Introduction

Non-suicidal self-injury (NSSI) behavior is a conscious infliction of harm on one’s body tissues [1], which is usually a coping strategy for dysfunctional emotional regulation. It is associated with higher risks of hospitalization, emergency visits, and suicidal thoughts and behaviors [2, 3], which leads to considerable family and individual burdens [4]. The lifetime prevalence for NSSI is approximately 22% in adolescents globally, with comorbid mental disorders doubling this risk [5, 6]. The prevalence of NSSI is particularly higher among adolescents with major depressive disorder (MDD), with female adolescents exhibiting a much higher prevalence than male adolescents [7]. Over the past decade, considerable investigations into various psychotropic medications and psychotherapies have been made. However, in a recent meta-analysis, the evidence for the pharmacological treatment of NSSI was reviewed, concluding that its effectiveness was uncertain [8]. Moreover, NSSI behaviors have shown promising improvements after long-term structured psychological interventions [9]. Considering that systemic psychotherapy requires extensive time, and skilled psychotherapists are scarce in underprivileged regions [10], there is an urgent need for efficient and readily accessible therapies for NSSI among depressed adolescents.

NSSI often occurs in youths with inconsistent maturity in the limbic and prefrontal cortices when undergoing acute stress [11, 12]. Patients with NSSI commonly exhibit lower amygdala-frontal connectivity, which is associated with NSSI episodes, reflecting challenges in regulating negative emotions [13]. In a cognitive interference task study, patients with NSSI showed decreased dorsolateral prefrontal cortex (DLPFC) activation but increased cingulate gyrus activation, which is linked to emotional dysregulation and higher impulsivity [14].

Repetitive transcranial magnetic stimulation (rTMS) is a noninvasive neuromodulation therapy, which is an auxiliary treatment for mental disorders [15]. Previous data showed the potential efficacy of rTMS in reducing NSSI. Findings from an open-label study revealed that an adjunctive 10 Hz one-session rTMS per day on the LDLPFC for 4 weeks significantly improved NSSI frequency in adolescents with MDD [16]. Intermittent theta burst stimulation (iTBS), a novel patterned form of rTMS, induces long-term potentiation-like plasticity, facilitating long-lasting cortical excitability [17]. Stimulating the LDLPFC with iTBS for 4–6 weeks has shown comparable efficacy for treatment-resistant depression (TRD). Compared with high-frequency rTMS (HF-rTMS), iTBS improves treatment efficiency by compressing the time course of every single stimulation session from 37.5 min to 3 min [18]. Stanford Neuromodulation Therapy (SNT), which delivers 10 sessions of prolonged iTBS daily for 5 days, has shown favorable tolerability, robust efficacy of high response, and remission rate on TRD within 1 week [19, 20]. The United States Food and Drug Administration (FDA) approved this protocol for treating TRD in September, 2022. iTBS makes it possible to deliver higher pulse doses with accelerated or prolonged patterns than HF-rTMS during a short period, contributing to the rapid remission of MDD [18, 21, 22]. Furthermore, iTBS targeting the DLPFC could regulate internal emotions and reduce suicidal ideation in adolescents with MDD [23], and neuromodulation that activates the prefrontal cortex (PFC) could help with behavioral addiction [24]. Therefore, administering iTBS to the LDLPFC may improve NSSI by regulating negative emotions and addictive behavior patterns.

In this study, we hypothesized that iTBS stimulation would be feasible and effective in reducing NSSI symptoms, suicidal ideation, and depressive and anxiety symptoms. According to our pilot study, adolescents experienced difficulty tolerating headaches related to stimulations of 10 sessions a day in the SNT protocol. In addition, previous studies on adolescents did not provide iTBS exceeding five sessions a day [25,26,27,28]. Thus, we lowered the daily amount and intensity of stimulation with our accelerated prolonged iTBS protocol, comprising five daily sessions for 5 consecutive days.

