Dialectical Behavior Therapy for Adolescents With Bipolar Disorder

Key Points

Question

Is dialectical behavior therapy (DBT) effective in treating mood symptoms, states, and suicidal behavior among adolescents with bipolar spectrum disorder?

Findings

In this randomized clinical trial of 100 adolescents with bipolar spectrum disorder, compared with those who received standard of care psychotherapy, youth who received 1 year of DBT had fewer suicide attempts over 1 year. The groups displayed similar improvement in depression and hypomania and mania.

Meaning

These findings support DBT as the first psychosocial intervention with demonstrated effects on suicidal behavior for adolescents with bipolar spectrum disorder.

This randomized clinical trial evaluates whether dialectical behavior therapy for adolescents with bipolar spectrum disorder is more effective than standard of care psychotherapy in decreasing suicide attempts over 1 year.

Abstract

Importance

Early-onset bipolar disorder conveys substantial risk for suicide. No psychosocial intervention for this population expressly targets suicidal behavior.

Objective

To determine whether dialectical behavior therapy (DBT) for adolescents with bipolar spectrum disorder is more effective than standard of care (SOC) psychotherapy in decreasing suicide attempts over 1 year.

Design, Settings, and Participants

Adolescents aged 12 to 18 years diagnosed with bipolar spectrum disorder were recruited from a specialty outpatient psychiatric clinic between November 2014 and September 2019. Independent evaluators conducted quarterly assessments over 1 year with participants and parents. Data were analyzed from March 2021 to November 2022.

Interventions

Participants were randomly assigned to 1 year of DBT (36 sessions; n = 47) or SOC psychotherapy (schedule clinically determined; n = 53). All youth received medication management via a flexible algorithm.

Main Outcomes and Measures

Primary outcomes included suicide attempts over 1 year and mood symptoms and states (depression and hypomania/mania). Secondary analyses included moderation of DBT effects by history of suicide attempt and mediation through emotion dysregulation.

Results

Of 100 included participants, 85 (85%) were female, and the mean (SD) age was 16.1 (1.6) years. Participants were followed up over a mean (SD) of 47 (14) weeks. Both treatment groups demonstrated significant and similar improvement in mood symptoms and episodes over 1 year (standardized depression rating scale slope, −0.17; 95% CI, −0.31 to −0.03; standardized mania rating scale slope, −0.24; 95% CI, −0.34 to −0.14). DBT and SOC participants reported similar suicide attempt rates at intake as measured on the Adolescent Longitudinal Follow-Up Evaluation (ALIFE; mean [SD] attempts, 2.0 [4.5] vs 1.8 [3.9], respectively; P = .80). DBT participants reported slightly more suicide attempts at intake as measured on the Columbia-Suicide Severity Rating Scale Pediatric Version (C-SSRS; mean [SD] attempts, 1.4 [3.6] vs 0.6 [0.9]; P = .02). DBT participants reported significantly fewer suicide attempts over follow-up compared with SOC participants via the ALIFE (mean [SD] attempts per follow-up period, 0.2 [0.4] vs 1.1 [4.3], controlling for baseline attempts: P = .03) and the C-SSRS (mean [SD] attempts per follow-up period, 0.04 [0.2] vs 0.10 [0.3], controlling for baseline attempts; P = .03). DBT was significantly more effective than SOC psychotherapy at decreasing suicide attempts over 1 year (ALIFE: incidence rate ratio [IRR], 0.32; 95% CI, 0.11-0.96; C-SSRS: IRR, 0.13; 95% CI, 0.02-0.78). Decreased rate of suicide attempts in DBT was moderated by presence of lifetime history of suicide attempt and time (IRR, 0.23; 95% CI, 0.13-0.44) and mediated by improvement in emotion dysregulation (IRR, 0.61; 95% CI, 0.42-0.89), particularly for those with high baseline emotion dysregulation (standardized β, −0.59; 95% CI, −0.92 to −0.26).

Conclusions and Relevance

In this randomized clinical trial, DBT demonstrated efficacy in decreasing suicide attempts among the high-risk population of adolescents with bipolar spectrum disorder.

