A Dose-Finding, Biomarker Validation, and Effectiveness Study of Transcranial Magnetic Stimulation for Adolescents With Depression

Abstract

Objective:

Research and clinical application of transcranial magnetic stimulation (TMS) for adolescents with major depressive disorder (MDD) has advanced slowly. Significant gaps persist in our understanding of optimized, age-specific protocols and dosing strategies. This study aimed to compare the clinical effects of 1 Hz versus 10 Hz TMS regimens and examine a biomarker-informed treatment approach with glutamatergic intracortical facilitation (ICF).

Method:

Participants with moderate-to-severe symptoms of MDD were randomized to 30 sessions of left prefrontal 1 Hz or 10 Hz TMS, stratified by baseline ICF measures. The primary clinical outcome measure was the Children’s Depression Rating Scale, Revised (CDRS-R). The CDRS-R and ICF biomarker were collected weekly.

Results:

Forty-one participants received either 1 Hz (n = 22) or 10 Hz (n = 19) TMS treatments. CDRS-R scores improved compared to baseline in both 1 Hz and 10 Hz groups. For participants with low ICF at baseline, the overall least squares means of CDRS-R scores over the 6-week trial showed that depressive symptom severity was lower for the group treated with 1 Hz TMS than for those who received 10 Hz TMS. There were no significant changes in weekly ICF measurements across the 6 weeks of TMS treatment.

Conclusion:

Low ICF may reflect optimal glutamatergic N-methyl-d-aspartate (NMDA) receptor activity that facilitates the therapeutic effect of 1 Hz TMS through long-term depression-like mechanisms on synaptic plasticity. The stability of ICF suggests that it is a tonic, trait-like measure of NMDA receptor-mediated neurotransmission, with potential utility to inform parameter selection for therapeutic TMS in adolescents with MDD.

INTRODUCTION

There are limited evidence-based treatments for major depressive disorder (MDD) in adolescents. Current first-line treatment options for MDD in adolescents include fluoxetine, escitalopram, and evidence-based psychotherapies,¹ yet forty percent of adolescents with MDD do not respond to standard interventions.^1–3 Interventions for treatment-resistant depression (TRD) in adolescents are markedly inadequate. Various forms of transcranial magnetic stimulation (TMS) are now cleared by the U.S. Food and Drug Administration (FDA) and are widely available for the treatment of TRD in adults.^4–6 Although there was a recent FDA clearance of TMS as adjunctive treatment for MDD in adolescents aged 15 years and older,⁷ the development, study, and clinical implementation of TMS for adolescents has been a protracted process.^8,9 Concerns about safety in the developing brain, methodological challenges unique to adolescents, and limited resources are formidable barriers for research and adaptation of therapeutic TMS for MDD in adolescents.^10,11

Prior TMS research in adolescents with MDD has yielded mixed findings. A multisite, randomized, sham-controlled trial of TMS for adolescents with MDD found no difference between active stimulation and sham treatment on depressive symptoms.⁸ More recent efforts examining adjunctive TMS with novel study designs have been more promising.¹² Elevated placebo response rates have been demonstrated consistently in clinical trials of various interventions for adolescents with MDD.^11,13,14 Notably, the structural, procedural, and technological aspects of TMS enhance nonspecific antidepressant effects in adolescent studies.^8,15,16 Early TMS research in adolescents, including the randomized controlled trial,⁸ applied standard 10 Hz TMS protocols with stimulation parameters similar to adult studies.¹⁶ Excitatory and inhibitory neurotransmission, which is posited to underlie the mechanisms of TMS, undergoes considerable development during adolescence.¹⁷ As a result, adolescents have increased neurophysiological heterogeneity compared to adults, which likely impacts the therapeutic effects of TMS.^18–21 Yet, the effects of manipulating stimulation parameters for TMS in youth has not been studied systematically, and no trials to date have examined how developmental neurophysiologic heterogeneity intersects with different forms of TMS stimulation to achieve clinical effects.

