Treatment of mixed depression with theta-burst stimulation (TBS): results from a double-blind, randomized, sham-controlled clinical trial

Abstract

Mixed depression is probably different in terms of clinical course and response to treatment. Repetitive transcranial magnetic stimulation (rTMS) is well established in non-mixed depression, and theta-burst stimulation (TBS) protocol is replacing conventional protocols because of noninferiority and reduced delivery time. However, TBS has not been adequately studied in mixed states. This study was a double-blind, six-week, sham-controlled, and randomized clinical trial of bilateral TBS targeting the right and left dorsolateral prefrontal cortex, respectively. Adults with bipolar and major depressive disorder experiencing an acute mixed depression were eligible if they had not benefited from a first- or second-line treatment for acute unipolar or bipolar depression recommended by the Canadian Network for Mood and Anxiety Treatments. Out of 100 patients included, 90 composed modified intention-to-treat sample, which was patients that completed at least one week of the intervention. There were no significant differences in Montgomery-Asberg depression rating scale score changes (least squares mean difference between groups at week 3, −0.06 [95% CI, − 3.39 to 3.51; P = 0.97] in favor of sham TBS). Response and remission rates per MADRS were also not statistically different among active and sham groups (35.7% vs. 43.7%, and 28.5% vs. 37.5% respectively at week 6, ps > 0.51). No other analyses from baseline to weeks 3 or 6 revealed significant time x group interaction or mean differences among groups in the mITT sample. Bilateral TBS targeting the DLPFC is not efficacious as an add-on treatment of acute bipolar and unipolar mixed depression. ClinicalTrials.govIdentifier: NCT04123301

Introduction

Mood disorders are highly prevalent conditions, associated with important burden and disability [1, 2]. Certain presentations are particularly difficult to identify, manage, and treat, such as mixed depression. A recent review [3] showed that the percentage of mixed features ranges from 4.3% [4] to 58.6% [5] in bipolar disorder (BD), and from 0% [4] to 34% [5] in major depressive disorder (MDD). This variability is explained by the fact that the DSM-5 mixed features specifier excludes the symptoms of distractibility, irritability, and psychomotor agitation, also known as “DIP symptoms” [6]. While mixed manic/hypomanic episodes are relatively easier to manage [7], mixed depression has been associated with greater depression severity [8], rapid cycling [9, 10], higher comorbidities with anxiety [11], impulsivity, and substance abuse [12, 13], worst sleep outcomes [14], higher relapse rates [15], refractoriness [16], and high suicide risk [12, 17].

Repetitive transcranial magnetic stimulation (rTMS) is a non-pharmacological treatment with proven effectiveness for unipolar depression [18]. Several data support the use of high-frequency rTMS over the left dorsolateral prefrontal cortex (DLPFC), and low frequency rTMS over the right DLPFC in treatment-resistant depression [19, 20]. Recently, a new form of rTMS, theta-burst stimulation (TBS) presented noninferiority and reduced delivery time compared to standard rTMS [21–25]. Analogously to rTMS, there are two modalities of TBS: intermittent [iTBS], with excitatory effects, and continuous [cTBS], with inhibitory effects [24], which are applied over the left and right DLPFC for depression. TBS and rTMS were less investigated for bipolar and mixed depression [26, 27]. In fact, studies using rTMS for mixed states are mostly restricted to patients with mixed manic episodes [27–31].

Taking into account these issues, the main objective of this randomized, sham-controlled trial was to evaluate the efficacy and safety of TBS as adjuvant therapy in BD or MDD patients in an acute depressive episode with mixed features after at least one previous failed trial. Our primary hypothesis is that active TBS would be superior to sham as an add-on treatment in improving depressive symptoms over a three-week treatment course.

Patients and methods

Study design

We conducted a single-center, double-blind, randomized, parallel-group, sham-controlled clinical trial (Clinicaltrials.gov identifier: NCT04123301) that lasted six weeks, comprising five consecutive days a week sessions for the first three weeks and then two days a week until week 6. The protocol was approved by the Ethics Committee of the University of Sao Paulo and was conducted in accordance with the Helsinki Declaration [32].