Trial Registration

ClinicalTrials.gov Identifier: NCT02003690

Introduction

Up to 50% of youth with bipolar spectrum disorder (BP) attempt suicide,^1,2,3,4 and psychological autopsy studies indicate that of all psychiatric diagnoses, BP imparts the greatest risk of death by suicide in youth.⁵ Expert practice parameters for early-onset BP include a combination of pharmacotherapy and psychotherapy.⁶ Several psychosocial interventions have demonstrated efficacy for mood stabilization and decreased recurrence risk for youth with BP.^7,8,9 Yet to our knowledge, no treatment expressly targets suicidal behavior in this population, and there are few report on suicidal outcomes. In a secondary data analysis from a randomized clinical trial of child- and family-focused cognitive behavioral therapy vs treatment as usual among participants aged 7 to 13 years with BP, Weinstein et al¹⁰ demonstrated improvement in suicidal ideation across treatment groups. However, suicide attempts were not examined. Conversely, trials that demonstrate treatment efficacy for decreasing adolescent suicidal ideation and behavior in transdiagnostic populations generally exclude youth with BP,¹¹ rendering a dearth of evidence to guide suicide prevention in this ultrahigh-risk group.

Dialectical behavior therapy (DBT)¹² is an evidence-based psychosocial treatment developed for adults with borderline personality disorder with demonstrated efficacy in decreasing suicidal behavior.^13,14 Randomized clinical trials and meta-analyses examining DBT for suicidal and self-harming adolescents support the efficacy of DBT over comparator treatments in reducing suicidal ideation, self-harm, and suicide attempts, rendering DBT a well-established treatment for suicidal youth.^15,16,17,18 Yet youth with BP were largely excluded from these trials despite their elevated risk status.

We assert DBT is well matched for the treatment of adolescents with BP, given their heightened suicide risk, as well as other shared treatment targets, including emotional dysregulation (a core feature of early-onset BP)¹⁹ and psychosocial functioning.^20,21 We made minor adaptations to the manual for DBT for adolescents by Miller et al²² via 2 open treatment development studies with adolescents with BP and demonstrated feasibility, acceptability, and preliminary evidence of clinical impact.²³ In a subsequent pilot randomized clinical trial, compared with standard of care (SOC) psychotherapy, DBT resulted in significantly greater improvement in targeted treatment domains, including suicidal ideation and depressive symptoms, over 1 year.²³

Herein, we describe results from a randomized clinical trial of DBT compared with SOC psychotherapy in a specialty clinic for adolescents with BP. All participants received pharmacotherapy from study-affiliated psychiatrists. We hypothesized adolescents who received DBT would exhibit fewer suicide attempts and greater mood stability over 1 year compared with adolescents who received SOC psychotherapy (primary outcomes). We planned a priori to test for moderation of DBT effects by history of suicidality and mediation through emotional dysregulation.

Methods

Participants

Participants were recruited from a specialty psychiatric clinic at the University of Pittsburgh Medical Center between November 2014 and September 2019. Eligibility criteria included (1) age of 12 years, 0 months, to 18 years, 11 months; (2) DSM-5 diagnosis of BP (I, II, or not otherwise specified [NOS]); (3) no developmental disability or autism spectrum disorder; (4) not engaged in other psychotherapy; (5) engaged in or willing to proceed with pharmacotherapy with a study-affiliated psychiatrist; and (6) at least 1 parent, guardian, or close contact willing to participate in DBT family skills training. The Institutional Review Board at the University of Pittsburgh approved all study procedures. Participating adolescents and parents reviewed all procedures with study staff to ensure understanding. Parents provided written informed consent and adolescents provided assent prior to initiation of study procedures. The trial was registered with ClinicalTrials.gov, and the trial protocol can be found in Supplement 1. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.

Procedures

Diagnostic Evaluation

Interviewers assessed current and past psychiatric disorders at intake via separate interviews with the parent and adolescent; summary scores were based on consensus. Mood symptoms and disorders were assessed using the Kiddie Schedule for Affective Disorders and Schizophrenia for School-Aged Children (KSADS-P) depression scale (DEP)²⁴ and mania scale (MRS).²⁵ Research-operationalized criteria for BP NOS were applied.^26,27 Non–mood disorders were assessed using the KSADS Present and Lifetime Version (KSADS-PL)²⁸ and borderline personality disorder via the Structured Interview for DSM-IV Personality.²⁹ Interviewers demonstrated good interrater reliability for BP subtype and borderline personality disorder (Kendall Ws >0.80), assessed annually on 2 randomly selected evaluations.

Randomization

Participants were randomized using a modification of the Efron biased coin toss procedure, an approach especially useful for ensuring treatment balance in smaller samples by increasing the likelihood that new arrivals are allocated to the underrepresented group without introducing deterministic bias to the procedure. Random assignment stratified on sex assigned at birth, age (younger than 15 years vs 15 years or older), and history of suicide attempt via KSADS and/or Columbia-Suicide Severity Rating Scale Pediatric Version (C-SSRS).³⁰

Pharmacotherapy

Study psychiatrists (not naive to treatment condition) applied a flexible but standardized best-practice algorithm based on treatment guidelines from the American Academy of Child and Adolescent Psychiatry for management of pediatric BP³¹ with an updated menu of reasonable choices and drug dosage table.

