2.7% 1.5% 327
269 him/her 328
270 Acts overly seductive or sexy, 1.7%a 2.6% 0.9% 329
271 flirts a lot 330
272 Attracted to unsuitable 0.7% 0.7% 0.6% 331
273 romantic partners 332
274 Emotions spiral out of control, 9.1%a 10.5 7.7% 333
275 has extremes of rage, 334
276 despair, excitement 335
277 336
Cannot think when upset, 6.6% 7.5% 5.8%
278 337
becomes irrational
279 338
280 Unable to soothe or 3.7% 4.0% 3.4% 339
281 comfort self 340
282 Lacks stable image of self, 3.5% 3.8% 3.1% 341
283 changes goals/values 342
284 Expresses emotions in an 11.5% 10.8% 12.2% 343
285 exaggerated dramatic way 344
286 Irritable, touchy, or quick to 7.3%a 8.9% 5.7% 345
287 “fly off the handle” 346
288 Angry and hostile 1.7% 2.2% 1.3% 347
289 348
Engages in self-harm behavior 2.9% 2.9% 2.9%
290 349
291 Note: N ¼ 2,139–2,141. 350
292 a
Sex differences were statistically significant at p < .05. 351
293 352

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 3

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 WERTZ et al.

We used this measure in our main analyses. Second, for means of poor outcomes at age 18 years among study
353 illustrative purposes and to approximate clinically significant members with a high versus lower adolescent borderline 412
354 levels of borderline symptoms, we created a dichotomous symptom score, defined as being at or above versus below 413
355 measure identifying study members scoring at or above the 95th percentile for borderline symptoms at age 12. 414
356 415
357
versus below the 95th percentile of the continuous Second, we tested whether borderline symptoms added in- 416
358 borderline symptom scale at age 12 (n ¼ 122, 5.7% of the cremental value to other behavioral and emotional problems 417
359 sample). The 5% (or 2 SD) cutoff was chosen a priori that study members experienced at age 12. We did this by 418
360 because it is consistent with previous approaches to iden- adding symptoms of conduct disorder, depression, and 419
361 tifying clinically significant borderline pathology using a anxiety at age 12 as additional covariates to test unique 420
362 421
dimensional measure,10 falls within the range of prevalence effects of adolescent borderline symptoms on poor out-
363 422
364 estimates reported for clinically significant borderline pa- comes. Third, we tested whether borderline symptoms were 423
365 thology in adolescents,20 and is consistent with estimates of influenced by familial risk. We did this by comparing cor- 424
366 the prevalence of borderline personality disorder in adults in relations in borderline symptoms among genetically iden- 425
367 the community.21 tical (MZ; n ¼ 594) and nonidentical (DZ; n ¼ 476) twin 426
368 pairs. We also formally analyzed genetic and environmental 427
369 428
Assessment of Outcomes at Age 18 Years influences on adolescent borderline symptoms using a
370 429
371 When study members were 18 years old, we collected in- univariate twin model.29 Twin models compare within-pair 430
372 formation on a variety of outcomes indicating psychosocial similarity for MZ twins, who are genetically identical, and 431
373 adjustment: personality functioning, mental health, func- DZ twins, who share on average half their segregating genes. 432
374 tional outcomes, and experiences of victimization (Table 2). This information can be used to estimate genetic (A), shared 433
375 434
We assessed outcomes using study members’ self-reports, environmental (C), and nonshared environmental (E) in-
376 435
377 reports by coinformants nominated by each twin (typi- fluences on a phenotype. C represents environmental factors 436
378 cally their cotwin and a parent), and official records. Out- that make members of a family similar, whereas E represents 437
379 comes and their assessment are described in Table 2. factors that make members of a family different and also in- 438
380 cludes error of measurement. Fourth, we compared poor 439
381 Covariates: Adolescent Behavioral and Emotional outcomes among genetically identical twins who differed in 440
382 Problems and Childhood Victimization 441
383
their borderline symptoms at age 12 to test whether adolescent 442
384 Symptoms of conduct disorder at age 12 were measured borderline symptoms predict poor outcomes over and above 443
385 using mothers’ and teachers’ reports of children’s behavioral familial influences—both genetic and environmental—shared 444
386 problems, using the Achenbach family of instruments and between identical twins growing up in the same family. Dif- 445
387 DSM-IV items as previously described.22-24 Consistent with ferences in borderline symptoms were operationalized as any 446
388 DSM-IV criteria, children with five or more symptoms were difference in the continuous symptoms score between identical 447
389 448
390
assigned a diagnosis of conduct disorder (5.5% of cohort). (MZ) twins. There were 462 MZ pairs who differed in their 449
391 Depression and anxiety at age 12 were assessed using chil- age 12 borderline symptom score. 450
392 dren’s self-reports on the Children’s Depression Inventory25 Poisson regression models were used for binary outcomes, 451
393 and the 10-item version of the Multidimensional Anxiety and linear regression models were used for continuous out- 452
394 Scale for Children,26 respectively. Scores of 20 or more on the comes. We chose Poisson over logistic regression models for 453
395 454
Children’s Depression Inventory were used to indicate clini- the binary outcomes to obtain risk ratios, which are a more
396 455
397 cally significant depressive symptoms25,27 (3.5% of cohort), easily interpretable measure of risk, particularly when out- 456
398 and scores of 13 or more (corresponding to the 95th comes are common. Standard errors were adjusted for the 457
399 percentile) on the Multidimensional Anxiety Scale for Chil- clustering of twins within families. Fixed-effects Poisson and 458
400 dren were used to indicate extreme anxiety28 (6.1% of cohort). linear regression models with robust standard errors were used 459
401 460
for the twin comparisons. Stata version 14.130 was used for
402 461
403
Statistical Analyses these analyses. Twin models were fitted using the structural 462
404 Our statistical analysis proceeded as follows. First, we tested equation modeling program OpenMx.31 463
405 associations between the continuous measure of age 12 464
406 borderline symptoms, standardized to mean (SD) 0 (1), and RESULTS 465
407 poor outcomes at age 18. We did this by predicting each Borderline Symptoms in 12-Year-Olds Predicted a 466
408 467
poor outcome from age 12 borderline symptoms. All Difficult Personality at Age 18
409 468
410 models included sex of study members. We illustrate the Study members with more borderline symptoms at age 12 469
411 results of these analyses by comparing percentages and showed a more difficult personality profile at age 18 470

4 www.jaacap.org Journal of the American Academy of Child & Adolescent Psychiatry

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 529
528
527
526
525
524
523
522
521
520
519
518
517
516
515
514
513
512
511
510
509
508
507
506
505
504
503
502
501
500
499
498
497
496
495
494
493
492
491
490
489
488
487
486
485
484
483
482
481
480
479
478
477
476
475
474
473
472
471
Volume - / Number - / - 2019
Journal of the American Academy of Child & Adolescent Psychiatry

TABLE 2 Description of Outcomes Measured During the Transition to Adulthood, at Age 18 Years

Measure Informant Description Prevalence Reference

Personality
Poor mental health
Mental disorder diagnoses
Conduct disorder
FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce
Coinformants At age 18, study members nominated two people “who knew them well.” — 40,41
Coinformants—mostly parents and cotwins—described each participant
using a 25-item version of the Big Five Inventory. 99.3% of study members
had data from at least one coinformant. 83% had data from two
coinformants. We standardized and averaged data from coinformants.
Participant Based on DSM-5 criteria, assessed as part of a computer-assisted module. 15.1% 42,43