Biomarker-stratified approaches to evaluating therapeutic mechanisms of TMS are needed to inform how stimulation parameters, such as pulse frequency, influence clinical outcomes within the context of neurophysiologic development. One promising marker, intracortical facilitation (ICF), is an index of cortical excitability measured noninvasively using paired-pulse TMS with quantification of evoked potentials on electromyography (Figure 1A–1B). Previous mechanistic research indicates that ICF measures glutamatergic N-methyl-d-aspartate (NMDA) receptor-mediated neurotransmission.^18,22–24 Prior work also has implicated glutamatergic ICF as a developmentally-unique marker of depressive symptoms in adolescents.¹⁸ Notably, this differs from adults, in whom deficient γ-aminobutyric acid (GABA)-ergic inhibition is a marker of MDD, suggesting that the pathophysiology of depression may evolve across the lifespan.²⁵ However, it is unknown if ICF predicts treatment response or if ICF measures change with therapeutic brain interventions. For example, ICF may be modulated differentially using high-frequency stimulation paradigms (e.g., 10 Hz TMS) that are thought to increase cortical excitability, versus lower frequencies (e.g., 1 Hz TMS) that enhance inhibition.^26,27

Figure 1.

Biomarker Measurement and Transcranial Magnetic Stimulation Treatment.

Note: A) Measurement of the intracortical facilitation (ICF) biomarker using a figure-of-eight TMS coil over the left primary motor cortex. B) Electromyography of paired-pulse ICF paradigm, with a subthreshold conditioning stimulus preceding suprathreshold test stimulus and the resulting facilitated motor evoked potential. C) TMS treatment with the coil placed over the left dorsolateral prefrontal cortex. D) Schematic of glutamatergic neurotransmission involving NMDA and AMPA ionotropic glutamatergic receptors underlying long-term potentiation (LTP) and long-term depression (LTD), proposed as potential mechanisms of 10 Hz and 1 Hz TMS, respectively.

With these considerations in mind, we conducted a double-blind, biomarker-stratified effectiveness trial of 1 Hz vs. 10 Hz TMS in adolescents with MDD. In our study, the treatment randomization was stratified by individual participants’ baseline ICF measurements. Aligning with the National Institute of Mental Health experimental medicine initiative,²⁸ the study treatment arms had identical TMS parameters, with the exception of stimulus frequency. Clinical trials most commonly deliver 10 Hz TMS to the left dorsolateral prefrontal cortex (DLPFC) and 1 Hz TMS to the right DLPFC.²⁹ Conceptually, the left DLPFC is thought to have decreased activity in patients with depression, while the right DLPFC has relative hyperactivity. It has been assumed that 10 Hz TMS has an excitatory effect for the hypoactive left DLPFC, and that 1 Hz stimulation has an inhibitory effect on increased activity in the right DLPFC, and respectively this underlies the antidepressant mechanisms of action for lateralized TMS treatments. However, these theories and assumptions have not been studied rigorously, and other research suggests that any TMS dosing has similar bilateral hemispheric effects.^9,29 In our study, both treatment arms (1 Hz and 10 Hz) targeted the left DLPFC (Figure 1C). This approach was selected to isolate effects of frequency in the TMS dosing in order to bolster the rigor of the study design. The left (rather than the right) DLPFC was selected as the left DLPFC is the most common cortical target used for TMS treatments for depression in clinical practice.²⁹

The primary aim was to assess whether TMS frequency (1 Hz vs. 10 Hz) had differential clinical effects (improvement in depressive symptoms) based on participants’ baseline glutamatergic neurophysiology (ICF). We hypothesized that participants with low ICF at baseline would have superior outcomes with 10 Hz TMS (speculating that an excitatory form of TMS would increase ICF), while participants with high ICF at baseline would have superior outcomes with 1 Hz TMS (conversely speculating that an inhibitory form of TMS would decrease ICF). A secondary aim was to examine ICF as an index of potential mechanisms for changes in depressive symptoms with 10 Hz and 1 Hz stimulation (Figure 1D). We hypothesized that changes in depressive symptom severity would have direct correlations with changes in weekly ICF measures with 1 Hz TMS treatment, via inhibitory mechanisms, and indirect correlations with changes in weekly ICF measures with 10 Hz TMS treatment, via excitatory mechanisms. Hence, the study framework was positioned to mitigate sham/placebo effects and to advance understanding of adolescent neurophysiology irrespective of clinical outcomes.

