Asarnow 2021

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

N EW R E S E A R C H

Dialectical Behavior Therapy for Suicidal Self-Harming


Youth: Emotion Regulation, Mechanisms, and Mediators
Joan Rosenbaum Asarnow, PhD, Michele S. Berk, PhD, Jamie Bedics, PhD, Molly Adrian, PhD,
Robert Gallop, PhD, Judith Cohen, MD, Kathryn Korslund, PhD, Jennifer Hughes, PhD,
Claudia Avina, PhD, Marsha M. Linehan, PhD, Elizabeth McCauley, PhD

Objective: This study evaluated mechanisms, mediation, and secondary/exploratory outcomes in our randomized controlled trial evaluating dia-
lectical behavior therapy (DBT) compared to individual and group supportive therapy (IGST). We expand on previously reported results indicating a
DBT advantage at posttreatment on planned suicide/self-harm outcomes, and greater self-harm remission (absence of self-harm, post hoc exploratory
outcome) during active-treatment and follow-up periods.
Method: This was a multi-site randomized trial of 173 adolescents with prior suicide attempts, self-harm, and suicidal ideation. Randomization was to
6 months of DBT or IGST, with outcomes monitored through 12 months. Youth emotion regulation was the primary mechanistic outcome.
Results: Compared to IGST, greater improvements in youth emotion regulation were found in DBT through the treatment-period [t(498) ¼ 2.36,
p ¼ .019] and 12-month study period (t(498) ¼ 2.93, p ¼ .004). Their parents reported using more DBT skills: posttreatment t(497) ¼ 4.12, p <
.001); 12-month follow-up t(497) ¼ 3.71, p < .001). Mediation analyses predicted to self-harm remission during the 6- to 12-month follow-up, the
prespecified outcome and only suicidality/self-harm variable with a significant DBT effect at follow-up (DBT 49.3%; IGST 29.7%, p ¼ .013). Im-
provements in youth emotion regulation during treatment mediated the association between DBT and self-harm remission during follow-up (months
612, estimate 1.71, CI 1.012.87, p ¼ .045). Youths in DBT reported lower substance misuse, externalizing behavior, and total problems at
posttreatment/6 months, and externalizing behavior throughout follow-up/12 months.
Conclusion: Results support the significance of emotion regulation as a treatment target for reducing self-harm, and indicate a DBT advantage on
substance misuse, externalizing behavior, and self-harm-remission, with 49.3% of youths in DBT achieving self-harm remission during follow-up.
Clinical trial registration information: Collaborative Adolescent Research on Emotions and Suicide; https://www.clinicaltrials.gov/;
NCT01528020
Key Words: suicide, self-harm, nonsuicidal self-injury, dialectical behavior therapy
J Am Acad Child Adolesc Psychiatry 2021;-(-):-–-.

uicide is currently the second leading cause of availability of telephone coaching 24 hours daily, and therapist
S death among adolescents and young adults in the
United States, and is responsible for more deaths
than any single medical illness.1 The need to reduce suicide
consultation teams. Based on the biosocial DBT theory, DBT
aims to strengthen skills that lead to improved emotion regu-
lation, as difficulties in emotion regulation are viewed as a
deaths has led to efforts to identify effective treatments for driver of suicidal and self-harm behaviors, which are viewed as
high-risk individuals and to clarify mechanisms contributing attempts to regulate intense and/or painful emotions.6 DBT
to treatment efficacy. One approach to this aim has involved also aims to improve skills in parents (eg validation, behavioral
treatment trials focusing on individuals with histories of management, emotion regulation), which are hypothesized to
previous suicide attempts (SAs) and nonsuicidal self-injury support DBT skill use and improved emotion regulation in
(NSSI), a reliable indicator of risk for fatal and their adolescents.
nonfatal SAs.2 Although definitions of emotion regulation vary, DBT
Dialectical behavior therapy (DBT) has emerged as a conceptualizes emotion regulation as involving the
promising treatment for this high-risk population.3-5 With following: sensitivity to emotional stimuli; intensity of
adolescents, this multi-component treatment includes indi- emotional reactions; and ability to regulate negative affective
vidual psychotherapy, multifamily group skills training, responses.6 Within biosocial theory, self-harm (SAs plus

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


Volume - / Number - / - 2021
ASARNOW et al.

NSSI, hereafter called self-harm) is conceptualized as an This secondary report expands on an earlier report of
emotion regulation strategy that occurs among individuals the US DBT trial.5,13 First, we report on mechanistic tar-
who have a biologically based sensitivity to emotions that gets. Consistent with the prediction that improvements in
interacts with adverse and/or invalidating experiences/in- youth emotion regulation would reduce the likelihood of
teractions (eg, “this is nothing to be upset about”) to yield self-harm, emotion regulation was the primary pre-specified
difficulties regulating emotional reactions.6 Self-harm pro- mechanistic target. Exploratory mechanistic targets
vides a strategy for regulating intense negative emotions and included: youth and parent DBT skills, given the emphasis
is strengthened due to negative reinforcement, relief from on DBT skills for improving emotion regulation and
painful emotions. For example, the youth who responds to reducing self-harm risk; and parent emotion regulation. The
intense distress by cutting is often distracted by the cutting, rationale for identifying emotion regulation as the primary
lessening the distress from painful emotions. This concep- mechanistic outcome was that even if DBT skills increased
tualization is consistent with other views of NSSI that in youth, if this did not lead to improved emotion regula-
emphasize emotion regulation, the intra- and interpersonal tion, we would not expect improvements in suicidality and
consequences of self-harm, and reinforcement patterns.7,8 self-harm. Similarly, improvements in DBT skills and
Despite extensive literature supporting the value of DBT emotion regulation in parents would not necessarily lead to
for high-risk adults,3 research on adolescents is limited. Two improvements in youth ability to regulate emotions. Sec-
randomized controlled trials (RCTs) have reported advantages ond, we examine whether the DBT treatment effect would
of DBT for reducing self-harm. A first RCT conducted in be mediated by improvement in youth emotion regulation.
Norway reported a significant advantage for DBT for reducing Third, we report on secondary clinical (youth symptoms,
self-harm frequency, compared to treatment as usual (TAU, functioning) and exploratory (parent distress/symptoms)
enhanced by standardizing session frequency to an average of at outcomes. These outcomes were pre-specified prior to study
least 1 weekly session). Results on emotion regulation were not implementation, and were selected because youths with
reported.4 When followed up at 1 year and again at 3 years, the suicidality commonly present with other significant mental
advantage on self-harm remained.9,10 Reduced hopelessness health and substance use symptoms in the context of
during treatment, mediated the effect of DBT on self-harm increased parental distress/symptoms.14
frequency at 3-years.10 A second RCT conducted in the We predicted that, compared to IGST, DBT would be
United States demonstrated that, when compared to individ- associated with significantly greater improvements on the
ual and group supportive psychotherapy (IGST) matched to mechanistic target of youth emotion regulation throughout
DBT to include both individual and group treatment com- the 12-month study observation period, and that improve-
ponents, DBT led to fewer SAs (primary outcome) and NSSI ments in youth emotion regulation during the 6-month
and total self-harm episodes (secondary outcomes), with a treatment period would be associated with an increased
significant advantage only at posttreatment.5 At both post- likelihood that youths would show self-harm remission (ie,
treatment/6 months and follow-up/12 months, DBT youths no self-harm at any point) during the follow-up period
were significantly more likely than IGST youths to show an (612 months). We selected self-harm remission as the
absence of self-harm at any point, a post hoc variable developed outcome for this analysis, as self-harm remission was signif-
as an indicator of clinically significant change, referred to icantly greater in the DBT condition during the follow-up
hereafter as self-harm remission. Self-harm remission was the period, providing an intervention effect to be mediated.
only examined suicide/self-harm outcome on which DBT
showed an advantage during the follow-up interval of 6 to 12 METHOD
months. Detailed descriptions of participants, assessments, treat-
Evaluation of emotion regulation or other mechanisms ments, and outcomes are available elsewhere.5 We focus
contributing to DBT’s efficacy for reducing self-harm can here on measures and procedures relevant to this report.
enhance understanding of processes that drive clinical im- The study was reviewed by each site’s local institutional
provements and ultimately lead to more effective treatments.11 review board and monitored by a Data Safety and Moni-
However, there is limited information on posited mechanistic toring Board. All youths and parents gave informed assent/
outcomes, particularly the degree to which improvements in consent (as appropriate).
emotion regulation mediate DBT treatment effects in ado-
lescents. Although there are some data to support DBT effects Participant Selection, Recruitment, and Enrollment
on emotion regulation in adults,6,12 additional evaluation of Participants (N ¼ 173) were recruited from January 2012
whether improvements in emotion regulation account for to August 2014 at 4 sites: University of Washington; Seattle
intervention benefits is needed. Children’s Hospital; Harbor-UCLA Medical Center; and
2 www.jaacap.org Journal of the American Academy of Child & Adolescent Psychiatry
Volume - / Number - / - 2021
DBT FOR SUICIDAL SELF-HARMING YOUTH