Participants were randomized using a computer-generated list in a 1:1 ratio and block randomization was performed allowing the permutation of the order and the size of the blocks. Allocation concealment consisted of sequentially numbered cards, which determined whether the active or sham TBS coil would be used. Thus, both participants and the personnel applying the stimulation sessions were blinded to the treatment group. Raters were also blinded to allocation group status.

Participants

All participants signed an informed consent form. We enrolled adults aged 18–65 years old diagnosed with BD type I, BD type II, or MDD in a moderate or severe acute depressive episode with mixed features according to modified DSM-5 criteria, i.e., including “DIP symptoms”. The diagnosis was confirmed using a Portuguese-validated version of the structured interview of DSM-IV (Structured Clinical Interview - SCID IV)[33] modified with DSM-5 mixed features criteria and allowing “DIP symptoms”.

The main inclusion criterion was presenting a Montgomery–Åsberg Depression Rating Scale [34, 35] score >19 at baseline. Also, patients had to present a Young Mania Rating Scale (YMRS) [36] score ≥1 on three or more items, according to criteria used in the International Mood Disorders Collaborative Project [4] and consistent with definitions of mixed depression [37–39]. Patients were necessarily using an appropriate first- or second-line pharmacological treatment for an acute MDD or BD depressive episode according to the Canadian Network for Mood and Anxiety Treatments (CANMAT) guidelines [7, 40]; therefore TBS was applied as an add-on treatment in patients’ refractory to at least one adequate treatment. Pharmacotherapy remained stable during the TBS intervention, and patients were using adequate pharmacological doses for at least four weeks prior to trial onset. Exclusion criteria were provided in the study protocol published previously [41].

Interventions

TBS sessions were performed using a MagPro X100 TMS device (Magventure, Lucernemarken, Denmark). Identical butterfly coils for active and sham stimulations were used. TBS was performed over the DLPFC bilaterally and the scalp localization was obtained through anatomical measurements (F3 and F4 positions determined by the 10–20 Electroencephalographic International System).

According to the inter-hemispheric asymmetry hypothesis right and left hemispheres have opposite effects on mood control and therapeutic applications of TMS should follow the paradigm of using high-frequency stimulation to activate the left DLPFC and low-frequency to suppress the right DLPFC in depression [42]. As there were no clinical trials evaluating TBS specifically in mixed depressive episodes, our clinical protocol was based on a study [43] that evaluated TBS in treatment-resistant MDD, showing that bilateral TBS stimulation (right cTBS and left iTBS) produced greater results in treating resistant major depressive episodes compared to iTBS over the left DLPFC. As mixed states are mostly mood states that are resistant to conventional treatments, we opted for using bilateral TBS in our study. Importantly, when this study was designed, the pivotal trial of Blumberger et al. [44] was not yet available.

The optimal dose for maximal outcome for TBS or TMS is not known; however evidence from clinical studies has suggested that TBS obeys a dose-response function, that is, a greater number of stimuli administered are capable of optimizing clinical results in patients with depression[45, 46].

The TBS doses varied from 18,000 [47, 48] to 36,000 pulses [49, 50] in the studies. In our study, cTBS was applied in the right DLPFC at the dose of 1,800 pulses per session [versus 21 sessions] and iTBS in the left DLPFC at the dose of 1,800 pulses per session [versus 21 sessions] resulting in a total applied dose of 75,600 pulses, much higher than the studies previously cited. The total study time was six weeks, as most studies in the area [44, 51, 52]. We defined three weeks as the primary endpoint thinking from a clinical point of view, as mixed depression is a condition associated with great distress and therefore considering that a faster response would be particularly desirable for this population.