METHOD

Study Design

The study aims were examined with a randomized, double-blind, biomarker-stratified design. This trial was prospectively registered (NCT03363919), an investigational device exemption was obtained from the FDA (G170212), and the study was reviewed and approved by the Mayo Clinic Institutional Review Board prior to any research activities. Participants and parents/guardians provided informed assent and consent before any study activities.³⁰

Participants

Participants were aged 12–18 years with moderate-to-severe MDD based on a structured diagnostic interview, the Mini-International Neuropsychiatric Interview (MINI/MINI-KID),^30–32 and depressive symptom severity corresponding to a score of 40 or greater on the Children’s Depression Rating Scale, Revised (CDRS-R).³³ The duration of the current depressive episode was at least 4 weeks but no greater than 3 years. Participants were not taking antidepressants or other psychotropic medications during the study. Participants who had insufficient benefit from a recent antidepressant trial underwent a taper. The medication taper had clinician supervision with a washout period of at least 1 week (4 weeks for fluoxetine). Primary or concurrent psychotic disorders, bipolar disorders, anorexia nervosa, bulimia nervosa, and substance use disorders within the past year were exclusionary. Standard precautions for prevention of induced seizures and general safety were implemented in the inclusion and exclusion criteria (please see the Supplement 1, available online).³⁴ Demographic information was ascertained with a clinical interview and prior medical records (please see Supplement 1, available online, for a detailed description).

Neurophysiology Biomarker

A paired-pulse TMS ICF biomarker of NMDA glutamatergic-mediated neurotransmission was used as the stratification (or moderator) variable for the primary aim and as the secondary neurobiological outcome. The treatment randomization (1 Hz vs. 10 Hz) was stratified by low (≤1.5) or high (>1.5) baseline ICF. The selection of the 15-milliseconds (ms) interstimulus interval and a cutoff value of 1.5 were determined by analysis of data from a previous sample of 71 adolescents (aged 12–18 years) with depression who underwent ICF measurement. In that sample, an ICF cutoff value of 1.5 for demarcating low vs. high ICF resulted in an approximate median split of the ICF distribution (Figure S1, available online). When ICF was measured at the 15-ms interval, there was no difference in depression severity (p = .89) between depressed youth with high ICF (>1.5) and those with low ICF (<1.5).

For the present study, ICF measures were collected at baseline and weekly (after every 5 TMS sessions) using methods consistent with prior published protocols.^18,22,24 In brief, surface electromyography (EMG) recorded right abductor pollicis brevis (APB) muscle activity while single- and paired-pulse TMS stimulation was applied to the contralateral (left) primary motor cortex using the Magstim BiStim² system, consisting of two magnetic stimulators and a 70-mm diameter figure-of eight coil (Magstim Company Ltd, Whitland, Carmarthenshire, UK). The TMS coil was held tangentially on the scalp at 45 degrees to the midline, moving the coil in 1-cm increments over the motor cortex to localize the optimal site for APB stimulation. The resting motor threshold (RMT),³⁴ defined as the minimum stimulation intensity that elicited a motor evoked potential (MEP) of >50 microvolts in 5 of 10 trials with a relaxed APB muscle, was measured first using single-pulse stimulation. Subsequently, ICF was measured using a subthreshold conditioning stimulus (CS) at 80% RMT that preceded stimulation with a suprathreshold test stimulus (TS), which was calibrated to produce an average MEP of 0.5 to 1.5 mV peak-to-peak amplitude on EMG in the contralateral APB muscle. The CS and TS were separated by 15 ms for this ICF biomarker. Change in MEP amplitude generated by the TS was expressed as a ratio of the mean unconditioned MEP amplitude.