Alcohol use disorder Participant Based on DSM-5 criteria, evaluated in face-to-face interviews using DIS. 27.8% 42,43
Cannabis use disorder Participant Based on DSM-5 criteria, evaluated in face-to-face interviews using DIS. 6% 42,43
Depression Participant Based on DSM-5 criteria, evaluated in face-to-face interviews using DIS. 20.1% 42,43
Generalized anxiety disorder Participant Based on DSM-5 criteria, evaluated in face-to-face interviews using DIS. The 7.4% 42,43
6-month symptom duration criterion was not required because of the
young age of the study sample.
Posttraumatic stress disorder Participant Based on DSM-5 criteria, evaluated in face-to-face interviews using DIS. 4.4% 42,43
Suicide attempts or self-harm Participant To assess suicide attempts, study members were asked whether they had 14.3% 28,44
tried to kill themselves or attempted suicide since age 12, using a life-
history calendar. If they answered positively, further questions were asked
to obtain details and establish intent to die. 3.8% of study members had
attempted suicide. To assess self-harm, study members were asked
whether they had ever tried to hurt themselves, to cope with stress or
emotional pain, since age 12, using a life-history calendar. Individuals who
endorsed self-harm were queried about methods. 10 behaviors were
probed (eg, cutting, burning, overdose), plus the option to describe any
other way they had hurt themselves. 13.6% of study members had harmed

OUTCOMES OF BORDERLINE SYMPTOMS AT AGE 12

Service use for behavioral or
emotional problems
www.jaacap.org
themselves.
Participant At age 18, study members reported whether they had accessed support 17.2% 45
services (eg, mental health professionals, medical doctors, or social
services), spent time in the hospital, or had taken medication for dealing
with emotional or behavioral problems in the past year.

Coinformant reports of poor
mental health
Poor functional outcomes
Low educational qualifications
Coinformants Coinformants completed a questionnaire that included 10 items querying 30.4%
about study members’ mental health within the previous 12 months
(example items: “Feels depressed, miserable, sad, or unhappy”; “Has
alcohol problems”). We created a binary measure indicating whether both
coinformants had endorsed one or more symptoms of poor mental health.
Participant Indicates whether study members reported that they did not obtain or 21.9%
scored a low average grade on their GCSE, a standardized examination
taken at the end of compulsory education at age 16 years

(continued )
5

588
587
586
585
584
583
582
581
580
579
578
577
576
575
574
573
572
571
570
569
568
567
566
565
564
563
562
561
560
559
558
557
556
555
554
553
552
551
550
549
548
547
546
545
544
543
542
541
540
539
538
537
536
535
534
533
532
531
530
 647
646
645
644
643
642
641
640
639
638
637
636
635
634
633
632
631
630
629
628
627
626
625
624
623
622
621
620
619
618
617
616
615
614
613
612
611
610
609
608
607
606
605
604
603
602
601
600
599
598
597
596
595
594
593
592
591
590
589
6

WERTZ et al.
TABLE 2 Continued

Measure Informant Description Prevalence Reference

NEET status Participant Indicates whether study members were NEET, based on reporting that they 11.6% 46
were not studying, working in paid employment, or pursuing a vocational
qualification or apprenticeship.
Cigarette smoking Participant Indicates whether study members reported that they were currently smoking 22.3%
www.jaacap.org

daily.
Risky sexual behavior Participant Indicates whether study members reported that they had engaged in two or 25.8% 47
more of the following risky sexual behaviors: having had sex before age
16; having had three or more sexual partners; practicing safe sex only
FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

sometimes or never; usually or always having sexual intercourse after a

Social isolation
Low life satisfaction
Official crime record
Journal of the American Academy of Child & Adolescent Psychiatry
night out involving a lot of alcohol and/or drug use; having been told by a
doctor that they had a sexually transmitted disease; and having had sexual
relations resulting in pregnancy.
Participant Study members were asked about their access to supportive relationships 20.0% 48,49
with family and friends using the MSPSS. The scale scoring was reversed,
and social isolation was defined as being among the 20% highest scoring
study members.
Participant Study members were asked about their life satisfaction using the Satisfaction 18.7% 49,50
With Life Scale. The scale scoring was reversed, and low life satisfaction
was defined as being among the 20% highest scoring study members.
Official records Official records of participants’ cautions and convictions beginning at age 10 10.2% 49
through age 19 were obtained through United Kingdom PNC record

searches conducted in cooperation with the Ministry of Justice.

Victimization during Participant Participants were interviewed about exposure to a range of adverse 51-53
adolescence experiences between 12 and 18 years using the JVQ-R2, adapted as a
clinical interview. Exposure to victimization was coded on a 3-point scale
(0, no exposure; 1, probable or less severe exposure; 2, definite or severe
exposure). Individuals who reported a definite or severe level of exposure
were coded as positive. Our adapted JVQ comprised 45 questions
covering 7 different forms of victimization: maltreatment (3.3%), neglect
(2.2%), sexual victimization (2.6%), family violence (12.1%), peer/sibling
victimization (15.6%), cyber-victimization (6.5%), and crime
Volume - / Number - / - 2019

victimization (19.3%).

Note: DIS ¼ Diagnostic Interview Schedule; GCSE ¼ General Certificate of Secondary Education; JVQ-R2 ¼ Juvenile Victimization Questionnaire, 2nd revision; MSPSS ¼ Multidimensional
Scale of Perceived Social Support; NEET ¼ not in education, employment, or training; PNC ¼ Police National Computer.
706
705
704
703
702
701
700
699
698
697
696
695
694
693
692
691
690
689
688
687
686
685
684
683
682
681
680
679
678
677
676
675
674
673
672
671
670
669
668
667
666
665
664
663
662
661
660
659
658
657
656
655
654
653
652
651
650
649
648
 OUTCOMES OF BORDERLINE SYMPTOMS AT AGE 12