UCLA Medical Center. Inclusion criteria were age 12 to 18 both conditions: DBT (n ¼ 384 sessions), mean 4.1 (4.0
years; 1 or more lifetime SA; 3 lifetime self-harm epi- considered adherent), SD 0.15; IGST (n ¼ 386 sessions),
sodes, 1 in 12-weeks before screening; 24 on the Suicidal mean 99.32% (>90% considered adherent), SD 3.64.
Ideation QuestionnaireJunior (SIQ-Jr),15 3 Borderline
Personality Disorder criteria on Structured Clinical Inter- Assessments
view for the DSM-IV, Axis II, SCID-II).16 Exclusion criteria Assessments were conducted at baseline, 3 months, 6
were primary problem psychosis, mania, anorexia, other life- months (posttreatment), 9 months, and 12 months. As-
threatening condition; IQ <70, Kauffman Brief Intelli- sessors were masked to treatment arm, and every effort was
gence Test17; youths not fluent in English; and parent not taken to conceal treatment arm, including separating work
fluent in English or Spanish. spaces for therapists and evaluators to prevent chance en-
counters with therapists during evaluations. After training,
Randomization
interview measures were co-rated by a designated “criterion
A computerized adaptive minimization procedure was used interviewer” until the assessor demonstrated 0.80 interrater
to randomly assign participants to 6 months of DBT or reliability, with 1 in 15 randomly selected interviews co-
IGST, with groups matched within sites on age, number of rated thereafter to ensure ongoing reliability.5
SAs, number of previous self-injuries, and psychiatric
medication use.5 Randomization status and sequence were Youth Self-harm, SAs, and NSSI. The Suicide Attempt
concealed from recruitment/assessment staff. Self-Injury Interview (SASII)20 measured the frequency,
intent, and severity of self-harm acts (suicide attempts,
Treatment Conditions NSSI). The SASII has demonstrated strong reliability and
Both treatments were manualized, used similar training and validity; interrater reliability was maintained throughout the
adherence protocols, and offered comparable treatment study at ICC  0.80 at the item level.5
exposure.
DBT Mechanistic Targets: Youths and Parents. Emotion
DBT. DBT included the following: weekly individual psy- regulation was assessed using the Difficulties in Emotion
chotherapy; multi-family group skills training; brief (gener- Regulation Scale (DERS; higher scores indicate more dys-
ally about 10-minutes) youth and parent telephone coaching regulation).21 Items assess the following: emotional aware-
focused on skill use/crisis support available 24 hours; and ness; emotional clarity; acceptance; goal-directed behavior;
weekly therapist teams.18,19 Parents were seen in the first impulse control; and emotion regulation strategies. Self-
session, with an optional 7 family sessions. Suicide risk was report on the Revised Ways of Coping Checklist
monitored regularly; increased risk triggered use of the (RWCCL, DBT-Scale) measured coping skills taught
Linehan Suicide-Risk Assessment and Management Protocol through DBT (emotion regulation, mindfulness, distress
(LRAMP).5 tolerance, interpersonal effectiveness).22
IGST. IGST emphasized acceptance, validation, and feelings Youth-Reported Psychopathology/Substance Use. The
of connectedness/belonging. IGST included individual Borderline Personality Features Scale for Children (BPFS)23
psychotherapy, adolescent supportive group therapy, as assessed borderline symptoms. The Center for Epidemio-
needed (7) parent-sessions, and weekly therapist team logical StudiesDepression Scale (CES-D) evaluated
consultation. Assessment and management of suicidal depression.24 Substance abuserelated impairment (here-
behavior followed the AACAP Practice Parameters. Thera- after described as substance misuse) was assessed using the
pists were available by telephone during office hours, and Drug Use Screening Inventory (DUSI).25 The Youth Self
crisis numbers were provided for 24-hour coverage. Report assessed internalizing (anxiety and depression),
Therapist Training/Quality Assurance
externalizing (behavior), and total problems.26 Raw scores
were used in analyses due to truncation of T scores at the
Therapists provided treatment in 1 study arm and attended a
lower range, allowing full consideration of the variation in
multi-day training led by the treatment developer (DBT,
scores.26
Linehan; IGST, Cohen). Within each condition, therapists
participated in weekly cross-site training/meetings and weekly Baseline Clinical Interviews. Baseline DSM-IV-TR di-
site team/consultation meetings (DBT) or group supervision agnoses are provided based on mood, anxiety, psychosis, and
(IGST). Treatment adherence was evaluated on randomly eating disorder modules from the Schedule for Affective Dis-
selected individual and group sessions using the DBT and orders and Schizophrenia for School-Age ChildrenPresent
IGST adherence scales, respectively.5 Adherence was strong in and Lifetime Version [K-SADS-PL])27 and Structured

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


Volume - / Number - / - 2021
ASARNOW et al.

Clinical Interview for DSM-II borderline personality disorder confounding variables. Accordingly, we implemented a
module.16 causal mediation model called the marginal mediation
model, which does not require the assumption of sequential
Parent-Reported Distress and Symptoms. Parent distress
ignorability, and connects the common mediation param-
and symptoms were assessed using the Brief Symptom In-
eters to causal parameters.34 The marginal mediation model
ventory Global Severity Index (BSI).28 The Borderline
produces the significance of the multiplicative paths from
Symptom List (BSL-23)29 assessed intensity of 23 symp-
(1) intervention to the outcome (path c), (2) intervention to
toms characteristic of borderline personality disorder.
the posited mediator (path a), and (3) the mediator to
Demographic Information. Parent report was obtained on outcome (path b), controlling for intervention under the
youth age, gender, race/ethnicity, and family income. marginal mediation model referred to as the natural indirect
effect (NIE). Path c’ corresponds to the controlled direct
Statistical Analyses effect (CDE) of the intervention on the outcome after the
Our analytic approach had 3 foci: (1) to evaluate whether mediator is introduced. The CDE and NIE account for the
DBT led to differential improvement on hypothesized direct and indirect effect, respectively, while addressing the
mechanistic variables; (2) to assess whether improvement in sequential ignorability violation.34 To maintain temporal
youth emotion regulation during the active treatment sequence between the mediatoroutcome relationship, the
period would mediate the differential treatment effect on outcome is self-harm remission during the follow-up period
self-harm remission at 12 months; and (3) to examine with MCMC imputation used to account for missing data.
mechanistic and secondary/exploratory outcomes, for which This same approach was used for analyses of treatment and
we used the same intent-to-treat approach used in our prior time effects on self-harm remission.
paper.5 For continuous measures, we used a 2-level hierar- We limit description to effects of p < .05, with some
chical linear model (HLM) in which change over time fol- marginal effects (p < .10) noted. To protect against type I
lowed a mathematical piecewise linear profile with 2 phases of errors with multiple tests, we carefully specified primary
change: baseline through posttreatment/6 months; and outcomes and contrasts within each aim, and emphasize
posttreatment/6 months through follow-up/12 months. effect sizes and confidence intervals. For analyses of mech-
These analyses included treatment group (DBT, IGST) as the anisms and secondary/exploratory measures, we imple-
between-subjects factor. Pairwise contrasts from the HLM mented the false discovery rate approach35 in sensitivity
models were used to evaluate between-group differences for analyses adjusting for the number of tests within each
change during active treatment, change during the follow-up domain and report these values (pFDR). Although this more
period, and on-average change over the entire longitudinal conservative approach considers the potential for inflation of
period.30 Analyses adjusted for site and assessed for differen- type I error rates, it also increases the risk of type II errors
tial treatment effects across site by including a site  treat- whereby “true” effects are attributed to chance.36
ment interaction. Site  treatment interactions were
nonsignificant. Because higher levels of treatment were pre- RESULTS
dicted in DBT versus IGST, we used pattern-mixture models As reported previously5 (see Table S1, available online), at
to evaluate evidence of an informative attrition mechanism, baseline the sample had a mean age of 14.89 years (SD 1.47
defined in this case as differential treatment rates leading to years), was 95% female, 28% Hispanic, 44% racial mi-
differences in outcomes.31 Sensitivity analyses using Markov nority, included a substantial proportion of youths with K-
Chain Monte Carlo (MCMC) imputation methods (PROC SADS-PLdefined current depressive disorders (83.81%)
MI, SAS 9.4)32 were also used on primary and secondary and anxiety disorders (54.10%) and SCID-IIdefined
models to determine the impact of attrition/missing data. borderline personality disorder (53.20%), with no statisti-
In assessing mediation, traditional mediation models cally significant between-group differences on demographic
assume that there are no unmeasured confounders affecting and clinical variables. Among the 173 RCT participants
the relationship between the mediator and outcome, (DBT n ¼ 86, IGST n ¼ 87), 164 completed 1 post-
defined as sequential ignorability.33 Because we defined the baseline assessment (see Figure S1, available online).
mediator as change in the variables in question from base- Completion rates were as follows: DBT n ¼ 84, IGST n ¼
line to posttreatment/6 months, participants were not ran- 80 at 6 months; DBT n ¼ 77, IGST n ¼ 69 at 12 months.
domized to levels of the mediator, raising the possibility that The number of youths missing all follow-up evaluations did
observed relations between mediators and subsequent out- not differ significantly across groups (DBT 2.3% [2/86];
comes might be spurious and depend on unknown IGST 8.0% [7/87], Fisher exact p ¼ .17).

