Group dialectical behavior therapy skills training versus group cognitive behavioral therapy for adults with ADHD: a randomized controlled trial

Abstract

Background

Attention-Deficit/Hyperactivity Disorder (ADHD) is a neurodevelopmental disorder characterized by inattention, hyperactivity, impulsivity, emotion dysregulation, and executive dysfunction. Current treatment guidelines recommend a multimodal approach. Cognitive Behavioral Therapy (CBT) has proven effective for alleviating core symptoms, improving executive functioning, and reducing emotional comorbidities, with benefits sustained for up to one year. Dialectical Behavior Therapy (DBT), an adaptation of CBT that integrates mindfulness and emotion regulation components, has also shown potential benefits in adult ADHD populations. However, no prior studies have directly compared CBT and DBT in Chinese adults with ADHD. This trial aimed to compare the efficacy of group DBT and group CBT in adults with ADHD across multiple dimensions, with follow-up assessments extending to six months post-treatment.

Methods

Ninety-eight adults with ADHD were randomly assigned to either the DBT (n = 49) or CBT (n = 49) group, and received 12 weeks of group-based intervention. Assessments were conducted at baseline (T0), weeks 4 (T1) and 8 (T2), post-treatment (T3), and at 3-month (T4) and 6-month (T5) follow-ups. Outcome measures included core symptoms (ADHD-RS), emotional symptoms (SAS and SDS), emotion regulation (DERS and ERQ), quality of life (WHOQOL-BREF), global functioning (GSES and SDS), and executive function assessed via both self-report (BRIEF-A) and laboratory-based tasks (TMT, SCWT, SST, and CPT-IP). Linear mixed models (LMM) were employed to examine group-by-time interaction effects.

Results

Both the DBT and CBT groups showed improvement over time in core ADHD symptoms and several secondary outcomes at post-treatment and follow-up. No significant group-by-time interactions were observed for core ADHD symptoms and the between-group effect sizes were small at post-treatment (d = 0.06, 95% CI [-0.33, 0.46]) and at 6-month follow-up (d = 0.17, 95% CI [-0.27, 0.61]). However, the confidence intervals were not fully contained within the prespecified non-inferiority margin (d = 0.40), and formal non-inferiority of DBT relative to CBT was therefore not established. Across secondary outcomes, most between-group differences were generally small, time-specific, and not consistently maintained over follow-up. For laboratory-based executive function measures, CBT showed a relative advantage on a spatial working memory task, whereas no stable between-group differences were observed for most other measures.

Conclusion

This is the first randomized controlled trial in China to directly compare DBT and CBT for adults with ADHD. The findings indicate that both interventions were associated with improvement across multiple symptom and functional domains, with generally small between-group differences over 6 months of follow-up. DBT may represent a potentially useful psychotherapeutic option for adults with ADHD. However, formal non-inferiority relative to CBT was not established, and the present findings do not support a conclusion of equivalence between the two treatments. Further studies with larger and more diverse samples are needed to obtain more precise estimates of treatment effects and to evaluate the generalizability of the present findings.

Trial Registration

ChiCTR2300072075, registered 2023.6.1.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-026-08130-w.

Background

Attention-deficit/hyperactivity disorder (ADHD) is a prevalent neurodevelopmental disorder characterized by persistent patterns of inattention, hyperactivity, impulsivity, and emotion dysregulation [1–3]. The estimated prevalence of ADHD in early adulthood is approximately 2.5% [4].Recent longitudinal studies suggest that childhood-onset ADHD rarely follows a simple linear path to remission. Although some individuals achieve stable full remission, more than half continue to meet diagnostic criteria at follow-up, and many others exhibit a fluctuating course characterized by alternating periods of remission and recurrence across development [5, 6]. Psychiatric comorbidity is highly prevalent in adult ADHD, with approximately 70–80% of affected individuals experiencing at least one co-occurring mental disorder [4]. Substance use, mood, anxiety, and personality disorders are among the most commonly reported comorbidities, and their prevalence is consistently higher than that observed in non-ADHD populations [7]. Beyond its core symptoms, adult ADHD is associated with broad and persistent impairments across educational, occupational, interpersonal, and health-related domains, contributing to poorer life success and greater psychosocial burden [8, 9]. Moreover, these impairments are linked to an elevated risk of adverse long-term outcomes, including premature mortality [10, 11], as well as suicidal spectrum behaviors [12].

Given its chronic course, high comorbidity, and broad functional burden, effective long-term management is particularly important. Although ADHD is highly heritable, with heritability estimates reaching up to 80% [4], psychosocial factors (e.g., early life adversity) also play a substantial role in its onset and progression [13, 14]. Therefore, for adults with ADHD, multimodal interventions targeting both core symptoms and associated domains, such as emotion regulation, personality development, neurocognitive deficits, depression, anxiety, and sleep disturbances, are essential to improve overall treatment outcomes [14, 15]. Pharmacological interventions, including psychostimulants and atomoxetine, primarily act on dopaminergic and noradrenergic pathways and have been shown to effectively alleviate the core symptoms of ADHD [16]. However, these medications demonstrate limited efficacy in addressing emotion dysregulation [17] and have little impact on broader outcomes such as QoL [15]. Moreover, their long-term effectiveness remains uncertain, and adverse effects contribute to higher discontinuation rates compared with placebo treatments [18]. Psychotherapy, as a fundamental component of ADHD management, has been shown to enhance the effectiveness of pharmacological interventions [15].

Cognitive behavioral therapy (CBT) is one of the primary non-pharmacological interventions and among the most extensively researched treatment modalities for ADHD [15]. CBT for adult ADHD was originally developed to address the residual symptoms and functional impairments that frequently persist despite pharmacotherapy [19, 20], and has since evolved into a structured, skills-based intervention targeting executive dysfunction, with a particular focus on organization, planning, time management, distractibility, and procrastination, while also addressing maladaptive self-beliefs and demoralization associated with repeated failure experiences [21, 22]. A growing body of research has demonstrated that CBT effectively improves core ADHD symptoms, executive functioning, and comorbid emotional difficulties, as well as overall QoL [23–26]. Moreover, these therapeutic benefits have been shown to persist for up to one year after treatment completion [27–30]. Our previous mechanistic studies further indicated that, in the short term, CBT alleviates core ADHD symptoms primarily through improvements in emotion regulation [26]. Over the longer term, changes in cognitive patterns mediate emotional improvements, which subsequently influence core symptoms. Additionally, enhancements in QoL occur indirectly through reductions in core and emotional symptoms and through the restructuring of maladaptive cognitions.

Parallel to the development of CBT for adult ADHD, Dialectical Behavior Therapy (DBT) emerged as another early, targeted intervention [31]. Originally developed by Marsha M. Linehan as a third-wave extension of CBT, DBT specifically targets emotion dysregulation through the integration of mindfulness, emotion regulation, distress tolerance, and interpersonal effectiveness skills [32, 33]. This framework is clinically relevant to adult ADHD, a condition frequently characterized by marked emotion dysregulation, impulsivity, and broader self-regulatory difficulties alongside executive dysfunction [31]. Supporting this therapeutic rationale, A recent meta-analysis identified third-wave components, as well as organizational strategies and problem-solving skills, as key CBT techniques for alleviating core ADHD symptoms [34]. Given its established use in borderline personality disorder and related conditions characterized by severe emotion dysregulation [32, 35–38], DBT was adapted by Hesslinger et al. into a structured intervention for adults with ADHD [31]. The primary mechanism of change in this model involves enhancing emotion regulation and executive functioning through mindfulness practice and behavioral analysis [15]. Previous studies have shown that DBT can significantly improve core ADHD symptoms [32, 39–41], emotional symptoms, executive function (EF), and QoL [40]. However, findings across studies have been inconsistent, with some reporting that DBT did not show significant advantages over control conditions in improving emotional symptoms [42, 43] or EF [44]. In summary, although DBT has been adapted for adults with ADHD since the early development of psychotherapy research in this field, its empirical base remains less extensive and less consistently replicated than that of CBT, particularly with regard to long-term outcomes [43].

To date, no randomized controlled trial (RCT) has directly compared DBT and CBT in adults with ADHD. Lopez-Pinar et al. conducted a meta-analysis of psychosocial interventions for adult ADHD and reported that the long-term efficacy of DBT might be lower than that of CBT, although this conclusion was largely influenced by a few large-sample studies [28]. In other clinical populations, Assmann et al. found that DBT and schema therapy demonstrated comparable efficacy in alleviating borderline personality disorder (BPD) symptoms, except for anger, where DBT showed superior outcomes at follow-up [45]. Similarly, Lammers et al. compared CBT and DBT in individuals with binge eating disorder and found that while CBT outperformed DBT in reducing overall eating disorder psychopathology at post-treatment and depressive symptoms at follow-up, both interventions significantly improved primary and secondary outcomes, demonstrating comparable overall efficacy [46].

Emerging evidence suggests that adult ADHD symptoms may be expressed and recognized differently across sociocultural contexts. In a large cross-cultural study spanning 42 countries, East Asian populations, including China, reported lower ADHD screening scores than Western populations [47]. Moreover, a Chinese psychometric study developed a brief adult ADHD self-report scale (ASRS-CSV) and identified a culturally specific pattern of core symptoms that differed from those highlighted in American and Japanese short forms [48]. These findings highlight the importance of examining adult ADHD and its treatment within specific cultural contexts rather than relying solely on evidence derived from Western populations.

Consistent with this perspective, our previous studies in Chinese samples with ADHD suggest that 46.37% of those diagnosed in childhood continued to meet full diagnostic criteria in early adulthood, and the persistence rate increased to 72.69% when residual symptoms and functional impairment were also considered; childhood hyperactivity/impulsivity and ADHD combined type predicted persistence, whereas higher IQ was protective [49]. Compared with healthy controls, Chinese adults with ADHD have also been found to show greater emotion dysregulation, characterized by poorer emotional control, less use of cognitive reappraisal, and greater reliance on expressive suppression [50]. Moreover, ADHD-related impairment in this population extends beyond core symptoms, and ADHD may influence QoL both directly and indirectly through executive dysfunction, maladaptive cognitions, and emotional symptoms [51, 52]. Against this background, Our previous studies have demonstrated that CBT is well-accepted among Chinese adults with ADHD and has significant efficacy in improving core symptoms, emotional symptoms, self-esteem, and QoL [24, 26, 53]. We also conducted a preliminary investigation showing that DBT effectively improved core symptoms, depressive symptoms, and emotion regulation. Among the treatment components, the emotion regulation and interpersonal effectiveness modules were perceived as the most beneficial [54]. However, no studies in China have examined the differential efficacy of CBT and DBT in adults with ADHD. Given the distinct therapeutic emphases of these two approaches, exploring their respective effects across various symptom dimensions could provide valuable insights for developing more individualized and targeted intervention strategies for adults with ADHD [15]. Therefore, this study aimed to examine whether group DBT is non-inferior to group CBT in improving core ADHD symptoms, and to explore potential differences between the two interventions regarding their effects on emotional symptoms, emotion regulation, EF, global functioning and QoL.

Method

Participants

Participants diagnosed with ADHD were recruited from the outpatient department of Peking University Sixth Hospital and online platforms between July 2023 and July 2024. Eligible participants were required to have remained either medication-naïve or on stable medication regimen for at least 1 month before treatment initiation. Permitted pharmacological treatments included extended-release methylphenidate or atomoxetine hydrochloride capsules [27].