compared with their peers with fewer borderline symptoms. the sample). Figure 1 shows personality profiles for in-
707 Specifically, coinformants who knew these study members dividuals with high borderline symptom scores versus their 766
708 well described them as more narrow-minded (low openness cohort peers with lower symptom scores. 767
709 to experience), antagonistic (low agreeableness), easily dis- 768
710 769
tressed (high neuroticism), and having poorer impulse Borderline Symptoms in 12-Year-Olds Predicted Poor
711 770
control (low conscientiousness) at age 18 (Table 3). To Mental Health at Age 18
712 771
713 approximate clinically significant levels of borderline Study members with more borderline symptoms at age 12 772
714 symptoms, we created a categorical measure by grouping experienced worse mental health at age 18 compared with 773
715 study members with high borderline symptom scores their peers with fewer symptoms (Table 3). They were more 774
716 likely to meet diagnostic criteria for a mental disorder, to 775
(operationalized as scoring at or above the 95th percentile of
717 776
718 the quantitative symptom scale at age 12; n ¼ 122, 5.7% of have attempted suicide or engaged in self-harm, and to have 777
719 778
720 TABLE 3 Borderline Symptoms (Measured on a Continuous Scale) of 12-Year-Olds Predict Outcomes at Age 18 779
721 780
722 Age 18 Outcome Model 1a Model 2b Model 3c 781
723 Personality b (95% CI) b (95% CI) b (95% CI) 782
724 Openness to experience L.08 (L0.13, L0.03) L.02 (L0.08, 0.04) .00 (L0.10, 0.11) 783
725 Conscientiousness L.16 (L0.21, L0.12) L.09 (L.14, L.03) .02 (L.11, .14) 784
726 Extraversion .06 (0.02, 0.11) .13 (0.07, 0.18) .02 (L0.11, 0.14) 785
727 Agreeableness L.28 (L0.32, L0.23) L.17 (L0.22, L0.11) L.08 (L0.17, 0.03) 786
728 Neuroticism .23 (0.19, 0.28) .19 (0.13, 0.25) .09 (L0.03, 0.22) 787
729 788
Poor mental health IRR (95% CI) IRR (95% CI) IRR (95% CI)
730 789
Conduct disorder 1.41 (1.31, 1.51) 1.17 (1.05, 1.31) 1.02 (0.81, 1.29)
731 790
732 Alcohol use disorder 1.12 (1.05, 1.19) 1.08 (0.99, 1.17) 0.99 (0.81, 1.22) 791
733 Cannabis use disorder 1.44 (1.24, 1.66) 1.09 (0.89, 1.32) 1.04 (0.73, 1.49) 792
734 Depression 1.18 (1.09, 1.28) 1.09 (0.99, 1.20) 0.95 (0.79, 1.15) 793
735 Generalized anxiety disorder 1.13 (0.97, 1.31) 1.02 (0.83, 1.23) 1.32 (0.90, 1.92) 794
736 Posttraumatic stress disorder 1.26 (1.05, 1.52) 1.15 (0.90, 1.47) 1.61 (0.86, 2.99) 795
737 Suicide attempts or self-harm 1.38 (1.27, 1.50) 1.26 (1.13, 1.40) 1.10 (0.89, 1.35) 796
738 Service use 1.31 (1.21, 1.41) 1.26 (1.14, 1.40) 1.18 (0.99, 1.40) 797
739 Coinformant report of poor 1.36 (1.29, 1.44) 1.21 (1.12, 1.31) 1.00 (0.87, 1.14) 798
740 799
mental health
741 800
Poor Functioning IRR (95% CI) IRR (95% CI) IRR (95% CI)
742 801
Low educational qualifications 1.40 (1.32, 1.49) 1.18 (1.08, 1.28) 1.13 (0.93, 1.38)
743 802
744 NEET status 1.35 (1.21, 1.50) 0.99 (0.86, 1.13) 1.12 (0.74, 1.68) 803
745 Cigarette smoking 1.40 (1.33, 1.49) 1.19 (1.09, 1.30) 1.00 (0.89, 1.11) 804
746 Risky sexual behavior 1.29 (1.22, 1.37) 1.17 (1.07, 1.26) 0.93 (0.78, 1.11) 805
747 Social isolation 1.23 (1.14, 1.33) 1.07 (0.97, 1.18) 1.14 (0.95, 1.37) 806
748 Low life satisfaction 1.27 (1.17, 1.36) 1.06 (0.96, 1.17) 1.14 (0.95, 1.37) 807
749 Official crime record 1.54 (1.40, 1.69) 1.19 (1.03, 1.36) 1.02 (0.84, 1.26) 808
750 Adolescent Victimization IRR (95% CI) IRR (95% CI) IRR (95% CI) 809
751 Maltreatment 1.89 (1.61, 2.22) 1.47 (1.09, 1.98) 1.18 (0.95, 1.46) 810
752 Neglect 1.84 (1.56, 2.17) 1.63 (1.23, 2.16) 1.10 (0.61, 1.98) 811
753 812
Sexual victimization 1.45 (1.22, 1.72) 1.00 (0.77, 1.29) 1.30 (0.80, 2.10)
754 813
Family violence 1.30 (1.18, 1.44) 1.14 (1.00, 1.31) 0.85 (0.66, 1.11)
755 814
756 Peer victimization 1.28 (1.17, 1.39) 1.17 (1.05, 1.31) 1.07 (0.87, 1.31) 815
757 Cyber-victimization 1.32 (1.16, 1.50) 1.27 (1.07, 1.51) 1.15 (0.73, 1.80) 816
758 Crime victimization 1.25 (1.16, 1.34) 1.16 (1.05, 1.29) 1.01 (0.84, 1.22) 817
759 818
Note: Statistically significant estimates (p > .05) are in bold. All outcome measures are described in Table 2. b ¼ standardized regression coefficient;
760 819
CI, confidence interval; IRR ¼ incidence rate ratio (interpretable as risk ratios); NEET ¼ not in education, employment, or training.
761 a 820
Model 1: Analyses adjust for sex. For models additionally adjusting for baseline measures where possible, see Table S1, available online.
762 b
Model 2: Analyses additionally adjust for age 12 behavioral and emotional problems. 821
763 c
Model 3: Estimates are within-monozygotic-twin-pair estimates, ie, analyses adjust for the influence of genes and environments shared between 822
764 genetically identical twins growing up in the same family. 823
765 824

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 7

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 WERTZ et al.

FIGURE 1 Personality Profiles for Individuals With High Borderline Symptom Scores Versus Their Cohort Peers With Lower
825 Symptom Scores Q5 884
826 885
Openness to experience
827 1
886
828 887
829 High age-12 borderline symptom score 888
0.5
830 889

z-scores
831 Lower age-12 borderline symptom score 890
832 0 891
833 Neuroticism
892
834 Conscientiousness 893
-0.5
835 894
836 895
-1
837 896
838 897
839 898
web 3C=FPO

840 899
841 900
842 901
843 902
844 Agreeableness Extraversion
903
845 904
846 Note: A high borderline symptom score at age 12 (operationalized as being at or above the 95th percentile for borderline symptoms at age 12) predicts a distinct per-
905
847 sonality profile at age 18, characterized by narrow-mindedness (low openness to experience), antagonism (low agreeableness), distress (high neuroticism), and poor im- 906
848 pulse control (low conscientiousness). All analyses are adjusted for study members’ sex. 907
849 908
850 909
851 used clinical and support services to cope with emotional fewer symptoms (Table 3). Adolescents with borderline 910
852 911
and behavioral problems. Findings of worse mental health symptoms experienced victimization both within and
853 912
854 were corroborated by coinformants (Table 3). These find- outside of their families through maltreatment, neglect, 913
855 ings are illustrated in Figure 2, which shows the prevalence family violence, bullying by peers, and as victims of crime 914
856 of mental health outcomes among study members with high (Table 3). These findings are illustrated in Figure 2, which 915
857 versus lower borderline symptom scores at age 12. shows the prevalence of victimization exposures among 916
858 study members with high versus lower borderline symptom 917
859 918
Borderline Symptoms in 12-Year-Olds Predicted Poor scores at age 12. Previous studies, including our own, have
860 919
Functioning at Age 18 shown that adolescent borderline symptoms are often pre-
861 920
862 Study members with more borderline symptoms at age 12 ceded by victimization during childhood.12 Victimization 921
863 experienced worse functioning at age 18 compared with during adolescence may therefore reflect continuing expo- 922
864 their peers with fewer symptoms (Table 3). They had sure to victimization rather than effects of borderline 923
865 poorer educational and economic outcomes, as indicated by 924
symptoms. However, even after statistically controlling for
866 925
867
educational failure and unemployment; engaged in more childhood victimization, borderline symptoms predicted 926
868 unhealthy behaviors, as indicated by cigarette smoking and adolescents’ risk of becoming victimized in adolescence 927
869 risky sexual activity; experienced lower well-being, as indi- (Table S1, available online). 928
870 cated by social isolation and dissatisfaction with life; and 929
871 were more likely to have broken the law, as indicated by Borderline Symptoms in 12-Year-Olds Were Correlated 930
872 931
having an official crime record (Table 3). These findings are With Behavioral and Emotional Problems, But This Did
873 932
illustrated in Figure 2, which shows the prevalence of poor Not Explain Away Associations With Most Poor
874 933
functional outcomes among study members with high Outcomes
875 934
876 versus lower borderline symptom scores at age 12. At age 12, study members who displayed more borderline 935
877 symptoms tended to also experience more symptoms of 936
878 Borderline Symptoms in 12-Year-Olds Predicted other behavioral and emotional problems, including 937
879 Becoming a Victim of Violence conduct disorder (r ¼ .56, 95% confidence interval [CI] 938
880 939
881
Study members with more borderline symptoms at age 12 [0.51, 0.62], p < .01), depression (r ¼ .27, 95% CI [0.21, 940
882 were more likely to become victims of violence during 0.32], p < .01), and anxiety (r ¼ .10, 95% CI [0.04, 0.15], 941
883 adolescence (age 12–18 years) compared to their peers with p < .01). More than half (55%) of study members with a 942