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


Volume - / Number - / - 2021
DBT FOR SUICIDAL SELF-HARMING YOUTH

Sensitivity analyses for the impact of attrition/missing achieve an additional participant with self-harm remission
data and attendance yielded results consistent with those for during the follow-up period, an additional 5.2 participants
our primary analyses, with no change in conclusions. To would need to be treated with DBT rather than IGST.
limit redundancy, we report the primary analyses below. Time effects were also observed indicating that remission
during treatment predicted remission status during follow-
Mechanistic Targets up [c2 (1) ¼ 17.58, p < .0001]. Sensitivity analyses us-
There was a significant treatment-by-time interaction on ing observed cases only and last observation carried forward/
emotion regulation, with the DBT group improving more worse-case imputation for missing data (as previously re-
during the active treatment period [06 months, ported)5 confirmed all reported effects.
difference ¼ 9.04, SE ¼ 3.835, t(498) ¼ 2.36, p ¼ .019,
pFDR ¼ .038], and throughout the full study period [012 Mediation
months, difference ¼ 10.75, SE ¼ 3.667, t(498) ¼ 2.93, p ¼ Analyses of mediation indicated that changes in youth
.004, pFDR ¼ 0.007]. (Table 137 shows group differences in emotion regulation during active treatment (06 months)
rates of change quantified through d effect sizes). Although mediated the effect of DBT on self-harm remission during
these group differences weakened during the follow-up follow-up (612 months). In Figure 2, effect sizes for all
period (612 months), the DBT group maintained their paths are illustrated as d.39 There was a significant inter-
gains and showed more improvement in emotion regulation vention effect on youth emotion regulation (path a)
(declining DERS) across the full 12-month period compared [t(171) ¼ 2.36, p ¼ .02, d ¼ 0.36, 95% CI ¼ 0.050.66],
to the IGST group (Figure 1). Parents in the DBT group and a significant effect of emotion regulation on outcome,
reported significantly greater use of DBT skills (RWCCL) at controlling for treatment condition, a natural indirect effect
the end of treatment [06 months, difference 0.24, SE ¼ (NIE; path b) [c2 (1) ¼ 4.02, p ¼ .045, odds ratio (OR) ¼
0.059, t(497) ¼ 4.12, p < .001, pFDR < .001), and at the last 1.71, 95% CI ¼ 1.012.87], which indicates that for each
study follow-up (12 months, difference ¼ 0.22, SE ¼ SD decrease in emotion regulation problems during treat-
0.058, t(497) ¼ 3.71, p < .001, pFDR < .001 ). Table S2 ment, there is a 70.6% increase in the odds of self-harm
provides estimated means and standard deviations, and remission at 12 months. After adjusting for the mediator,
Table S3 gives rates of change for the DBT and IGST groups the direct effect of treatment on follow-up/12-month self-
(both available online). harm was nonsignificant [c2 (1) ¼ 0.36, p ¼ .55]. The size
Significant time effects were also observed, with the of the direct effect for treatment as an odds ratio is 1.29, with
strongest time effects generally observed during the acute a 95% CI of 0.57 to 2.91. Therefore, after controlling for
treatment period and a leveling off during the follow-up change in emotion regulation, the increase in the odds of self-
period (Table 1, Figure 1). Specifically, both DBT and harm remission during follow-up reduced from 122% to
IGST youths showed significant improvements in emotion 29%. Sensitivity analyses yielded similar results as follows:
regulation (lower DERS). However, within-group effects in only observed data [significant indirect effectNIE, c2 (1) ¼
the DBT group were large (d ¼ 1.02 and 1.33 during the 4.02, p ¼ .045, OR ¼ 1.74, 95% CI ¼ 1.023.01,
treatment (06 months) and full study (012 months) nonsignificant direct effect c2(1) ¼ 1.32, p ¼ .25]; and using
periods, respectively, compared to moderate effects in IGST, the last observation carried forward (LOCF)/worse-case
d ¼ 0.61 and 0.78 (see Tables S2 and S3, available on- imputation [significant natural indirect effectNIE, c2
line). Youth skills (RWCCL) also improved over time in (1) ¼ 4.12, p ¼ .044, OR ¼ 1.75, 95% CI ¼ 1.023.02,
both groups (see Table S2, available online). In contrast, nonsignificant direct-effect c2 (1) ¼ 0.63, p ¼ .43].
whereas DBT parents reported increased skill use, IGST Exploratory probes of whether changes in parent or
parents reported significantly less DBT skill use over time youth skills during treatment mediated the DBT effect
(see Figure 1 and Tables S2 and S3, available online). on 12-month self-harm remission provided no evidence
of mediation for parent skills. A marginal mediation ef-
Self-Harm Remission fect was detected for youth skills [NIE, path b, c2
Expanding on our prior report,5 the self-harm remission (1) ¼2.84, p ¼ .092, OR ¼ 1.44, 95% CI ¼
rate for DBT was significantly higher compared to IGST for 0.952.21]. Sensitivity analysis confirmed these results
both the treatment [c2 (1) ¼ 5.64, p ¼ .018] and follow-up using the observed data [natural indirect effect-NIE, c2
periods [c2 (1) ¼ 6.21, p ¼ .013] (Table 2). Results (1) ¼ 2.92, p ¼ .087, OR ¼ 1.53, 95% CI ¼
indicate that, on average, 5.6 participants would need to be 0.884.53] and using LOCF/worse-case imputation
treated with DBT instead of IGST for 1 additional partic- (indirect effect-NIE, c2 (1) ¼ 2.96, p ¼ .085, OR ¼
ipant to achieve self-harm remission at 6 months. To 1.43, 95% CI ¼ 0.894.53].
Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 5
Volume - / Number - / - 2021
ASARNOW et al.

TABLE 1 Between-Group Differences in Rates of Change Quantified Through d Effects Sizes with 95% Confidence Intervals
(Group-by-Time Interactions) Over Each Study Period Plus Main Effects for Change Over Time

