Main

Borderline personality disorder (BPD) is associated with an exceptionally high risk of suicide, with an average of more than three lifetime suicide attempts1 and a completed suicide rate of 10% (ref. 2). BPD and major depressive disorder (MDD) are frequently comorbid, which further elevates the risk of suicide3. At least one study showed that more severe baseline depression lowers the response to evidence-based psychotherapies for BPD, such as dialectical behaviour therapy (DBT)4. Further, available evidence does not generally support the use of medication to treat symptoms of BPD5. As such, although reducing suicide risk is a major area of focus in the clinical management of BPD, healthcare practitioners often have few treatment options for patients who remain at elevated risk for suicide following inadequate response to first-line interventions. Consequently, a concerningly high number of individuals with BPD repeatedly present to the emergency department and are subsequently hospitalized for suicide risk, which in the long term can increase the risk of suicide in BPD6. Taken together, there is an urgent need to identify additional safe and effective interventions to reduce suicide risk in BPD.

Non-invasive brain stimulation interventions to treat depression and reduce the associated risk of suicide offer the promise of new treatments for suicidal and depressed individuals with BPD, although few studies have investigated their safety and effectiveness in this population. For example, although electroconvulsive therapy (ECT) is the most well-researched and efficacious brain stimulation intervention for severe, treatment-resistant depression (TRD)7, few trials have investigated the efficacy of ECT for MDD in BPD, and limited findings suggest that antidepressant responsiveness to ECT in BPD, and in BPD traits, ranges from slightly8 to substantially9 less robust compared with patients without BPD. Further, ECT carries well-documented cognitive adverse effects that include autobiographical memory impairment.

Magnetic seizure therapy (MST) is emerging as a safer alternative to ECT for the treatment of TRD10 and associated suicidal ideation11. As in ECT, the hypothesized mechanism of action in MST is the induction of a seizure. However, MST uses magnetic fields to induce an electrical field, which results in focal stimulation. An MST-induced seizure produces effects in more superficial cortical areas, which may explain why patients experience fewer cognitive side effects12. A systematic review of eight MST trials found that a clinical response is achieved in 40–60% of participants with TRD13, which is consistent with findings from a large clinical trial of MST for MDD10. Further, a recent pilot study comparing 100 Hz MST with ECT found no differences in antidepressant effectiveness between the two approaches12. Crucially, MST does not appear to be associated with the cognitive side effects observed after ECT14, which is a pertinent advantage for individuals with BPD who show a range of cognitive deficits, including in episodic memory15.

MST has yet to be investigated in BPD, although recent findings in BPD samples suggest that a related brain stimulation technique, repetitive transcranial magnetic stimulation (rTMS), may reduce levels of BPD severity and depression commonly observed in suicidal individuals with BPD16,17,18,19. In this Article, we present the results from an open-label clinical trial investigating the feasibility, and clinical and cognitive effects, of a 5-week course of conjoint MST plus DBT (n = 9) compared with DBT alone (n = 10) in suicidal patients with comorbid BPD and treatment-resistant depression. We hypothesized that treatment in the MST + DBT arm would result in substantial rates of response and remission on measures of suicidal ideation and depression. Moreover, we hypothesized that MST would have limited, if any, effects on performance on standard neuropsychological measures of attention, memory and executive functioning.

Results

Participant characteristics and medications are displayed in Tables 1 and 2, respectively. A Pearson correlation matrix of baseline symptom scores and demographics is presented in Supplementary Table 1.

Table 1 Participant demographics and clinical characteristics
Full size table
Table 2 Medications by type taken during the active treatment course, during pre-treatment only and from post-treatment to follow-up
Full size table

A CONSORT flow diagram is displayed in Fig. 1. Our initial recruitment target was N = 30. Altogether, 21 out of 62 screened participants met eligibility criteria and provided written informed consent to participate. Due to the impact of COVID-19, additional recruitment beyond March 2020 was precluded. Our final sample size was N = 19 (9 MST + DBT and 10 DBT).