8 www.jaacap.org Journal of the American Academy of Child & Adolescent Psychiatry

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 OUTCOMES OF BORDERLINE SYMPTOMS AT AGE 12

FIGURE 2 Prevalence of Mental Health Outcomes Among Study Members With High Versus Lower Borderline Symptom Scores
943 at Age 12 1002
944 1003
Poor mental health
945 1004
946 Conduct disorder 1005
947 High age-12 borderline symptom score
1006
Alcohol use disorder
948 Lower age-12 borderline symptom score 1007
949 Cannabis use 1008
disorder
950 1009
Depression
951 1010
952 Generalized anxiety 1011
disorder
953 1012
954 PTSD 1013
955 Suicide attempts or 1014
self-harm
956 1015
957 Service use 1016
958 Co-informant report 1017
959 of poor mental… 1018
960 Poor functional outcomes 1019
961 Low educational 1020
962 qualifications 1021
963 NEET 1022
964 1023
Cigarette smoking
965 1024
966 Risky sexual 1025
behavior
967 1026
968 Social isolation 1027
969 Low life satisfaction 1028
970 1029
971 Official crime record 1030
972 1031
Adolescent victimization
973 1032
974 Maltreatment 1033
975 1034
Neglect
976 1035
977 Sexual victimization 1036
978 1037
Family violence
979 1038
980 Peer/sibling 1039
victimization
981 1040
web 3C=FPO

Cybervictimization
982 1041
983 Crime victimization 1042
984 1043
985 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1044
986 Prevalence of outcome 1045
987 1046
988 Note: 12-year-olds with a high borderline symptom score (operationalized as being at or above the 95th percentile for borderline symptoms at age 12) experience worse 1047
outcomes at age 18 compared with their cohort peers with a lower borderline symptom score. Error bars indicate 95% confidence intervals. All analyses adjust for study
989 1048
members’ sex. All outcome measures are described in Table 2. NEET ¼ not in education, employment, or training; PTSD ¼ posttraumatic stress disorder.
990 1049
991 1050
992 1051
993 high (ie, at or above the 95th percentile) borderline symp- anxiety at age 12 when predicting poor outcomes (Table 3). 1052
994 tom score at age 12 met clinical criteria for at least one of Borderline symptoms continued to predict most outcomes 1053
995 these problems compared with 10% of study members with independently of correlated problems, particularly a difficult 1054
996 a lower symptom score (ie, below the 95th percentile). We personality at age 18, and most of the poor functional 1055
997 tested whether borderline symptoms added incremental outcomes and experiences of victimization. Some associa- 1056
998 1057
999
value to behavioral and emotional problems when predict- tions between age 12 borderline symptoms and age 18 poor 1058
1000 ing poor outcomes by statistically controlling for continuous outcomes were explained away by co-occurring behavioral 1059
1001 symptom scores of conduct disorder, depression, and and emotional problems at age 12, most notably 1060

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 9

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 WERTZ et al.

associations with nearly all diagnoses of mental disorders at methods, including official records. Differences in outcomes
1061 age 18. However, even within the psychiatric outcomes were striking: young people with the highest borderline 1120
1062 domain, borderline symptoms in 12-year-olds continued to symptom scores at age 12 were nearly three times more 1121
1063 predict adverse outcomes, including conduct disorder, sui- likely to engage in suicidal and self-harming behavior, to 1122
1064 1123
1065
cide attempts and self-harm, service use, and coinformant find themselves without training or job opportunities, to 1124
1066 reports of poor mental health (Table 3). have a criminal record, and to have experienced victimiza- 1125
1067 tion compared with their cohort peers with lower symptom 1126
1068 Borderline Symptoms in 12-Year-Olds Developed scores. These findings show that adolescent borderline 1127
1069 Against a Backdrop of Familial Risk, Which Accounted symptoms observed as early as at age 12 forecast meaningful 1128
1070 for Most Associations With Most Poor Outcomes at 1129
individual differences in young people’s lives.
1071 Age 18 1130
1072 Second, although many 12-year-olds experienced behav- 1131
1073
Twins growing up in the same families resembled each ioral and emotional problems alongside their borderline 1132
1074 other in their borderline symptom scores at age 12 (r ¼ .49, symptoms, borderline symptoms added incremental value to 1133
1075 95% CI [0.42, 0.55], p < .01), suggesting familial risk for predicting most poor outcomes over and above these other 1134
1076 borderline symptoms. Comparing correlations between problems, indicating that the later-life impairments associated 1135
1077 members of genetically identical (MZ) and nonidentical 1136
1078
with adolescent borderline pathology are insufficiently 1137
(DZ) twin pairs revealed that familial risk was entirely ge- described by these problems. Notably, behavioral and
1079 1138
1080 netic, as indicated by MZ correlations that were twice as emotional problems of 12-year-olds accounted for associations 1139
1081 high as DZ correlations (rMZ ¼ .66, 95% CI [0.60, 0.73], with nearly all of their psychiatric diagnoses at age 18, but did 1140
1082 p < .01; rDZ ¼ .28, 95% CI [0.19, 0.37], p < .01). We not account for associations with many other adverse out- 1141
1083 formally analyzed twin correlations using a univariate twin 1142
comes. This finding illustrates that psychiatric diagnoses do
1084 model and obtained a heritability estimate (A) of 0.66 (95% 1143
1085 not capture the full scope of life challenges associated with 1144
1086
CI [0.58, 0.70]), indicating that 66% of individual differ- adolescent borderline symptoms and shows that it is impor- 1145
1087 ences in borderline symptoms at age 12 were explained by tant to look beyond psychiatric status when testing the pre- 1146
1088 genetic influences (Table S2, available online). The dictive validity of adolescent borderline symptoms. 1147
1089 remainder was accounted for by environmental influences Third, borderline symptoms of 12-year-olds were under 1148
1090 not shared between family members (E) (estimate: E ¼ considerable genetic influence, and genetically identical
1149
1091 1150
0.34; 95% CI [0.30, 0.38]). There were no shared envi- twins of children with elevated borderline symptoms were at
1092 1151
1093 ronmental influences (C) (estimate: C ¼ 0.00; 95% CI increased risk for poor outcomes even if they did not have 1152
1094 [0.00, 0.07]). If genetic influences affect both borderline equally elevated borderline symptoms themselves. This 1153
1095 symptoms at age 12 and poor outcomes at age 18, adoles- finding raises three issues. First, it raises the question of why 1154
1096 cent borderline symptoms may be an expression of shared 1155
twins with the same genetic susceptibility do not share
1097 genetic risk for poor outcomes, rather than an influential 1156
1098 similar borderline symptoms. Our previous study pointed to 1157
1099
factor in itself. Our findings supported this hypothesis: twins’ unique environmental experiences as a possible 1158
1100 genetically identical twins who differed in borderline explanation: we reported that twins in the same families 1159
1101 symptoms experienced similar levels of poor outcomes at experienced different levels of harsh parental treatment and 1160
1102 age 18 (Table 3). This finding suggests that borderline that adolescents’ genetic vulnerability interacted with harsh 1161
1103 symptoms in 12-year-olds predict poor outcomes at age 18 1162
1104
parental treatment in the etiology of borderline symp- 1163
because borderline symptoms and poor outcomes are toms.12 This finding is consistent with diathesis-stress
1105 1164
1106 manifestations of shared genetic risk. models of borderline personality, which propose that it is 1165
1107 the interaction between children’s genetically influenced, 1166
1108 DISCUSSION early emerging temperamental difficulties and an invalid- 1167
1109 1168
Our follow-up of 12-year-olds with borderline symptoms to ating, abusive, and ineffective caregiving environment that
1110 1169
1111 age 18 revealed three main findings. First, at a time in life increases risk for borderline problems (and other poor 1170
1112 when young people take a leap toward greater social, eco- outcomes) in a transactional process across develop- 1171
1113 nomic, and personal maturity, study members with a his- ment.12,32 Second, our findings indicate that adolescent 1172
1114 tory of borderline symptoms were held back by psychosocial borderline symptoms reflect broader genetic risk for poor 1173
1115 difficulties. Difficulties were evident in numerous areas outcomes, rather than being the cause of these outcomes. 1174
1116 1175
1117
(personality; psychopathology; vocational, health, and social This finding does not undermine the prognostic significance 1176
1118 functioning; and experiences of victimization); observed by of adolescent borderline symptoms. However, it suggests 1177
1119 multiple informants; and assessed through multiple that adolescents remain at risk for adverse psychosocial 1178