Patients and methods

A randomized, sham-controlled trial was conducted at the Second Xiangya Hospital of Central South University between July 1, 2022, and November 19, 2023. This study was approved by the National Clinical Medical Research Center Ethics Committee of the Second Xiangya Hospital of Central South University and registered at ClinicalTrials.gov (NCT05384405). Participants assented to their inclusion in the study, and written informed consent was obtained from their legal guardians. This study adhered to the Consolidated Standards of Reporting Trials reporting requirements.

Participants

We recruited patients from the inpatient and outpatient psychiatric departments. Eligibility criteria were adolescents aged ≥12 and <18 years, a primary diagnosis of current MDD assessed using the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM-5), at least three self-harm behaviors that met the DSM-5 criteria for NSSI in the previous month, a Hamilton Depression Rating Scale 17-item (HAM-D17) score ≥18, and at least one caregiver to supervise the patient for 3 months.

The exclusion criteria were current substance abuse, comorbid bipolar disorder (Young Mania Rating Scale score >12), or any other psychosis deemed to be of primary pathology; organic brain disease; metallic implants or objects in the body; having a history of seizure; daily use of lorazepam or equivalents >2 mg; receiving systematic psychotherapy within the last 3 months; receiving systematic rTMS or electroconvulsive therapy in the last month; and other medical conditions considered inappropriate by researchers. Patients were withdrawn from the study if their antidepressant prescriptions changed.

Randomization and blinding

The participants were randomly assigned to the active or sham treatment group in a 1:1 ratio using block randomization generated by SAS (https://welcome.oda.sas.com/home). A staff member not involved in the study held the online code database and created an opaque, sealed envelope. Each envelope contained a unique randomization identification number and treatment assignment code for each patient, which was provided to the treatment operators after enrollment. The treatment operators were aware of the treatment assignment and were restricted from participating in other trial steps. Clinical evaluators and data analysts were isolated in different clinical areas and blinded to the treatment assignment. Patients were also blinded to the groupings and instructed not to discuss their allocations with each other or the evaluators.

Intermittent theta burst stimulation protocol

All participants received iTBS treatments supplied by a MagPro X100 stimulator equipped with a B65 fluid-cooled coil (MagVenture, Farum, Denmark). We chose to activate the LDLPFC because its activation has been shown to improve NSSI ideation and behaviors in previous studies [16]. The target stimulation site for the LDLPFC was located 5 cm ahead of the motor cortex, based on published research that focused on improving depressive symptoms [29, 30] and reducing suicidality [28, 31] using rTMS in adolescents with MDD. According to our pilot study and previous iTBS studies for adolescents, we lowered the daily amount and intensity of stimulation referring to SNT. The determined stimulation comprised five sessions, each containing 1800 pulses patterned into 5 Hz bursts, administered at fixed intervals of 1 h daily for 5 consecutive days (Fig. 1). The stimuli were delivered at an intensity of 80% resting motor threshold, which was determined as the lowest intensity required to elicit at least five out of ten contractions of the right abductor pollicis brevis muscle when stimulating the left motor cortex before the first session every day. Sham stimulation was provided to the control group with the coil tilted 90° to the scalp, which produced noise and vibratory sensations similar to real stimuli.

Fig. 1: Stimulation protocols of accelerated and prolonged iTBS.
Fig. 1: Stimulation protocols of accelerated and prolonged iTBS.
Full size image

iTBS intermittent theta-burst stimulation, NSSI non-suicidal self-injury, OSI Ottawa self-injury inventory.

Assessments

The primary outcome of this trial was the Deliberate Self-Harm Inventory-Adolescent Revised version (DSHI-AR) score at 4-week follow-up [32], which is based on common time points established in previous studies and supported by previous research findings [20, 33]. The DSHI-AR score was acquired at baseline, the day after the 5-day treatment completion, and at the 2-week and 4-week follow-ups to calculate the frequency and severity of NSSI. Secondary measurements were changes in the Deliberate Self-Harm Ideation Scale for Adolescents Revised (DSHI-AR ideation), Ottawa Self-Injury Inventory (OSI) addiction subscale [34], HAM-D17 [35], Hamilton Anxiety Rating Scale (HAMA) [36], Beck Scale for Suicide Ideation (BSI) [37], Pittsburgh Sleep Quality Index (PSQI) [38], and Barratt Impulsiveness Scale-11 (BIS-11) [39] scores rated at baseline, the day after the 5-day treatment completion, and 2 and 4 weeks after the intervention. Detailed information regarding the assessments is provided in the study protocol (see Supplements).