Clinical Assessments

Clinical assessments were collected at screening, baseline, and weekly (after every 5 TMS treatments). Participants were excluded if their baseline CDRS-R score had a decrease of 25% or greater from the screening score. Clinical assessment measurements were collected by the principal investigator (PI; PEC) or trained clinical raters supervised by the PI. Pre- and post-treatment safety assessments included auditory thresholds, height, weight, blood pressure, heart rate, and the National Institutes of Health Toolbox Cognition Battery.³⁵ The PI and all raters were blind to treatment assignment and were not allowed in the treatment suite during therapeutic TMS sessions.

TMS Protocol

Participants were stratified using baseline ICF testing and then randomized to either 1 Hz TMS (2400 continuous pulses per session) or 10 Hz TMS (trains of 40 pulses over 4 seconds, with 36 seconds between pulse trains, for 2400 pulses per session). Both 1 Hz and 10 Hz treatments were provided using the NeuroStar device (Neuronetics, Inc., Malvern, PA, USA). Stimulation in both arms was applied to the left DLPFC at 120% RMT (determined by visual inspection of APB movement). When participants did not tolerate TMS stimulation at 120% RMT at the first session, the stimulation was applied at the highest tolerable intensity. The TMS intensity was then escalated to 120% RMT within the first week. This titration approach for tolerability is consistent with prior research protocols and clinical practice. Sessions in each treatment arm had identical stimulus intensity (120% RMT), session duration (40 minutes), number of pulses (2400), and treatment location (left DLFPC), differing only in stimulus frequency (1 Hz vs. 10 Hz). The TMS coil was localized to the left DLFPC using the Beam F3 method³⁶ based on available resources, practical factors, and an intent to examine TMS as commonly delivered in clinical practice.²⁹ Prior work also suggests that the Beam F3 approach reasonably approximates magnetic resonance imaging-guided structural neuronavigation.³⁷

Blinding Procedures

Participants and study team members who completed the clinical assessments were blinded to treatment assignment. Adolescent participants and parents had no prior TMS exposure or experience and were not informed of treatment assignment. Participants did not discuss stimulation sessions with clinical raters. All adolescents, parents, and raters were asked to guess which type of TMS (1 Hz vs. 10 Hz) was delivered at the end of treatment.

Outcome Variables

The primary and secondary outcomes were depressive symptom severity (CDRS-R total score) and ICF, respectively, measured weekly over the 6-week trial. Both the CDRS-R total score (two-way random effects ICC = 0.86; 95% CI: 0.68 to 0.93) and the ICF-15 (two-way random effects ICC = 0.66; 95% CI: 0.43 to 0.82) demonstrated good test-retest reliability across the 6 weeks.

The CDRS-R consists of 17 items. The first 14 items of the CDRS-R are rated from the patient’s and parent’s responses, with the clinician providing a summary score considering scores of both parent and child interviews. The last three items are based on rater observations (facial affect, speech, and hypoactivity). Items are rated on five-point (items 4, 5, and 16) or seven-point (items 1–3, 6–15, and 17) Likert-type scales. Total scores range from 17 to 113 (with a greater score representing greater depressive symptom severity).

Independent Variable and Covariates

The primary independent variable was treatment with either 10 Hz TMS or 1 Hz TMS. Biomarker status [high baseline ICF (>1.5) vs. low baseline ICF (≤1.5)] was a moderator variable in the primary CDRS-R model. Age (years), sex at birth, number of previous failed medication trials, baseline CDRS-R total score (for the primary CDRS-R model), baseline ICF (for the secondary ICF model), and CDRS-R total score as a time-varying covariate (for the secondary ICF model) were included as covariates in the respective models. These variables, selected a priori, were included as covariates in the models to bolster precision in evaluating the effect of TMS treatment on the outcomes.

Statistical Analysis

Demographic and clinical characteristics for the sample of youth were described using the sample mean and standard deviation for continuous variables and the frequency and percentage for categorical variables. Differences between the characteristics of the two groups [10 Hz TMS (n = 19) vs. 1 Hz TMS (n = 22)] were identified using the two-independent sample t-test with the Satterthwaite method for unequal variances (continuous variables) and Fisher’s exact test (categorical variables).