10 www.jaacap.org Journal of the American Academy of Child & Adolescent Psychiatry

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 OUTCOMES OF BORDERLINE SYMPTOMS AT AGE 12

outcomes even after symptom reduction.10 Third, if weakness of discordant twin analyses is their higher likeli-
1179 borderline symptoms are not the cause of poor outcomes hood of false-negative findings because the limited variation 1238
1180 but are on the pathway from genetic risk to poor outcomes, within twin pairs magnifies the impact of measurement 1239
1181 more work is needed to understand how genetic risk in- error and reduces the precision of estimates.36 Third, our 1240
1182 1241
1183
fluences both borderline symptoms and poor outcomes. A study does not contain a measure of borderline symptoms at 1242
1184 hypothesis consistent with diathesis-stress models of age 18, so we were unable to test the continuity of 1243
1185 borderline personality is that genetic risk begins to manifest borderline symptoms. However, our findings show that 18- 1244
1186 early in life, as a difficult temperamental profile character- year-olds with a history of borderline symptoms show a 1245
1187 ized by high negative affect, poor impulse control, and high personality profile characterized by emotional and inter- 1246
1188 1247
emotional sensitivity. A child’s difficult temperament sub- personal lability that is typical of adults with borderline
1189 1248
1190 sequently increases risk for borderline pathology as well as personality disorder.37 Fourth, although a considerable 1249
1191 for other poor outcomes, particularly when it is met by an portion of 12-year-olds with borderline symptoms went on 1250
1192 invalidating caregiving environment.32 to experience poor outcomes, there were also adolescents 1251
1193 Our work expands on previous literature in three ways. who bucked this trend and fared well despite their symptom 1252
1194 First, although several studies have investigated the clinical history. Follow-on work is needed to investigate factors that 1253
1195 1254
1196
and psychosocial outcomes of borderline personality disor- predict variability in poor outcomes associated with 1255
1197 der, a recent systematic review concluded that many of these borderline symptoms. Fifth, our study members are still 1256
1198 studies have limitations, such as sampling bias, high rates of young, so it is unclear how persistent their psychosocial 1257
1199 attrition, and a narrow range of psychosocial outcomes.11 difficulties will be. However, many of the outcomes we 1258
1200 Our study overcomes some of these limitations because measured—low educational qualifications, cigarette smok- 1259
1201 1260
our cohort is nationally representative, follow-up of partic- ing, personality dysfunction, having a criminal record, risky
1202 1261
1203 ipants has occurred with extremely low attrition, and we sexual behavior—are still meaningful because they represent 1262
1204 report associations with a wide range of clinical and psy- barriers to leading a prosperous and healthy adult life. 1263
1205 chosocial outcomes. Second, there are very few studies Moreover, previous work testing associations with some of 1264
1206 testing associations between borderline pathology and the same outcomes we report, such as attainment and social 1265
1207 exposure to victimization, particularly in adolescence. Our support, has shown that adolescent borderline symptoms 1266
1208 1267
1209
study extends the literature by showing that borderline predict these outcomes up to age 33 years.10 1268
1210 symptoms in 12-year-olds predict exposure to different Our findings have implications for health professionals 1269
1211 types of victimization, both inside and outside the home, working with adolescents who display borderline symptoms. 1270
1212 during adolescence. Third, in addition to reporting that First, our findings support the assessment of adolescents’ 1271
1213 adolescents’ borderline pathology predicts poor outcomes, borderline symptoms in addition to other emotional and 1272
1214 1273
we find that these associations do not persist after ac- behavioral disorders if borderline pathology is suspected.
1215 1274
1216 counting for familial influences shared between identical Some clinicians are thought to prefer assessing only 1275
1217 twins growing up in the same family. Although several emotional and behavioral disorders in adolescents present- 1276
1218 studies have tested outcomes of adolescent borderline ing with borderline symptoms, perhaps to avoid a stigma- 1277
1219 symptoms using a twin design,33 to our knowledge our tizing diagnosis of personality disorder.6,38 However, our 1278
1220 study is the first to apply this approach to a wide range of findings and findings of others show that adolescents’ 1279
1221 1280
1222
clinical and psychosocial outcomes. borderline symptoms provide independent prognostic in- 1281
1223 Our findings should be interpreted in light of some formation. Second, our findings argue in favor of early ac- 1282
1224 limitations. First, we did not make a formal diagnosis of cess to treatment for adolescents with borderline symptoms 1283
1225 borderline personality disorder. Without a replication in and against a wait-and-see approach.39 Psychological treat- 1284
1226 adolescents with a diagnosis of borderline personality dis- ments for adult patients with borderline personality disorder 1285
1227 1286
order, we cannot be sure that our findings generalize to this have been adapted for use with adolescents and show
1228 1287
1229 population. However, our measure captures core diagnostic promise for improving symptoms.7 In addition to treatment 1288
1230 features of borderline personality disorder (affective insta- for personality pathology, adolescents with borderline 1289
1231 bility, cognitive disturbance, impulsivity, and interpersonal symptoms need access to support services that help reduce 1290
1232 dysfunction), and our previous study showed that the the risk for future poor functioning, such as educational 1291
1233 etiological factors, comorbidity, sex differences, and herita- support services. Third, our findings show that adolescents’ 1292
1234 1293
1235
bility of our measure of borderline symptoms are compa- borderline symptoms signal a longer-term need for care. 1294
1236 rable to results from studies of borderline personality Even if symptoms decrease after treatment, adolescents 1295
1237 disorder in community samples.10,12,34,35 Second, a general remain at risk for adverse outcomes because symptoms 1296