Inclusion criteria were as follows: (1) Age ≥ 18 years, with no gender restrictions; (2) Diagnosis of adult ADHD according to DSM-5 criteria, confirmed by an attending psychiatrist or higher-level clinician, with a Clinical Global Impression-Severity (CGI-S) score ≥ 3 (indicating at least mild impairment); (3) Either not currently taking medication or stably taking extended-release methylphenidate or atomoxetine hydrochloride (dosage fluctuation < 10% within the past month). Medication status and stability were monitored using online self-report questionnaires administered at baseline, weekly during treatment, and at each follow-up assessment (every 3 months), covering current medication type, dosage, frequency, and adverse effects. These reports were reviewed by the therapists in both intervention groups. When necessary, this information was further confirmed with participants in person. Exclusion criteria included: (1) Current diagnosis of the following psychiatric disorders: psychotic disorders (to avoid confounding related to differing clinical priorities and symptom mechanisms), acute episodes of bipolar disorder, depressive disorders with psychotic features or high risk of suicide/self-harm, severe panic disorder, substance dependence, autism spectrum disorder, or antisocial personality disorder; (2) Intelligence quotient (IQ) below 90 (to ensure the feasibility of some cognitive assessments, consistent with prior studies [19, 29, 53]); (3) Presence of significant suicidal ideation; (4) Unstable physical conditions or inadequately managed medical illnesses, as determined by the research team (e.g., hypertension, diabetes, angina, active hepatitis, etc.); (5) Ongoing participation in other psychological treatments. Withdrawal and termination criteria were as follows: (1) Voluntary withdrawal by the participant; (2) Development of new psychiatric disorders other than ADHD during the study period; (3) Attendance at fewer than seven of the twelve group sessions; (4) Participation in other psychological treatments or training programs during the study; (5) Significant symptom fluctuations requiring major medication adjustments during the study period (to minimize confounding from substantial pharmacological changes).

Sample size

The sample size was calculated using G*Power software, with α set at 0.05 (two-tailed) and 1-β at 0.80. The primary outcome measure was the self-report version of the ADHD Rating Scale (ADHD-RS) score. According to previous literature, the effect size for non-inferiority studies should be set at half of that observed in prior controlled studies [55]. A review of previous trials indicated that the effect size of CBT ranged from 0.76 (95% CI: 0.21–1.31) [56] to 0.84 (95% CI: 0.50–1.18) [57]. Therefore, Cohen’s d was set at 0.40, yielding a required sample size of n₁ = n₂ = 14 per group. Based on our previous clinical trials, we estimated a treatment dropout rate of approximately 25% and a follow-up dropout rate of approximately 20% [58]. After adjusting for these anticipated attrition rates, the final required sample size was calculated to be 38 participants per group, resulting in a total of 76 participants across both groups.

Randomization, treatment allocation, and blinding

Participants were sequentially numbered and randomly assigned to treatment groups a 1:1 ratio using a randomization sequence generated in Microsoft Excel by a member of the research team. Treatment allocation was implemented by another designated research team member according to the pre-generated sequence who was not an independent researcher but was not involved in outcome assessment. Because of the nature of the psychotherapy interventions, therapists were aware of the treatment condition and of the non-inferiority design. Participants were aware of their treatment assignment (CBT or DBT), but were not informed of the non-inferiority hypothesis. Participant allocation strictly followed the study protocol, adhering to predefined inclusion and exclusion criteria to ensure sample integrity. Clinical and cognitive assessors were blinded to group assignments throughout the study to minimize assessment bias.

Procedures

This RCT was conducted in accordance with the CONSORT 2010 guidelines. The study protocol was registered at Peking University Sixth Hospital (registration ID: ChiCTR2300072075). The study adhered to the principles of the Declaration of Helsinki, and ethical approval was obtained from the Ethics Committee of Peking University Sixth Hospital. All participants provided written informed consent prior to enrollment. Assessments were conducted at six time points: baseline (T0), week 4 (T1), week 8 (T2), post-treatment (T3), 3-month follow-up (T4), and 6-month follow-up (T5). At baseline, participants underwent a diagnostic interview and a comprehensive clinical evaluation, including assessments of intelligence, ADHD symptom severity, emotional symptoms, ecological EF, global functioning, QoL, and neurocognitive task performance. Diagnostic interviews were conducted by psychiatrists at the attending level or higher. Self-report questionnaires were administered via secure online surveys, while neurocognitive tasks were performed by trained psychiatry postgraduates with established inter-rater reliability. Clinical symptom assessments were conducted at all time points (T1–T5), whereas neurocognitive tasks were administered only at T3–T5.

As illustrated in Fig. 1, a total of 288 participants were recruited from outpatient clinics and online ADHD communities. Among these, 98 met the eligibility criteria, completed baseline assessments, and were randomly assigned to either the DBT group (n = 49) or the CBT group (n = 49). At the 6-month follow-up, 45 participants in the DBT group (94%) and 44 participants in the CBT group (90%) completed all assessments. Participants were instructed to maintain stable medication use during the treatment phase, and any necessary treatment adjustments due to clinical deterioration were required to be reported to the therapist in advance. No restrictions were imposed on medication dosage or concomitant treatments during the follow-up period. At each follow-up assessment, participants were asked to report any medication changes that had occurred during the preceding 3 months.

Fig. 1.

Study flow chart

Intervention methods

DBT treatment protocol

The DBT treatment protocol was developed based on the Linehan’s original DBT model [59], tailored to the emotion dysregulation characteristics of adults with ADHD, and incorporated selected elements from Hesslinger’s eighth session [31]. Additionally, the protocol integrated the clinical experience of the research team in treating adults with ADHD. The program comprised modules on mindfulness skills, behavioral analysis, emotion regulation, distress tolerance, and interpersonal effectiveness. It also included focused discussions and behavioral analyses related to procrastination, addictive behaviors, and psychiatric comorbidities, which are particularly relevant to individuals with ADHD. Details of the intervention structure are presented in Table 1.

Table 1.

Treatment Protocol

Week	DBT Treatment Protocol		CBT Treatment Protocol
	Intervention Modules	sessions	Intervention Modules	sessions
1	Psycho-education	Overview of DBT group skills; Psychoeducation on adult ADHD	Psycho-education	Psychoeducation and Interpersonal Skills
2	Basic Mindfulness	“What it is” and “How to do it” techniques	Organization and Planning Skills	Goal Setting and Social Network
3	Behavior Chain Analysis	ADHD-related problem behaviors; Integration of Behavior Chain Analysis Techniques		Organizing Multiple Tasks
4	Emotion Regulation	Recognize, label and describe emotions; verify facts and problem-solving		Problem-Solving and Major Tasks Management
5		ABC skills and PLEASE skills		File Organization
6	Advanced Mindfulness	Integrating mindfulness with current emotions; core techniques: spiritual awareness, wise mind, balance	Managing Distraction	Environmental Modification Measuring Attention Span and Strategies for Delaying Distraction
7	Midpoint Review	Review previously learned skills; analyze and summarize factors hindering treatment progress	Adaptive Thinking	ADHD Cognitive Model Introduction
8	Comorbidity and Procrastination	Psychoeducation on adult ADHD, depression, addiction and procrastination		Recognizing and Adjusting Maladaptive Thoughts
9	Distress Tolerance	Crisis survival skills; identifying crisis situations; acceptance and cognitive reframing		Breaking the Thought–Emotion Negative Cycle
10	Interpersonal Effectiveness	Setting interpersonal goals		Identifying Intermediate Beliefs and Understanding Core Beliefs
11		Maintaining skills; self-respect strategies	Coping with Procrastination	Applying Learned Skills to cope with Procrastination
12	Review and Outlook	Consolidate group skills and engage in a therapeutic farewell ritual	Relapse Prevention	Reviewing All Modules and Preparing for Post-Treatment Life

CBT Treatment Protocol

The CBT protocol used in this study was adapted from Safren et al.’s program [19] and modified to better suit the Chinese population [24, 26, 53]. Compared with the original version, the main adaptations, based on our prior clinical experience, included reducing the distractibility module to one session and expanding the cognitive module to four sessions. The intervention focused on enhancing organizational and planning skills, reducing distractibility, promoting adaptive thinking, and developing cognitive-behavioral coping strategies. It also incorporated targeted techniques for managing procrastination, improving emotion regulation, and facilitating behavioral modification. Details of the intervention structure are presented in Table 1.

Throughout the trial, each treatment condition was delivered by a fixed pair of therapists. All CBT groups were conducted by the same two therapists, and all DBT groups were conducted by another two therapists; therapists were not shared across conditions. In each condition, one therapist served as the group leader responsible for guiding the sessions, while the other acted as a co-leader, responsible for recording observations, providing supplementary input, and offering timely feedback. Both CBT and DBT were delivered in small groups of 8–12 participants. All therapists had received standardized psychotherapy training through programs at Peking University Sixth Hospital and the Chinese Psychological Society.

Group therapy sessions were conducted once per week at a fixed time, with each session lasting 120 min. Each session included a review of previously covered content and homework, followed by the introduction of new material, skills practice, and group discussion. Homework was assigned after each session and reviewed in the subsequent session by the therapists as part of routine treatment implementation. Participants’ difficulties and engagement with homework were discussed during these reviews, and, when necessary, homework-related problems were further addressed in individual one-to-one discussions. The structured program was designed to facilitate skill acquisition, peer interaction, and self-reflection. Participants who were unable to attend a session were required to notify the group in advance; relevant materials were provided for self-study, and any questions were addressed in subsequent sessions to maintain group cohesion. To monitor and enhance treatment fidelity, all sessions were audio-recorded. After each session, the two therapists reviewed the recording together to reflect on session implementation and process. In addition, all therapists attended monthly group supervision throughout the trial to support treatment fidelity and therapist competence.

Outcome measures

Diagnostic interview and IQ evaluations

Adult ADHD was assessed using the Conners’ Adult ADHD Diagnostic Interview for DSM-IV (CAADID) [60]. The diagnostic criteria of the DSM-5 were applied, whereby a diagnosis was established if at least five symptoms were endorsed in either the inattentive or the hyperactive–impulsive dimension. Baseline psychiatric comorbidities were evaluated using the Structured Clinical Interview for DSM-IV (SCID), with SCID-I employed to assess Axis I psychiatric disorders [61] and SCID-II to assess Axis II personality disorders [62]. Intellectual functioning was measured with the Chinese version of the Wechsler Adult Intelligence Scale-Revised (WAIS-RC) [63]. The full-scale IQ (FIQ) score was used as the primary indicator of general cognitive ability.

Primary outcome measures

ADHD Rating Scale (ADHD-RS)

The primary outcome measure was the severity of core ADHD symptoms, assessed using the self-report version of the ADHD-RS. The ADHD-RS consists of 18 items rated on a 4-point Likert scale ranging from 0 (never) to 3 (very often) [64]. Higher total scores reflect greater symptom severity of ADHD. In the present sample, the ADHD-RS demonstrated good internal consistency across assessment points; detailed Cronbach’s α coefficients are provided in Supplementary Table S1.

Secondary outcome measures

All remaining clinical, cognitive, and exploratory measures were treated as secondary outcomes.

Emotional symptoms

The Self-Rating Depression Scale (SDS) is a self-report instrument used to assess subjective depressive symptoms experienced over the past week. Higher scores indicate more severe depressive symptoms. The Chinese version of the SDS has demonstrated good reliability and validity [65]. Similarly, the Self-Rating Anxiety Scale (SAS) is a self-report measure that evaluates subjective anxiety symptoms over the past week. Higher scores reflect greater severity of anxiety symptoms. The reliability and validity of the Chinese version of the SAS have been well established [66]. In the present sample, the SDS and SAS demonstrated good to excellent internal consistency across assessment points; detailed Cronbach’s α coefficients are provided in Supplementary Table S1.

Emotion regulation

The Emotion Regulation Questionnaire (ERQ) is a self-report scale used to assess individual differences in emotion regulation strategies. It comprises two subscales: cognitive reappraisal (CR) and expressive suppression (ES). The Chinese version of the ERQ has demonstrated good reliability and validity [67]. The Difficulties in Emotion Regulation Scale (DERS) is a self-report measure that assesses deficits in emotion regulation across six domains: emotional acceptance, strategy use, impulse control, emotional awareness, goal-directed behavior, and emotional clarity. The Chinese version of the DERS has shown good internal consistency and structural validity [68]. In the present sample, the ERQ-CR demonstrated good internal consistency across assessment points, whereas the expressive suppression subscale (ERQ-ES) showed overall acceptable to good internal consistency; the DERS Total score demonstrated excellent internal consistency across assessment points; detailed Cronbach’s α coefficients are provided in Supplementary Table S1.