The participants were asked to guess which treatment group they were assigned to after the final stimulation session to assess the integrity of the blinding.

Safety

Adverse events were documented after each session using an adverse event record form (AERF) (see Supplements for details), and pain intensity resulting from the intervention was assessed using a visual analog scale (VAS) after every five sessions. The stimuli were delivered appropriately and gradually increased to the targeted intensity over the first five sessions to ensure patient tolerance.

Sample size

The findings of our unpublished pilot study comprising 14 participants independent of this study revealed that the mean change in the DSHI-AR score for the active treatment group from baseline to 4-week follow-up was –12.56 points, whereas the mean change for the control group was 2.09 points. The sample size was calculated assuming a difference of –14 points between groups, with 90% power and α ≤ 0.05. The analysis revealed that 42 participants (21 per group) were required. Considering a possible dropout rate of 30%, a final sample size of 60 participants (30 per group) was estimated.

Statistical analysis

Statistical analyses were conducted blindly for treatment allocation using the Statistical Package for Social Sciences Statistics for Windows, version 26.0 (SPSS Inc., Chicago, IL, USA). A Shapiro–Wilk test was used to assess normality. For baseline analysis, t-tests or Fisher’s exact tests were used to determine the significant between-group differences in demographic and clinical characteristics identified as normally distributed variables or categorical variables, respectively. Mann–Whitney U tests were used for non-normally distributed variables. Intention-to-treat (ITT) analyses for the primary analysis of DSHI-AR scores 4 weeks after the intervention were performed using random intercept mixed models. Similar mixed linear models within maximum likelihood estimation were used for secondary outcomes including DSHI-AR ideation, OSI-addiction, HAM-D17, HAMA, BSI, suicidal attempts, BIS-11, and PSQI scores from baseline to the 4-week follow-up. The fixed effects of time, group, their interaction, age, sex, duration of NSSI, baseline fluoxetine equivalent, baseline HAM-D17 scores, and other corresponding baseline outcome measures were assessed. Unstructured covariance structures were used for all analyses based on the Akaike information criterion for optimum fitting model determination, in addition to cases in which the models failed to converge. Post hoc multiple comparisons were all Bonferroni-corrected. Per-protocol (PP) set analyses involving participants who completed all treatments and follow-ups were also performed to verify the stability of the results. Additionally, prespecified sensitivity analyses of the primary and secondary outcomes were performed by excluding male and psychotic symptoms subgroups. Spearman’s rank correlation tests were employed in the whole sample to determine the relationships between changes in outcomes at the end of the intervention, 2- and 4-week follow-ups, as depicted by Origin (Pro2021; OriginLab, Northampton, MA, USA). Simple mediation models were used to evaluate whether the efficacy of iTBS on NSSI was mediated by changes in depression and anxiety symptoms along with suicidal ideation and addiction in both groups. All models were adjusted for the baseline dependent variable and estimated the direct, indirect, and total effect sizes with 95% confidence intervals (CIs) calculated through bootstrapping using PROCESS v3.3 in SPSS. A two-sided p-value of < 0.05 was considered statistically significant. Data are presented as means (standard deviations [SDs]) or means (95% CIs).

Results

Participant characteristics

Figure 2 illustrates the recruitment and trial process. After screening 79 participants, 60 adolescents (mean age: 14.2 [1.5] years; 55 [91.7%] females) were randomly assigned to either the active (n = 30) or sham (n = 30) treatment group. No statistically significant differences were observed between the treatment groups in demographic, clinical variables, medication types and daily dosage (Table 1 and Supplementary Table 1). According to the defined daily dose method, the baseline doses of antipsychotic and antidepressant drugs were converted to the chlorpromazine and fluoxetine equivalents, respectively [40, 41], and no significant differences were found between the two groups in either case. Twenty-seven participants completed all 25 iTBS sessions in the active treatment group, and 24 participants completed the sessions in the sham treatment group. Sixty participants were included in the ITT and sensitivity analyses. We also performed a per-protocol (PP) analysis involving 46 participants to test the reliability.