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 11

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 WERTZ et al.

partly reflect genetic risk for future difficulties. Adolescents King’s College London, London, UK, and the National & Specialist CAMHS
1297 should be monitored and supported accordingly, particu- Clinic for Trauma, Anxiety and Depression, South London & Maudsley NHS 1356
Foundation Trust, London, UK.
1298 larly during the transition to adulthood when they face 1357
1299 The Environmental Risk (E-Risk) Longitudinal Twin Study is funded by the 1358
discharge from child and adolescent mental health services. Medical Research Council (MRC; grant G1002190). Additional support was
1300 1359
1301
Fourth, our findings imply that young people with a history provided by the Eunice Kennedy Shriver National Institute of Child Health and
1360
Human Development (NICHD; grant HD077482) and the Jacobs Foundation. Q9
1302 of borderline symptoms may turn up on the doorsteps of 1361
The authors are grateful to the study mothers and fathers, the twins, and the twins’
1303 many services, including mental health care services, un- teachers for their participation. The authors thank CACI Inc., Google Street View, 1362
1304 employment offices, sexual health centers, courts, emer- and the United Kingdom Ministry of Justice for assistance with data and members 1363
1305 of the E-Risk team for their dedication, hard work, and insights. 1364
gency departments, and social services. The breadth of poor
1306 Disclosure: Dr. Arseneault is the Mental Health Leadership Fellow for the United 1365
outcomes among these young people requires an integrated Kingdom Economic and Social Research Council (ESRC). Dr. Belsky has received
1307 1366
1308 treatment approach that involves coordination across mul- support by an Early-Career Research Fellowship from the Jacobs Foundation.
1367
Dr. Fisher has received support from a British Academy Mid-Career Fellowship
1309 tiple social and support services. (MDy170005). Dr. Richmond-Rakerd is a postdoctoral fellow with the Carolina 1368
1310 Consortium on Human Development and the Center for Developmental Science at 1369
the University of North Carolina-Chapel Hill. Drs. Wertz, Caspi, Danese, Matthews,
1311 and Moffitt and Mr. Ambler report no biomedical financial interests or potential
1370
Accepted July 11, 2019.
1312 conflicts of interest. 1371
1313 Drs. Wertz, Caspi, Richmond-Rakerd, and Moffitt are with Duke University, 1372
Durham, NC. Drs. Caspi and Moffitt are also with the Duke University Medical Correspondence to Jasmin Wertz, PhD, Department of Psychology and Neuro-
1314 Center, Durham, NC. Drs. Caspi, Ambler, Arseneault, Danese, Fisher, Mat- science, Duke University, Box 104410, Durham, NC, 27708; e-mail: jasmin.wertz@ 1373
1315 thews, and Moffitt are with Social, Genetic & Developmental Psychiatry, Insti- duke.edu 1374
1316 tute of Psychiatry, Psychology and Neuroscience, King’s College London, 0890-8567/$36.00/ª2019 American Academy of Child and Adolescent Psychiatry. 1375
1317 London, UK. Dr. Ambler is also with Dunedin Multidisciplinary Health and Published by Elsevier Inc. This is an open access article under the CC BY license 1376
Development Unit, University of Otago, New Zealand. Dr. Belsky is with (http://creativecommons.org/licenses/by/4.0/).
1318 Columbia University Mailman School of Public Health, New York, NY. Dr. 1377
1319 Danese is also with the Institute of Psychiatry, Psychology and Neuroscience, https://doi.org/10.1016/j.jaac.2019.07.005 1378
1320 1379
1321 1380
1322 REFERENCES 1381
1323 1. Black DW, Pfohl B, Blum N, et al. Attitudes toward borderline personality disorder: a 16. Odgers CL, Caspi A, Bates CJ, Sampson RJ, Moffitt TE. Systematic social observation of 1382
survey of 706 mental health clinicians. CNS Spectr. 2011;16:67-74. children’s neighborhoods using Google Street View: a reliable and cost-effective method.
1324 2. Chanen AM, McCutcheon LK. Personality disorder in adolescence: the diagnosis that J Child Psychol Psychiatry. 2012;53:1009-1017.
1383
1325 dare not speak its name. Personal Ment Health. 2008;2:35-41. 17. Odgers CL, Caspi A, Russell MA, Sampson RJ, Arseneault L, Moffitt TE. Supportive 1384
1326 3. Griffiths M. Validity, utility and acceptability of borderline personality disorder diagnosis parenting mediates neighborhood socioeconomic disparities in children’s antisocial 1385
1327 in childhood and adolescence: survey of psychiatrists. Psychiatrist. 2011;35:19-22. behavior from ages 5 to 12. Dev Psychopathol. 2012;24:705-721. 1386
4. Laurenssen EMP, Hutsebaut J, Feenstra DJ, Van Busschbach JJ, Luyten P. Diagnosis of 18. Achenbach T. The Achenbach System of Empirically Based Assessment (ASEBA):
1328 1387
personality disorders in adolescents: a study among psychologists. Child Adolesc Psy- Development, Findings, Theory, and Applications. Burlington, VT: University of Ver-
1329 chiatry Ment Health. 2013;7(3). mont, Research Center for Children, Youth, and Families; 2009. 1388
1330 5. Westen D, Shedler J, Durrett C, Glass S, Martens A. Personality diagnoses in adoles- 19. Bradley R, Conklin CZ, Westen D. The borderline personality diagnosis in adolescents: 1389
1331 cence: DSM-IV Axis II diagnoses and an empirically derived alternative. Am J Psychiatry. gender differences and subtypes. J Child Psychol Psychiatry. 2005;46:1006-1019. 1390
1332 2003;160:952-966. 20. Zanarini MC, Horwood J, Wolke D, Waylen A, Fitzmaurice G, Grant BF. Prevalence of 1391
6. Miller AL, Muehlenkamp JJ, Jacobson CM. Fact or fiction: diagnosing borderline per- DSM-IV borderline personality disorder in two community samples: 6,330 English 11-
1333 sonality disorder in adolescents. Clin Psychol Rev. 2008;28:969-981. year-olds and 34,653 American adults. J Pers Disord. 2011;25:607-619.
1392
1334 7. Sharp C, Fonagy P. Practitioner review: borderline personality disorder in adolescence— 21. Grant BF, Chou SP, Goldstein RB, et al. Prevalence, correlates, disability, and comor- 1393
1335 recent conceptualization, intervention, and implications for clinical practice. J Child bidity of DSM-IV borderline personality disorder: results from the Wave 2 National 1394
1336 Psychol Psychiatry. 2015;56:1266-1288. Epidemiologic Survey on Alcohol and Related Conditions. J Clin Psychiatry. 2008;69: 1395
8. Sharp C, Tackett JL. Introduction: an idea whose time has come. In: Handbook of 533-545.
1337 1396
Borderline Personality Disorder in Children and Adolescents. New York, NY: Springer 22. Achenbach TM. Manual for the Child Behavior Checklist/4-18 and 1991 Profile.
1338 New York; 2014:3-8. Burlington, VT: University of Vermont Department of Psychiatry; 1991. 1397
1339 9. Chanen AM, Jovev M, Jackson HJ. Adaptive functioning and psychiatric symptoms 23. Achenbach TM. Manual for the Teacher’s Report Form and 1991 Profile. Burlington, 1398
1340 in adolescents with borderline personality disorder. J Clin Psychiatry. 2007;68: VT: University of Vermont Department of Psychiatry; 1991. 1399
1341 297-306. 24. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 1400
10. Winograd G, Cohen P, Chen H. Adolescent borderline symptoms in the community: 4th ed, text rev. Washington, DC: American Psychiatric Association; 2000.
1342 prognosis for functioning over 20 years. J Child Psychol Psychiatry. 2008;49:933-941. 25. Kovacs M. Children’s Depression Inventory (CDI) Manual. North Tonawanda, NY:
1401
1343 11. Winsper C, Marwaha S, Lereya ST, Thompson A, Eyden J, Singh SP. Clinical and Multi-Health Systems; 1992. 1402
1344 psychosocial outcomes of borderline personality disorder in childhood and adolescence: a 26. March J. Manual for the Multidimensional Anxiety Scale for Children. Toronto, CA: 1403
1345 systematic review. Psychol Med. 2015;45:2237-2251. Multi-Health Systems; 1997. 1404
12. Belsky DW, Caspi A, Arseneault L, et al. Etiological features of borderline personality 27. Rivera-Medina C, Bernal G, Rossello G. The Children Depression Inventory (CDI) and
1346 1405
related characteristics in a birth cohort of 12-year-old children. Dev Psychopathol. 2012; the Beck Depression Inventory (BDI): their validity as screening measures for major
1347 24:251-265. depression in a group of Puerto Rican adolescents. Int J Clin Health Psychol. 2005;5: 1406
1348 13. Bornovalova MA, Hicks BM, Iacono WG, McGue M. Stability, change, and heritability 485-498. 1407
1349 of borderline personality disorder traits from adolescence to adulthood: a longitudinal 28. Fisher HL, Moffitt TE, Houts RM, Belsky DW, Arseneault L, Caspi A. Bullying vic- 1408
1350 twin study. Dev Psychopathol. 2009;21:1335. timisation and risk of self harm in early adolescence: longitudinal cohort study. BMJ. 1409
14. Jaffee SR, Ambler A, Merrick M, et al. Childhood maltreatment predicts poor economic 2012;344:e2683.
1351 and educational outcomes in the transition to adulthood. Am J Public Health. 2018; 29. Neale M, Cardon L. Methodology for Genetic Studies of Twins and Families. Boston,
1410
1352 108(9):1142-1147. MA: Kluver Academic Publishers; 1992. 1411
1353 15. Moffitt TE; E-Risk Study Team. Teen-aged mothers in contemporary Britain. J Child 30. StataCorp. Stata Statistical Software: Release 14. College Station, TX: StataCorp 1412
1354 Psychol Psychiatry. 2002;43:727-742. LP; 2015. 1413
1355 1414