DBT

Master’s-level therapists delivered DBT using the DBT for Adolescents with BP manual³² based on recommendations by Miller et al²² on DBT for suicidal adolescents. Adaptations for youth with BP are based on our prior work^23,33 and include 36 sessions (18 individual DBT sessions and 18 family skills training sessions conducted with the family unit) delivered over 1 year. DBT therapists attended a 2-day study training and read Dialectical Behavior Therapy With Suicidal Adolescents²² and the study treatment manual.³² Therapists achieved 6 of 8 consecutive training case sessions rated at or above adherence prior to treating participants.

SOC Psychotherapy

Therapists with experience treating youth with BP conducted SOC psychotherapy. All received intensive training and supervision in manualized, empirically supported approaches for early-onset mood disorders and BP, including psychoeducational, cognitive behavioral, family-focused, interpersonal, family systems, and supportive therapy; they had no DBT training. This comparator condition was selected to balance scientific rigor with external validity while maximizing potential to demonstrate outcomes relative to current usual SOC practices.

Treatment Fidelity

Session recordings (DBT and SOC psychotherapy) were rated for treatment fidelity using the DBT Adherence Coding Scale.³⁴ Independent raters naive to treatment assignment underwent training with the scale developer. A total of 15% of sessions were randomly selected quarterly for fidelity coding; 108 of 113 DBT sessions (95.6%) were rated at or above DBT fidelity (4.0 Global score or higher; mean [SD; range] score, 4.1 [0.1; 3.5-4.4]), whereas 68 of 72 SOC psychotherapy sessions (94.4%) were rated below (indicative of non-DBT; mean [SD; range] score, 3.5 [0.3; 2.5-4.2]). The scale developer co-coded a random sample of 10% of sessions annually to ensure ongoing interrater reliability.

Measures

Participants and parents completed assessments quarterly with an independent evaluator naive to treatment condition. Evaluators were trained to 80% reliability on all study measures and maintained adequate reliability (0.8 or greater) on a random sample of 3 recorded study assessments annually.

Demographic Characteristics

At intake, socioeconomic status was established using the Hollingshead criteria.³⁵ Participants self-reported sex assigned at birth and race and ethnicity via a questionnaire. Race categories included American Indian or Alaska Native, Asian, Black, Native Hawaiian or Other Pacific Islander, White, and other race (open response); ethnicity categories included Hispanic or non-Hispanic. Sexual orientation was assessed on the Adolescent Longitudinal Follow-Up Evaluation (ALIFE) Psychosocial Functioning Scale (PSF).³⁶

Clinical Variables

Age at BP onset was when the adolescent first met DSM-5 criteria for hypomania/mania or research-operationalized criteria for BPNOS on the KSADS-PL. History of psychiatric hospitalizations was documented on the KSADS-PL. Psychotropic medications prescribed at intake was documented on the ALIFE Psychotropic Treatment Scale.³⁷ Emotional dysregulation was assessed via the self-reported Difficulties With Emotion Regulation Scale (DERS).³⁸

Lifetime history of suicide attempt was positive if the participant endorsed an actual interrupted and/or aborted suicide attempt on the C-SSRS³⁰ at intake; any self-injurious behavior coded on the ALIFE in the 3 months preceding intake met the suicide attempt criteria; and/or the participant endorsed a lifetime attempt via the KSADS DEP suicidal acts items during the most severe past episode and/or the KSADS Summary Lifetime Diagnostic Checklist suicide attempt item.³⁹

Suicide Attempts

Suicide attempts over follow-up were assessed using the C-SSRS.⁴⁰ We administered the ALIFE Self-Injurious/Suicidal Behavior Scale³⁹ to gather temporal anchors for each behavior and detailed information on method, intent, and medical threat using KSADS DEP ratings for these items.²⁴ In keeping with our prior work,³⁹ suicide attempt on the ALIFE was defined as any self-injurious behavior that included definite intent and/or mild lethality.

Mood State and Symptom Severity

Mood state since prior follow-up was assessed using the ALIFE Psychiatric Status Rating (PSR) scales to yield week-by-week ratings (scale range, 1-6) corresponding to DSM-5 criteria.^41,42 Severity of depressive and hypomanic/manic symptoms during the worst week in the month preceding each assessment were rated using the KSADS DEP²⁴ and MRS.²⁵

SOC Psychotherapy Content

SOC psychotherapy therapists completed the Therapy Procedures Checklist⁴³ following each session to indicate the frequency of specific psychotherapy techniques delivered, yielding 4 approach subscales: cognitive, behavioral, family systems, and psychodynamic.