EF

The Behavior Rating Inventory of Executive Function—Adult Version (BRIEF-A) is a self-report instrument used to assess EF impairments in adults. It consists of 75 items rated on a three-point Likert scale and yields two main indices: the Behavioral Regulation Index (BRI) and the Metacognition Index (MI). The BRI comprises subdomains of inhibition, shifting, emotional control, and self-monitoring, while the MI includes initiation, working memory, planning/organization, organization of materials, and task monitoring. The total score, known as the Global Executive Composite (GEC), represents the overall level of executive dysfunction, with higher scores indicating greater impairment. The Chinese version of the BRIEF-A has been validated for reliability and validity [69]. In the present sample, the BRI, MI, and GEC all demonstrated excellent internal consistency across assessment points; detailed Cronbach’s α coefficients are provided in Supplementary Table S1.

The Continuous Performance Test—Identical Pairs (CPT-IP) assesses sustained attention and vigilance. Participants are presented with a rapid sequence of flashing numbers on a standard laptop equipped with CPT-IP software. They are instructed to respond as quickly as possible by clicking the left mouse button whenever two consecutive numbers are identical, and to refrain from responding otherwise. The test comprises three difficulty levels (2-digit, 3-digit, and 4-digit sequences), each consisting of 150 stimuli, with 20% (30 trials) being identical pairs. The signal detection index (dʹ) is calculated, with higher values indicating better attention and vigilance [70].

The Stroop Color and Word Test (SCWT) evaluates inhibitory control and selective attention through three subtests: rapid color naming, rapid word reading, and a color-word interference task in which participants must name the ink color of words while ignoring their semantic content. Reaction time and accuracy are used to assess cognitive interference and inhibitory control [71]. In this study, only the third section of the SCWT was administered.

The Spatial Span Test (SST) measures spatial working memory using an irregularly arranged array of nine cubes. During the task, participants are asked to reproduce a sequence of blocks tapped by the examiner, either in the same order (forward) or in reverse (backward). Testing continues until the participant fails to accurately reproduce the sequence [72]. In the present study, the total SST score (SS Total) —calculated as the sum of forward and backward spans—was used as the index of spatial working memory.

The Trail Making Test (TMT) assesses processing speed and cognitive flexibility. It comprises two parts: TMT-A, in which participants connect numbers (1–25) in ascending order, and TMT-B, in which they alternate between numbers (1–13) and letters (A–I) in ascending order. Performance is evaluated by total completion time, with shorter times reflecting better EF [73].

QoL

The Brief Version of the World Health Organization Quality of Life Scale (WHOQOL-BREF) is a self-report instrument designed to assess quality of life across four domains: physical health, psychological well-being, social relationships, and environmental factors. Each item is rated on a five-point Likert scale, with higher scores indicating better QoL. The Chinese version of the WHOQOL-BREF has demonstrated good reliability and validity [74]. In the present sample, the WHOQOL physical, psychological, and environmental domains demonstrated overall acceptable to good internal consistency across assessment points, while the social relationships domain showed somewhat greater variability, ranging from borderline acceptable to good; detailed Cronbach’s α coefficients are provided in Supplementary Table S1.

Global Functioning

The General Self-Efficacy Scale (GSES) measures an individual’s perceived self-efficacy, or belief in their ability to cope effectively with challenging situations. It comprises 10 items rated on a 4-point Likert scale, with higher scores reflecting greater self-efficacy [75]. In the present sample, the GSES Total score demonstrated excellent internal consistency across assessment points; detailed Cronbach’s α coefficients are provided in Supplementary Table S1.

The Sheehan Disability Scale (SDS) is a brief self-report measure used to evaluate functional impairment across three key domains: work/school, social life, and family life/home responsibilities. Each domain is rated on a 10-point Likert scale, with higher scores indicating greater levels of impairment [76]. In the present sample, the Sheehan Disability Scale showed variable internal consistency across assessment points, with acceptable reliability observed at some time points; detailed Cronbach’s α coefficients are provided in Supplementary Table S1.

According to the trial registration (ChiCTR2300072075), the study was designed to assess both treatment effectiveness and potential mechanisms. In the present manuscript, we primarily report the main clinical outcomes and laboratory-based EF measures, while the status of all prespecified secondary outcomes is provided in Supplementary Table S2. TMT and SST were not prespecified in the trial registration and are therefore reported as exploratory outcomes.

Statistical analysis

All data were double-entered and cross-checked prior to statistical analysis. Baseline group equivalence was assessed for general demographic and clinical characteristics. For continuous variables, independent samples t-tests were used to compare groups, whereas the Mann-Whitney U test was applied for non-normally distributed data. For categorical variables, χ² tests were performed. Linear Mixed Model (LMM) was employed to analyze changes in scale scores over time. Follow-up time points (T1-T5), treatment group, and their interaction were modeled as fixed effects, while individual differences and time were treated as random effects. Treatment efficacy indicators across multiple dimensions were compared between the DBT and CBT groups from T0 to T5. When a significant group × time interaction was observed, separate LMM analyses were conducted within each group to further explore the time effect. To control for multiple comparisons, false discovery rate (FDR)-adjusted post hoc pairwise comparisons were performed using the Benjamini-Hochberg procedure. All analyses followed the intention-to-treat (ITT) principle, and missing data were handled using multiple imputation, generating 20 imputed datasets, and LMM estimates were pooled using Rubin’s rules. Effect sizes were derived from the corresponding model-based pairwise contrasts rather than calculated directly from raw observed means and standard deviations. For between-group comparisons, Cohen’s d was used to represent the standardized difference between DBT and CBT at each assessment point. For within-group comparisons, within-group effect sizes were used to represent the magnitude of change over time within each treatment group. Both indices were approximated from the corresponding model-based pairwise contrasts using the associated t statistics and degrees of freedom. Statistical analyses were performed using SPSS version 26 and R version 4.4.2. All tests were two-tailed, with the level of statistical significance set at α = 0.05. Effect sizes were interpreted as small (0.2), medium (0.5), and large (0.8).

Results

Baseline characteristics

There were no statistically significant differences between the DBT and the CBT groups in demographic or clinical characteristics at baseline (Tables 2 and 3).

Table 2.

Demographic characteristics of the two groups

Characteristic	DBT group	CBT group	χ2/t value	p value
Female(n%)	31(63%)	34(69%)	0.41	0.520
Age, year(SD)	28.02(7.74)	28.08(5.24)	0.046	0.964
Total IQ (SD)	121.51(8.29)	123.10(9.61)	0.877	0.383
Education, year (SD)	16.51(2.83)	17.20(2.84)	1.213	0.228
ADHD subtype			1.020	0.600
Inattentive subtype	25(51%)	24(49%)
Hyperactive-impulsive type	0	1(2%)
Combined subtype	24(49%)	24(49%)
medication	20(41%)	28(57%)	2.613	0.106
methylphenidate	17(35%)	26(53%)
tomoxetine	3(6%)	1(2%)
combined	0	1(2%)
Other psychotropic medications	7(14%)	5(10%)	0.380	0.538
Anxiolytics and antidepressants	5(10%)	5(10%)	0.000	1.000
Antipsychotic medications	2(4%)	0	0.510	0.475
Mood stabilizers	4(8%)	0	2.346	0.126
Other medications	1(2%)	1(2%)	0.000	1.000
Axis I comorbidity	35(71%)	30(61%)	1.142	0.285
Current Major depressive disorder	7(14%)	8(16%)	0.079	0.779
Previous Major depressive disorder	24(49%)	22(45%)	0.164	0.686
Bipolar disorder	9(18%)	5(10%)	1.333	0.248
dysthymia	5(10%)	1(2%)	1.598	0.206
Anxiety disorders	8(16%)	10(20%)	0.272	0.602
Panic disorder	5(10%)	3(6%)	0.544	0.461
Social anxiety disorder	4(8%)	2(4%)	0.178	0.673
Generalized Anxiety Disorder	5(10%)	3(6%)	0.136	0.712
Specific Phobia Disorder	1(2%)	3(6%)	0.261	0.610
Unspecified Anxiety Disorder	1(2%)	1(2%)	0.000	1.000
Obsessive-Compulsive Disorder	4(8%)	2(4%)	0.178	0.673
Axis II comorbidity	16(33%)	13(27%)	0.441	0.507
Avoidant personality disorder	8(8%)	2(4%)	2.784	0.095
Obsessive-compulsive personality disorder	3(6%)	7(14%)	1.002	0.317
Passive-aggressive personality disorder	5(10%)	4(8%)	0.000	1.000
Depressive personality disorder	3(6%)	5(10%)	0.136	0.712
Paranoid personality disorder	0	2(4%)	0.510	0.475
Borderline personality disorder	2(4%)	3(6%)	0.000	1.000

Table 3.

Clinical characteristics of the two groups

Characteristic	DBT group	CBT group	χ2/t value	p value
ADHD-RS	32.84 (8.44)	32.12 (9.18)	-0.401	0.689
Inattention	18.53 (3.66)	18.45 (4.95)	-0.093	0.926
Hyperactivity/ impulsivity	14.31 (5.87)	13.67 (5.33)	-0.558	0.578
SAS	36.82 (8.07)	37.57 (9.43)	0.426	0.671
SDS	41.06 (10.25)	41.76 (9.55)	0.347	0.730
ERQ
Cognitive reappraisal	26.08 (6.34)	26.61 (7.17)	0.388	0.699
Expressive suppression	14.47 (6.11)	13.53 (6.06)	-0.763	0.447
DERS	99.31 (19.86)	95.00 (24.50)	-0.956	0.342
WHOQOL-BREF
Physical domain	11.58 (2.51)	11.80 (3.00)	0.397	0.692
Psychological domain	11.93 (2.68)	11.71 (3.18)	-0.367	0.715
Social relation domain	12.60 (3.51)	12.54 (3.68)	-0.075	0.940
Environmental domain	13.54 (2.34)	13.65 (2.83)	0.214	0.831
GSES	24.04 (7.36)	25.88 (7.02)	1.264	0.209
Sheehan Disability Scale	16.08 (5.59)	16.69 (6.25)	0.511	0.610
BRIEF-A GEC	144.06(26.27)	147.67(24.91)	-0.698	0.487
BRI	57.06 (13.19)	57.12 (13.42)	-0.153	0.879
MI	90.27 (14.37)	87.08 (15.63)	-1.059	0.292
TMT
TMT-A	23.94(8.217)	25.35(10.299)	-0.748	0.456
TMT-B	60.82(38.018)	63.84(31.764)	-0.427	0.671
SS Total	17.57(2.693)	18.18(3.346)	-0.998	0.321
SCWT-CW	48.67(12.034)	46.51(10.911)	0.932	0.354
CPT-IP
2-digit dʹ	3.56(0.636)	3.63(0.691)	-0.560	0.577
3-digit dʹ	3.14(0.816)	3.13(0.872)	0.041	0.967
4-digit dʹ	2.34(0.908)	2.09(0.863)	1.415	0.160

ADHD-RS ADHD Rating Scale; SAS Zung Self-rating Anxiety Scale; SDS Zung Self-rating Depression Scale; ERQ Emotion Regulation Questionnaire; DERS Difficulties in Emotion Regulation Scale; GSES General Self-Efficacy Scale; WHOQOL- BREF Brief Version of the World Health Organization Quality of Life; BRIEF-A Behavior Rating Inventory of Executive Function-Adult Version; BRI Behavioral Regulation Index; MI Metacognition Index; GEC Global Executive Composite; TMT Trail Making Test; SS Spatial span test; SCWT-CW Stroop Color and Word Test-Color Word Test; CPT-IP Continuous Performance Test-Identical Pairs

Treatment adherence

The mean number of sessions attended was 10.49 ± 2.33 in the DBT group and 9.93 ± 1.90 in the CBT group. At the end of treatment, adherence rates were 95.92% for DBT and 93.88% for CBT. At the three-month follow-up, adherence remained high at 95.92% in the DBT group and 91.84% in the CBT group, while at the six-month follow-up, adherence was 91.84% and 89.80%, respectively. No significant between-group differences were observed in adherence or follow-up participation at any time point (P > 0.05).