12 www.jaacap.org Journal of the American Academy of Child & Adolescent Psychiatry

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 OUTCOMES OF BORDERLINE SYMPTOMS AT AGE 12

31. Boker S, Neale M, Maes H, Wilde M, Spiegel M. OpenMx: an open source extended 43. Robins LN, Cottler L, Bucholz KK, Compton W. Diagnostic Interview Schedule for
structural equation modeling framework. Psychometrika. 2011;76:306-317. DSM-IV. St Louis, MO: Washington University School of Medicine; 1995.
1415 32. Crowell SE, Beauchaine TP, Linehan MM. A biosocial developmental model of 44. Baldwin JR, Arseneault L, Caspi A, et al. Adolescent victimization and self-injurious 1474
1416 borderline personality: elaborating and extending linehan’s theory. Psychol Bull. 2009; thoughts and behaviors: a genetically sensitive cohort study. J Am Acad Child Adolesc 1475
1417 135:495-510. Psychiatry. 2019;58:506-513. 1476
33. Bornovalova MA, Hicks BM, Iacono WG, McGue M. Longitudinal twin study of 45. Moffitt TE, Caspi A, Taylor A, et al. How common are common mental disorders?
1418 1477
borderline personality disorder traits and substance use in adolescence: developmental Evidence that lifetime prevalence rates are doubled by prospective versus retrospective
1419 change, reciprocal effects, and genetic and environmental influences. Personal Disord ascertainment. Psychol Med. 2010;40:899-909. 1478
1420 Theory Res Treat. 2013;4:23-32. 46. Goldman-Mellor S, Caspi A, Arseneault L, et al. Committed to work but vulnerable: self- 1479
1421 34. Lenzenweger MF, Lane MC, Loranger AW, Kessler RC. DSM-IV personality disorders perceptions and mental health in NEET 18-year olds from a contemporary British 1480
1422 in the National Comorbidity Survey Replication. Biol Psychiatry. 2007;62:553-564. cohort. J Child Psychol Psychiatry. 2016;57:196-203. 1481
35. Torgersen S, Kringlen E, Cramer V. The prevalence of personality disorders in a com- 47. Ramrakha S, Caspi A, Dickson N, Moffitt TE, Paul C. Psychiatric disorders and risky
1423 munity sample. Arch Gen Psychiatry. 2001;58:590-596. sexual behaviour in young adulthood: cross sectional study in birth cohort. BMJ. 2000;
1482
1424 36. Boardman JD, Fletcher JM. To cause or not to cause? That is the question, but identical 321:263-266. 1483
1425 twins might not have all of the answers. Soc Sci Med. 2015;127:198-200. 48. Zimet GD, Dahlem NW, Zimet SG, Farley GK. The Multidimensional Scale of 1484
1426 37. Saulsman LM, Page AC. The five-factor model and personality disorder empirical Perceived Social Support. J Pers Assess. 1988;52:30-41. 1485
literature: a meta-analytic review. Clin Psychol Rev. 2004;23:1055-1085. 49. Wertz J, Agnew-Blais J, Caspi A, et al. From childhood conduct problems to poor
1427 1486
38. Shiner RL, Allen TA. Assessing personality disorders in adolescents: seven guiding functioning at age 18 years: examining explanations in a longitudinal cohort study. J Am
1428 principles. Clin Psychol Sci Pract. 2013;20:361-377. Acad Child Adolesc Psychiatry. 2018;57:54-60. 1487
1429 39. Chanen A, Sharp C, Hoffman P, Global Alliance for Prevention and Early Intervention 50. Diener E, Emmons RA, Larsen RJ, Griffin S. The Satisfaction With Life Scale. J Pers 1488
1430 for Borderline Personality Disorder. Prevention and early intervention for borderline Assess. 1985;49:71-75. 1489
1431 personality disorder: a novel public health priority. World Psychiatry. 2017;16:215-216. 51. Finkelhor D, Hamby SL, Turner HA, Ormrod RK. The Juvenile Victimization Ques- 1490
40. Benet-Martınez V, John OP. Los Cinco Grandes across cultures and ethnic groups: tionnaire (JVQ-R2), 2nd rev. Durham, NH: Crimes Against Children Research Cen-
1432 multitrait multimethod analyses of the Big Five in Spanish and English. J Pers Soc ter; 2011.
1491
1433 Psychol. 1998;75:729-750. 52. Hamby SL, Finkelhor D, Ormrod RK, Turner HA. The Comprehensive JV Adminis- 1492
1434 41. Richmond-Rakerd LS, Caspi A, Arseneault L, et al. Adolescents who self-harm and tration and Scoring Manual. Durham, NH: Crimes Against Children Research 1493
1435 commit violent crime: testing early-life predictors of dual harm in a longitudinal cohort Centre; 2004. 1494
study. Am J Psychiatry. 2019;176:186-195. 53. Fisher HL, Caspi A, Moffitt TE, et al. Measuring adolescents’ exposure to victimization:
1436 1495
42. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, The Environmental Risk (E-Risk) Longitudinal Twin Study. Dev Psychopathol. 2015;
1437 5th ed. Washington, DC: American Psychiatric Association; 2013. 27:1399-1416. 1496
1438 1497
1439 1498
1440 1499
1441 1500
1442 1501
1443 1502
1444 1503
1445 1504
1446 1505
1447 1506
1448 1507
1449 1508
1450 1509
1451 1510
1452 1511
1453 1512
1454 1513
1455 1514
1456 1515
1457 1516
1458 1517
1459 1518
1460 1519
1461 1520
1462 1521
1463 1522
1464 1523
1465 1524
1466 1525
1467 1526
1468 1527
1469 1528
1470 1529
1471 1530
1472 1531
1473 1532

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 13

Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 WERTZ et al.