The bilateral TBS sessions were applied in the following order: cTBS over the right DLPFC followed by iTBS over the left DLPFC. The following parameters were used: for cTBS, bursts of three pulses at 50 Hz (20-ms interval between stimuli) were applied continuously for 120 sec totaling 1800 pulses in the right DLPFC; for iTBS, bursts of three pulses at 50 Hz (20-ms interval between stimuli) were applied for 2 sec, being repeated every 10 sec for a total time of 570 sec; also totaling 1800 pulses in the left DLPFC. We used a 80% intensity of the motor threshold, which was the minimum necessary stimulus to generate a visible muscular contraction in the index finger in three out of five trials, according to the safety parameters published [53].

Assessments

The primary outcome was the comparison between active and sham groups regarding the change in MADRS scores from baseline to week 3 of intervention. Secondary outcomes included the comparisons between active and sham groups regarding response rate (defined as 50% or greater reduction in MADRS scores) from baseline to weeks 3 and 6; remission rate (defined as MADRS score lower than 11) [34] in weeks 3 and 6; changes in other scales such as the Hamilton Anxiety Scale (HAM-A)[54], Global Clinical Impression of Severity (GCI-S)[55], Global Assessment of Functioning (GAF) [56], the World Health Organization questionnaire of quality of life–brief version [57] and the Barratt Impulsivity Scale [58].

Frequency of treatment-emergent manic switch (TEMS), worsening of depression and other adverse events were assessed. Blinding efficacy was assessed at the end of week 6 by asking participants and the personnel applying the stimulation sessions about their allocation group.

Clinical assessments were conducted weekly until week 3 and thereafter a final assessment was conducted at week 6. Adverse events were evaluated daily during the first week and then weekly until week 6.

Statistical analyses

Analyses were performed using the lme4 [59], mice [60] and emmeans [61] packages of R version 3.6.3 [62]. The overall significance level was set at 0.05.

The sample size calculation was based in a clinical trial of treatment-resistant unipolar depression that found a reduction of 52.5% of the depression scale for the active group and a reduction of 17.4% for the sham group (F/X² = 6.166) [43]. Considering alfa and beta values of 0.05 and 0.1, respectively, we estimated 82 participants. A dropout rate of 20% was calculated, similar to recent studies in this field [25]. Thus, we enrolled 100 patients, 50 participants in each group.

The primary efficacy analysis was performed in the modified intention-to-treat – mITT – sample,

which were patients that completed at least one week of the intervention. Other assessments included the intention-to-treat (ITT) and per protocol (PP)samples. Continuous outcomes with more than two measurements were analyzed using 2-level linear mixed-effects models with the restricted maximum likelihood variance estimator. Continuous outcomes with two measurements were analyzed using a repeated-measures ANOVA, with group, time and their interaction as factors. The interaction between group and time is reported for these models. Frequency of TEMS and adverse events were compared among groups using Fisher’s exact test or the χ² test. To verify blinding integrity, we asked, at week 6, for participants and the personnel applying the stimulation sessions to guess whether the allocation group was active on a 0-100 scale; guessing scores were compared using a Chi-Square test between “real group” and “guessed group” by the participant or staff member.

Results

Participants

Out of 481 volunteers, 331 were screened and 231 were excluded due to several reasons. Out of 100 patients included, 90 and 63 of them composed mITT and PP samples, respectively. The dropouts that occurred before completing one trial week were eight in the active group and two in the sham group, all due to difficulties to reach the stimulation center. The dropouts that occurred before completing week three were 22 in the active group (n = 42) and nine in the sham group (n = 48) (p = 0.002). These significant unequal dropouts before week 3 were related to difficulties to reach the stimulation center. Dropouts due to adverse effects were listed in Fig. 1. The dropout rates that occurred in the whole six weeks of the study were not statistically different among active and sham groups (X-squared = 1.5444, df = 1, p value = 0.214). In terms of the number of subjects who dropped out from the study due to worsening of depression or suicide risk, there was no statistically significant difference between active (four dropouts, n = 48) and sham (three dropouts, n = 42) groups (OR = 1, CI 95% [0.17, 5.71], p = 1).

Fig. 1. Flow-chart diagram of screening, enroll and randomization.