In the DBT group, two participants withdrew after the first session, and another two were lost to follow-up at the six-month assessment. In the CBT group, two participants experienced medication changes (one doubled the dose of methylphenidate, and another discontinued sertraline), one participant withdrew before completing the final session, and one participant was excluded for attending only six sessions. Additionally, one participant was lost to follow-up at six months. Because all assessments required in-person attendance, some participants missed individual assessment time points due to scheduling conflicts but completed subsequent assessments, ensuring inclusion in the longitudinal analysis.

As shown in Table 4, completers (n = 89) and non-completers (n = 9) were broadly comparable in baseline demographic and clinical characteristics. Categorical variables were compared using Fisher’s exact test or Pearson’s chi-square test, as appropriate, and continuous variables were compared using the Mann-Whitney U test. No statistically significant between-group differences were observed in demographic characteristics, psychiatric comorbidity, ADHD subtype distribution, medication status, or baseline measures of core ADHD symptoms, anxiety and depressive symptoms, emotion dysregulation, executive dysfunction, quality of life, self-efficacy, and functional impairment (all P > 0.05). Overall, these findings do not indicate that study non-completion was systematically related to the measured baseline variables.

Table 4.

Baseline characteristics of completers and non-completers

Characteristic	Overall (N = 98)	Completers (n = 89)	non-completers (n = 9)	p value
Demographic and clinical characteristics
Treatment group				1.000
CBT	49 (50.0%)	44 (49.4%)	5 (55.6%)
DBT	49 (50.0%)	45 (50.6%)	4 (44.4%)
Axis I comorbidity				1.000
Yes	65 (66.3%)	59 (66.3%)	6 (66.7%)
No	33 (33.7%)	30 (33.7%)	3 (33.3%)
Axis II comorbidity				0.120
Yes	29 (29.6%)	24 (27.0%)	5 (55.6%)
No	69 (70.4%)	65 (73.0%)	4 (44.4%)
Sex				0.480
Male	33 (33.7%)	29 (32.6%)	4 (44.4%)
Female	65 (66.3%)	60 (67.4%)	5 (55.6%)
ADHD subtype				0.903
Combined subtype	48 (49.0%)	44 (49.4%)	4 (44.4)
Hyperactive-impulsive type	1 (1.0%)	1 (1.1%)	0 (0.0)
Inattentive subtype	49 (50.0%)	44 (49.4%)	5 (55.6%)
Current ADHD medication				0.738
Yes	48 (49.0%)	43 (48.3%)	5 (55.6%)
No	50 (51.0%)	46 (51.7%)	4 (44.4%)
Other psychotropic medication				0.595
Yes	12 (12.2%)	12 (13.5%)	0 (0.0)
No	86 (87.8%)	77 (86.5%)	9 (100.0%)
Age, years	27 [24–31]	28 [24–31]	26 [21–27]	0.108
IQ	123 [117–128]	123 [118–128]	119 [116–126]	0.522
Education, years	16 [15.25-18]	16 [16–18]	16 [14–18]	0.778
ADHD-RS	32.5 [28–39]	32 [28–39]	36 [29–39]	0.471
Inattention	18 [16-21.75]	18 [16–21]	18 [17–25]	0.364
Hyperactivity/ impulsivity	15 [11–19]	15 [11–19]	14 [13–18]	0.763
Additional self-report measures
SAS	35.5 [31–42]	36 [31–43]	34 [31–38]	0.349
SDS	42 [34.25-47]	42 [35–47]	41 [34–47]	0.721
WHOQOL-BREF
Physical domain	11.43 [9.71–13.71]	11.43 [9.71–13.14]	14.29 [9.71–14.29]	0.542
Psychological domain	11.67 [10–14]	11.33 [10.67-14]	12 [7.33-14]	0.490
Social relation domain	12 [10.67–14.67]	12 [10.67–14.67]	12 [9.33-16]	0.965
Environmental domain	13.5 [12-15.38]	13.5 [12–15]	13.5 [11.5–16.5]	0.684
GSES	26 [20-29.75]	26 [20–30]	21 [20–27]	0.313
Sheehan Disability Scale	16 [12–20]	16 [13–20]	14 [12–19]	0.494
ERQ-CR	27 [22–30]	27 [22–31]	23 [22–28]	0.427
ERQ-ES	14 [9.25-18]	14 [9–18]	14 [12–18]	0.666
DERS	97 [79-113.75]	97 [81–113]	96 [77–124]	0.931
BRIEF-A GEC	146.5 [129-162.5]	147 [128–163]	145 [134–153]	1.000
BRI	56.5 [47–64]	57 [47–64]	53 [49–63]	0.980
MI	89.5 [78.25–100]	89 [78–100]	90 [80–99]	0.946

ADHD-RS ADHD Rating Scale; SAS Zung Self-rating Anxiety Scale; SDS Zung Self-rating Depression Scale; ERQ Emotion Regulation Questionnaire; DERS Difficulties in Emotion Regulation Scale; GSES General Self-Efficacy Scale; WHOQOL- BREF Brief Version of the World Health Organization Quality of Life; BRIEF-A Behavior Rating Inventory of Executive Function-Adult Version; BRI Behavioral Regulation Index; MI Metacognition Index; GEC Global Executive Composite;

Categorical variables are presented as n (%), and continuous variables are presented as median [interquartile range]. Fisher’s exact test was used for 2 × 2 categorical comparisons, Pearson’s chi-square test was used for multi-category comparisons, and the Mann-Whitney U test was used for continuous variables

In accordance with the study protocol, treatment regimens were not restricted during the post-treatment follow-up period. Medication changes during follow-up are summarized in Table 5, with detailed records of medication regimen adjustments presented in Supplementary Table S3.

Table 5.

Medication changes from T3 (end of treatment) to T5 (6-month follow-up)

Variable	DBT group, n (%)	CBT group, n (%)	p value
Any medication change	23 (46.9%)	22 (44.9%)	1.000
Any ADHD medication change	18 (36.7%)	21 (42.9%)	0.680
ADHD medication initiation	9 (18.4%)	4 (8.2%)	0.233
ADHD medication discontinuation	13 (26.5%)	12 (24.5%)	1.000
ADHD medication dose/frequency or regimen adjustment	4 (8.2%)	10 (20.4%)	0.147
Any other psychotropic medication	10 (20.4%)	6 (12.2%)	0.413
Other psychotropic medication initiation	3 (6.1%)	3 (6.1%)	1.000
Other psychotropic medication discontinuation	4 (8.2%)	3 (6.1%)	1.000

1. Counts are participant-level counts: each participant was counted once per row if they experienced at least one change of that type during the period

2. ADHD medications included extended-release methylphenidate or atomoxetine hydrochloride. Other psychotropic medications included antipsychotics, mood stabilizers, antidepressants, and anxiolytics

3.p values were calculated using Fisher’s exact test

Primary outcomes

The LMM results presented below are pooled estimates based on 20 multiply imputed datasets combined using Rubin’s rules. No additional covariates were included in the LMM analyses. Cohen’s d values are also reported for between-group comparisons to indicate the magnitude of group differences, and within-group effect sizes are reported for within-group comparisons to reflect the magnitude of change over time.

LMM analyses revealed no significant group × time interaction effects for the ADHD-RS total score across all assessment points (Table 6). At the end of treatment (T3), the between-group difference was small (d = 0.06 [-0.33, 0.46]), and a similarly small difference was observed at the 6-month follow-up (d = 0.17 [-0.27, 0.61]). Although no significant between-group differences were detected, the 95% confidence intervals were not fully contained within the prespecified non-inferiority margin of 0.40. Therefore, formal non-inferiority could not be established.

Table 6.

Mixed-Effects Linear Model Results for ADHD and Emotional Symptoms Across Assessment Points, with Post Hoc Group Comparisons to Baseline