Effect of During treatment During follow-up


Measure interest (06 mo) (612 mo) Total change (012 mo)
Youth mechanistic outcomes
Emotion Regulation Treatment 0.36 (0.05 to 0.66)* 0.06 (e0.24 to 0.36) 0.45 (0.14 to 0.75)**
(DERS)LPrimary contrast
Mechanistic Outcome
Time e0.81 (e0.96 to e0.66)*** e0.17 (e0.32 to e0.01)* e1.04 (e1.19 to e0.89)***
Skills (RWCCL) Treatment e0.15 (e0.45 to 0.15) 0.05 (e0.25 to 0.35) e0.10 (e0.41 to 0.20)
contrast
Time 0.44 (0.29 e 0.59)*** 0.15 (e0.01 e 0.30)* 0.60 (0.45 e 0.76)***
Parent mechanistic outcomes
Emotion Treatment 0.13 (e0.18 to 0.43) e0.24 (e0.55 to 0.06) e0.13 (e0.43 to 0.17)
Regulation (DERS) contrast
Time e0.14 (e0.30 to 0.01) e0.11 (e0.27 to 0.04) e0.27 (e0.42 to e0.11)**
Skills (RWCCL) Treatment e0.62 (e0.93 to e0.31)*** 0.07 (e0.23 to 0.37) e0.58 (e0.88 to e0.27)***
contrast
Time 0.20 (0.05 e 0.35)** e0.21 (e0.37 e e0.06)** e0.03 (e0.18 e 0.13)
Youth clinical outcomes: secondary trial outcomes
Borderline (BPFS-C) Treatment 0.22 (e0.08 to 0.52) 0.00 (e0.30 to 0.30) 0.23 (e0.07 to 0.53)
contrast
Time e0.50 (e0.65 to e0.35)*** e0.26 (e0.41 e0.82 (e0.97 to e0.67)***
to e0.10)**
Depression (CES-D) Treatment 0.01 (e0.29 to 0.31) 0.05 (e0.25 to 0.35) 0.07 (e0.23 to 0.37)
contrast
Time e0.65 (e0.80 to e0.50)*** e0.12 (e0.27 to 0.04) e0.79 (e0.94 to e0.64)***
Substance Treatment 0.33 (0.02 to 0.63)* e0.17 (e0.47 to 0.13) 0.15 (e0.15 to 0.45)
misuse (DUSI) contrast
Time e0.01 (e0.17 to 0.14) e0.03 (e0.18 to 0.12) e0.04 (e0.20 to 0.11)
Internalizing (YSR-I) Treatment 0.15 (e0.15 to 0.45) e0.11 (e0.41 to 0.19) 0.01 (e0.29 to 0.31)
contrast
Time e0.58 (e0.73 to e0.43)*** e0.12 (e0.27 to 0.03) e0.63 (e0.78 to e0.48)***
Externalizing (YSR-E) Treatment 0.36 (0.06 to 0.67)* e0.05 (e0.35 to 0.26) 0.31 (0.00 to 0.61)*
contrast
Time e0.22 (e0.37 to e0.06)** e0.17 (e0.32 to e0.02)* e0.39 (e0.54 to e0.24)***
Total problems (YSR-TP) Treatment 0.29 (e0.02 to 0.59)* e0.11 (e0.41 to 0.19) 0.14 (e0.17 to 0.44)
contrast
Time e0.54 (e0.69 to e0.39)*** e0.17 (e0.32 to e0.01)* e0.66 (e0.81 to e0.51)***
Social adjustment (SAS) Treatment e0.06 (e0.36 to 0.24) e0.01 (e0.31 to 0.29) e0.07 (e0.37 to 0.23)
contrast
Time e0.46 (e0.61 to e0.31)*** e0.16 (e0.31 to e0.01)* e0.61 (e0.76 to e0.46)***
Parent clinical outcomes: exploratory trial outcomes
Borderline (BSL) Treatment 0.12 (e0.19 to 0.42) e0.10 (e0.40 to 0.20) 0.02 (e0.28 to 0.32)
contrast
Time e0.25 (e0.4 to e0.1)*** e0.10 (e0.26 to 0.05) e0.38 (e0.53 to e0.23)***
Symptom severity (BSI) Treatment 0.23 (e0.07 to 0.53) e0.10 (e0.40 to 0.20) 0.14 (e0.16 to 0.44)
contrast
Time e0.15 (e0.30 to 0.00)* e0.12 (e0.27 to 0.03) e0.28 (e0.43 to e0.13)***
37
Note: d (95% CI) are given. Thresholds for small, medium, and large effects are 0.20, 0.50, and 0.80, respectively. BPFS-C ¼ Borderline Personality
Features ScaleChildren; BSI ¼ Brief Symptom Inventory; BSL ¼ Borderline Symptom List; CES-D ¼ Center for Epidemiologic StudiesDepression
Scale; DERS ¼ Difficulties in Emotion Regulation Scale; DUSI ¼ Drug Use Screening Inventory; RWCCL ¼ Revised Ways of Coping Checklist; SAS ¼
Social Adjustment ScaleSelf Report; YSR ¼ Youth Self-Report subscales: E ¼ externalizing, I ¼ internalizing, TP ¼ total problems.
*
p < .05; **p < .01; ***p < .001, based on tests of derived estimates.

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


Volume - / Number - / - 2021
DBT FOR SUICIDAL SELF-HARMING YOUTH

FIGURE 1 Change in Dialectical Behavior Therapy and Individual and Group Supportive Therapy Groups From Baseline Through
Active Treatment (06 Months) and Follow-up (612 Months) Periods
Youth Emotion Regulation (DERS) Youth Substance Misuse (DUSI)
130 27
125
25
120
115 23
110
21
105
100 19
95
17
90
85 15
0 3 6 9 12 0 3 6 9 12
Time in Treatment (Months) Time in Treatment (Months)
Treatment

Parent DBT Skills (RWCCL-DSS) Youth Externalizing Problems (YSR-E)


2.25 27

2.2
25
2.15
23
2.1

2.05 21

2
19
1.95
17
1.9

1.85 15
0 3 6 9 12 0 3 6 9 12
Time in Treatment (Months) Time in Treatment (Months)

Note: Plots show the estimated means for each treatment group over time, based on the hierarchical linear models. DBT ¼ dialectical behavior therapy; DERS ¼ Difficulties
in Emotion Regulation Scale; DUSI ¼ Drug Use Screening Inventory; IGST ¼ individual and group supportive therapy; RWCCL-DSS ¼ Revised Ways of Coping Checklist
Dialectical Behavior Therapy Skills Subscale; YSR-E ¼ Youth Self-Reportexternalizing scale (raw scores).

Secondary and Exploratory Outcomes SE ¼ 1.66, t(254) ¼ 2.06, p ¼ .041, pFDR ¼ .27; YSR total
The DBT group showed greater improvement during the problems posttreatment/6 months, difference ¼ 3.17, SE ¼
active treatment period (06 months) on DUSI substance 1.59, t(170) ¼ 1.99, p ¼ .048, pFDR ¼ .096]. Effect sizes
misuse, YSR externalizing behavior problems, and YSR total were in the small range (0.290.36). The DBT advantage
problems, and on YSR externalizing problems through 12 survived FDR adjustment for decreasing substance misuse
months (Table 1, Figure 1) [DUSI posttreatment/6 (DUSI) and YSR externalizing problems at posttreatment/6
months, difference ¼ 9.47, SE ¼ 3.60, t(170) ¼ 2.63, p ¼ months.
.009, pFDR ¼ .023; YSR externalizing problems posttreat- Significant time effects were also observed, with
ment/6 months, difference 3.86, SE ¼ 1.61, t(170) ¼ 2.39, improvement in both groups, and the greatest change
p ¼ .018, pFDR ¼ .046; 012 months, difference 3.33, occurring during the treatment period (see Table 1 and

TABLE 2 Self-Harm Remission Among Youths Randomized to Dialectical Behavior Therapy (DBT) and Individual and Group
Supportive Therapy (IGST) During the Treatment and Follow-up Periods

Self-harm remission
Treatment effect DBT IGST OR 95% CI c2 (1 df) p NNT
Treatment period 0-6 months 34 (44.2%) 18 (27.3%) 2.12 1.14L3.94 5.64 .018 5.617
Follow-up period 6-12 months 34 (49.3%) 19 (29.7%) 2.17 1.18L3.97 6.21 .013 5.248

Note: Percentages are unadjusted numbers based on number of youths with available data for the examined time period: treatment period 77 DBT, 66
IGST; follow-up period 69 DBT, 64 IGST; data for both treatment and follow-up periods 69 DBT, 64 IGST. Presented values differ from those reported
in McCauley et al. (2018),5 which were adjusted based on last observation carried forward (LOCF) or worse-case imputation for participants with
complete missing data in the respective period. Presented logistic regression models used multiple imputation (Markov Chain Monte Carlo [MCMC]).
Treatment effects were confirmed in sensitivity analyses using the following: only observed data (06 months, OR ¼ 2.11, 95% CI ¼ 1.044.26, c2 ¼
4.31, p ¼ .038; 612 months, OR ¼ 2.30, 95% CI ¼ 1.134.70, c2 (1) ¼ 5.23, p ¼ .022); and LOCF/worse case imputation38 (06 months, OR ¼ 2.28,
95% CI ¼ 1.214.30, c2 ¼ 6.53, p ¼ .011, 612 months, OR ¼ 2.21, 95% CI ¼ 1.19–4.09, c2¼ 6.32, p ¼ .012). NNT ¼ number needed to treat; OR ¼
odds ratio.

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


Volume - / Number - / - 2021
ASARNOW et al.