Dropouts for each reason are provided beside each allocation group.

The groups were similar in terms of demographic characteristics, diagnosis, clinical course, scales scores, pharmacological treatments and previous electroconvulsive treatment at baseline. The whole sample consisted of 48% of patients with MDD, 33% with BD type II and 19% with BD type I. Fifty-eight patients (64%) were using at least one first-line treatment according to CANMAT guidelines and <1% had electroconvulsive therapy previously (Table 1). Unfortunately, we did not collect data regarding the duration of the current acute depressive episode. The overall prevalence of any anxiety disorder, any substance use disorder and any personality disorder were 61%, 5.5%, and 7.8%, and did not differ between groups (all ps > 0.24).

Table 1.

Baseline clinical and demographic characteristics in the mITT sample.

	Sham (n = 48+)	Active (n = 42)	Total (n = 90)	P value
Demographic characteristics
Age, mean [SD]	38.0 (10.85)	40.8 (9.98)	39.4 (10.4)	0.196
Gender, fem (%)	33 (68.8)	31 (73.8)	64 (82.2)	0.768
Ethnicity, Caucasian (%)	30 (62.5)	29 (69.0)	59 (65.5)	0.848
Marital status, married (%)	38 (79.1)	31 (73.8)	69 (76.7)	0.727
Number of children, mean [SD]	0.54 (0.99)	0.67 (0.87)	0.65 (0.93)	0.529
Years at school, mean [SD]	15.33 (3.87)	14.31 (4.18)	14.82 (4.02)	0.231
Employment status, not employed (%)	28 (58.3)	23 (54.8)	51 (56.7)	0.898
Diagnosis and clinical course
Bipolar type I, n (%)	10 (20.8)	7 (16.7)	17 (18.9)	0.815
Bipolar type II, n (%)	19 (39.6)	11 (26.2)	30 (33.3)	0.262
Major depressive disorder, n (%)	19 (39.6)	24 (57.1)	43 (47.8)	0.146
Recurrent depression, n (%)	42 (87.5)	36 (85.7)	78 (86.7)	0.127
Melancholic depression, n (%)	44 (91.6)	42 (100.0)	86 (95.5)	0.120
Atypical depression, n (%)	3 (0.06)	0 (0.0)	3 (0.03)	0.245
Previous psychotic depression, n (%)	5 (0.1)	4 (0.09)	9 (0.1)	1.0
Previous hospitalization due to depression, n (%)	9 (18.8)	10 (23.8)	19 (21.1)	0.108
Baseline scales scores
MADRS, mean [SD]	34.54 (5.71)	35.43 (6.45)	34.98 (6.08)	0.491
YMRS, mean [SD]	9.71 (3.02)	10.10 (3.27)	9.90 (3.14)	0.561
CGI-S, mean [SD]	4.58 (0.74)	4.60 (0.73)	4.59 (0.73)	0.939
GAF, mean [SD]	34.62 (10.07)	34.19 (10.32)	34.40 (10.19)	0.84
HAM-A, mean [SD]	20.67 (7.16)	21.83 (7.39)	21.25 (7.27)	0.449
WHOQol – brief, mean [SD]	67.88 (10.22)	65.86 (13.83)	66.87 (12.02)	0.43
BIS, mean [SD]	68.38 (9.33)	68.26 (9.48)	68.32 (9.40)	0.955
Comorbidities
Any anxiety disorder, n (%)	31 (64.5)	24 (57.1)	55 (61.1)	0.613
Any substance use disorder, n (%)	3 (6.2)	2 (4.7)	5 (5.5)	1
Any personality disorder, n (%)	2 (4.2)	5 (12.0)	7 (7.8)	0.245
Pharmacological treatments
Bipolar type I
First line treatmentsa, n (%)	8 (16.7)	6 (14.3)	14 (15.5)	0.984
Second line treatmentsb, n (%)	3 (6.2)	1 (2.4)	4 (0.04)	0.62
Bipolar type II
First line treatmentsc, n (%)	0 (0.0)	1 (2.4)	1 (0.01)	0.467
Second line treatmentsd, n (%)	19 (39.6)	9 (21.4)	28 (31.1)	0.104
Major depressive disorder
First line treatmentse, n (%)	19 (39.6)	24 (57.1)	43 (47.7)	0.146
Second line treatmentsf, n (%)	0 (0.0)	4 (9.5)	4 (0.04)	0.044
Previous neuromodulation treatment
Electroconvulsive therapy, n (%)	4 (8.3)	2 (4.8)	6 (0.06)	0.681