Fig. 1: CONSORT flow diagram.
figure 1

Study protocol from eligibility assessment to follow-up for participants allocated to MST + DBT (left) and DBT only (right).

Full size image

Symptom outcomes

Primary outcomes of suicidal ideation on the Modified Scale for Suicide Ideation (MSSI)20 and clinician-rated depression on the Hamilton Rating Scale for Depression – 24 (HRSD-24)21 and additional outcomes of self-reported depression on the Quick Inventory of Depressive Symptomatology Self-Report (QIDS-SR)22 and BPD symptom severity on the Zanarini Rating Scale for BPD (ZAN-BPD)23 are displayed in Fig. 2. Individual symptom trajectories are shown in Fig. 3.

Fig. 2: Multilevel-modelling post hoc tests of primary outcomes of suicide ideation and clinician-assessed depression severity and additional outcomes of BPD symptom severity and self-reported depression.
figure 2

a–d, Total mean symptom scores by treatment arm for MSSI suicidality (a), HRSD-24 depression (b), ZAN-BPD severity (c) and QIDS-SR depression (d). Data are presented as mean values. Error bars represent standard error. All tests were two-tailed. Outcomes at baseline and post-treatment comprise n = 9 independent MST + DBT participants and n = 10 independent DBT-only participants. Follow-up outcomes include n = 8 independent participants in each of MST + DBT and DBT only. *P = 0.03; ***P < 0.001, uncorrected.

Full size image
Fig. 3: Individual symptom trajectories of suicidal ideation, clinician-assessed depression severity, total BPD symptom severity and self-reported depression.
figure 3

a–d, Individual symptom trajectories by treatment arm: suicidality (a), clinician-rated depression (b), total BPD symptom severity (c) and self-reported depression (d).

Full size image

Suicidal ideation

At post-treatment, only the MST + DBT arm showed improvements in suicidal ideation, as indicated by a significant interaction effect (beta (β) = −12.03 [95% confidence interval (CI): –19.24, –4.82]; standard error (SE) = 3.75; t[31] = −3.21; P < 0.01). Post hoc tests showed that improvements in MST + DBT at post-treatment (β = 13.33; SE = 2.72; t(31) = 4.90; P < 0.001) were sustained at follow-up (β = 7.74; SE = 2.84; t(31) = 2.73; P = 0.03). Within the DBT-only condition, no pairwise contrasts reached statistical significance (P > 0.05), although a modest improvement from baseline to follow-up was observed (β = 6.51; SE = 2.80; t(31) = 2.32; P = 0.07). There were no interaction effects between baseline BPD symptom severity and the effects of time or condition (P < 0.05) and no three-way interaction effects among BPD symptom severity, time and condition (P < 0.05).

Depression

At post-treatment, only the MST + DBT arm showed improvements in clinician-rated HRSD-24 depression, as indicated by a marginal interaction effect (β = −7.32 [95% CI: –14.10, –0.54]; SE = 3.53; t[31] = –2.08; P = 0.05). Post hoc tests showed significant improvements in MST + DBT at post-treatment (β = 11.22; SE = 2.56; t(31) = 4.39; P < 0.001), which were not sustained at follow-up (β = 4.90; SE = 2.67; t(31) = 1.84; P = 0.17). Within the DBT-only condition, no pairwise contrasts for HRSD-24 depression reached statistical significance (P > 0.05). There were no interaction effects between baseline BPD symptom severity and the effects of time or condition (P < 0.05) and no three-way interaction effects among BPD symptom severity, time and condition (P < 0.05).

In terms of self-reported depressive symptoms on the QIDS-SR, no significant time, condition or interaction effects were found (P > 0.05).