		assessment point ×treatment type interaction					DBT (n = 49)		CBT (n = 49)
		b	t	df	p	cohen’s d [95%CI]	Mean (SD)	within-group effect size[95%CI]	Mean (SD)	within-group effect size[95%CI]
ADHD-RS	T1	-0.348	-0.238	567.32	0.812	0.05 [-0.37, 0.47]	26.20(6.52)	-0.88 [-1.15, -0.61]***#	25.84(8.96)	-0.69 [-0.92, -0.47]***#
	T2	-0.975	-0.655	548.52	0.513	-0.03 [-0.39, 0.33]	24.03(8.39)	-1.05 [-1.29, -0.80]***#	24.29(9.36)	-0.84 [-1.07, -0.62]***#
	T3	-0.197	-0.131	559.21	0.896	0.06 [-0.33, 0.46]	22.25(7.44)	-1.33 [-1.59, -1.07]***#	21.73(8.61)	-1.17 [-1.40, -0.93]***#
	T4	-0.836	-0.524	564.00	0.601	-0.02 [-0.45, 0.41]	22.24(6.96)	-1.37 [-1.66, -1.08]***#	22.37(7.26)	-1.18 [-1.45, -0.91]***#
	T5	0.500	0.285	560.39	0.776	0.17 [-0.27, 0.61]	23.42(7.82)	-1.16 [-1.46, -0.86]***#	22.21(6.17)	-1.27 [-1.58, -0.95]***#
Inattention	T1	0.305	0.367	567.25	0.714	0.10 [-0.32, 0.51]	15.12(3.68)	-0.93 [-1.25, -0.61]***#	14.73(4.39)	-0.80 [-1.04, -0.55]***#
	T2	0.791	0.934	547.14	0.351	0.19 [-0.17, 0.56]	14.04(4.51)	-1.09 [-1.38, -0.81]***#	13.17(4.58)	-1.11 [-1.35, -0.86]***#
	T3	0.567	0.663	551.52	0.508	0.15 [-0.23, 0.54]	12.92(3.94)	-1.47 [-1.79, -1.16]***#	12.27(4.55)	-1.30 [-1.55, -1.05]***#
	T4	0.066	0.073	558.10	0.942	0.04 [-0.39, 0.47]	13.10(3.39)	-1.54 [-1.90, -1.18]***#	12.95(4.02)	-1.22 [-1.50, -0.94]***#
	T5	0.480	0.481	552.84	0.631	0.15 [-0.30, 0.59]	13.25(4.21)	-1.34 [-1.69, -0.99]***#	12.68(3.34)	-1.37 [-1.70, -1.03]***#
Hyperactivity/ impulsivity	T1	-0.653	-0.778	565.59	0.437	0.00 [-0.41, 0.40]	11.08(5.08)	-0.59 [-0.80, -0.37]***#	11.10(5.69)	-0.47 [-0.68, -0.26]***#
	T2	-1.766	-2.066	544.38	0.039*	-0.20 [-0.58, 0.17]	9.99(5.32)	-0.77 [-0.98, -0.56]***#	11.12(5.97)	-0.45 [-0.66, -0.24]***#
	T3	-0.764	-0.885	552.27	0.377	-0.02 [-0.42, 0.37]	9.32(5.12)	-0.90 [-1.12, -0.69]***#	9.46(5.32)	-0.79 [-1.02, -0.56]***#
	T4	-0.902	-0.979	556.57	0.328	-0.06 [-0.49, 0.37]	9.15(4.87)	-0.96 [-1.19, -0.72]***#	9.42(4.22)	-0.88 [-1.15, -0.62]***#
	T5	0.020	0.020	559.89	0.984	0.14 [-0.28, 0.56]	10.18(4.93)	-0.76 [-1.02, -0.50]***#	9.53(4.25)	-0.86 [-1.16, -0.56]***#
SAS	T1	1.120	0.854	566.66	0.394	0.04 [-0.36, 0.44]	37.02(8.17)	0.02 [-0.20, 0.25]	36.65(9.56)	-0.10 [-0.29, 0.10]
	T2	0.805	0.606	560.24	0.545	0.01 [-0.38, 0.40]	35.50(8.76)	-0.16 [-0.38, 0.06]	35.45(9.14)	-0.23 [-0.43, -0.03]*
	T3	1.070	0.792	557.07	0.428	0.04 [-0.37, 0.44]	34.12(8.05)	-0.34 [-0.57, -0.10]**#	33.80(8.98)	-0.41 [-0.61, -0.20]***#
	T4	1.827	1.244	533.37	0.214	0.12 [-0.25, 0.49]	35.98(9.15)	-0.10 [-0.33, 0.14]	34.91(9.14)	-0.29 [-0.50, -0.07]*#
	T5	0.808	0.499	550.64	0.618	0.01 [-0.43, 0.45]	34.40(7.98)	-0.30 [-0.58, -0.02]*	34.35(7.53)	-0.38 [-0.64, -0.11]**#
SDS	T1	1.385	0.934	567.53	0.351	0.08 [-0.36, 0.52]	41.12(8.49)	0.01 [-0.21, 0.23]	40.43(8.67)	-0.15 [-0.37, 0.08]
	T2	0.833	0.558	565.52	0.577	0.01 [-0.38, 0.41]	38.95(9.83)	-0.21 [-0.42, 0.00]*	38.81(8.95)	-0.32 [-0.54, -0.09]**#
	T3	2.477	1.640	562.04	0.102	0.19 [-0.20, 0.58]	38.98(9.90)	-0.21 [-0.41, 0.00]	37.20(9.03)	-0.49 [-0.72, -0.26]***#
	T4	1.614	1.016	560.12	0.310	0.09 [-0.28, 0.46]	39.43(10.02)	-0.16 [-0.38, 0.06]	38.51(9.93)	-0.33 [-0.56, -0.11]**#
	T5	1.062	0.624	564.13	0.533	0.04 [-0.40, 0.48]	38.16(8.76)	-0.30 [-0.55, -0.06]*	37.80(8.53)	-0.44 [-0.70, -0.17]**#
ERQ
Cognitive reappraisal	T1	-0.987	-0.777	568.18	0.437	-0.24 [-0.63, 0.16]	26.55(6.46)	0.07 [-0.20, 0.35]	28.07(6.40)	0.21 [-0.04, 0.47]
	T2	0.464	0.363	564.66	0.716	-0.01 [-0.42, 0.40]	28.54(6.67)	0.38 [0.10, 0.65]**#	28.60(5.57)	0.31 [0.03, 0.59]*
	T3	0.625	0.485	563.63	0.628	0.01 [-0.39, 0.42]	29.75(6.00)	0.59 [0.30, 0.88]***#	29.65(6.60)	0.44 [0.18, 0.70]***#
	T4	1.116	0.831	558.94	0.406	0.10 [-0.32, 0.51]	30.07(6.21)	0.64 [0.34, 0.93]***#	29.49(6.11)	0.43 [0.15, 0.71]**#
	T5	-0.101	-0.071	561.03	0.944	-0.09 [-0.49, 0.30]	29.67(6.87)	0.54 [0.24, 0.84]***#	30.30(6.63)	0.53 [0.24, 0.82]***#
Expressive suppression	T1	-0.140	-0.151	564.80	0.880	0.14 [-0.25, 0.53]	15.12(5.61)	0.11 [-0.11, 0.33]	14.32(6.08)	0.13 [-0.08, 0.34]
	T2	-1.358	-1.456	560.19	0.146	-0.07 [-0.47, 0.32]	13.81(5.57)	-0.11 [-0.34, 0.11]	14.23(5.67)	0.12 [-0.10, 0.34]
	T3	-0.315	-0.335	557.76	0.738	0.12 [-0.30, 0.54]	13.45(5.51)	-0.17 [-0.40, 0.05]	12.83(5.05)	-0.12 [-0.36, 0.11]
	T4	-0.822	-0.839	555.25	0.402	0.02 [-0.40, 0.44]	13.41(5.32)	-0.18 [-0.42, 0.05]	13.29(4.91)	-0.04 [-0.29, 0.20]
	T5	-0.070	-0.067	548.62	0.947	0.16 [-0.24, 0.57]	13.16(4.87)	-0.24 [-0.50, 0.03]	12.30(5.68)	-0.21 [-0.46, 0.04]
DERS	T1	-7.399	-2.154	567.79	0.032*	-0.16 [-0.60, 0.27]	95.22(17.15)	-0.22 [-0.48, 0.04]	98.31(20.40)	0.15 [-0.06, 0.36]
	T2	-4.960	-1.431	565.57	0.153	-0.03 [-0.43, 0.36]	91.77(19.41)	-0.38 [-0.63, -0.14]**#	92.42(21.70)	-0.11 [-0.32, 0.10]
	T3	-3.712	-1.058	566.07	0.290	0.03 [-0.36, 0.42]	86.36(19.81)	-0.65 [-0.90, -0.41]***#	85.77(21.71)	-0.40 [-0.61, -0.19]***#
	T4	-2.910	-0.778	563.02	0.437	0.07 [-0.32, 0.46]	87.18(21.32)	-0.59 [-0.84, -0.34]***#	85.78(21.02)	-0.40 [-0.63, -0.18]***#
	T5	-1.328	-0.326	566.37	0.745	0.15 [-0.28, 0.57]	84.94(20.53)	-0.71 [-0.99, -0.43]***#	81.96(19.95)	-0.58 [-0.84, -0.33]***#

ASRS ADHD Rating Scale; SAS Zung Self-rating Anxiety Scale; SDS Zung Self-rating Depression Scale; ERQ Emotion Regulation Questionnaire; DERS Difficulties in Emotion Regulation Scale; T1 = Week 4; T2 = Week 8; T3 = Post-treatment; T4 = 3-month follow-up; T5 = 12-month follow-up; *** p < 0.001 compared with baseline; ** p < 0.01 compared with baseline; * p < 0.05 compared with baseline

^# p-values were adjusted using the multiple comparisons using the false discovery rate (FDR) correction (Benjamini–Hochberg procedure)

All results are pooled estimates based on 20 multiply imputed datasets combined using Rubin’s rules

For the inattention subscale, no significant group × time interaction effects were observed across all assessment points. Between-group effect sizes at each assessment point were small, with no clear evidence of a stable group difference over time.

For the hyperactivity/impulsivity subscale, no significant group × time interaction effects were observed at any assessment point except T2, where a nominally significant group × time effect was found (b = -1.766, t = -2.066, p = 0.039); however, this result did not remain significant after FDR correction. The between-group effect size at T2 was small (d = -0.20 [-0.58, 0.17]).

Post hoc pairwise comparisons showed significant within-group reductions from baseline (T0) in both treatment groups. At the end of treatment (T3), large within-group effect sizes were observed for the ADHD-RS total score in both the DBT group (d = -1.33 [-1.59, -1.07], p < 0.001) and the CBT group (d = -1.17 [-1.40, -0.93], p < 0.001), all of which remained significant after FDR correction. Similar within-group reductions were observed for the inattention subscale (DBT: d = -1.47 [-1.79, -1.16], p < 0.001; CBT: d = -1.30 [-1.55, -1.05], p < 0.001) and the hyperactivity/impulsivity subscale (DBT: d = -0.90 [-1.12, -0.69], p < 0.001; CBT: d = -0.79 [-1.02, -0.56], p < 0.001). At the 3-month and 6-month follow-ups, these within-group improvements were generally maintained, with large within-group effects persisting across all outcomes. Detailed results are presented in Table 6; Fig. 2.

Fig. 2.

Trajectories of core symptoms across follow-up in the CBT and DBT groups. Note: ADHD-RS = Attention-Deficit/Hyperactivity Disorder Rating Scale; T0 = pre-treatment; T1 = week 4; T2 = week 8; T3 = post-treatment; T4 = 3-month follow-up; T5 = 6-month follow-up. Values are presented as pooled estimated means [95% CIs] based on 20 multiply imputed datasets combined using Rubin’s rules

Secondary outcomes

Emotional symptoms

LMM analyses revealed no significant group × time interaction effects for either SAS or SDS scores across all assessment points (Table 6). Between-group effect sizes were small throughout, with no evidence of stable group differences over time. For SAS scores, significant within-group reductions after FDR correction were observed from T3 to T5 in the CBT group (d = -0.38 [-0.64, -0.11] to -0.41 [-0.61, -0.20]), and at T3 in the DBT group (d = -0.34 [-0.57, -0.10]), whereas the reduction at T5 in the DBT group was nominal only. For SDS scores, significant within-group reductions after FDR correction were observed from T2 to T5 in the CBT group (d = -0.32 [-0.54, -0.09] to -0.44 [-0.70, -0.17]), whereas reductions in the DBT group were nominal and did not remain significant after correction. Detailed results are presented in Table 6; Fig. 3.

Fig. 3.

Trajectories of emotional symptoms across follow-up in the CBT and DBT groups. Note: SAS = Self-Rating Anxiety Scale; SDS = Self-Rating Depression Scale; T0 = pre-treatment; T1 = week 4; T2 = week 8; T3 = post-treatment; T4 = 3-month follow-up; T5 = 6-month follow-up. Values are presented as pooled estimated means [95% CIs] based on 20 multiply imputed datasets combined using Rubin’s rules

Emotion regulation

LMM analyses revealed no significant group × time interaction effects for ERQ-CR or ERQ-ES. For ERQ-CR scores, significant within-group increases after FDR correction were observed from T2 to T5 in the DBT group (d = 0.38 [0.10, 0.65] to 0.64 [0.34, 0.93]), and from T3 to T5 in the CBT group (d = 0.44 [0.18, 0.70] to 0.53 [0.24, 0.82]), whereas the increase at T2 in the CBT group was nominal only.

For ERQ-ES scores, no significant group × time interaction effects were observed across all assessment points, and neither group showed significant within-group changes after FDR correction.

For DERS scores, a nominally significant group × time interaction was observed at T1 (b = -7.399, t = -2.154, p = 0.032); however, this effect did not remain significant after FDR correction and the between-group effect size at T1 was small (d = -0.16 [-0.60, 0.27]). No further group × time interaction effects or between-group differences were observed at later assessments.

Post hoc comparisons showed significant within-group reductions after FDR correction were observed from T2 to T5 in the DBT group (d = -0.38 [-0.63, -0.14] to -0.71 [-0.99, -0.43]), and from T3 to T5 in the CBT group (d = -0.40 [-0.61, -0.19] to -0.58 [-0.84, -0.33]). Detailed results are presented in Table 6; Fig. 4.

Fig. 4.

Trajectories of emotion regulation across follow-up in the CBT and DBT groups. Note: ERQ = Emotion Regulation Questionnaire, CR = Cognitive reappraisal, ES = Expressive suppression; DERS = Difficulties in Emotion Regulation Scale; T0 = pre-treatment; T1 = week 4; T2 = week 8; T3 = post-treatment; T4 = 3-month follow-up; T5 = 6-month follow-up. Values are presented as pooled estimated means [95% CIs] based on 20 multiply imputed datasets combined using Rubin’s rules. *: P < 0.05; **: P < 0.01; ***: P < 0.001

QoL

No significant group × time interaction effects were observed for any of the four WHOQOL-BREF domains (Table 7). Between-group effect sizes were generally small across all assessment points, with no evidence of stable group differences over time. Post hoc within-group comparisons showed improvement in both groups, although the onset of significant change differed across domains.

Table 7.