Tables S2, and S3, available online). We saw improvements Moreover, improvements in emotion regulation at post-
at posttreatment and during the full 12-month period on: treatment significantly predicted greater self-harm remission
youth borderline symptoms (BPFS), depression (CES-D), at 12 months/follow-up and mediated the effect of DBT on
social adjustment (SAS-SR), YSR internalizing, external- self-harm remission at 12 months (such that the treatment
izing, and total problems; and parent borderline symptoms effect on self-harm remission was no longer significant after
(BSL) and BSI severity. The exception was youth substance adjusting for improvements in emotion regulation during
misuse which increased posttreatment in IGST, and treatment). These findings support the significance of
declined at follow-up (Figure 1). Despite this general emotion regulation as a therapeutic target in treatments for
improvement, a substantial number of youths in both self-harming youths and improvements in emotion regula-
groups scored above clinical cut scores: depression (CES- tion as a therapeutic change mechanism in DBT. Results are
D  24) 6 months DBT 47.1% (32/68), IGST 50.9% (29/ also consistent with accumulating evidence that dysregu-
57), 12 months DBT 44.4% (32/72), IGST 53.6%, (30/ lated emotions are associated with self-harm, particularly
56); externalizing symptoms (YSR T>63) 6 months DBT NSSI.40
36.8% (25/68), IGST 47.5% (29/61), 12 months DBT Another mechanistic target in DBT theory is DBT
30.4% (21/69), IGST 31.6% (18/57); internalizing symp- skills. Our finding that DBT was more effective than IGST
toms (YSR> 63) 6 months DBT 45.6% (31/68), IGST in improving parent skills, presumably due to their partic-
70.5% (43/61), 12 months DBT 47.8% (33/69), IGST ipation in multi-family skills training and some individual
57.9% (33/57); total problems (YSR>63) 6 months DBT sessions, indicates that parents receiving DBT used more
47.1% (32/68), IGST 62.3% (38/61), 12 months DBT DBT skills, which may have enhanced their abilities to
47.8% (33/69), IGST 50.9% (29/57) (numbers vary support skill use in their youths when they struggled with
because of missing data). painful emotions and self-harm urges. Although differential
treatment effects were not significant for youth skills, there
DISCUSSION was a trend-level signal that improvement in youth DBT
To our knowledge, CARES is the first randomized skills partially mediated the DBT advantage on self-harm
controlled trial with adolescents to demonstrate that DBT remission. Collectively, these results provide some support
led to significantly greater improvements on youth emotion for DBT theory, which emphasizes the value of emotion
regulation, a DBT “mechanistic target” through which regulation and DBT skills in treatments for self-harm.
DBT is hypothesized to lead to reduced self-harm. Although youths in both treatments improved on
secondary/exploratory clinical outcomes, DBT was supe-
FIGURE 2 Path Diagram of Mediation Model for Self-Harm rior to IGST for decreasing substance misuse, external-
Remission during 6- to 12-Month Follow-up Through izing problems, and total problems, an indicator of overall
Improvement in Youth Emotion Regulation During severity. In contrast to the pattern of declining substance
Treatment 0 to 6 Months misuse among DBT youths at posttreatment/6 months,
substance misuse increased in IGST. Given our finding
that improvements in emotion regulation mediated DBT
effects, the reduction among DBT youths in substance
misuse and externalizing behaviors may reflect improved
emotion regulation that mitigates the urge/impulse to use
substances or engage in dysregulated behavior. The
continuing substance misuse in both conditions may be
related to the trend towards increasing use of substances
during adolescence.41 From a clinical perspective, these
findings, combined with data indicating elevated rates of
death by alcohol or drug poisoning among youths with
self-harm histories,42-44 underscore the importance of
continuing attention to substance use in this high-risk
population.
Although the Norwegian DBT trial did not report on
Note: DBT ¼ dialectical behavior therapy; IGST ¼ individual and group supportive substance misuse and externalizing behaviors, a DBT
therapy.
advantage was observed on borderline symptoms and inter-
*p < .05.
viewer- but not self-rated depression at posttreatment,
8 www.jaacap.org Journal of the American Academy of Child & Adolescent Psychiatry
Volume - / Number - / - 2021
DBT FOR SUICIDAL SELF-HARMING YOUTH

although groups showed similar improvements on these problems, follow-up externalizing behavior). The FDR
outcomes at 1- and 3-year follow-ups.4,9,10 Our weaker effect adjustment increases the likelihood of type II error, falsely
on borderline symptoms may have been related to our concluding that there is no significant effect when a true
stronger comparator treatment, relative to the TAU condition difference occurs.36 Statistical power was limited for detect-
in the Norwegian trial, and our requirement of more severe ing small effects (eg, site effects).36 Replication is needed to
baseline suicidality and self-harm (prior SA, SIQ-Jr 24, 3 confirm study results. Study treatments were evaluated in an
self-harm episodes versus 2 self-harm episodes). Contrasting RCT; results might differ with treatments delivered under
severity levels on measures used at baseline, our youths had routine practice conditions. Finally, treatment effects beyond
more severe suicidal ideation (SIQ-Jr, 57.06 versus 36.91), our 12-month follow-up remain to be determined.
higher rates of BPD (SCID, 53.2% versus 20.5%), and In conclusion, the collective results of the CARES trial
higher rates of current depressive disorder (K-SADS-PL support the value of DBT for youths with severe suicidality
83.81% vs 59.8%). The 2 trials also used somewhat different and self-harm, particularly at the 6-month/posttreatment
DBT protocols. We offered 6 months of DBT using the point. However, the DBT advantage weakened during the
Linehan manual19; the Norwegian trial included 19 weeks follow-up period when study treatments were withdrawn. At
of DBT using the Miller, Rathus, and Linehan manual.18 the 12-month follow-up point, only the post hoc self-harm
Although a DBT advantage for reducing self-harm was remission variable showed a DBT advantage, not the plan-
observed in both this and the Norwegian trial, self-harm ned suicide/self-harm frequency outcomes. These data sug-
continued in many youths receiving DBT. In our trial, gest that DBT may be particularly beneficial for changing
50.7% of DBT youths engaged in self-harm during the 6- behavior to a point at which youth refrain from self-harm, an
to 12-month follow-up period. DBT youths in the Nor- important point because self-harm remission during treat-
wegian trial reported a mean of 9.0, 5.5, and 6.3 self-harm ment was associated with self-harm remission during follow-
episodes during the treatment, first follow-up year, and up. This pattern is consistent with other data indicating that
second follow-up year respectively. Although SAs are rare, early cessation of suicidal behavior is associated with
limiting power even in samples with elevated SA risk, continued remission of suicidal behavior.48 Our results on
neither study found a significant DBT advantage on SAs mediation and mechanistic outcomes suggest that improve-
during posttreatment follow-up intervals.5,10 Other, less ments in emotion regulation contributed to the DBT effect
intensive treatments have shown promise,37,44-46 including on self-harm remission, perhaps due to increased DBT-skill
an open trial of a 12-week, Internet-delivered adolescent use—a healthy, safe way to manage painful emotions.
emotion regulation therapy indicating that improvements in Thus, although results are promising, it is premature to draw
emotion regulation mediated reduced self-harm.47 These conclusions about how to most effectively treat youths pre-
collective results suggest the need for treatment algorithms senting with suicidal ideation, attempts, and repeat self-harm.
to inform clinical decisions regarding extending, intensi- The continuing self-harm and other mental health problems
fying, and/or augmenting treatment with the goal of opti- found in both our DBT and IGST groups during the follow-
mizing treatment response. up period underscores the need for continuing care and
Study limitations include our small number of male adaptive algorithms for guiding clinical decision making to
participants, the group most likely to die by suicide, sug- provide individual youths and families with treatment that is
gesting the need for alternative sampling/eligibility criteria. most likely to be beneficial. Future research is needed to
The primary CARES study outcome was suicide events, further elucidate mechanisms contributing to recovery in
underscoring the need for cautious interpretation and repli- youths at very high risk for suicide/SAs, to clarify strategies
cation of our mediation results, which focused on DBT ef- for optimizing treatment to prevent continuing and/or
fects on self-harm remission, a post hoc outcome developed recurrent risk, and ultimately to prevent the tragedy of death
as an indicator of clinically significant change. Although there by suicide. The question for the future is: can we do better,
was sample attrition over time, sensitivity analyses for the and if so, how?
impact of attrition/missing data produced results consistent
with the primary analyses, enhancing confidence in reported Accepted January 22, 2021.
findings. We examined multiple measures, and implemented
Drs. Asarnow and Avina are with the University of California, Los Angeles. Dr.
the false discovery rate (FDR) method to consider whether Berk is with Stanford University, California. Dr. Bedics is with California
results survived a correction for possible chance findings with Lutheran University, Thousand Oaks, California. Drs. Adrian, Korslund,
McCauley, and Linehan are with the University of Washington, Seattle. Dr.
multiple comparisons. Although most outcomes survived this Gallop is with the West Chester University of Pennsylvania, West Chester. Dr.
Cohen is with the Drexel University College of Medicine, Philadelphia, Penn-
adjustment (youth DERS, parent skills; substance misuse, sylvania. Dr. Hughes is with the University of Texas Southwestern, Dallas.
posttreatment externalizing behavior), others did not (total
Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 9
Volume - / Number - / - 2021
ASARNOW et al.