Fig. 2: Flow chart of study.
Fig. 2: Flow chart of study.
Full size image

iTBS intermittent theta-burst stimulation, ITT intention to treat.

Table 1 Baseline demographic and clinical characteristics of randomized participants.
Full size table

Efficacy

Primary outcomes

The primary outcome was the DSHI-AR score, measured at the end of 4 weeks after completing the 5-day intervention. ITT analyses revealed greater reductions in NSSI symptoms in the active group compared with the sham group (time-by-group interaction: F = 5.7, df = 3, 34.5, p = 0.003). The estimated between-group mean difference in the DSHI-AR scores at 0, 2, and 4 weeks post-treatment were −13.05 (95% CI, −21.18–−4.93; p = 0.002), −18.04 (95% CI, −28.40–−7.67; p = 0.004), and −18.66 (95% CI, −28.35–−8.97; p < 0.001), respectively (Table 2).

Table 2 ITT analyses of primary and secondary measurements.
Full size table

Linear mixed models also revealed significant main effects of the interaction between time and group on the DSHI-AR-behavior subscale score (F = 5.2, df = 3, 31.0, p = 0.005) and DSHI-AR-severity subscale score (F = 3.3, df = 3, 32.4, p = 0.031). Compared with participants in the sham treatment group, those in the active treatment group displayed significantly greater reductions in DSHI-AR behavior and DSHI-AR severity subscale scores at all time points following the completion of the intervention (Bonferroni-corrected p < 0.05) (Table 2).

Secondary outcomes

All secondary outcomes were analyzed using the ITT sample (Table 2). The DSHI-AR ideation scale exhibited a greater reduction from baseline to the 4-week follow-up in the active treatment group compared with the sham treatment group (time-by-group interaction: F = 4.5, df = 3, 157.8, p = 0.005). Similarly, BSI (time by group interaction: F = 7.1, df = 3, 157.2, p < 0.001), HAM-D17 (time-by-group interaction: F = 5.2, df = 3, 154.2, p = 0.002), and HAMA (time-by-group interaction: F = 4.4, df = 3, 496.0, p = 0.005) scores showed more improvement in the active treatment group than in the sham treatment group at all time points following the intervention. Additionally, the OSI addiction subscale scores showed greater reductions in the active treatment group than in the sham treatment group at the 2-week (estimated mean difference [EMD], −4.59; 95% CI, −8.48–−0.71; p = 0.021) and 4-week (EMD, −6.16; 95% CI, −10.58–−1.74; p = 0.007) follow-ups. However, none of the time-by-group interactions on suicidal attempts, BIS-11, and PSQI were statistically significant (Table 2).

The correlation analysis showed that reductions in DSHI-AR scores were associated with decreases in HAM-D17(r = 0.41, p = 0.006), HAMA (r = 0.50, p = 0.001), BSI (r = 0.32, p = 0.032), and OSI addiction (r = 0.37, p = 0.012) scores from baseline to the 4-week follow-up (Table 3). Changes in DSHI-AR behavior and DSHI-AR severity subscale scores also correlated significantly with improvements in HAM-D17, HAMA, BSI, and OSI addiction scores at the 4-week follow-up (Table 3). Reductions in DSHI-AR significantly correlated with changes in OSI addiction scores at the 1-week (r = 0.29, p = 0.040) and 2-week (r = 0.38, p = 0.007) follow-ups (Supplementary Fig. 1). Similar significant correlations were also observed between decreased NSSI symptoms and improvements in HAM-D17 (r = 0.47, p < 0.001), HAMA (r = 0.45, p = 0.001), and BSI (r = 0.44, p = 0.001) scores at the 2-week follow-up (Supplementary Fig. 1).