SD Standard deviation, MADRS Montgomery–Åsberg depression rating scale, YMRS Young mania rating scale, CGI-S Clinical global impression-severity of illness, GAF Global assessment of functioning, HAM-A Hamilton anxiety rating scale, WHOQol-brief World Health Organization quality of life questionnaire (brief version), BIS Barratt Impulsivity Scale, SSRI Selective serotonin reuptake inhibitor.

P values represent the significance of Chi-Square or Fisher exact tests for categorical and t-tests for continuous variables. SD standard deviation.

^alithium or quetiapine or lamotrigine or lurasidone or lithium/divalproex + lamotrigine or lithium/divalproex + lurasidone.

^bolanzapine + fluoxetine or divalproex or lithium/divalproex + SSRI or lithium/divalproex + bupropion.

^cquetiapin.

^dlithium or lamotrigine or bupropion or sertraline or venlafaxine.

^eagomelatine or bupropion or citalopram or desvenlafaxine or duloxetine or escitalopram or fluoxetine or fluvoxamine or mirtazapine or paroxetine or sertraline or venlafaxine or vortioxetine.

^ftricyclic antidepressant or trazodone or quetiapine.

Main findings

Primary outcome

Linear mixed-effect models showed no clinical superiority of active stimulation, as no significant differences were observed in MADRS scores between the two groups (Fig. 2). The least squares mean difference in MADRS scores at week 3 was −0.06, 95%CI [−3.39, 3.51], p = 0.97, in favor of sham stimulation (Table 2).

Fig. 2. Change in MADRS score over time.

Circles represent means in each group, bars represent ±1 standard deviation from the mean. In orange, participants using active-TBS; in gray participants using sham-TBS. The y-axis indicates the values of the MADRS score in the mITT sample. On the x-axis the values at each time of the assessment are indicated (baseline, first week, second week, third week, and sixth week).

Table 2.

Linear mixed models for imputed mITT sample.