Mixed-effects linear model results for QoL and global functioning across assessment points, with post Hoc group comparisons to baseline

		assessment point ×treatment type interaction					DBT (n = 49)		CBT (n = 49)
		b	t	df	p	cohen’s d [95%CI]	Mean (SD)	within-group effect size[95%CI]	Mean (SD)	within-group effect size[95%CI]
WHOQOL-BREF
Physical domain	T1	0.397	0.928	551.62	0.354	0.06 [-0.34, 0.47]	12.42(2.85)	0.31 [0.09, 0.54]**#	12.25(2.59)	0.16 [-0.05, 0.37]
	T2	-0.187	-0.432	532.10	0.666	-0.14 [-0.52, 0.24]	12.39(2.92)	0.30 [0.07, 0.52]**#	12.80(2.88)	0.34 [0.14, 0.54]**#
	T3	-0.284	-0.648	511.84	0.517	-0.19 [-0.61, 0.22]	13.03(2.57)	0.57 [0.33, 0.81]***#	13.54(2.72)	0.61 [0.39, 0.82]***#
	T4	0.090	0.197	486.99	0.844	-0.05 [-0.43, 0.34]	13.07(2.89)	0.55 [0.32, 0.78]***#	13.20(2.79)	0.48 [0.26, 0.70]***#
	T5	-0.201	-0.413	500.91	0.680	-0.17 [-0.60, 0.27]	13.28(2.72)	0.65 [0.39, 0.91]***#	13.70(2.34)	0.71 [0.45, 0.96]***#
Psychological domain	T1	-0.491	-1.130	539.09	0.259	-0.10 [-0.52, 0.32]	12.29(2.69)	0.13 [-0.10, 0.36]	12.56(2.68)	0.29 [0.08, 0.49]**#
	T2	-0.815	-1.862	539.49	0.063	-0.22 [-0.62, 0.18]	12.71(2.85)	0.28 [0.06, 0.50]*#	13.31(2.60)	0.55 [0.34, 0.76]***#
	T3	-0.714	-1.604	522.24	0.109	-0.18 [-0.59, 0.22]	13.20(2.60)	0.48 [0.25, 0.71]***#	13.70(2.78)	0.66 [0.45, 0.87]***#
	T4	-0.699	-1.455	461.90	0.146	-0.17 [-0.56, 0.21]	12.89(2.64)	0.36 [0.11, 0.61]**#	13.37(2.94)	0.54 [0.32, 0.76]***#
	T5	-0.894	-1.744	537.57	0.082	-0.28 [-0.72, 0.16]	13.42(2.27)	0.60 [0.31, 0.88]***#	14.10(2.50)	0.83 [0.58, 1.08]***#
Social relation domain	T1	-0.352	-0.656	563.69	0.512	-0.10 [-0.52, 0.32]	13.20(3.17)	0.18 [-0.04, 0.40]	13.49(2.96)	0.28 [0.06, 0.51]*#
	T2	-0.822	-1.499	535.55	0.134	-0.25 [-0.66, 0.16]	13.27(3.22)	0.20 [-0.03, 0.43]	14.04(2.99)	0.44 [0.22, 0.67]***#
	T3	-0.250	-0.451	540.53	0.652	-0.06 [-0.46, 0.33]	13.82(2.73)	0.39 [0.15, 0.63]**#	14.02(3.54)	0.41 [0.20, 0.62]***#
	T4	-0.463	-0.788	552.86	0.431	-0.14 [-0.55, 0.27]	13.78(2.95)	0.36 [0.11, 0.61]**#	14.19(2.84)	0.50 [0.25, 0.75]***#
	T5	-0.634	-0.981	546.02	0.327	-0.21 [-0.64, 0.22]	13.81(2.59)	0.39 [0.10, 0.69]**#	14.39(2.95)	0.55 [0.28, 0.82]***#
Environmental domain	T1	-0.563	-1.454	551.20	0.147	-0.27 [-0.68, 0.14]	13.57(2.16)	0.01 [-0.22, 0.25]	14.25(2.77)	0.21 [0.02, 0.40]*
	T2	-0.308	-0.775	517.10	0.439	-0.17 [-0.57, 0.24]	13.97(2.24)	0.19 [-0.06, 0.43]	14.39(2.75)	0.26 [0.07, 0.46]**#
	T3	-0.105	-0.260	537.42	0.795	-0.08 [-0.44, 0.28]	14.34(2.27)	0.35 [0.10, 0.59]**#	14.56(3.15)	0.30 [0.11, 0.49]**#
	T4	-0.389	-0.860	511.00	0.390	-0.20 [-0.61, 0.20]	14.36(2.31)	0.35 [0.09, 0.62]**#	14.87(2.61)	0.44 [0.21, 0.68]***#
	T5	-0.379	-0.737	536.48	0.462	-0.21 [-0.66, 0.23]	14.50(1.97)	0.44 [0.11, 0.78]**#	14.99(2.58)	0.49 [0.23, 0.76]***#
GSES	T1	-0.350	-0.365	563.02	0.716	-0.32 [-0.69, 0.06]	24.22(6.80)	0.02 [-0.16, 0.21]	26.40(7.02)	0.07 [-0.12, 0.26]
	T2	0.407	0.420	561.87	0.675	-0.24 [-0.68, 0.19]	25.08(5.64)	0.16 [-0.05, 0.36]	26.51(6.16)	0.10 [-0.11, 0.30]
	T3	-0.994	-1.007	557.95	0.314	-0.52 [-0.97, -0.06]	25.40(4.62)	0.22 [0.00, 0.44]	28.23(6.24)	0.35 [0.15, 0.56]***#
	T4	-0.234	-0.219	555.63	0.826	-0.33 [-0.71, 0.06]	25.75(6.18)	0.25 [0.03, 0.47]*	27.82(6.49)	0.29 [0.07, 0.51]*#
	T5	-0.238	-0.201	558.89	0.841	-0.34 [-0.75, 0.06]	26.62(5.75)	0.39 [0.14, 0.64]**#	28.69(6.30)	0.42 [0.17, 0.67]**#
SDS	T1	0.813	0.815	568.16	0.416	0.04 [-0.38, 0.46]	15.76(4.37)	-0.06 [-0.34, 0.21]	15.56(5.42)	-0.20 [-0.43, 0.04]
	T2	1.245	1.237	563.51	0.217	0.13 [-0.29, 0.55]	15.35(4.64)	-0.14 [-0.41, 0.13]	14.72(5.05)	-0.35 [-0.59, -0.10]**#
	T3	0.377	0.371	560.87	0.711	-0.04 [-0.44, 0.35]	13.18(5.44)	-0.53 [-0.78, -0.27]***#	13.42(5.02)	-0.58 [-0.83, -0.33]***#
	T4	2.317	2.176	555.62	0.030*#	0.31 [-0.08, 0.70]	14.49(5.77)	-0.28 [-0.54, -0.02]*	12.78(5.12)	-0.68 [-0.94, -0.42]***#
	T5	1.556	1.367	554.65	0.172	0.19 [-0.26, 0.64]	12.40(4.77)	-0.71 [-1.01, -0.40]***#	11.45(5.16)	-0.92 [-1.19, -0.64]***#

GSES General Self-Efficacy Scale; WHOQOL- BREF Brief Version of the World Health Organization Quality of Life; SDS Sheehan Disability Scale; T1 = Week 4; T2 = Week 8; T3 = Post-treatment; T4 = 3-month follow-up; T5 = 12-month follow-up; *** p < 0.001 compared with baseline; ** p < 0.01 compared with baseline; * p < 0.05 compared with baseline

# p-values were adjusted using the multiple comparisons using the false discovery rate (FDR) correction (Benjamini–Hochberg procedure)

All results are pooled estimates based on 20 multiply imputed datasets combined using Rubin’s rules

For the physical domain, significant within-group gains after FDR correction were observed from T1 to T5 in the DBT group (d = 0.31 [0.09, 0.54] to 0.65 [0.39, 0.91]) and from T2 to T5 in the CBT group (d = 0.34 [0.14, 0.54] to 0.71 [0.45, 0.96]). For the psychological domain, significant within-group improvements after FDR correction were observed from T2 to T5 in the DBT group (d = 0.28 [0.06, 0.50] to 0.60 [0.31, 0.88]) and from T1 to T5 in the CBT group (d = 0.29 [0.08, 0.49] to 0.83 [0.58, 1.08]). For social relationships, significant within-group gains after FDR correction were observed from T3 to T5 in the DBT group (d = 0.39 [0.15, 0.63] to 0.39 [0.10, 0.69]) and from T1 to T5 in the CBT group (d = 0.28 [0.06, 0.51] to 0.55 [0.28, 0.82]). For the environmental domain, significant within-group improvements after FDR correction were observed from T3 to T5 in the DBT group (d = 0.35 [0.10, 0.59] to 0.44 [0.11, 0.78]) and from T2 to T5 in the CBT group (d = 0.26 [0.07, 0.46] to 0.49 [0.23, 0.76]). Detailed results are presented in Table 7; Fig. 5.

Fig. 5.

Trajectories of quality of life across follow-up in the CBT and DBT groups. Note: WHOQOL-BREF = the Brief Version of the World Health Organization Quality of Life Scale; WHOQOL-BREF domains — PH = Physical Health; PSY = Psychological Health; SR = Social Relationships; ENV = Environment. T0 = pre-treatment; T1 = week 4; T2 = week 8; T3 = post-treatment; T4 = 3-month follow-up; T5 = 6-month follow-up. Values are presented as pooled estimated means [95% CIs] based on 20 multiply imputed datasets combined using Rubin’s rules

Global functioning

For self-efficacy measured by GSES scores, LMM analyses revealed no significant group × time interaction effects at any assessment point (Table 7). Between-group effect sizes were small, with no evidence of stable group differences over time. Post hoc comparisons showed significant within-group improvements in both treatment groups, with significance after FDR correction observed at T5 in the DBT group (d = 0.39 [0.14, 0.64]) and from T3 to T5 in the CBT group (d = 0.35 [0.15, 0.56] to 0.42 [0.17, 0.67]); the T4 change in the DBT group was only nominally significant.

For functional impairment measured by SDS scores, a significant group × time effect was observed at T4 (b = 2.317, t = 2.176, p = 0.030), favoring CBT, and this effect remained significant after FDR correction. The between-group effect size at T4 was small (d = 0.31 [-0.08, 0.70]), and the between-group contrast was not statistically significant. Post hoc pairwise comparisons showed significant within-group reductions after FDR correction were observed from T3 to T5 in the DBT group (d = -0.53 [-0.78, -0.27] to -0.71 [-1.01, -0.40]) and from T2 to T5 in the CBT group (d = -0.35 [-0.59, -0.10] to -0.92 [-1.19, -0.64]), whereas the reduction at T4 in the DBT group was only nominally significant. Detailed results are presented in Table 7; Fig. 6.

Fig. 6.

Trajectories of self-reported global functioning across follow-up in the CBT and DBT groups. Note: GSES = General Self-Efficacy Scale, SDS=Sheehan Disability Scale; T0 = pre-treatment; T1 = week 4; T2 = week 8; T3 = post-treatment; T4 = 3-month follow-up; T5 = 6-month follow-up. Values are presented as pooled estimated means [95% CIs] based on 20 multiply imputed datasets combined using Rubin’s rules. *: P < 0.05; **: P < 0.01; ***: P < 0.001

EF

LMM analyses revealed no significant group × time interaction effects for BRIEF-A total scores, the BRI, or the MI across all assessment points (Table 8). Between-group effect sizes were small throughout, with no evidence of stable group differences over time.

Table 8.