Table 3 Spearman rank correlations between changes in DSHI-AR and other clinical variables from baseline to 4-week follow-up.
Full size table

The effectiveness of iTBS treatment on NSSI symptoms was partially mediated by depression (−5.77; 95% CI, −9.03–−3.06), anxiety (−5.91; 95% CI, −9.40–−3.05), suicidal ideation (−5.33; 95% CI, −8.37–−2.65), and NSSI addiction (−4.76; 95% CI, −8.46–−2.02) (Supplementary Fig. 2). The direct effects of iTBS on NSSI remained significant after controlling for mediating variables (Supplementary Fig. 2).

PP analyses (Supplementary Table 2, Supplementary Fig. 3) were generally consistent with the results of the primary, secondary outcomes and the mediation effects in ITT analyses.

Sensitivity analysis restricting the data to female participants or those without psychotic symptoms produced similar results of primary analysis (Supplementary Fig. 4).

Blinding integrity

In total, 51 participants attempted to guess the treatment they received. Of these, 17 participants (63.0%) in the active treatment group and 11 (45.7%) in the sham treatment group believed they had received active treatment; 4 participants (14.8%) in the active treatment group and 5 (20.8%) in the sham group believed that they had received sham treatments. The remaining participants were uncertain of their treatment assignments. Fisher’s exact test compared participants’ correct guesses regarding their treatment group assignment, revealing no significant difference (p = 0.468) (Supplementary Table 3). No association was found between the speculated allocation and changes in DSHI-AR scores (Beta = −0.26; 95% CI, −4.32–3.80; p = 0.901) (Supplementary Table 4).

Safety

Participants reported no severe adverse events, such as seizures, cramps, or psychomotor disturbances throughout the study (Table 4). Although 23 participants (76.7%) in the active treatment group and 17 (56.7%) in the sham treatment group reported adverse events, the difference was not statistically significant (p = 0.171, chi-square test). The most common adverse event in the first 10 sessions was headaches, reported by 53.3% of participants in the active treatment group and 30.0% in the sham treatment group (p = 0.115). The severity of the headache gradually decreased and was tolerable, with an average daily severity score of 2.81 and 1.95 in the active and sham treatment groups, respectively (p = 0.148). No statistically significant differences were observed in the incidences of other adverse events, including tenderness, fatigue, nausea, sleepiness, and dysphoria, between the two groups.

Table 4 Adverse events reported after any iTBS treatment session.
Full size table

Discussion

This randomized sham-controlled study investigated the efficacy and tolerability of iTBS on NSSI in adolescents with MDD. Participants who underwent active iTBS treatments demonstrated significantly greater reductions in NSSI frequency, severity, ideation, and addiction, as well as suicidal thoughts and depressive and anxiety symptoms. Furthermore, the intervention was well-tolerated by patients, as none of them withdrew because of side effects during the 5-day treatment course. In the active treatment group, 53.5% of participants reported headaches, comparable to the 57% reported in the FDA-approved SNT iTBS protocol [20]. The results support our hypothesis that prolonged accelerated iTBS targeting the DLPFC is feasible, effective, and safe for alleviating NSSI symptoms in adolescents with MDD.

Previous studies have spotlighted on the efficacy of psychotropic medications and psychotherapies for adolescent NSSI. In a meta-analysis of 29 clinical trials, treatments encompassing medications, psychotherapies, and combined treatments did not significantly change NSSI occurrences among children and adolescents with psychosis or self-injurious behaviors (pooled RR = 1.30 [0.92, 1.82]) [42]. Medications involving selective serotonin reuptake inhibitors, selective norepinephrine reuptake inhibitors, antipsychotics, and alpha-2 adrenergic agonists did not significantly reduce the occurrence of self-harm in adolescents (pooled RR = 1.11 [0.73, 1.68]) [43]. Mentalization-based therapy, emotion regulation group therapy, cognitive behavioral therapy, and dialectical behavior therapy exhibited significant effects on NSSI, with a pooled standardized mean difference of −0.53 [−0.82, −0.25] among adolescents with a borderline personality disorder or MDD [44]. Nevertheless, psychotherapies typically involve longer treatment durations, spanning 8–52 weeks, and necessitate a standardized workflow led by experienced psychotherapists [45,46,47]. In contrast, the iTBS fast-track protocol can be completed in 1 week and is available in most public mental health care institutes, making it a more efficient treatment option.