BPD symptom severity

ZAN-BPD total symptom severity

At post-treatment, only the MST + DBT arm showed an improvement in total BPD symptom severity, as indicated by a marginal but non-significant interaction effect (β = −5.47 [95% CI: –10.93, –0.001]; SE = 2.84; t[31] = −1.92; P = 0.06). Pairwise contrasts showed significant improvements in MST + DBT participants at post-treatment (β = 5.67; SE = 2.06; t(31) = 2.75, P = 0.03), which were not sustained at follow-up (β = 2.80; SE = 2.15; t(31) = 1.30, P = 0.40). Within the DBT-only condition, no pairwise contrasts reached statistical significance (P > 0.05).

ZAN-BPD subscale severity

At post-treatment, a significant interaction effect was found on the interpersonal ZAN-BPD subscale, suggesting that only the MST + DBT arm showed improvements in interpersonal BPD symptoms (β = −1.89 [95% CI: −3.60, −0.18]; SE = 0.89; t(31) = −2.12; P = 0.04). Pairwise contrasts showed significant improvements in MST + DBT participants at post-treatment, relative to baseline (β = 1.89; SE = 0.65; t(31) = 2.9; P = 0.02), which were not sustained at follow-up (β = 1.22; SE = 0.67; t(31) = 1.81; P = 0.18). Within the DBT-only condition, pairwise contrasts did not reach statistical significance (P > 0.05). Regarding other ZAN-BPD subscales (affective, cognitive and impulsive symptoms), no significant time, treatment condition or time × treatment condition interaction effects were found (P > 0.05).

Depression response and suicidal ideation remission

Of the total of 19 participants in the treatment, 1 MST + DBT participant and 1 DBT-only participant exhibited an alleviation of depressive symptoms (as defined by a 50% reduction from baseline in HRSD-24 score). One MST + DBT and three DBT-only participants experienced remission from suicidal ideation (as defined by a score of zero on the MSSI on 2 consecutive weekly assessments), although the three DBT-only participants who remitted exhibited either complete remission or substantial reduction in suicidal ideation from enrolment to their baseline assessment, suggesting that their decreases in suicidal ideation were due to factors not directly related to DBT, such as an increase in hopefulness or perceived support, or extraneous factors.

Neuropsychological performance

Participants in both conditions demonstrated stable neuropsychological performance from baseline to post-intervention. Two significant time-by-condition interaction effects were observed for Brief Visuospatial Memory Test (BVMT) Learning (F[1,16] = 7.14; P = 0.02) and BVMT Retention (F[1,16] = 5.64; P = 0.03), with post hoc tests suggesting that within the DBT-only group, participants scored better on BVMT Learning (F[1,8] = 4.50; P = 0.07) and worse on BVMT Retention (F[1,8] = 3.68; P = 0.09) at post-treatment relative to baseline. Main effects of time were found on Delis Kaplan Executive Functioning Category Switch Total Correct (F[1,16] = 7.98; P = 0.01) and Total Switch (F[1,16] = 5.07; P = 0.04), indicating that participants in both conditions performed worse at post-treatment relative to pre-treatment. No other significant effects were found (P > 0.05), including no significant main or interaction effects in autobiographical memory (P > 0.05), suggesting that MST + DBT did not impact cognitive performance relative to DBT. All neuropsychological test scores are provided in Supplementary Table 2.

Safety

Among the 21 participants who signed consent, there were three serious adverse events that were deemed unrelated to MST treatment, including attempted suicide before completing study-related treatments (n = 1) and ureteroscopic surgeries related to a pre-existing kidney condition (n = 2).

DBT attendance

Across the 5-week intervention, participants in MST + DBT attended a mean of 4.4 (±1.0) individual DBT therapy sessions and 4.3 (±1.7) DBT skills training sessions. DBT-only participants attended a mean of 4.9 (±1.0) individual therapy sessions and 4.7 (±1.3) skills training sessions. There were no significant between-group differences in the number of sessions attended for individual DBT therapy (t = 0.94: P = 0.36) or DBT skills training (t = 0.52; P = 0.61).

Electroencephalogram seizure duration

Seizure duration was calculated as the mean duration of the second treatment and the last treatment. The mean (s.d.) seizure duration was 65.72 (±58.3) s.