Mixed-Effects linear model results on executive function, with post Hoc group comparisons to baseline

		assessment point ×treatment type interaction					DBT (n = 49)		CBT (n = 49)
		b	t	df	p	cohen’s d [95%CI]	Mean (SD)	within-group effect size[95%CI]	Mean (SD)	within-group effect size[95%CI]
BRIEF-A total
BRI	T1	-0.732	-0.350	564.99	0.726	-0.06 [-0.47, 0.35]	57.00(12.95)	0.00 [-0.23, 0.22]	57.80(12.77)	0.05 [-0.17, 0.27]
	T2	0.199	0.094	563.98	0.925	0.01 [-0.36, 0.38]	53.53(15.07)	-0.25 [-0.46, -0.04]*#	53.40(12.25)	-0.29 [-0.52, -0.06]*#
	T3	-0.167	-0.078	563.67	0.938	-0.02 [-0.44, 0.40]	49.86(11.65)	-0.58 [-0.82, -0.34]***#	50.09(12.11)	-0.55 [-0.78, -0.32]***#
	T4	0.225	0.097	564.07	0.922	0.01 [-0.39, 0.42]	49.44(12.45)	-0.59 [-0.85, -0.34]***#	49.28(11.28)	-0.63 [-0.89, -0.37]***#
	T5	-1.362	-0.530	564.68	0.597	-0.12 [-0.52, 0.28]	48.18(11.60)	-0.71 [-1.00, -0.42]***#	49.61(12.55)	-0.58 [-0.86, -0.30]***#
MI	T1	-3.026	-1.052	569.02	0.293	0.01 [-0.41, 0.43]	86.39(14.54)	-0.27 [-0.54, 0.01]	86.23(15.34)	-0.06 [-0.31, 0.20]
	T2	-2.369	-0.818	568.43	0.414	0.05 [-0.33, 0.43]	81.73(18.56)	-0.51 [-0.76, -0.27]***#	80.91(14.08)	-0.41 [-0.68, -0.14]**#
	T3	-1.972	-0.673	566.99	0.501	0.07 [-0.30, 0.45]	76.78(16.11)	-0.88 [-1.15, -0.62]***#	75.57(16.79)	-0.71 [-0.96, -0.46]***#
	T4	-2.654	-0.858	561.58	0.391	0.03 [-0.36, 0.42]	75.78(17.18)	-0.92 [-1.18, -0.64]***#	75.25(15.35)	-0.76 [-1.04, -0.48]***#
	T5	-5.118	-1.537	566.02	0.125	-0.12 [-0.54, 0.30]	73.99(16.87)	-1.04 [-1.34, -0.74]***#	75.92(15.11)	-0.73 [-1.03, -0.42]***#
GEC	T1	-3.758	-0.801	568.72	0.424	-0.02 [-0.44, 0.40]	143.39(25.45)	-0.16 [-0.41, 0.10]	144.03(25.18)	-0.01 [-0.26, 0.24]
	T2	-2.171	-0.459	568.52	0.646	0.03 [-0.34, 0.40]	135.26(32.36)	-0.42 [-0.64, -0.19]***#	134.31(24.17)	-0.39 [-0.65, -0.13]**#
	T3	-2.139	-0.447	567.68	0.655	0.04 [-0.35, 0.43]	126.63(25.97)	-0.81 [-1.07, -0.55]***#	125.65(27.34)	-0.69 [-0.93, -0.44]***#
	T4	-2.429	-0.480	567.44	0.632	0.03 [-0.37, 0.42]	125.22(27.90)	-0.83 [-1.10, -0.57]***#	124.53(25.03)	-0.76 [-1.03, -0.49]***#
	T5	-6.481	-1.179	567.39	0.239	-0.13 [-0.54, 0.28]	122.17(26.37)	-0.98 [-1.28, -0.68]***#	125.53(25.77)	-0.71 [-1.01, -0.42]***#
TMT
TMT-A	T3	2.11	1.49	373.27	0.137	0.42 [0.05, 0.80]	23.31 (10.29)	-0.20 [-0.39, -0.01]*	19.79 (5.64)	-0.59 [-0.87, -0.31]***#
	T4	1.07	0.69	367.03	0.493	0.45 [-0.05, 0.95]	20.57 (6.44)	-0.56 [-0.81, -0.30]***#	18.10 (4.45)	-0.88 [-1.21, -0.56]***#
	T5	1.48	0.84	365.25	0.401	0.46 [0.05, 0.86]	20.79 (7.53)	-0.51 [-0.78, -0.23]***#	17.90 (4.85)	-0.90 [-1.26, -0.53]***#
TMT-B	T3	5.35	0.95	377.29	0.343	0.29 [-0.08, 0.66]	60.02 (35.72)	-0.11 [-0.34, 0.12]	51.65 (19.69)	-0.30 [-0.56, -0.04]*#
	T4	6.47	1.07	375.95	0.285	0.55 [0.02, 1.08]	51.17 (20.86)	-0.47 [-0.78, -0.16]**#	41.68 (12.27)	-0.68 [-0.97, -0.38]***#
	T5	2.55	0.38	376.14	0.702	0.36 [-0.16, 0.88]	46.57 (16.86)	-0.68 [-1.04, -0.31]***#	41.00 (14.06)	-0.69 [-1.02, -0.37]***#
SS Total	T3	-1.71	-2.57	362.63	0.010*	-0.44 [-0.88, 0.01]	17.95 (2.58)	-0.08 [-0.38, 0.23]	19.05 (2.43)	0.57 [0.21, 0.94]**#
	T4	-2.15	-3.11	318.34	0.002**#	-0.56 [-0.99, -0.14]	18.28 (2.58)	0.03 [-0.29, 0.35]	19.81 (2.88)	0.80 [0.46, 1.15]***#
	T5	-0.94	-1.36	360.09	0.174	-0.12 [-0.52, 0.29]	19.63 (2.79)	0.47 [0.15, 0.78]**#	19.96 (2.92)	0.85 [0.50, 1.20]***#
SCWT-CW	T3	-2.34	-1.4	368.9	0.163	-0.43 [-0.84, -0.02]	48.63 (11.06)	0.19 [-0.02, 0.40]	53.13 (9.95)	0.40 [0.19, 0.61]***#
	T4	-0.79	-0.43	364.3	0.669	-0.27 [-0.66, 0.11]	50.04 (11.86)	0.31 [0.08, 0.54]**#	52.99 (9.75)	0.39 [0.16, 0.63]**#
	T5	1.48	0.71	371.07	0.481	-0.07 [-0.48, 0.34]	52.34 (12.40)	0.50 [0.25, 0.75]***#	53.02 (7.46)	0.43 [0.14, 0.73]**#
CPT-IP
2-digit dʹ	T3	-0.19	-1.14	304.57	0.255	-0.16 [-0.54, 0.22]	3.62 (0.77)	-0.02 [-0.33, 0.28]	3.73 (0.64)	0.27 [-0.10, 0.64]
	T4	-0.09	-0.5	270.97	0.615	-0.03 [-0.59, 0.52]	3.81 (0.47)	0.30 [-0.12, 0.72]	3.83 (0.54)	0.45 [0.03, 0.88]*
	T5	-0.25	-1.28	331.05	0.201	-0.24 [-0.65, 0.16]	3.52 (0.83)	-0.15 [-0.51, 0.20]	3.69 (0.59)	0.21 [-0.23, 0.66]
3-digit dʹ	T3	-0.27	-1.41	304.4	0.159	-0.28 [-0.71, 0.16]	3.26 (0.83)	0.08 [-0.25, 0.41]	3.47 (0.68)	0.45 [0.08, 0.81]*
	T4	-0.14	-0.67	208.8	0.504	-0.10 [-0.56, 0.36]	3.29 (0.80)	0.12 [-0.26, 0.49]	3.37 (0.81)	0.29 [-0.07, 0.65]
	T5	-0.39	-1.8	285.29	0.073	-0.36 [-0.77, 0.05]	2.99 (0.98)	-0.24 [-0.60, 0.12]	3.32 (0.84)	0.22 [-0.14, 0.58]
4-digit dʹ	T3	-0.02	-0.09	289	0.925	-0.32 [-0.77, 0.12]	2.37 (0.79)	0.34 [0.03, 0.65]*	2.64 (0.89)	0.34 [0.02, 0.65]*
	T4	-0.01	-0.04	221.79	0.968	-0.27 [-0.69, 0.14]	2.46 (0.95)	0.41 [0.09, 0.73]*	2.72 (0.97)	0.40 [0.09, 0.71]*
	T5	-0.14	-0.65	197.05	0.515	-0.43 [-0.88, 0.03]	2.26 (0.93)	0.19 [-0.16, 0.55]	2.66 (0.92)	0.35 [0.01, 0.69]*

BRIEF-A Behavior Rating Inventory of Executive Function-Adult Version; BRI Behavioral Regulation Index; MI Metacognition Index; GEC Global Executive Composite; TMT Trail Making Test; SS Spatial span test; SCWT-CW Stroop Color and Word Test-Color Word Test; CPT-IP Continuous Performance Test-Identical Pairs; T1 = Week 4; T2 = Week 8; T3 = Post-treatment; T4 = 3-month follow-up; T5 = 12-month follow-up; *** p < 0.001 compared with baseline; ** p < 0.01 compared with baseline; * p < 0.05 compared with baseline

# p-values were adjusted using the multiple comparisons using the false discovery rate (FDR) correction (Benjamini–Hochberg procedure)

All results are pooled estimates based on 20 multiply imputed datasets combined using Rubin’s rules

For BRIEF-A total scores, post hoc pairwise comparisons showed significant and sustained within-group improvements after FDR correction from T2 to T5 in both treatment groups, with effect sizes ranging from d = -0.42 [-0.64, -0.19] to -0.98 [-1.28, -0.68] in the DBT group and from d = -0.39 [-0.65, -0.13] to -0.76 [-1.03, -0.49] in the CBT group. Comparable within-group patterns were observed for BRI and MI scores. Detailed results are presented in Table 8; Fig. 7.

Fig. 7.

Trajectories of self-reported executive function across follow-up in the CBT and DBT groups. Note: BRIEF-A = Behavior Rating Inventory of Executive Function—Adult Version. BRI=Behavioral Regulation Index, MI=Metacognition Index, GEC=Global Executive Composite; T0 = pre-treatment; T1 = week 4; T2 = week 8; T3 = post-treatment; T4 = 3-month follow-up; T5 = 6-month follow-up. Values are presented as pooled estimated means [95% CIs] based on 20 multiply imputed datasets combined using Rubin’s rules

For laboratory-based EF tasks, LMM analyses showed no significant group × time interactions for most measures, including TMT-A, TMT-B, SCWT-CW, and CPT-IP two-, three-, and four-digit tasks. Between-group effect sizes were generally small, with no evidence of stable group differences over time. A significant between-group difference was observed only for SS Total at T3 (b = -1.71, t = -2.57, p = 0.010) and T4 (b = -2.15, t = -3.11, p = 0.002), favoring CBT, and the effect at T4 remained significant after FDR correction. The between-group effect sizes were small to moderate at T3 (d = -0.44 [-0.88, 0.01]) and moderate at T4 (d = -0.56 [-0.99, -0.14]).

Although no significant between-group differences were observed at most time points across most laboratory-based EF measures, post hoc pairwise comparisons showed that significant within-group improvements emerged earlier in the CBT group for TMT, SCWT-CW, and SS Total. In the CBT group, significant within-group improvements after FDR correction were observed from T3 to T5 for TMT-A (d = -0.59 to -0.90), TMT-B (d = -0.30 to -0.69), and SCWT-CW (d = 0.40 to 0.43), whereas corresponding improvements in the DBT group were observed from T4 to T5 for TMT-A (d = -0.51 to -0.56), TMT-B (d = -0.47 to -0.68), and SCWT-CW (d = 0.31 to 0.50). Detailed results are presented in Table 8; Figs. 8 and 9.

Fig. 8.