This research was supported by grants from the National Institute of Mental Generation Foundation. Dr. Berk has reported receiving grant money from
Health (NIMH; R01MH090159 and R01MH93898). The content is solely the re- NIMH and AFSP. Dr. Adrian has reported receiving grant money from NIMH
sponsibility of the authors and does not necessarily represent the official views and AFSP. She has reported receiving support from the Seattle Children’s
of the National Institutes of Health. Hospital Foundation. Dr. Gallop has reported receiving grant money from
NIMH and AFSP. Dr. Cohen has reported receiving grant money from NIMH,
Dr. Gallop served as the statistical expert for this research. the Eunice Kennedy Shriver National Institute of Child Health and Human
Author Contributions Development, and SAMHSA. She has reported receiving royalties from Guil-
Conceptualization: Asarnow, Berk, Bedics, Gallop, Cohen, Linehan, McCauley ford Press, UpToDate, and the Medical University of South Carolina.
Data curation: Asarnow, Berk, Bedics, Adrian, Gallop, Korslund, Hughes, Avina, Dr. Korslund has reported receiving a salary for her role as Clinical Director of
Linehan, McCauley THIRA Health, LLC, a DBT-based partial hospital and intensive outpatient
treatment program. She has reported holding shares in MODRE, Inc., which
Formal analysis: Asarnow, Bedics, Gallop
owns THIRA Health, LLC. She has reported receiving consulting fees for DBT
Funding acquisition: Asarnow, Berk, Gallop, Linehan, McCauley
and DBT adherence consultation on federally and internationally funded
Investigation: Asarnow, Berk, Bedics, Adrian, Gallop, Cohen, Korslund, research. She has been a trainer for Behavioral Tech, LLC, a training company
Hughes, Avina, Linehan, McCauley providing DBT training for mental health professionals. Dr. Hughes has served
Methodology: Asarnow, Berk, Bedics, Adrian, Gallop, Cohen, Korslund, as Youth Aware of Mental Health (YAM) trainer, has consulted for Mental
Hughes, Avina, Linehan, McCauley Health in Mind International, and has served as a board member for the
Project administration: Asarnow, Berk, Linehan, McCauley American Psychological Association (APA) Division 53, Society for Clinical Child
Resources: Cohen, Korslund, Linehan, McCauley and Adolescent Psychology (SCCAP). She has received a stipend for publishing
Software: Gallop the newsletter as part of her appointed position as the Chair and is current
Supervision: Asarnow, Berk, Gallop, Cohen, Korslund, Hughes, Avina, Linehan, Past-Chair of the Association for Behavioral and Cognitive Therapies (ABCT)
McCauley Child and Adolescent Depression Special Interest Group (unpaid position).
Validation: Berk, Adrian, Gallop, Cohen, Korslund, McCauley She has reported receiving royalties from Guilford Press. Dr. Avina has re-
Visualization: Bedics, Gallop ported receiving grant money from NIMH for her role on the study. Dr. Linehan
Writing e original draft: Asarnow, Bedics, Gallop, McCauley has reported receiving grant money from NIMH and AFSP. She has reported
Writing e review and editing: Asarnow, Berk, Bedics, Adrian, Gallop, Cohen, receiving royalties from Guilford Press for books that she has written on dia-
Korslund, Hughes, Avina, McCauley lectical behavior therapy, royalties for training materials from Behavioral Tech
LLC, and compensation for dialectical behavior therapy online programs and
ORCID books. Dr. McCauley has received grant or research support from NIMH, the
Joan Rosenbaum Asarnow, PhD: https://orcid.org/0000-0002-1610-044X Institute of Education Sciences - US Department of Education, AFSP, the
Molly Adrian, PhD: https://orcid.org/0000-0002-6415-4434 Scooty Fund, and the University of Washington. She has served as a consultant
Judith Cohen, MD: https://orcid.org/0000-0002-2219-9802 to King County Public Health—School-Based Mental Health Programs, School
Claudia Avina, PhD: https://orcid.org/0000-0001-6333-9683 Mental Health, Ontario. She has received honoraria for trainings on Behavioral
Elizabeth McCauley, PhD: https://orcid.org/0000-0001-9763-9262 Activation with Adolescents and for school-based mental health providers on a
Brief Intervention for School Clinicians (BRISC). She has received book royalties
The authors wish to thank the children, families, and colleagues who made this
from Guilford Press for Behavioral Activation with Adolescents: A Clinician’s
project possible. They thank David Brent, PhD, of the University of Pittsburgh,
Guide and Academic Media Solutions for a psychiatry textbook. She has
Donald Guthrie, PhD, retired, Seattle, Washington, and Cheryl King, PhD, of served on the speakers’ bureau of the University of Washington/Seattle Chil-
the University of Michigan, who participated on their Data Safety and Man-
dren’s Hospital. Dr. Bedics has reported no biomedical financial interests or
agement Board.
potential conflicts of interest.
Disclosure: Dr. Asarnow has reported receiving grant support from NIMH, the Correspondence to Joan Rosenbaum Asarnow, PhD, UCLA School of Medi-
Substance Abuse and Mental Health Services Administration (SAMHSA), the
cine, Psychiatry, UCLA Semel Institute, 760 Westwood Plaza, Los Angeles, CA
American Foundation for Suicide Prevention (AFSP), and the American Psy-
90024-1759; e-mail: jasarnow@mednet.ucla.edu
chological Foundation. She has reported receiving support from the Associa-
tion for Child and Adolescent Mental Health, the Society of Clinical Child and 0890-8567/$36.00/ª2021 American Academy of Child and Adolescent
Adolescent Psychology, and consulting on quality improvement interventions Psychiatry
for depression and suicidal/self-harm behavior. She has served on the Scientific
Council for AFSP and the Scientific Advisory Board for the Klingenstein Third https://doi.org/10.1016/j.jaac.2021.01.016

REFERENCES
1. Kochanek K, Murphy S, Xu J, Arias E. Deaths: final data for 2017. Natl Vital Stat Rep. 9. Mehlum L, Ramberg M, Tørmoen AJ, et al. Dialectical behavior therapy compared with
2019;68:1-77. enhanced usual care for adolescents with repeated suicidal and self-harming behavior:
2. Ribeiro J, Franklin J, Fox K, et al. Self-injurious thoughts and behaviors as risk factors for outcomes over a one-year follow-up. J Am Acad Child Adolesc Psychiatry. 2016;55:
future suicide ideation, attempts, and death: a meta-analysis of longitudinal studies. 295-300.
Psychol Med. 2016;46:225-236. 10. Mehlum L, Tørmoen A, Ramberg M, et al. Long term effectiveness of dialectical
3. Linehan MM, Comtois KA, Murray AM, et al. Two-year randomized controlled trial behavior therapy versus enhanced usual care for adolescents with self-harming and sui-
and follow-up of dialectical behavior therapy vs therapy by experts for suicidal behaviors cidal behavior. J Child Psychol Psychiatry. 2019;60:1112-1122.
and borderline personality disorder. Arch Gen Psychiatry. 2006;63:757-766. 11. Kazdin AE. Evidence-based treatment research: advances, limitations, and next steps. Am
4. Mehlum L, Tørmoen A, Ramberg M, et al. Dialectical behavior therapy for adolescents Psychol. 2011;66:685-698.
with repeated suicidal and self-harming behavior: a randomized trial. J Am Acad Child 12. Goodman M, Carpenter D, Tang CY, et al. Dialectical behavior therapy alters emotion
Adolesc Psychiatry. 2014;53:1082-1091. regulation and amygdala activity in patients with borderline personality disorder.
5. McCauley E, Berk MS, Asarnow JR, et al. Efficacy of dialectical behavior therapy for J Psychiatr Res. 2014;57:108-116.
adolescents at high risk for suicide a randomized clinical trial. JAMA Psychiatry. 2018;75: 13. Adrian M, McCauley E, Berk MS, et al. Predictors and moderators of recurring self-harm
777-785. in adolescents participating in a comparative treatment trial of psychological in-
6. Neacsiu AD, Bohus M, Linehan MM. Dialectical behavior therapy: an intervention terventions. J Child Psychol Psychiatry. 2019;60:1123-1132.
for emotion dysregulation. In: Handbook of Emotion Regulation. 2nd ed. New York, 14. Asarnow JR, Baraff LJ, Berk M, et al. Pediatric emergency department suicidal patients:
NY: Guilford; 2014:491-507. two-site evaluation of suicide ideators, single attempters, and repeat attempters. J Am
7. Chapman AL, Gratz KL, Brown MZ. Solving the puzzle of deliberate self-harm: the Acad Child Adolesc Psychiatry. 2008;47:958-966. https://doi.org/10.1097/CHI.
experiential avoidance model. Behav Res Ther. 2006;44:371-394. 0b013e3181799ee8.
8. Nock MK, Prinstein MJ. A functional approach to the assessment of self-mutilative 15. Reynolds W. Suicidal Ideation QuestionnaireJunior. Lutz, FL: Psychological Assess-
behavior. J Consult Clin Psychol. 2004;72:885-890. ment Resources; 1987.