Statistical Analysis

We contrasted treatment groups on suicide attempts reported at baseline and over follow-up using scaled Poisson regression. We then modeled all outcomes with repeated measures using linear and generalized linear mixed models fitting random intercepts to account for within-participant clustering over repeated measures and random slopes where Wald tests indicated significant variation in fixed-effects slopes by participant. Predictors in each model included treatment, time, baseline values of the outcome modeled, and all second-order and third-order interactions. We standardized all continuous moderator variables and reported standardized effect sizes. Power calculations estimate that mixed models had 80% power to detect approximately medium-sized direct treatment effects (d = 0.39; odds ratio, 2.03) and second-order interactions (standardized slope difference, 0.63; odds ratio, 2.38), and large-sized third-order interactions (standardized slope difference, 0.75; odds ratio, 3.00).

Given that we used 2 measures of suicide attempt outcomes (C-SSRS and ALIFE) at each time point that use similar yet distinct definitions of suicide attempt (κ = 0.3), to enhance power, minimize the burden of multiple testing, and reduce bias from instrument measurement or calibration, we used a nested mixed Poisson regression model to analyze doubly repeated measures of suicide attempt counts. Effect sizes were estimated via incidence rate ratios (IRRs). Power calculations estimate that mixed Poisson regressions with doubly repeated measures had 80% power to detect approximately medium-sized direct treatment effects (IRR, 1.47) and second-order interactions (IRR, 1.71) and large-sized third-order interactions (IRR, 2.59).

We tested treatment effects on mean PSR mood scores by follow-up interval using mixed linear regression, percentage of weeks per follow-up period in subthreshold/threshold mood states using mixed gamma regression, and odds of subthreshold/threshold mood symptoms using mixed logistic regression. We tested the effects of treatment on longitudinal DEP and MRS scores using mixed linear regression.

We tested the effects of treatment on DERS total scores over follow-up using mixed linear regression. We next tested the effects of lagged measures of DERS total scores on doubly repeated measures of suicide attempt counts using mixed Poisson regression. A moderated mediation model tested whether DBT improvements in DERS mediated the effect of treatment on future suicide attempts while accounting for moderating effects of time and baseline DERS. We then estimated indirect effects differentially by baseline DERS (1 SD greater than mean, equal to the mean, and 1 SD less than mean) and time. Because the model predicts suicide attempts as a function of lagged DERS, the moderated mediation model predicts suicide attempts at time points 6 to 12 months using DERS scores from time points 3 to 9 months.

We assessed the impact of 2 outliers in the ALIFE suicide attempts data set by rerunning models after winsorizing those participants. To account for the multitude of tests performed, we further present P values adjusted for false discovery rate (FDR) grouped by hypothesis type (direct treatment effect, treatment-by-time, treatment-by-baseline, and 3-way interaction). P values were calculated using t, χ², and F tests, as appropriate. Significance was set at P < .05, and all P values were 2-tailed. We fit linear and generalized linear mixed models using SAS version 9.4 (SAS Institute) and moderated mediation models using Mplus version 5.21 (Muthén and Muthén).

Results

Of 100 included participants, 85 (85%) were female, and the mean (SD) age was 16.1 (1.6) years. One participant (1%) was Asian, 17 (17%) were Black, 6 (6%) were multiracial, 74 (74%) were White, and 2 (2%) were another race; 7 (7%) were Hispanic and 93 (93%) were non-Hispanic (Table 1). A total of 14 had BPI, 28 had BPII, and 58 had BPNOS. We screened 199 individuals to yield a sample of 100 (Figure 1). On average, participants were retained in the study for 47 weeks. Eight were lost to follow-up (4 undergoing DBT and 4 undergoing SOC psychotherapy) and 28 withdrew from the study (16 undergoing DBT and 12 undergoing SOC psychotherapy). There were no significant differences in demographic or clinical factors between retained vs withdrawn participants, except participants with BPII vs BPNOS were slightly more likely to withdraw (20 of 42 [47.6%] vs 16 of 58 [27.6%]; P = .03). Between-group comparisons were consistent when excluding those who withdrew, indicating minimal influence of withdrawals on analyses.

Table 1. Demographic and Baseline Clinical Characteristics of Adolescents With Bipolar Spectrum Disorder.