Depression severity (measured by CDRS-R total score) over the 6-week treatment period was the primary continuous outcome measure. A linear mixed model analysis of repeated measures was used to evaluate the TMS treatment (low frequency 1 Hz TMS vs. high frequency 10 Hz TMS) by biomarker status (high baseline ICF vs. low baseline ICF) interaction effect on depression severity (CDRS-R total) over the 6-week trial. The mixed model contained fixed effects terms for TMS treatment, baseline ICF status, time (categorical), TMS treatment × baseline ICF × time interaction, TMS treatment × baseline ICF interaction, TMS treatment × time interaction, and baseline ICF × time interaction. Baseline CDRS-R total, age, sex at birth, and number of previously failed medication trials were included as covariates in the model. Restricted maximum likelihood estimation along with Type 3 tests of fixed effects were used with the Kenward-Roger correction applied to the autoregressive (AR1) covariance structure. Least squares (LS) means (adjusted TMS group means) were estimated as part of the mixed model in order to interpret the TMS group effect (LS mean difference between groups). Simple TMS group effects at each time point as well as within-group change over the 6 weeks were also assessed. Cohen’s d was calculated and interpreted as the effect size estimator.

Intracortical facilitation (ICF) over the 6-week study period was the secondary continuous outcome measure. The change in ICF over time was compared between the TMS treatment groups using a linear mixed model analysis of repeated measures similar to that described above. The mixed model contained fixed effects terms for TMS treatment, time, and TMS treatment × time interaction. Baseline ICF, time-varying CDRS-R total, age, sex at birth, and number of previously failed medication trials were included as covariates in the model.

A separate linear mixed model repeated measures analysis like that described above was also implemented to examine the relationship between ICF (as the outcome) and depressive symptom severity (CDRS-R total as a time-varying covariate) over 6 weeks of TMS treatment. The model contained fixed effects terms for TMS treatment, time-varying CDRS-R total, time, and the TMS treatment × time-varying CDRS-R total interaction. Baseline CDRS-R total, baseline ICF, age, sex at birth, and number of previously failed medication trials were included as covariates in the model. The parameter estimates (regression coefficients) were interpreted from the solution for fixed effects in the mixed model analysis for those who received low frequency (1 Hz) TMS treatment and for those who received high frequency (10 Hz) TMS treatment.

Statistical analyses were conducted using SAS software, version 9.4 (SAS Institute, Inc., Cary, NC). The level of significance was set at α = 0.05 (two-tailed), and the False Discovery Rate (FDR) procedure³⁸ was used to control false positives over the multiple tests.

RESULTS

Participant Characteristics

The study randomized 47 adolescents to 1 Hz or 10 Hz TMS, and 46 received at least one TMS treatment. Five youth (four receiving 10 Hz and one receiving 1 Hz TMS) exited the study prior to week 1 assessments. Data from the remaining 41 adolescents (1 Hz, n = 22; 10 Hz, n = 19) were analyzed (details available in Supplement 2 [CONSORT diagram], and Figure S2, available online). Of the 41 youth, 60.98% were female (at birth), 82.92% were white, and the mean age was 15.78 ± 1.51 years (range 12–18 years). The mean duration of the current depressive episode was 14.90 ± 8.52 months, and 83% of participants had a family history of MDD. The average baseline CDRS-R total score was 58.44 ± 6.73, indicating moderately severe symptomatology. The mean baseline ICF was 1.58 ± 0.49. Demographic and clinical characteristics of the sample are summarized in Table 1. The 10 Hz and 1 Hz treatment groups did not differ on any demographic or clinical characteristics.

Table 1.