	Sham		Active		Least squares mean difference (95% CI)		Group and time interaction	P value
	N	Mean [SD]	N	Mean [SD]
Primary outcome
MADRS, baseline to week 3
MADRS, baseline	48	34.54 (5.71)	42	35.43 (6.45)	0.0629 (−3.3867 to 3.5125)		F1,15.24 = 0.002120649	0.9715
MADRS, week 3	41	17.98 (10.04)	28	19.36 (11.16)
Secondary outcomes
MADRS, baseline to week 3 (stratified for MDD)
MADRS, baseline	19	33.95 (6.48)	24	33.88 (7.37)	−0.101 (−4.8834 to 4.6814)		F1,33.24 = 1.146865	0.9673
MADRS, week 3	16	16.44 (7.94)	19	20.11 (11.15)
MADRS, baseline to week 3 (stratified for BD)
MADRS, baseline	29	34.93 (5.22)	18	37.5 (4.37)	0.531 (−4.3102 to 5.3722)		F1,17.84 = 0.6010476	0.8308
MADRS, week 3	25	18.96 (11.22)	9	17.78 (11.68)
MADRS, baseline to week 6
MADRS, week 6	36	14.11 (11.83)	28	16.5 (11.98)	0.332 (−3.1764 to 3.8404)		F1,37.54 = 0.2181534	0.8532
YMRS, baseline to week 3 and 6
YMRS, baseline	48	9.71 (3.02)	42	10.1 (3.27)	0.282 (−0.9038 to 1.4678)		F1,116.59 = 0.2275535	0.6426
YMRS, week 3	41	4.15 (3.42)	28	4.54 (3.52)
YMRS, week 6	36	3.53 (4.18)	28	3.57 (3.65)	0.231 (−0.9254 to 1.3874)		F1,111.35 = 0.001526855	0.6965
HAM-A, baseline to week 3 and 6
HAM-A, baseline	48	20.67 (7.16)	42	21.83 (7.39)	1.45 (−0.9804 to 3.8804)		F1,58.06 = 0.0146615	0.245
HAM-A, week 3	41	9.9 (7.4)	28	12.61 (8.71)
HAM-A, week 6	36	9.06 (8.22)	28	11.43 (7.99)	1.46 (−1.0292 to 3.9492)		F1,93.45 = 0.09531674	0.2548
CGI-S, baseline to week 3 and 6
CGI-S, baseline	48	4.58 (0.74)	42	4.6 (0.73)	−0.0126 (−0.2987 to 0.2735)		F1,121.38 = 0.09647006	0.9311
CGI-S, week 3	41	3.63 (1.04)	28	3.71 (1.08)
CGI-S, week 6	36	2.78 (1.33)	28	3.25 (1.29)	−1.37 (−5.0548 to 2.3148)		F1,127.03 = 0.08313419	0.4702
GAF, baseline to week 3 and 6
GAF, baseline	48	34.63 (10.07)	42	34.19 (10.32)	−1.46 (−5.0076 to 2.0876)		F1,96.52 = 0.9209387	0.4222
GAF, week 3	41	52.29 (13.14)	28	47.79 (15.41)
GAF, week 6	36	61.47 (17.08)	28	57.36 (18.51)	−1.37 (−5.0548 to 2.3148)		F1,127.03 = 0.08313419	0.4702
WHOQol, baseline to week 6
WHOQol, baseline	48	67.88 (10.22)	42	65.86 (13.83)	−1.94 (−7.1732 to 3.2932)	F1,25.95 = 0.000878861		0.4688
WHOQol, week 6	36	75.56 (14.06)	28	74.93 (19.91)
BIS, baseline to week 6
BIS, baseline	48	68.38 (9.33)	42	68.26 (9.48)	−1.94 (−7.1732 to 3.2932)		F1,169.09 = 0.13873	0.6528
BIS, week 6	36	64.14 (10.92)	28	63 (8.69)

SD Standard Deviation, MADRS Montgomery–Åsberg depression rating scale, YMRS Young mania rating scale, CGI-S Clinical global impression-severity of illness, GAF Global assessment of functioning, HAM-A Hamilton anxiety rating scale, WHOQol-brief World Health Organization quality of life questionnaire (brief version), BIS Barratt Impulsivity Scale. P values represent the significance of Chi-Square or Fisher exact tests for categorical and t-tests for continuous variables. SD standard deviation.

Secondary outcomes

Response and remission rates per MADRS were also not statistically different among active and sham groups (35.7% vs. 43.7%, and 28.5% vs. 37.5% respectively at week 6, ps > 0.51). No other analyses from baseline to weeks 3 or 6 revealed significant time x group interaction or mean differences among groups in the mITT sample, including changes in YMRS score, CGI-S, GAF score, HAM-A score, WHOQol score, BIS score, and response and remission rates at week 3 (Tables 2 and 3). Results in ITT and PP samples are provided in Supplementary material.

Table 3.

MADRS response and remission in mITT sample.

	Sham	Active	χ2	P value
MADRS, baseline to week 3
Response	21/48	10/42	3.1108	0.0777
Remission	14/48	7/42	1.3202	0.2506
MADRS, baseline to week 6
Response	21/48	15/42	0.31436	0.575
Remission	18/48	12/42	0.452	0.5014

MADRS Montgomery–Åsberg depression rating scale, mITT modified intention-to-treat sample, ITT Intention-to-treat sample, PP per protocol sample.