TABLE S1 Borderline Symptoms (Measured on a Continuous Scale) at Age 12 Predict Poor Outcomes at Age 18 Even After
1533 Adjusting for Poor Adjustment During Childhood 1592
1534 a b
1593
1535 Age 18 Outcome Model 1 Model 2 1594
1536 Personality b (95% CI) b (95% CI) 1595
1537 Openness to experience L.08 (L0.13, L0.03) L.07 (L0.12, L0.02) 1596
1538 Conscientiousness L.16 (L0.21, L0.12) L.14 (L0.19, L0.09) 1597
1539 Extraversion .06 (0.02, 0.11) .07 (0.02, 0.12) 1598
1540 Agreeableness L.28 (L0.32, L0.23) L.24 (L0.29, L0.20) 1599
1541 Neuroticism .23 (0.19, 0.28) .23 (0.19, 0.28) 1600
1542 Poor Mental Health IRR (95% CI) IRR (95% CI) Q6 1601
1543 1602
Conduct disorder 1.41 (1.31, 1.51) 1.30 (1.18, 1.42)
1544 1603
Alcohol use disorder 1.12 (1.05, 1.19) N/A
1545 1604
1546 Cannabis use disorder 1.44 (1.24, 1.66) N/A 1605
1547 Depression 1.18 (1.09, 1.28) 1.14 (1.05, 1.24) 1606
1548 Generalized anxiety disorder 1.13 (0.97, 1.31) 1.12 (0.97, 1.30) 1607
1549 Posttraumatic stress disorder 1.26 (1.05, 1.52) N/A 1608
1550 Suicide attempts or self-harm 1.38 (1.27, 1.50) 1.31 (1.19, 1.43) 1609
1551 Service use 1.31 (1.21, 1.41) 1.27 (1.18, 1.38) 1610
1552 Coinformant report of poor mental health 1.36 (1.29, 1.44) N/A 1611
1553 Poor Functioning IRR (95% CI) IRR (95% CI) Q7 1612
1554 Low educational qualifications 1.40 (1.32, 1.49) 1.23 (1.15, 1.31) 1613
1555 1614
NEET status 1.35 (1.21, 1.50) N/A
1556 1615
Cigarette smoking 1.40 (1.33, 1.49) N/A
1557 1616
1558 Risky sexual behavior 1.29 (1.22, 1.37) N/A 1617
1559 Social isolation 1.23 (1.14, 1.33) 1.07 (0.97, 1.18) 1618
1560 Low life satisfaction 1.27 (1.17, 1.36) N/A 1619
1561 Official crime record 1.54 (1.40, 1.69) N/A 1620
1562 Adolescent Victimization IRR (95% CI) IRR (95% CI) 1621
1563 Maltreatment 1.89 (1.61, 2.22) 1.66 (1.39, 1.98) 1622
1564 Neglect 1.84 (1.56, 2.17) 1.59 (1.34, 1.88) 1623
1565 Sexual victimization 1.45 (1.22, 1.72) 1.26 (1.07, 1.50) 1624
1566 Family violence 1.30 (1.18, 1.44) 1.20 (1.09, 1.33) 1625
1567 1626
Peer victimization 1.28 (1.17, 1.39) 1.25 (1.14, 1.36)
1568 1627
Cyber-victimization 1.32 (1.16, 1.50) 1.29 (1.12, 1.48)
1569 1628
1570 Crime victimization 1.25 (1.16, 1.34) 1.19 (1.10, 1.28) 1629
1571 Note: Statistically significant estimates (p > .05) are in bold. b ¼ standardized regression coefficient; CI¼confidence interval; IRR¼incidence rate ratio 1630
1572 (interpretable as risk ratios); N/A ¼ not applicable. Q8 1631
1573 a
Model 1: Models are adjusted for sex only. 1632
1574 b 1633
Model 2: Models are additionally adjusted for the baseline measurement of the respective outcome (eg, when predicting personality functioning at
1575 age 18, we adjusted for personality ratings by examiners at age 12; when predicting attainment at age 18, we adjusted for educational attainment as 1634
1576 rated by teachers at age 12; when predicting victimization during adolescence, we adjusted for childhood physical or sexual victimization by an adult 1635
1577 as assessed through mothers’ reports, as previously described.1,2 Adjusting for the baseline measurement was not always possible (eg, because 1636
1578 outcomes could not or were not measured during childhood, such as risky sexual behavior). 1637
1579 1638
1580 1639
1581 1640
1582 1641
1583 1642
1584 TABLE S2 Genetic and Environmental Influences on Borderline Symptoms at Age 12 1643
1585 1644
1586 Variance Components 1645
1587 A (95% CI) C (95% CI) E (95% CI) 1646
1588 Age 12 borderline symptoms 0.66 (0.58, 0.70) 0.00 (0.00, 0.07) 0.34 (0.30, 0.38) 1647
1589 1648
1590 Note: A ¼ additive genetic influences; C ¼ shared environmental influences; E ¼ nonshared environmental influences; CI ¼ confidence interval. 1649
1591 1650

13.e1 www.jaacap.org Journal of the American Academy of Child & Adolescent Psychiatry
Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce

 OUTCOMES OF BORDERLINE SYMPTOMS AT AGE 12

SUPPLEMENTAL REFERENCES
1651 1. Danese A, Moffitt TE, Arseneault L, et al. The origins of cognitive deficits in victimized 2. Marzi SJ, Sugden K, Arseneault L, et al. Analysis of DNA methylation in young people: 1710
children: implications for neuroscientists and clinicians. Am J Psychiatry. 2017;174: limited evidence for an association between victimization stress and epigenetic variation in
1652 349-361. blood. Am J Psychiatry. 2018;175:517-529.
1711
1653 1712
1654 1713
1655 1714
1656 1715
1657 1716
1658 1717
1659 1718
1660 1719
1661 1720
1662 1721
1663 1722
1664 1723
1665 1724
1666 1725
1667 1726
1668 1727
1669 1728
1670 1729
1671 1730
1672 1731
1673 1732
1674 1733
1675 1734
1676 1735
1677 1736
1678 1737
1679 1738
1680 1739
1681 1740
1682 1741
1683 1742
1684 1743
1685 1744
1686 1745
1687 1746
1688 1747
1689 1748
1690 1749
1691 1750
1692 1751
1693 1752
1694 1753
1695 1754
1696 1755
1697 1756
1698 1757
1699 1758
1700 1759
1701 1760
1702 1761
1703 1762
1704 1763
1705 1764
1706 1765
1707 1766
1708 1767
1709 1768

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 13.e2
Volume - / Number - / - 2019

FLA 5.6.0 DTD
JAAC2620_proof
4 September 2019
1:42 am
ce