Characteristic	No. (%)
	DBT	SOC psychotherapy
Total, No.	47	53
Age, mean (SD), y	16.1 (1.8)	16.2 (1.5)
Racef
Asian	<10e	<10e
Black	11 (23.4)	<10e
Multiracial	<10e	<10e
White	31 (65.9)	43 (81.1)
Undisclosed	<10e	<10e
Sex assigned at birth
Female	40 (85.1)	45 (84.9)
Male	7 (14.9)	8 (15.1)
Sexual minority, No./total No. (%)a	10/37 (27)	21/45 (47)
SES, mean (SD)b	3.6 (1.1)	3.9 (1.1)
Bipolar typec
BPI	<10e	8 (15.1)
BPII	15 (31.9)	13 (24.5)
BPNOS	26 (55.3)	32 (60.3)
Age at BP onset, mean (SD), y	13.3 (3.2)	13.7 (2.6)
Comorbid conditionsc
ADHD	14 (29.8)	14 (26.4)
Any anxiety disorder	34 (72.3)	41 (77.4)
GAD	23 (48.9)	24 (45.3)
Social phobia	<10e	12 (22.6)
Panic disorder	<10e	<10e
PTSD	<10e	<10e
Substance abuse	<10e	<10e
ODD	<10e	<10e
Borderline personality disorderd	<10e	<10e
Psychotropic medications prescribed, mean (SD)	2 (1.5)	2.2 (1.3)
K-SADS DEP score, mean (SD)	25.4 (11.7)	24.9 (12.0)
K-SADS MRS score, mean (SD)	19.1 (12.2)	21.4 (13.0)
History of psychiatric hospitalization	26 (55.3)	20 (37.7)
History of suicide attempt	29 (61.7)	33 (62.3)

Abbreviations: ADHD, attention-deficit/hyperactivity disorder; BPI, bipolar spectrum disorder I; BPII, bipolar spectrum disorder II; BPNOS, bipolar spectrum disorder not otherwise specified; DBT, dialectical behavior therapy; DEP, depression scale; GAD, generalized anxiety disorder; K-SADS, Kiddie Schedule for Affective Disorders and Schizophrenia for School-Aged Children; MRS, mania scale; PTSD, posttraumatic stress disorder; ODD, oppositional defiant disorder; SES, socioeconomic status; SOC, standard of care.

^a Adolescent Longitudinal Follow-Up Evaluation Psychosocial Functioning Scale.

^b Hollingshead-Redlich criteria.³⁵

^c Diagnosed via K-SADS Present and Lifetime version semistructured interview.

^d Structured Interview for DSM-IV Personality Disorders.

^e Values less than 10 were suppressed because of identifiability.

^f Race data were collected by self-report via a questionnaire.

Figure 1. Study CONSORT Diagram.

DBT indicates dialectical behavior therapy; SOC, standard of care.

Treatment Attendance and Content

Adolescents undergoing DBT attended more sessions than those undergoing SOC psychotherapy (F_1,94 = 13.81; P < .001). Over 1 year, DBT participants attended a mean (SD; range) of 23.8 (16.9; 1-58) sessions (14.3 individual DBT sessions, 9.8 family skills training sessions), whereas SOC psychotherapy participants attended a mean (SD; range) of 13.1 (11; 1-47) sessions. SOC psychotherapy therapists most commonly used cognitive strategies (mean [SD], 1.8 [0.7]), followed by family (1.69 [0.74]), psychodynamic (1.54 [0.25]), and behavioral (1.13 [0.28]) approaches.

Primary Outcomes

Suicide Attempts

On average, DBT and SOC psychotherapy participants reported similar suicide attempt rates at intake as measured on the ALIFE (mean [SD] attempts, 2.0 [4.5] vs 1.8 [3.9], respectively; scaled Poisson χ²₁ = 0.06; P = .80); however, DBT participants reported slightly more suicide attempts at intake as measured on the C-SSRS (mean [SD] attempts, 1.4 [3.6] vs 0.6 [0.9]; scaled Poisson χ²₁ = 5.52; P = .02). Over follow-up, DBT participants reported significantly fewer suicide attempts over follow-up compared with SOC psychotherapy participants via both the ALIFE (mean [SD] attempts per follow-up period, 0.2 [0.4] vs 1.1 [4.3], controlling for baseline attempts: scaled Poisson χ²₁ = 4.93; P = .03) and the C-SSRS (mean [SD] attempts per follow-up period, 0.04 [0.2] vs 0.10 [0.3], controlling for baseline attempts: scaled Poisson χ²₁ = 4.97; P = .03), indicating a significant treatment effect on the overall rate of suicide attempts over follow-up.