P values represent the significance of Chi-Square test. χ2, Chi-Square.

The blinding integrity assessed through Pearson’s Chi-squared test with Yates’ continuity correction was guaranteed in relation to the participants (X-squared = 4.7661e-31, df = 1, p value = 1), but it was not preserved in relation to the staff (X-squared = 5.5815, df = 1, p value = 0.01815). TEMS was not statistically different among active and sham groups (p = 0.5). TMS side effects were not different significant among active and sham groups.

Discussion

To our knowledge, this is the first randomized, double-blind, sham-controlled clinical trial assessing the efficacy, safety, and tolerability of bilateral TBS for the treatment of major depressive episodes with mixed features of BD and MDD. Our primary hypothesis that active TBS would be superior to sham TBS was not demonstrated. Moreover, secondary outcomes as response and remission rates and changes in other scales were not different among active and sham groups from baseline to weeks 3 or 6. Active and sham TBS were similar in the rate of adverse events. Our results were similar for both BD type I, BD type II, and MDD patients.

We enrolled a real-world sample without excluding other comorbidities and using at least one first or second-line treatment according to CANMAT guidelines for BD or MDD. We included patients who had failed at least one treatment trial, although most had >3 depressive episodes throughout their lives. Although it is not recommended the use of antidepressants for bipolar mixed depression, they are widely used for treatment-resistant MDD, many of them are mixed states, so that we allowed MDD patients to receive conventional antidepressants, as there is only one trial [63] assessing the treatment of MDD with mixed features. Blinding integrity was guaranteed in relation to the study participants, but it was not preserved in relation to the personnel. The personnel guessed the group in which the patients were due to physical characteristics observed during stimulation, such as the noise of the active and sham coils. The personnel did not have access to the clinical evaluations performed and, therefore, were not aware of whether the patients were improving or worsening with the treatment, they only applied the stimulation sessions. Therefore, this finding does not affect the internal validity of the study since the clinical evaluations of the primary and secondary outcomes were carried out by an independent team.

According to the interhemispheric asymmetry theory, right and left hemispheres have opposite effects on mood control and therapeutic applications of TMS should follow this paradigm [64]. The few neuroimaging studies in mixed states to date support the hypothesis of lateralization of brain abnormalities in relation to depressive and manic symptoms, suggesting that neurofunctional abnormalities, preferentially located in the frontal and limbic areas of the right hemisphere, may be associated with the depressive component whereas abnormalities in similar regions on the left would be associated with the manic component [65]. An open and uncontrolled study demonstrated that rTMS over the right DLPFC over three weeks led to a response rate of 46% and 15% in depressive and manic scales, respectively, in BD type I mixed state [27]. On the other hand, a double-blind and sham-controlled clinical trial demonstrated that bilateral TBS (right cTBS and left iTBS) produced greater results in treatment-resistant major depressive episodes than unilateral stimulation [49]. As a result, we opted to use bilateral TBS (right cTBS and left iTBS) protocol in mixed depression treatment.

Mixed states are difficult to treat and are often hidden in treatment-resistant samples [66–71]. In particular, the treatment of depressive symptoms in mixed depression represents a clinical dilemma, mainly because conventional antidepressant medications commonly worsen instability and intra-episodic mood changes [72–74]. Most rTMS clinical trials were carried on in treatment-resistant depression samples and different protocols have been reported as effective and safe, with a low risk of (hypo)manic switches, suggesting a likely mood-stabilizing effect [75]. Thus, we hypothesized that mixed depressions would be adequately treated with rTMS. However, our results did not corroborate this hypothesis. Furthermore, the findings in the complete ITT analysis (see Supplementary material) suggested that active TBS could be inferior compared to sham TBS in treating mixed depression, i.e., active TBS could worsen mixed depression as conventional antidepressants do. Moreover, although TEMS did not differ among groups, the two manic switches during the trial occurred in the active group, suggesting that TBS could actually worsen manic symptoms.