Time to reorientation

Time to reorientation was calculated as the mean time of the second treatment and the last treatment. The mean (s.d.) reorientation time was 9.08 (±3.07) min.

Discussion

The current findings provide initial evidence to support the feasibility of conjoint MST and DBT for the treatment of moderate to severe suicidal ideation in individuals with BPD and TRD. We observed large, clinically meaningful reductions in suicidal ideation following conjoint treatment with MST and DBT, with improvements sustained at four-month follow-up. Significant reductions in clinician-rated depression and BPD interpersonal symptom severity were also seen from baseline to post-intervention in the MST + DBT condition, although these gains were not sustained at follow-up. All participants were evaluated from pre- to post-treatment using a neuropsychological battery of gold-standard measures spanning major domains of cognitive functioning most relevant to non-invasive neurotherapeutics24. Given the impact of ECT on memory, the battery was selected to allow for comprehensive assessment of verbal and non-verbal memory. There were no differential effects on cognitive performance observed between treatment arms. The intervention was feasible to implement across a multidisciplinary treatment team, and MST treatments were well tolerated across participants with no serious adverse events related to the treatment.

Chronic suicidality in BPD is one of the most challenging symptoms of the disorder, precluding patients from fully engaging in psychotherapy and reducing the willingness of clinicians to work with BPD due to beliefs that clinical management is too difficult25. Although individuals with BPD are at an exceptionally high risk of suicide, with estimates of up to 84% attempting suicide at least once26, there is a paucity of research on biological treatments to reduce suicide risk in BPD. Current evidence points to DBT and mentalization-based therapy as the most effective psychotherapy interventions for suicidal behaviour in BPD27, both of which require specialized clinician training and where reductions in suicidal ideation are slow and often observed only after months of treatment28. The innovative design of the present clinical trial is optimal in the sense that severely suicidal individuals concurrently engaged in a convulsive treatment while participating in DBT, a treatment combination that led to a reduction in suicidal ideation over 5 weeks with sustained and clinically meaningful improvements (average MSSI score reduction from the ‘severe’ range to ‘mild to moderate’). We propose that the success of this conjoint treatment would be unlikely if ECT were instead chosen due to its adverse side effects on cognition. Even brief ECT (ultra-brief pulse width) is associated with post-treatment amnesia across randomized controlled trials (RCTs)29. This is probably due to the differences in induced field strength between ECT and MST: MST uses a rapid, high-intensity, time-varying magnetic field that limits seizure spread30, resulting in stimulation that is not impeded by the skull and is five to ten times more focal, relative to right unilateral ultra-brief pulse ECT31,32. Taken together with previous findings of poorer ECT outcomes in depressed patients with BPD8,9, ECT even being more readily accessible than MST may not be the most optimal approach to combine with psychotherapy. However, since findings suggest that there are less-efficacious treatment options for individuals with depression and comorbid personality disorder features33, and since ECT is one of the most effective interventions for depression34, all potential treatment avenues for individuals with depression and BPD should be empirically investigated. As such, a clinical trial of ultra-brief pulse ECT plus psychotherapy versus MST plus psychotherapy could empirically investigate their comparative efficacy. Alternatively, non-invasive brain stimulation more widely used across treatment settings, such as rTMS, should be studied. To further prolong the decline in suicidal ideation observed in the current study, an MST tapering phase that consists of completing a full course of DBT could be considered.

Of note, rather than employing the Colombia Suicide Severity Rating Scale (C-SSRS35), suicidal ideation in the current study was assessed using the MSSI, a suicide-specific measure with excellent psychometric properties20 that has been tested across inpatient20 and outpatient36,37 samples. Whereas the C-SSRS has been recommended by the Food and Drug Administration to monitor for treatment-emergent suicidal ideation in clinical trials38, our study required participants to experience suicidal ideation at baseline. The MSSI was chosen as the primary outcome measure as it provides increased specificity and additional range of items (compared with the C-SSRS) and a continuous total score with a wide range from 0 to 54 that is ideal for statistical analysis in a small sample. Moreover, the MSSI has established severity thresholds, which ensured standardized assessment of the suicidal ideation inclusion criteria.