When modeling doubly repeated measures (ALIFE and C-SSRS at each visit) of suicide attempts over follow-up to enable estimation of group-by-time interactions, the 2-way interaction between treatment group and time was significant (IRR, 0.67; 95% CI, 0.49-0.92; P = .01), and the 3-way interaction between group, time, and number of intake suicide attempts (past 3 months) was significant (IRR, 0.23; 95% CI, 0.13-0.44; P < .001), indicating that suicide attempts declined to a greater extent over time in DBT vs SOC psychotherapy, particularly among those with more suicide attempts at intake (Figure 2). Both interactions remain significant when parameterizing intake suicide attempts dichotomously or using lifetime history of suicide attempts (Table 2). The 3-way interaction remained highly significant after winsorizing 2 outlier participants on the ALIFE, but the group-by-time interaction became nonsignificant (IRR, 0.84; 95% CI, 0.58-1.11; P = .18). After adjusting for FDR, only the 3-way interaction remains significant (Table 2). A sensitivity analysis of the ALIFE data in which we excluded participants with intent less than 3 points and lethality of 3 or more points (ie, those that could be argued to include nonsuicidal self-injury) yielded slightly more significant results.

Figure 2. Interaction Between Group (Dialectical Behavior Therapy [DBT] vs Standard of Care [SOC] Psychotherapy), Time (in Months), and Lifetime History of Suicide Attempt.

The shaded area indicates 95% CIs.

Table 2. Treatment Effect Models.

Poisson regressions	Treatment effects	Measure (95% CI)	P value	FDR-adjusted P value
ALIFE suicide attempts (full follow-up rate)	Direct treatment effect	IRR, 0.32 (0.11 to 0.96)	.03	.18
C-SSRS suicide attempts (full follow-up rate)	Direct treatment effect	IRR, 0.13 (0.02 to 0.78)	.03	.18
Mixed Poisson regression
Suicide attempts (doubly repeated measures)	Direct treatment effect	IRR, 0.97 (0.42 to 2.25)	.94	.94
	Treatment × baseline	IRR, 0.61 (0.17 to 2.18)	.44	.63
	Treatment × time	IRR, 0.67 (0.49 to 0.92)	.01	.10
	Treatment × time × baseline	IRR, 0.23 (0.13 to 0.44)	<.001	<.001
Mixed linear regressions
K-SADS DEP score	Direct treatment effect	Slope coefficient, −0.06 (−0.43 to 0.31)	.75	.93
	Treatment × baseline	Slope coefficient, −0.05 (−0.41 to 0.31)	.79	.79
	Treatment × time	Slope coefficient, 0.06 (−0.22 to 0.35)	.66	.90
	Treatment × time × baseline	Slope coefficient, 0.22 (−0.07 to 0.52)	.13	.33
K-SADS MRS score	Direct treatment effect	Slope coefficient, −0.03 (−0.38 to 0.31)	.85	.93
	Treatment × baseline	Slope coefficient, −0.16 (−0.50 to 0.18)	.35	.63
	Treatment × time	Slope coefficient, 0.02 (−0.19 to 0.22)	.86	.90
	Treatment × time × baseline	Slope coefficient, 0.05 (−0.16 to 0.25)	.67	.72
Mean ALIFE depression score	Direct treatment effect	Slope coefficient, 0.19 (−0.20 to 0.59)	.33	.70
	Treatment × baseline	Slope coefficient, 0.14 (−0.24 to 0.52)	.52	.65
	Treatment × time	Slope coefficient, −0.15 (−0.54 to 0.23)	.43	.90
	Treatment × time × baseline	Slope coefficient, 0.21 (−0.18 to 0.60)	.28	.47
Mean ALIFE hypomania/mania score	Direct treatment effect	Slope coefficient, 0.04 (−0.23 to 0.31)	.78	.93
	Treatment × baseline	Slope coefficient, −0.15 (−0.42 to 0.13)	.29	.63
	Treatment × time	Slope coefficient, −0.02 (−0.19 to 0.16)	.84	.90
	Treatment × time × baseline	Slope coefficient, 0.16 (−0.01 to 0.34)	.10	.33
DERS score	Direct treatment effect	Slope coefficient, −0.16 (−0.49 to 0.16)	.31	.70
	Treatment × baseline	Slope coefficient, −0.50 (−0.82 to −0.18)	.003	.03
	Treatment × time	Slope coefficient, −0.21 (−0.39 to −0.03)	.02	.10
	Treatment × time × baseline	Slope coefficient, −0.16 (−0.35 to 0.03)	.09	.33
Mixed gamma regressions
Percentage of time with subthreshold or worse depressive symptoms	Direct treatment effect	Slope coefficient, 0.18 (−0.08 to 0.44)	.18	.70
	Treatment × baseline	Slope coefficient, 0.05 (−0.23 to 0.33)	.71	.79
	Treatment × time	Slope coefficient, −0.12 (−0.39 to 0.15)	.39	.90
	Treatment × time × baseline	Slope coefficient, 0.18 (−0.11 to 0.47)	.22	.44
Percentage of time with subthreshold or worse hypomanic/manic symptoms	Direct treatment effect	Slope coefficient, 0.05 (−0.23 to 0.32)	.74	.93
	Treatment × baseline	Slope coefficient, 0.25 (−0.05 to 0.55)	.11	.55
	Treatment × time	Slope coefficient, 0.07 (−0.22 to 0.36)	.63	.90
	Treatment × time × baseline	Slope coefficient, −0.16 (−0.48 to 0.16)	.33	.47
Mixed logistic regressions
Odds of subthreshold or worse depressive symptoms	Direct treatment effect	OR, 1.65 (0.52 to 5.17)	.39	.70
	Treatment × baseline	OR, 0.49 (0.10 to 2.49)	.39	.63
	Treatment × time	OR, 0.68 (0.23 to 2.02)	.49	.90
	Treatment × time × baseline	OR, 1.73 (0.39 to 7.63)	.47	.59
Odds of subthreshold or worse hypomanic/manic symptoms	Direct treatment effect	OR, 1.54 (0.55 to 4.32)	.41	.70
	Treatment × baseline	OR, 0.45 (0.13 to 1.57)	.21	.63
	Treatment × time	OR, 0.95 (0.40 to 2.24)	.90	.90
	Treatment × time × baseline	OR, 1.21 (0.41 to 3.55)	.72	.72