Several mechanisms may contribute to the therapeutic effects of iTBS on NSSI. In our analysis, the changes in NSSI symptoms were correlated with changes in depression, anxiety, and suicidal ideation. Previous studies have found that patients with MDD would exhibit NSSI behaviors in the presence of negative emotions and suicidal thoughts [48]. Thus, iTBS may diminish the motives behind NSSI through transient emotional regulation and suicidal thoughts inhibition, thereby disrupting the occurrence and maintenance of NSSI. Stimulation targeting the DLPFC has consistently been demonstrated to alleviate depression, anxiety, and suicidality symptoms. Stimulation protocols utilizing 1 800 pulses per iTBS session for 10–20 sessions (totaling 18 000–36 000 pulses) have shown significant antidepressant, anti-anxiety, and anti-suicidal effects in adolescents with MDD [25,26,27,28].

NSSI has addictive features, and its repetitive pattern is related to its severity and negative emotional experience [49]. Adolescents with NSSI show reduced activation in the striatum and orbitofrontal cortex and decreased reactivity to reward in the limbic and midbrain cortices when making risky decisions, leading to increased frequency of reward-seeking behavior [50]. In the present study, the reductions in NSSI addiction induced by iTBS may have also contributed to a decrease in NSSI frequency. Himelein et al. found that 76.8% of individuals who engaged in repetitive self-harm were at least mildly addicted [49]. The addictive features of NSSI may be associated with dysfunction in the mesocorticolimbic dopamine reward and the endogenous opioid systems [51, 52]. Previous studies have shown that HF-rTMS at 10 or 20 Hz with 1000–10 000 pulses over the LDLPFC can effectively relieve cravings in addiction disorders, such as alcohol, cannabis, nicotine, and methamphetamine addiction, eating disorders, and gambling disorders [24]. A meta-analysis also found that neuromodulations targeting the left rather than the right DLPFC may be more effective in reducing craving in addction disorders [53].

Typically, structured interviews or brief self-reported scales, such as the 9-item DSHI and the Functional Assessment of Self-Mutilation were used to assess NSSI in previous interventional studies [46, 47]. We used the DSHI-AR, a modified Likert scale, and an extended DSHI version comprising 19 items. This scale encompasses a broader range of NSSI forms, allowing for a more comprehensive analysis of NSSI frequency and severity.

Nevertheless, this study had some limitations. First, the participants were only recruited from one center. Hence, further populations from different institutions are needed to ascertain the validity and reliability of the results. Second, due to funding limitations, we were constrained in implementing specially designed sham pads for the sham control group and ultimately chose the “coil tilting 90-degree” method as an alternative. This approach, although commonly adopted by former studies [54, 55], might not fully mask the absence of tactile and sensory feedback, potentially reducing the placebo effect in the sham group. Consequently, this could lead to bias in reporting outcomes. Furthermore, the current study emphasized short-term symptom improvement, with insufficient evaluation of target population and long-term treatment effects. Hence, future research should address this by extending the follow-up period and incorporating multi-dimensional assessments, such as quality-of-life measures, global functional outcomes, and patient-reported outcomes, to provide a more comprehensive and nuanced evaluation of treatment efficacy. Finally, owing to the unclear neuroanatomical correlations with NSSI symptoms in current neuroimaging studies, we did not use neuronavigation [56, 57]. In the future, the iTBS protocol and neuroimaging techniques such as functional magnetic resonance imaging and electroencephalography can be integrated to elucidate the underlying neural pathways of NSSI and explore an optimal neuroimaging marker for improving treatment efficiency.

In conclusion, iTBS targeting the LDLPFC in adolescents with MDD and NSSI significantly reduced NSSI behavioral frequency and severity after 5 days of treatment compared with the same course of sham stimulation. This study highlights important advances in the use of iTBS for MDD and demonstrates the potential of iTBS as an efficacious and tolerable strategy for treating NSSI in adolescents with MDD. Further research is needed to validate these findings across multiple centers and extend the follow-up period to assess long-term efficacy.