Regarding mechanisms of action, initial research suggests that MST induces cortical neuroplasticity via mechanisms related to long-term potentiation39, but further studies should investigate the specific biological pathways leading to reduction in depression and suicidal ideation after MST. Suicidality across psychiatric disorders is associated with disruptions in the frontolimbic and frontotemporal circuits subserving emotion regulation and cognitive control40. MDD is also associated with greater instability of connectivity in the default mode network (DMN)41 implicated in self-referential processing, rumination and social cognition42. Finally, interpersonal symptoms in BPD (for example, rejection sensitivity) have been associated with functional disruptions in the medial prefrontal cortex (MPFC)43, a region in the DMN involved in social cognition44. In the present study, MST was delivered frontally with the maximum electrical field activation in the dorsomedial PFC (DMPFC), which is a common shared hub among frontolimbic, frontotemporal and DMN networks. Modified functioning in these affective, executive and social cognitive-related pathways could be one of the potential mechanisms underlying symptom improvements after MST treatment.

Moreover, there is a strong association between interpersonal distress and suicidal ideation in BPD1,45,46. Relationship conflict and low perceived warmth in others have been shown to precipitate increases in suicidality in BPD1,46. Although it is not possible to determine whether improvements in interpersonal BPD symptoms were a cause or consequence of reductions in suicidal ideation, it is plausible that improvements in these two domains were not independent.

Pertaining to depression severity, although a significant reduction in clinician-rated depressive symptoms was observed from pre- to post-treatment, these gains were not sustained at follow-up. No improvements were observed in self-reported depression. These results are surprising given that recent preliminary findings show that rTMS, a less-intense form of magnetic stimulation, yields substantial alleviation of MDD symptoms in individuals with BPD17. Notably, in our trial, moderate to severe suicidal ideation on the MSSI was required to meet eligibility criteria. This level of baseline severity could be related to the less-robust depression response observed, whereby participants with greater suicidal ideation at baseline may require longer to experience improvements in MDD symptoms. As we did observe a 12-point mean reduction in HRSD-24 depression from baseline to post-treatment in the MST + DBT arm, there is a need for the study of maintenance treatments on depression and suicidal ideation, which could extend clinical benefits. Indeed, rates of depression relapse following ECT are high without maintenance treatment, which may indicate that additional treatments are similarly necessary to prolong MST-related treatment effects in severely ill individuals.

In addition, although participants in the DBT-only arm of this study did not exhibit a significant clinical response, a modest reduction in suicidality was observed at follow-up (P = 0.07), which is consistent with findings showing slower remission for suicide-related symptoms after interventions for BPD (for example, DBT)28. As such, these results do not suggest that DBT was ineffective, but rather the combination of MST + DBT was associated with a more rapid reduction in suicidality compared with 5 weeks of DBT alone. It is important to emphasize that participants in this study received MST in conjunction with DBT, which may be superior to either treatment alone.

Finally, our findings warrant future investigation into the utility of combining MST with more generalist or accessible treatments for BPD, such as community-based emotion regulation skills groups. In fact, the study of brain stimulation for BPD is an emerging research area47,48. Whether various types of brain stimulation interventions may be combined with psychotherapy for BPD is an important future direction for BPD treatment research. Overall, the quality of evidence for ECT in BPD is low; most findings are derived from case studies or retrospective chart review and suggest that ECT is less efficacious in BPD47. However, there is preliminary evidence that transcranial direct current stimulation may target affective dysregulation by altering the executive processes involved in the cognitive control of emotions in BPD47. Finally, four sham-controlled pilot RCTs have investigated rTMS for BPD16,17,18,19, with preliminary findings suggesting that stimulation of the dorsolateral PFC or DMPFC may improve core BPD symptoms, although some improvements have also been noted after sham stimulation16,17, emphasizing the need for large, adequately powered RCTs.