Abbreviations: ALIFE, Adolescent Longitudinal Follow-Up Evaluation; C-SSRS, Columbia-Suicide Severity Rating Scale; DEP, depression scale; DERS, Difficulties With Emotion Regulation Scale; FDR, false discovery rate; IRR, incidence rate ratio; K-SADS, Kiddie Schedule for Affective Disorders and Schizophrenia for School-Aged Children; MRS, mania scale; OR, odds ratio.

Mood State and Symptoms

Both treatment groups demonstrated significant improvement in mood states and symptoms over time. There were no significant direct or moderated effects of treatment on mean DEP or MRS scores, mean ALIFE depression, or hypomania/mania scores, percentage time in subthreshold/threshold mood states (Figure 3), or odds of subthreshold/threshold mood symptoms (all direct, second-order, and third-order interaction effect) (Table 2).

Figure 3. Percentage of Weeks in Threshold Mood States Over 1 Year for Adolescents Randomized to Receive Dialectical Behavior Therapy (DBT) vs Standard of Care (SOC) Psychotherapy.

Percentage of weeks meeting criteria for DSM-IV depressive episode (A) or hypomanic/manic episode (B) via Adolescent Longitudinal Follow-Up Evaluation Psychiatric Status Rating (PSR). The shaded area indicates 95% CIs.

Emotional Dysregulation

DBT participants demonstrated greater reduction in DERS over time than SOC psychotherapy participants; among those with higher baseline DERS, those who received DBT were less likely to have higher DERS over follow-up (Table 2). The third-order interaction effect was nonsignificant. After adjusting for FDR, only the group by baseline interaction remained significant. Further, there was a strong association between lagged DERS scores and current suicide attempts (standardized slope coefficient, 0.54; 95% CI, 0.30-0.78; P < .001).

Moderated Mediation

Given the significant treatment effect on DERS scores and suicide attempts over follow-up (moderated by time and baseline severity) and significant effects of lagged DERS scores on current suicide attempts, we combined the suicide attempts and DERS models above into a moderated mediation model to test the indirect temporal effect of treatment on rates of suicide attempts through DERS improvement. The indirect effect of DBT on suicide attempts as mediated by DERS became larger in magnitude over time (eFigure in Supplement 2; negative slopes indicate more reduction in suicide attempt rates in DBT vs SOC psychotherapy over time), and the overall indirect effect was significant among participants with baseline DERS scores at or above the mean but nonsignificant among participants with baseline DERS 1 SD below the mean (demonstrated by upper limits of 95% CIs overlapping zero).

Discussion

Given the high risk of suicide in those with early-onset BP and the lack of evidence-based treatments targeting suicidal behavior in this population, we examined whether DBT, an evidence-based treatment for suicidal behavior, decreases rates of suicide attempt among adolescents with BP. Youth received either DBT or SOC psychotherapy (delivered by clinicians highly trained in evidence-based treatments for this population but not DBT); both groups received pharmacotherapy. Findings indicate youth who received DBT had fewer suicide attempts over 1 year. Further, suicide attempts declined to a greater extent over time among those receiving DBT compared with those receiving SOC psychotherapy, particularly among participants with a recent and/or lifetime history of suicide attempt. Decreased risk of suicide attempt in DBT was mediated by improvement in emotion dysregulation, particularly for those with high baseline emotion dysregulation. Depression and hypomania/mania similarly improved in both groups.