Trajectories of laboratory-based executive function across follow-up in the CBT and DBT groups: TMT and SS. Note: TMT = Trail Making Test; SS = Spatial Span Test; T0 = pre-treatment; T1 = week 4; T2 = week 8; T3 = post-treatment; T4 = 3-month follow-up; T5 = 6-month follow-up. Values are presented as pooled estimated means [95% CIs] based on 20 multiply imputed datasets combined using Rubin’s rules. *: P < 0.05; **: P < 0.01; ***: P < 0.001

Fig. 9.

Trajectories of laboratory-based executive function across follow-up in the CBT and DBT groups: SCWT and CPT. Note: SCWT = Stroop Color and Word Test - Color Word Test; CPT = Continuous Performance Test-Identical Pairs; T0 = pre-treatment; T1 = week 4; T2 = week 8; T3 = post-treatment; T4 = 3-month follow-up; T5 = 6-month follow-up. Values are presented as pooled estimated means [95% CIs] based on 20 multiply imputed datasets combined using Rubin’s rules

To improve consistency with the trial registration, brief results for additional prespecified self-report secondary outcomes that were completed and analyzed in the trial are provided in Supplementary Table S4.

Discussion

This study is the first RCT to directly compare the efficacy of DBT and CBT for adults with ADHD in China. Across multiple symptom and functional domains, both interventions were associated with clinically meaningful improvement over time. For the primary outcomes, no significant between-group differences were observed for the ADHD-RS total score or the inattention subscale, and the nominally significant time-specific effect observed for the hyperactivity/impulsivity subscale did not remain significant after FDR correction. Across secondary outcomes, between-group differences were generally small, time-specific, and not consistently maintained over follow-up. The only interaction effect that remained significant after FDR correction was observed for functional impairment at T4; however, the corresponding between-group effect size was small, and the between-group contrast was not statistically significant. For laboratory-based EF tasks, most measures showed no significant group × time interactions. Although CBT showed a relative advantage on the spatial span task at post-treatment and the 3-month follow-up, no stable between-group differences were observed for most other measures. Taken together, the present findings suggest that both interventions were associated with improvement over time, with neither treatment demonstrating broad superiority over the other through the 6-month follow-up.

Primary outcomes

Regarding the primary outcome, although we hypothesized that DBT would demonstrate non-inferiority to CBT, formal non-inferiority was not established, as the confidence intervals were not fully contained within the prespecified non-inferiority margin (d = 0.40). Accordingly, the present findings should not be interpreted as evidence of equivalence between the two interventions. Nevertheless, both treatments showed large and sustained within-group reductions in core ADHD symptoms over follow-up. These findings are broadly consistent with previous RCTs, in populations with BPD [45] and eating disorder [46], in which both CBT and DBT have been associated with clinical improvement in the same diagnostic context.

This limited precision in estimating the between-group effect may partly reflect the sample size of the present trial, attrition over follow-up [77], and the substantial clinical heterogeneity of adults with ADHD [78]. In addition, variability in individual response to psychotherapy may also have contributed to the imprecision of the average between-group estimates [79].

Secondary outcomes

Contrary to our expectation of detecting potential differences between DBT and CBT across secondary outcomes, between-group differences were limited in magnitude and were not consistently maintained over time. Even when time-specific signals were observed, they did not translate into robust or sustained between-group differences. Overall, this pattern suggests that neither intervention demonstrated broad superiority across most secondary domains.

The lack of significant group differences in the primary and most secondary outcomes may be related to several factors. First, there is considerable theoretical overlap between the two interventions. Most psychological treatments for ADHD are rooted in cognitive-behavioral paradigms, which posit that neurobiological deficits in attention, EF, and inhibitory control lead to recurrent failures and frustration in adulthood. These difficulties contribute to negative self-perceptions, low self-esteem, emotion dysregulation, and maladaptive behaviors, ultimately impairing social functioning [20, 80]. Within this framework, both treatments primarily target symptom management and behavioral regulation [81]. Second, both interventions incorporate multiple treatment components that may operate through partially overlapping mechanisms [82]. CBT emphasizes cognitive restructuring and compensatory skills such as organization, planning, and distraction management. DBT, as a “third-wave” form of cognitive-behavioral therapy, builds upon the CBT framework while incorporating specific modules on emotion regulation, mindfulness, interpersonal effectiveness, and distress tolerance [57]. Although these emphases differ, they may still converge on some shared clinical processes, which could partly account for the limited between-group differences observed in the present study. The dialectical approach aims to balance acceptance and change by integrating conflicting perspectives, thereby reducing dichotomous or “black-and-white” thinking [41]—one of the core maladaptive cognitive patterns targeted in CBT. Moreover, CBT’s cognitive module, the “thought-emotion cycle,” promotes self-monitoring and recognition of emotional and behavioral cues, which enhances impulse control and emotion regulation [57]. Similarly, the distraction management module encourages participants to identify their optimal attention span and practice delayed response strategies while maintaining “non-judgmental awareness,” a central element of mindfulness. Third, common therapeutic factors such as therapeutic alliance, patient motivation and expectations, and group cohesion may also have contributed to symptom improvement in both treatments [83, 84]. Treatment adherence—reflected in attendance and frequency of skill application—and perceived acceptability may also influence treatment efficacy [85]. Offering multiple therapeutic options for adults with ADHD may help improve person-treatment fit and support treatment engagement.

An interesting finding was the relative advantage of CBT on the spatial span task. However, in the absence of consistent between-group differences on other laboratory-based EF measures, this finding should be interpreted cautiously. It may indicate a more specific advantage of CBT for spatial working memory–related performance, rather than a broader advantage for executive functioning as a whole.

Two possible explanations may account for this pattern. First, laboratory-based EF tasks assess relatively circumscribed cognitive processes and may not capture the full range of executive functioning. EF is commonly conceptualized as encompassing both “cool” and “hot” domains, with cool EF referring primarily to cognitive flexibility, working memory, and inhibitory control, and hot EF referring more to regulation under emotionally or motivationally salient conditions [86]. The tasks used in the present study mainly assessed cool EF and did not adequately capture hot EF processes. This may have reduced sensitivity to potential differences between the two interventions in domains more closely related to emotion regulation.

Second, discrepancies between self-report and laboratory-based assessments may reflect differences in the specific dimensions of EF being measured. Laboratory tasks primarily assess the capacity to perform structured cognitive operations under controlled conditions, whereas ecological measures of EF provide a more accurate reflection of everyday functioning and self-regulatory capacity. The parallel pattern of improvement on self-report executive function measures in both groups suggests that differences in everyday, real-world executive functioning may be limited [87].

Strengths and Limitations

To our knowledge, this is the first trial in China to directly compare the efficacy of DBT and CBT for adults with ADHD. By evaluating multiple symptom and functional domains and incorporating a six-month follow-up, this study contributes comparative evidence to the literature on psychosocial interventions for adult ADHD and may help inform future work on culturally adapted interventions in China.

Despite these strengths, several limitations should be considered when interpreting the findings. First, the sample largely comprised individuals with relatively high IQs, which may limit the generalizability of the results to the broader adult ADHD population [88]. In addition, participants with severe or unstable clinical conditions were excluded, which may further limit the applicability of the findings to more complex or clinically fluctuating patients seen in routine practice. Second, blinding of therapists was not feasible, which may have introduced expectancy bias. In addition, because only one therapist pair was assigned to each intervention, therapist effects were completely confounded with treatment effects. Therefore, any observed between-group differences may reflect not only treatment-specific effects but also therapist-specific factors such as therapeutic characteristics, style, and competence [89–91]. This represents a fundamental design constraint on causal inference regarding the relative efficacy of the two interventions. Third, external factors such as lifestyle changes, medication adjustments, or workplace demands may have influenced symptom fluctuations, making it difficult to isolate the specific effects of the interventions. Fourth, treatment engagement and adherence also warrant consideration. Although most participants remained engaged through the six-month follow-up, formal quantitative data on homework completion, treatment acceptability and frequency of skill use in daily life were not collected [83, 85]. Fifth, not all prespecified secondary outcomes were included in the present report, and the extensive repeated assessments may have increased participant burden. Although the status of all prespecified secondary outcomes and additional supplementary results are now provided to improve transparency, this discrepancy between trial registration and reporting, together with the assessment burden, should be considered when interpreting the findings. Finally, the laboratory assessments primarily measured “cool” EFs and did not sufficiently capture “hot” executive domains related to emotion regulation, which may have obscured potential between-group differences in affectively relevant executive control [86].

Future research directions

Future research should pursue several important directions. First, given the substantial heterogeneity of ADHD, further studies are needed to examine individual characteristics associated with treatment response and to clarify principles of treatment matching, which may help inform more individualized intervention strategies [15]. Second, it remains important to identify the active components of psychosocial interventions and to clarify how specific treatment components contribute to change over time [34, 82]. Although the present study did not detect significant between-group differences for most outcomes, some symptom and functional domains appeared to differ in the timing or pattern of improvement, which may warrant further investigation into potential treatment mechanisms. Third, future trials should more explicitly consider common therapeutic factors, therapist effects, patient engagement, and long-term adherence in order to better distinguish treatment-specific effects from other influences on outcome [85, 89–91]. Fourth, the rapid advancement of digital mental health technologies offers opportunities to increase accessibility, particularly for individuals in resource-limited areas or those unable to participate in face-to-face therapy [92]. Finally, future work may also benefit from considering strengths-based perspectives, including positive psychological resources such as resilience and self-efficacy, to broaden understanding of adaptive functioning in adults with ADHD [81].

Conclusion

This RCT provides novel evidence on the comparative efficacy of DBT and CBT for adults with ADHD in China. The findings indicate that both interventions were associated with clinically meaningful improvement across multiple symptom and functional domains over the 6-month follow-up. CBT showed a relative advantage on a spatial working memory task, whereas most other between-group differences were generally small and not consistently maintained over follow-up. However, because the confidence intervals for the primary outcome were not fully contained within the prespecified non-inferiority margin, formal non-inferiority of DBT relative to CBT could not be established, and the present findings should not be interpreted as evidence of equivalence between the two interventions. DBT may represent a potentially useful psychotherapeutic option for adults with ADHD. Future research in larger and more diverse samples is needed to obtain more precise estimates of relative treatment effects and to further examine mechanisms of change and long-term maintenance processes.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADHD

Attention-Deficit/Hyperactivity Disorder

ADHD-RS

ADHD Rating Scale

CBT

Cognitive Behavioral Therapy

DBT

Dialectical Behavior Therapy

DERS

Difficulties in Emotion Regulation Scale

EF

Executive Function

ERQ

Emotion Regulation Questionnaire

GSES

General Self-Efficacy Scale

ITT

Intention-to-Treat

LMM

Linear Mixed Model

QoL

Quality of Life

Author contributions

Li-Qian Zhang: Conceptualization, Writing – original draft, Writing – review & editing, Methodology, Formal analysis. Mei-Rong Pan: Conceptualization, Methodology, Investigation, Funding acquisition, Writing – review & editing. Min Domin: Investigation, Methodology, Supervision. Xin Kong: Formal analysis, Data curation, Investigation. Zi-Yan Yu: Formal analysis, Data curation, Investigation. Shi-Yu Zhang: Data Curation, Investigation. Hai-Mei Li: Investigation, Data curation. Lu Liu: Supervision, Methodology, Project administration. Yu-Feng Wang: Validation, Supervision, Project administration. Qiu-Jin Qian: Conceptualization, Investigation, Writing – review & editing.

Funding

This work was supported by the Key Attending Psychiatrist Program of Peking University Sixth Hospital (BDLYYLZL2023-03).

Data availability

Data available from the corresponding authors upon reasonable request.

Declarations

Ethics approval and consent to participate

This RCT was conducted in accordance with the CONSORT 2010 guidelines. The study protocol was registered at Peking University Sixth Hospital (registration ID: ChiCTR2300072075). The study adhered to the principles of the Declaration of Helsinki with the Medical Research Involving Human Subjects Act (WMO)., and ethical approval was obtained from the Ethics Committee of Peking University Sixth Hospital. All participants provided written informed consent prior to enrollment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.