Notably, our finding that MST + DBT had broad effects on suicidal ideation, depression and interpersonal BPD symptoms converges with findings from two recent 5 Hz (ref. 18) and 20 Hz (ref. 17) rTMS trials that also targeted the bilateral DMPFC, leading to active rTMS-related improvements in interpersonal and behavioural BPD symptoms18 and in depression severity17. As such, the DMPFC warrants further exploration as a potential stimulation site to target various BPD symptoms, given its functional role in social44 and non-social cognition49. Of note, participants in the current study were unique in their severity of suicidal ideation. It remains to be explored whether other non-convulsive therapies can reduce moderate to severe suicidal ideation in BPD.

Limitations and future directions

The open-label case-controlled design, small sample size and women-only participants in the current study are limitations, and investigation using a sham-controlled RCT with a gender-diverse sample is warranted. As participants in this study self-selected their treatment arm, patient expectations regarding the effectiveness of conjoint MST + DBT over DBT alone may have moderated or mediated the between-condition treatment effects observed; MST treatment involves a complex intervention with an advanced technology that has the potential for large expectation and non-specific effects. A sham-controlled design that includes DBT in both arms may address this critical limitation. Finally, although the ZAN-BPD is a standard measure used in clinical research, it assesses symptoms across a 1-week period. Given the 5 weeks between post-treatment and follow-up in the current study, a measure that is sensitive to change across a longer period (for example, The Borderline Personality Disorder Severity Index-IV50) would have been optimal.

Conclusions

This study represents a step towards addressing the long-standing problem of suicidality in BPD. The combination of MST + DBT led to a rapid and clinically meaningful reduction in suicidal ideation by 5 weeks in participants who did not respond to first-line treatments for depression. MST did not affect cognitive processes. Our findings provide evidence to support MST as a conjoint treatment for suicidal ideation in severely symptomatic patients with BPD and treatment-resistant depression.

Methods

Study design

This study was conducted at the Centre for Addiction and Mental Health (CAMH) between October 2017 and March 2020, with follow-up assessments ending in June 2020 (NCT03361826). The study protocol was approved by Health Canada and the CAMH and University of Toronto ethics boards (053-2015) and complied with all relevant ethics regulations. All participants provided written informed consent to participate. Study aims were to investigate the feasibility and clinical and cognitive effects of MST for suicidal ideation in BPD with comorbid TRD. No statistical methods were used to pre-determine sample size, but our sample size is similar to previous publications17,18,19. Moderate to severely suicidal participants were recruited using convenience and voluntary response sampling with a case-control design, comparing individuals receiving MST + DBT with a matched patient group receiving DBT only. Participants were enrolled by postdoctoral-level study staff under the supervision of the principal investigator. Across 5 weeks, participants in both arms received 1 h of weekly individual DBT and 1 h of weekly DBT skills training focused on distress tolerance. Participants in the MST + DBT group additionally received up to 15 MST treatments (3 treatments per week) across the 5-week protocol. Participants completed weekly symptom assessments from baseline to post-intervention to monitor the severity of primary outcomes in suicidal ideation and clinician-rated depression. Additional symptom outcomes included self-reported depression and BPD symptom severity. Participants returned for a follow-up symptom assessment at four months. To investigate cognitive effects of MST, all participants completed neuropsychological testing before and after the 5-week intervention. Participants also completed pre- and post-treatment functional magnetic resonance imaging scans to explore potential biomarkers of treatment response. Due to the impact of COVID-19 halting recruitment and resulting in a smaller sample size, functional magnetic resonance imaging data are too underpowered to detect reliable effects and will not be analysed or reported. As the result of an oversight, ethnicity data were not collected.

MST protocol

A MagVenture MagPro MST with Cool TwinCoil device with 100% machine output at 25 Hz was used, consistent with findings suggesting that 25 Hz is associated with the greatest reduction in suicidality in TRD11. Each circular coil was centred over the dorsolateral PFCs, bilaterally. This placement results in maximum e-field activation in the DMPFC of both hemispheres51. See Supplementary Information for the full MST and anaesthesia protocol.