Youth who received DBT had fewer suicide attempts over 1-year follow-up and greater decline in suicide attempts over 1 year compared with youth who received SOC psychotherapy. To our knowledge, this is the only study to date to demonstrate a treatment effect on suicide attempts among adolescents with BP. Of note, SOC psychotherapy—consisting of evidence-based treatment components delivered by highly trained therapists—had minimal impact on suicide risk. DBT may thus be considered within a clinical staging approach to treatment for early-onset BP⁴⁴ for youth at increased risk of suicidal behavior.

Data provide particularly strong support for DBT among adolescents with BP with a history of suicide attempt. Yet for up to 60% of individuals, their first suicide attempt is lethal,⁴⁵ and for those who survive, risk of death increases with each successive attempt. Thus, the capacity to predict first attempt among youth with BP and provide potentially preventive intervention like DBT offers compelling future direction. Our group developed a risk calculator that reliably predicts 1-year risk of suicide attempt in those with early-onset BP.⁴⁶ Application in clinical practice to identify youth at risk for onset of suicidal behavior affords potential for personalized and life-saving medicine.

DBT was associated with greater improvement in emotion dysregulation than SOC psychotherapy. Emotion dysregulation characterizes many psychiatric presentations in adolescence, including borderline personality disorder, for which DBT has demonstrated efficacy. Given the challenge of differential diagnosis of BP and borderline personality disorder, we assessed both via reliable methods. Only 8% of participants met criteria for co-occurring borderline personality disorder at intake, rendering findings unlikely attributable to misdiagnosis and/or co-occurring borderline personality disorder.

The Linehan biosocial theory¹² asserts difficulties with emotion regulation underlie problem behaviors, including suicidal behavior. Prior DBT studies across diagnoses and age groups support improved emotion regulation as a mechanism underlying enhanced outcomes (including depression, anxiety, and substance use) in DBT.⁴⁷ Asarnow et al⁴⁸ demonstrated that improvement in emotion regulation mediated self-harm remission with DBT among suicidal, self-harming adolescents. Our findings further contribute empirical support for improvement in emotion regulation as a mechanism through which DBT exerts its effects on decreased suicidal behavior.

DBT and SOC psychotherapy were similarly associated with improvement in depression and hypomania/mania over 1 year. These findings are in keeping with meta-analyses across diagnostic populations showing DBT has comparable efficacy with comparator psychotherapies in depressive improvement.⁴⁹ Indeed, SOC psychotherapy clinicians had training and supervision in other evidence-based treatments for adolescent BP that share common core components with DBT that are efficacious for mood management (eg, psychoeducation, problem solving).⁵⁰ Although youth randomized to SOC psychotherapy received fewer psychotherapy sessions on average than those randomized to DBT, the SOC psychotherapy treatment dose remains substantially greater than that received by most youth with BP in the community⁵¹ and also greater than that received in the community by this sample in the year preceding study entry.⁵² As such, although less intensive than DBT, SOC psychotherapy in this study was more rigorous and intensive than what youth with BP generally receive in the community. Thus, findings support DBT as an effective treatment for mood management in adolescents with BP, on par with other well-established psychosocial treatment approaches for this population.

Limitations

It is important to consider findings within the context of study limitations. The sample predominately includes individuals assigned female sex at birth, non-Hispanic White, and from middle-class backgrounds—common among treatment-seeking samples⁵³ but not representative of adolescents with BP.⁵⁴ Whether findings extend to more diverse samples remains unknown, and disparities in BP diagnosis and treatment represents a critical focus for future work.⁵⁵ Additionally, SOC psychotherapy was not manualized; fidelity checks confirmed SOC psychotherapy sessions were below acceptable DBT competence but did not evaluate SOC psychotherapy quality. The controlled nature of the study optimized scientific rigor but fails to reflect some realities of community practice. Forthcoming data from a community implementation field trial will address issues of generalizability.

Conclusions

Adolescents with BP at elevated risk of suicidal behavior and those exhibiting high emotion dysregulation stand to benefit most from DBT’s effects on decreasing suicide attempt risk through enhancing emotion regulation. Coupled with improvement in mood episodes and symptoms in DBT on par with other evidence-based strategies for this population, DBT offers promise in a stepped care model of treatment for early-onset BP.

Supplement 1.

Trial Protocol and Statistical Analysis Plan

Supplement 2.

eFigure. Indirect Effect of Group (Dialectical Behavior Therapy [DBT] vs Standard of Care Psychotherapy [SOC]) on Suicide Attempts as Mediated by Emotional Dysregulation (Difficulties With Emotion Regulation Scale [DERS]) and Moderated by Time

Supplement 3.

Data Sharing Statement