Earlier studies had already indicated that patients with mixed depression presentation were similar to BD patients in terms of early-onset, recurrence, positive family history of bipolar disorder and refractoriness [76, 77]. Although mixed states were commonly described in BD, recent research proved that mixed depression could also occur in MDD [78–80]. Several systematic reviews [20, 81–84] evaluated the efficacy of TMS in the treatment of major depressive episodes of mixed samples of BD and MDD patients and TMS is approved by the FDA for the treatment of major depressive episodes regardless of primary diagnosis. Nevertheless, rTMS randomized and sham-controlled clinical trials exclusively in bipolar samples are scarce. Although there is preliminary evidence of the efficacy of high-frequency rTMS [85, 86] and high-frequency deep TMS [87] in BD, the data in regard to novel protocols such as TBS have been inconclusive [88]. Indeed, there is a risk of extending data from MDD dominant samples to BD when novel protocols such as TBS have evidence of efficacy only in MDD. In consonance of growing evidence that TBS is less effective in bipolar depression compared to MDD [89–91], our data indicate that TBS is also less effective in mixed depression compared to no-mixed depressive episodes treatments.

TMS trials in manic episodes have been done with high-frequency rTMS in the right prefrontal cortex; however, the definitive evidence of the effectiveness of TMS in the treatment of mania is not yet available [92–97]. Another feasible way to treat manic symptoms could be iTBS delivery in the right DLPFC. Nevertheless, it is unknown whether stimulating the right DLPFC in a mixed depression could improve manic symptoms but worsen depression. Further studies are needed to elucidate these issues.

One first potential limitation of this trial is the absence of use of MRI-guided neuronavigation in every session - an approach that is not feasible or cost-efficient for most rTMS clinics. However, it has been previously demonstrated that the DLPFC target used in this trial can be accurately localized without MRI via a scalp-based measurement known as BeamF3 [98]. Thus, the present findings can be generalized more broadly to rTMS clinics where MRIguidance is unavailable. A second limitation is the short duration of this clinical trial that comprised a six-week intervention. Despite this, the most recent researches with TBS had two-week follow-up [47–50, 99, 100]. Further studies can confirm whether our findings can be generalized for longer follow-up clinical trials. Here, we opted to describe the main results of our study according to pre-established outcomes reported previously (e.g., clinicaltrials.gov). Notwithstanding, it is indeed important to further explore, in post-hoc approaches, possible predictors associated with our findings. This will be examined in future studies.

In conclusion, we have found that active bilateral TBS was not superior compared to sham as an effective add-on treatment in moderate to severe mixed depression of BD I, BD II or MDD. Our findings show that TBS does not have a mood-stabilizing effect and may even worsen mood stability in mixed depression which can guide clinicians’ decision with regard to the best TMS protocols in mixed depressed samples.

Funding

This study received funding from the Research foundation support agency of the state of Sao Paulo (FAPESP): 2017/19237-1. Period: July, 2018 to June, 2020.

Author contributions

Conceived and designed the clinical trial: DFT, ARB, RAM. Contributed to the acquisition, analysis, or interpretation of data for the work: DFT, PS, CGRS, DHM, LCLV, IK, LB, PMF, ARB, RAM. Wrote the first draft of the manuscript: DFT. Contributed to the writing of the manuscript: ARB, RAM.

Competing interests

DFT worked as a speaker and produced scientific source during the last two years for the following pharmaceutical companies: Cristália, Aché, Torrent, Abbott, Lundbeck. ARB reports grants from SaoPaulo Research State Foundation (FAPESP, 2017/50223-6, 2018/10861-7, 2019/06009-6), Newton Advanced Fellowship (12/1010), Brazilian National Council of Scientific Development Productivity Support (PQ-1B) and University of Sao Paulo Medical School Productivity Support (PIPA-A). . RAM reports grants from Sao Paulo Research State Foundation (FAPESP) and from the National Research Council (CNPq). Other authors declare no conflicts of interest.

Supplementary information

The online version contains supplementary material available at 10.1038/s41386-021-01080-9.