Clinical assessments and neuropsychological testing

Participants were assessed for BPD, MDD and other comorbidities using the International Personality Disorders Examination – BPD Section52 and the Structured Clinical Interview for DSM-IV53. The Antidepressant Treatment History Form54 was used to quantify antidepressant resistance. The Antidepressant Treatment History Form rates each medication trial on a scale from 1 to 5, with scores ≥3 indicating an adequate trial at least 4 weeks in length, and with higher scores suggesting greater antidepressant resistance55.

Primary outcomes of MSSI suicide ideation and HRSD-24 depression, and additional outcomes of the QIDS-SR22 and the clinician-administered ZAN-BPD23, were administered at screening/baseline and then once weekly from baseline to post-treatment and again at four-month follow-up. The Young Mania Rating Scale56 was administered once weekly during the 5-week treatment protocol as a safety measure to monitor for the emergence of manic symptoms.

Neuropsychological testing was carried out at baseline and post-treatment to assess cognitive functioning in the domains of auditory–verbal learning and episodic memory (California Verbal Learning Test Second Edition57), autobiographical memory (Autobiographical Memory Test58), visual–spatial learning and episodic memory (BVMT Revised59), auditory–verbal working memory and visual attention/psychomotor speed (Digit Span and Coding subtests, respectively, from the Weschler Adult Intelligence Scale Fourth Edition60), executive functioning (Delis Kaplan Executive Functioning61 Verbal Fluency, Color Word Interference and Sorting subtests) and visual attention/processing speed and cognitive flexibility (Trail Making Test Parts A and B, respectively62). The Weschler Test of Adult Reading63 was used to assess for the study exclusion criteria of a reading level below eighth grade.

Statistical analyses

Statistical analyses were completed using the ‘nlme’, ‘multcomp’, and ‘emmeans’ packages in R programming (version 4.1.2). Multilevel models evaluated changes in primary outcome measures of MSSI and HRSD-24 scores, and in the QIDS-SR and ZAN-BPD scores, at post-treatment and four-month follow-up relative to baseline. The assumptions of normality and homoscedasticity of the residuals of each multilevel model were formally tested using the Shapiro–Wilk and Levene’s F test, respectively. Residuals of the model of MSSI scores did not violate the assumption of normality (W = 0.98; P = 0.34) but displayed some heteroscedasticity (F [18,35] = 2.43; P = 0.01). The residuals for the model of HRSD-24 scores did not violate the assumption of normality (W = 0.98; P = 0.69) or homoscedasticity (F [18,35] = 1.50; P = 0.15). The residuals for the model of ZAN-BPD scores did not violate the assumption of normality (W = 0.98; P = 0.89) or homoscedasticity (F [18,35] = 1.83; P = 0.06). On the basis of a previously proposed four-factor structure of the ZAN-BPD64, changes in ZAN-BPD subscale scores within the domains of affective, cognitive, impulsive and interpersonal symptoms were also explored. Within each multilevel model, effects of time, treatment condition and time × treatment condition interactions were examined. Initial fixed- and random-effects models, as well as models with a CAR1 autoregressive correlation structure, were built and compared to test the best-fitting model for each outcome, which was selected according to the Akaike and Bayesian information criteria. To explore whether BPD symptom severity impacted the results, associations between baseline BPD symptom severity and the effects of time, treatment condition or time × treatment condition interactions were explored. No other covariates were examined. A linear random-effects model with a time × condition interaction term fit best for all variables. Post hoc pairwise comparisons were examined within each treatment arm. Due to the preliminary nature of the findings, effects are reported as significant at an uncorrected alpha threshold of P < 0.05. All tests were two-tailed.

Suicidal remission and depression response rates were also examined and defined as a score of zero on the MSSI for 2 consecutive weeks and ≥50% reduction in HRSD-24 scores from baseline, respectively.

Neuropsychological data were analysed using mixed ANOVA models examining effects of time, treatment condition and time × treatment condition interaction effects.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.