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


Volume - / Number - / - 2021
DBT FOR SUICIDAL SELF-HARMING YOUTH

16. First M, Gibbon M, Spitzer R, Williams J, Benjamin L. Structured Clinical Interview for 34. VanderWeele TJ. Explanation in causal inference: developments in mediation and
DSM-IV Axis II Personality Disorders SCID-II. Washington, DC: American Psychi- interaction. Int J Epidemiol. 2016;45:1904-1908.
atric Publishing; 1997. 35. Hochberg Y, Benjamini Y. More powerful procedures for multiple significance testing.
17. Kaufman A, Kaufman N. Kaufman Brief Intelligence Test, Second Edition (KBIT-2). Stat Med. 1990;9:811-818.
Circle Pines, MN: American Guidance Service; 2004. 36. Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology. 1990;
18. Miller AL, Rathus JH, Linehan MM. Dialectical Behavior Therapy with Suicidal Ado- 1:43-46.
lescents. New York, NY: Guilford Press; 2007. 37. Cohen J. Statistical Power Analysis for the Behavioural Sciences. Hillside, NJ: Lawrence
19. Linehan MM. Cognitive-Behavioral Treatment of Borderline Personality Disorder. New Earlbaum; 1988.
York, NY: Guilford Press; 1993. 38. Wood AM, White IR, Hillsdon M, Carpenter J. Comparison of imputation and
20. Linehan MM, Comtois KA, Brown MZ, Heard HL, Wagner A. Suicide Attempt Self- modelling methods in the analysis of a physical activity trial with missing outcomes. Int J
Injury Interview (SASII): development, reliability, and validity of a scale to assess sui- Epidemiol. 2005;34:89-99.
cide attempts and intentional self-injury. Psychol Assess. 2006;18:303-312. 39. Lenhard W, Lenhard A. Calculation of Effect Sizes. Dettelbach, Germany: Psycho-
21. Weinberg A, Klonsky ED. Measurement of emotion dysregulation in adolescents. Psy- metrica; 2016. https://doi.org/10.13140/RG.2.1.3478.4245.
chol Assess. 2009;21:616-621. 40. Wolff JC, Thompson E, Thomas SA, et al. Emotion dysregulation and non-suicidal self-
22. Neacsiu AD, Rizvi SL, Vitaliano PP, Lynch TR, Linehan MM. The dialectical behavior injury: a systematic review and meta-analysis. Eur Psychiatry. 2019;59:25-36.
therapy ways of coping checklist: Dev Psychometr Properties. J Clin Psychol. 2010;66: 41. Johnston LD, Miech RA, O’Malley PM, Bachman JG, Schulenberg JE, Patrick ME.
563-582. Monitoring the Future: National Survey Results on Drug Use 1975-2018. Overview: key
23. Crick NR, Murray-Close D, Woods K. Borderline personality features in childhood: a findings on adolescent drug use. Institute of Social Research. 2019;1-104.
short-term longitudinal study. Dev Psychopathol. 2005;17:1051-1070. 42. Hawton K, Bergen H, Kapur N, et al. Repetition of self-harm and suicide following self-
24. Radloff LS. The CES-D scale: a self-report depression scale for research in the general harm in children and adolescents: findings from the Multicentre Study of Self-harm in
population. Appl Psychol Meas. 1977;1:385-401. England. J Child Psychol Psychiatry Allied Discip. 2012;53:1212-1219.
25. Kirisci L, Mezzich A, Tarter R. Norms and sensitivity of the adolescent version of the 43. King CA, Arango A, Kramer A, et al. Association of the Youth-Nominated Support Team
drug use screening inventory. Addict Behav. 1995;20:149-157. Intervention for suicidal adolescents with 11- to 14-year mortality outcomes: secondary
26. Achenbach TM, Rescorla LA. Manual for the ASEBA School-Age Forms & Profiles. analysis of a randomized clinical trial. JAMA Psychiatry. 2019;76:492-498.
Burlington, VT: University of Vermont, Research Center for Children, Youth, & 44. Morgan C, Webb RT, Carr MJ, et al. Incidence, clinical management, and mortality risk
Families; 2001. following self harm among children and adolescents: cohort study in primary care. BMJ.
27. Kaufman J, Birmaher B, Brent D, et al. Schedule for Affective Disorders and Schizo- 2017;359:j4351.
phrenia for School-Age ChildrenPresent and Lifetime Version (K-SADS-PL): initial 45. Iyengar U, Snowden N, Asarnow JR, Moran P, Tranah T, Ougrin D. A further look at
reliability and validity data. J Am Acad Child Adolesc Psychiatry. 1997;36:980-988. therapeutic interventions for suicide attempts and self-harm in adolescents: an updated
28. Derogatis L, Melisaratos N. The Brief Symptom Inventory: an introductory report. systematic review of randomized controlled trials. Front Psychiatry. 2018;9:583.
Psychol Med. 1983;13:595-605. 46. Kaess M, Edinger A, Fischer-Waldschmidt G, Parzer P, Brunner R, Resch F. Effec-
29. Bohus M, Kleindienst N, Limberger MF, et al. The short version of the Borderline tiveness of a brief psychotherapeutic intervention compared with treatment as usual for
Symptom List (BSL-23): development and initial data on psychometric properties. adolescent nonsuicidal self-injury: a single-centre, randomised controlled trial. Eur Child
Psychopathology. 2009;42:32-39. Adolesc Psychiatry. 2020;29:881-891.
30. Kenward MG, Roger JH. Small sample inference for fixed effects from restricted 47. Bjureberg J, Sahlin H, Hedman-Lagerl€of E, et al. Extending research on Emotion
maximum likelihood. Biometrics. 1997;53:983-997. Regulation Individual Therapy for Adolescents (ERITA) with nonsuicidal self-injury
31. Hedeker D, Gibbons RD. Longitudinal Data Analysis. Hoboken, NJ: John Wiley & disorder: open pilot trial and mediation analysis of a novel online version. BMC Psy-
Sons; 2006. chiatry. 2018;18:1-13.
32. Yuan Y. Multiple imputation using SAS Software. J Stat Softw. 2011;45:1-25. 48. Babeva KN, Klomhaus AM, Sugar CA, Fitzpatrick O, Asarnow JR. Adolescent suicide
33. Ten Have TR, Joffe MM. A review of causal estimation of effects in mediation analyses. attempt prevention: predictors of response to a cognitive-behavioral family and youth
Stat Methods Med Res. 2010;21:77-107. centered intervention. Suicide Life Threat Behav. 2020;50:56-71.

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


Volume - / Number - / - 2021
ASARNOW et al.

FIGURE S1 CONSORT Diagram

Telephone Screening 246 Excluded:

N=576 59 Not eligible

187 Declined further assessment

157 Excluded:
In-Person
Eligibility/Baseline 135 Not eligible

Assessment (n=330) 21 Withdrew

1 Protocol deviation

173 Randomized

DBT n=86 IGST n=87


66 Completed Treatment 48 Completed Treatment

Assessments Completed1 Assessments Completed


77 Post-treatment/6-months 66 Post-treatment/6-months

69 Follow-Up/12-months 64 Follow-Up/12-months

69 Both post-treatment and follow-up 64 Both post-treatment and follow-up

Note: Adapted from McCauley et al., 20185 (https://doi.org/10.1001/jamapsychiatry.2018.1109; Reproduced with permission from JAMA Psychiatry. 2021. 75(8): 777-785.
Copyrightª(2018) American Medical Association. All rights reserved). DBT ¼ Dialectical Behavior Therapy; IGST ¼ Individual and Group Supportive Therapy. 1Assessments
completed based on Suicide Attempt Self-Injury Interview49, available self-harm data for the assessment period.

11.e1 www.jaacap.org Journal of the American Academy of Child & Adolescent Psychiatry
Volume - / Number - / - 2021
DBT FOR SUICIDAL SELF-HARMING YOUTH

TABLE S1 Sample Description at Baseline

Variable DBT group (n ¼ 86) IGST group (n ¼ 87) Total (N ¼ 173)


Sex, female 82 (95.30%) 81 (94.19%) 163 (94.80%)
Age 14.77 (1.50) 15.04 (1.43) 14.89 (1.47)
Race/ethnicity
White 50 (58.14%) 47 (55.29%) 97 (56.39%)
Hispanic 23 (26.70%) 24 (28.24%) 48 (27.49%)
African American 7 (8.14%) 5 (5.88%) 12 (7.02%)
Asian American 4 (4.65%) 6 (7.06%) 10 (5.85%)
Native American 1 (1.16%) 0 (0.00%) 1 (0.58%)
Other 1 (1.16%) 3 (3.53%) 4 (2.34%)
Marital status of parents
Married 44 (57.14%) 38 (52.05%) 82 (54.67%)
Single, divorced, separated 31 (40.26%) 32 (43.84%) 63 (42.00%)
Widowed 1 (1.30%) 2 (2.74%) 3 (2.00%)
Other 1 (1.30%) 1 (1.37%) 1 (1.33%)
Education of parents
Less than high school 7 (8.86%) 5 (6.76%) 12 (7.84%)
High school graduate or GED 10 (12.66%) 9 (12.16%) 19 (12.42%)
Some college or tech. school 12 (16.46%) 15 (20.27%) 28 (18.30%)
College graduate 49 (62.03%) 45 (60.81%) 94 (61.44%)
Income
<$15,00 8 (11.94%) 7 (10.14%) 15 (11.03%)
$15,000-$29,999 4 (5.97%) 5 (7.25%) 9 (6.62%)
$30,000-49,999 8 (11.00%) 17 (23.90%) 25 (17.40%)
$50,000D 52 (71.20%) 42 (59.20%) 95 (65.30%)
Current disorder
Depressive disorder, K-SADS-PL 68 (79.10%) 77 (88.50%) 145 (83.81%)
Anxiety disorder, K-SADS-PL 42 (48.80%) 51 (59.30%) 93 (54.10%)
Borderline personality disorder, SCID 43 (50.00%) 49 (56.3%) 92 (53.20%)

Note: Data are given as frequency (%) or as mean (SD).


Adapted from McCauley et al.5 (https://doi.org/10.1001/jamapsychiatry.2018.1109. Reproduced with permission from JAMA Psychiatry. 2021;75:777-785. [Copyrightª 2018
American Medical Association. All rights reserved]). DBT ¼ Dialectical Behavior Therapy; IGST ¼ Individual and Group Supportive Therapy; KSADS-PL ¼ Schedule for
Affective Disorders and Schizophrenia for School-Age ChildrenPresent and Lifetime Version; SCID ¼ Structured Clinical Interview for the Diagnostic and Statistical
Manual of Mental Disorders.

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 11.e2
Volume - / Number - / - 2021
ASARNOW et al.

TABLE S2 Estimated Means and Standard Errors for Both Treatment Groups (Dialectical Behavior Therapy and Individual and
Group Supportive Therapy) at Each Assessment Point on Mechanistic and Secondary/Exploratory Clinical and Functioning
Outcomes

Baseline 3-Month 6-Month 9-Month 12-Month


Youth Mechanistic Outcomes
Measure Treatment Mean SE Mean SE Mean SE Mean SE Mean SE
DERS DBT 126.79 2.37 114.31 2.35 101.83 3.01 99.09 2.66 96.34 3.03
IGST 125.88 2.34 117.92 2.39 109.96 3.13 108.07 2.77 106.19 3.24
RWCCL DBT 1.36 0.06 1.51 0.05 1.66 0.06 1.70 0.05 1.74 0.06
IGST 1.26 0.06 1.36 0.05 1.47 0.07 1.53 0.06 1.58 0.07
Parent mechanistic outcomes
DERS DBT 68.73 2.15 67.14 1.86 65.55 1.91 65.62 1.85 65.69 2.14
IGST 71.70 2.15 71.07 1.88 70.45 1.97 68.54 1.90 66.63 2.24
RWCCL DBT 1.96 0.05 2.06 0.04 2.16 0.05 2.11 0.05 2.06 0.05
IGST 2.06 0.05 2.03 0.04 2.01 0.05 1.97 0.05 1.93 0.06
Youth clinical and functioning outcomes
BPFS-C DBT 78.69 1.40 74.42 1.37 70.16 1.68 68.21 1.51 66.26 1.72
IGST 76.97 1.41 74.26 1.38 71.54 1.74 69.56 1.56 67.58 1.84
CES-D DBT 36.00 1.24 31.18 1.14 26.36 1.49 25.19 1.27 24.02 1.56
IGST 36.82 1.25 32.09 1.16 27.37 1.57 26.62 1.35 25.87 1.71
DUSI DBT 22.71 2.77 18.01 2.94 19.51 2.98
IGST 21.35 2.78 25.33 3.09 22.15 3.21
YSR-I DBT 33.79 1.12 25.06 1.52 24.82 1.68
IGST 34.15 1.12 28.28 1.59 25.44 1.79
YSR-E DBT 23.21 1.11 18.75 1.29 17.37 1.23
IGST 20.28 1.11 20.65 1.35 17.84 1.32
YSR-TP DBT 94.01 2.87 72.63 3.84 69.73 4.13
IGST 91.99 2.88 81.19 4.02 72.27 4.42
SAS DBT 2.87 0.06 2.56 0.07 2.45 0.07
IGST 2.97 0.06 2.63 0.07 2.50 0.08
Parent clinical outcomes
BSL DBT 0.66 0.06 0.47 0.07 0.44 0.05
IGST 0.66 0.06 0.54 0.07 0.46 0.06
BSI Global DBT 0.62 0.06 0.48 0.06 0.45 0.05
IGST 0.59 0.06 0.57 0.06 0.48 0.05

Note: YSR raw scores reported. BPFS-C ¼ Borderline Personality Features Scale for Children; BSI Global ¼ Brief Symptom Inventory Global Severity
Index; BSL ¼ Borderline Symptom List; CES-D ¼ Center for Epidemiologic Studies Depression Scale; DBT ¼ Dialectical Behavior Therapy; DERS ¼
Difficulties in Emotion Regulation Scale; DUSI ¼ Drug Use Screening Inventory; IGST ¼ Individual and Group Supportive Therapy; RWCCL ¼ Revised
Ways of Coping Checklist; SAS ¼ Social Adjustment Scale; YSR-E ¼ YSR-externalizing scale; YSR-I ¼ Youth Self-Report-internalizing scale; YSR-TP ¼
YSR-total problems scale.

11.e3 www.jaacap.org Journal of the American Academy of Child & Adolescent Psychiatry
Volume - / Number - / - 2021
DBT FOR SUICIDAL SELF-HARMING YOUTH

TABLE S3 Within-Group Differences in Rates of Change Quantified through Cohen’s d Effects Sizes With 95% Confidence
Bounds Over Each Study Time Period for Variables With Treatment by Time Interactions Reaching the Level of p < .05

Measure Effect of interest During active treatment (06 mo) During follow-up (612 mo) Total change (012 mo)
Youth mechanistic outcomes
DERS DBT e1.02 (e1.24 to e0.81)*** e0.21 (e0.42 to 0.01)# e1.33 (e1.55 to e1.12)***
IGST e0.61 (e0.83 to e0.40)*** e0.13 (e0.34 to 0.09) e0.78 (e0.99 to e0.56)***
Parent mechanistic outcomes
RWCCL DBT 0.53 (0.31 to 0.74)*** e0.28 (e0.49 to e0.06)* 0.25 (0.03 to 0.46)*
IGST e0.10 (e0.31 to 0.11) e0.18 (e0.40 to 0.03)# e0.30 (e0.51 to e0.08)**
Youth clinical and functioning outcomes
DUSI DBT e0.18 (e0.40 to 0.03)# 0.06 (e0.16 to 0.27) e0.13 (e0.34 to 0.09)
IGST 0.15 (e0.07 to 0.36) e0.11 (e0.32 to 0.11) 0.03 (e0.19 to 0.24)
YSR-E DBT e0.41 (e0.63 to e0.20)*** e0.15 (e0.37 to 0.06) e0.57 (e0.79 to e0.35)***
IGST e0.03 (e0.25 to 0.18) e0.18 (e0.40 to 0.03)# e0.23 (e0.44 to e0.01)*
YSR-TP DBT e0.70 (e0.92 to e0.49)*** e0.12 (e0.33 to 0.10) e0.76 (e0.98 to e0.55)***
IGST e0.38 (e0.60 to e0.17)*** e0.21 (e0.43 to 0.00)* e0.56 (e0.78 to e0.35)***

Note: Data are Cohen d (95% CI) (thresholds for small, medium, and large effects are 0.20, 0.50, and 0.80, respectively.37
Information is provided only on variables for which significant intervention effect was found. DBT ¼ Dialectical Behavior Therapy; DERS ¼ Difficulties in
Emotion Regulation Scale; DUSI ¼ Drug Use Screening Inventory; IGST ¼ Individual and Group Supportive Therapy; RWCCL ¼ Revised Ways of
Coping Checklist; YSR-E ¼ Youth Self-Reportexternalizing scale; YSR-TP ¼ Youth Self Reporttotal problems scale.
*
p < .05, **p < .01, ***p < .001, #p < .10, based on tests of derived estimates.

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 11.e4
Volume - / Number - / - 2021

You might also like