Demographic and clinical characteristics of the overall sample and of 10 Hz vs. 1 Hz TMS groups

Participant Characteristic	Study Sample (N = 41)	10 Hz TMS (n = 19)	1 Hz TMS (n = 22)	p
Demographics
Age in years, M ± SD	15.78 ± 1.51	15.89 ± 1.05	15.68 ± 1.83	.6460
Sex at birth, % (n)				.7899
Female	60.98 (25)	63.16 (12)	59.09 (13)
Male	39.02 (16)	36.84 (7)	40.91 (9)
Gender, % (n)				.9615
Female	51.22 (21)	52.63 (10)	50.00 (11)
Male	39.02 (16)	36.84 (7)	40.91 (9)
Transgender or Nonbinary	9.76 (4)	10.53 (2)	9.09 (2)
Ethnicity, % (n)				.6486
Hispanic or Latino	12.20 (5)	15.79 (3)	9.09 (2)
Not Hispanic or Latino	87.80 (36)	84.21 (16)	90.91 (20)
Race, % (n)				.6681
African American	4.88 (2)	5.26 (1)	4.55 (1)
Asian	12.20 (5)	5.26 (1)	18.18 (4)
White	82.92 (34)	89.48 (17)	77.27 (17)
Clinical Characteristics
CDRS-R total at baseline, M ± SD	58.44 ± 6.73	58.26 ± 7.05	58.59 ± 6.61	.8788
SHAPS total at baseline, M ± SD	30.61 ± 8.16	30.05 ± 8.24	31.09 ± 8.25	.6899
BDI total at baseline, M ± SD	30.36 ± 11.94	30.31 ± 13.35	30.41 ± 10.89	.9805
ICF at baseline, M ± SD	1.58 ± 0.49	1.63 ± 0.49	1.55 ± 0.51	.5879
Number of TMS sessions completed (out of 30 total), M ± SD	28.00 ± 5.43	28.31 ± 4.75	27.72 ± 6.05	.7341
Number of previous failed medication trials, M ± SD	1.41 ± 1.16	1.15 ± 1.06	1.63 ± 1.21	.1919
Number of lifetime depressive episodes, M ± SD	1.22 ± 0.47	1.21 ± 0.42	1.22 ± 0.52	.9120
Duration of current depressive episode in months, M ± SD	14.90 ± 8.52	14.68 ± 8.35	15.09 ± 8.85	.8811
Duration of lifetime MDD in months, M ± SD	38.63 ± 21.72	38.73 ± 21.55	38.54 ± 22.38	.9780
Family history of MDD, % (n)	82.93 (34)	84.21 (16)	81.82 (18)	.8389
Tanner stages of puberty, % (n)				.7453
Stage 3	2.44 (1)	5.26 (1)	0.00 (0)
Stage 4	53.66 (22)	52.63 (10)	54.55 (12)
Stage 5	43.90 (18)	42.11 (8)	45.45 (10)

Note: Continuous variables expressed as sample mean ± standard deviation (M ± SD). Categorical variables expressed as % (n). p-value (2-tailed) associated with the test of TMS treatment group differences (10 Hz vs. 1 Hz) on each characteristic.

Abbreviations: BDI = Beck Depression Inventory; CDRS-R = Children’s Depression Rating Scale, Revised; ICF = intracortical facilitation (with 15-ms interstimulus interval); MDD = major depressive disorder; SHAPS = Snaith–Hamilton Pleasure Scale; TMS = transcranial magnetic stimulation.

Depressive Symptom Severity

TMS Treatment Simple Effects.

Within treatment groups, the pattern of adjusted least squares means over the 6-week trial within each TMS treatment group revealed a significant improvement (decrease) in depressive symptom severity (adjusted CDRS-R total scores) for both the 1 Hz TMS group (p < .0001; p_FDR = .0001; d = 1.016) and for the 10 Hz TMS group (p < .0001; p_FDR = .0001; d = 0.932) (Figure 2A). However, the pattern of the overall least squares TMS treatment group means over the 6-week trial showed that depressive symptom severity (adjusted CDRS-R total scores) was lower for the 1 Hz TMS group than for the 10 Hz TMS group [41.76 (SE = 0.88) vs. 44.54 (SE = 0.96), p = .0372; d = 0.657; Table 2, Figure 2A].

Figure 2.

Depression Severity by Treatment Group (1 Hz and 10 Hz Transcranial Magnetic Stimulation) and Biomarker Status (Low vs. High Baseline Intracortical Facilitation) Across 6 Weeks of Treatment.

Note: Mixed model results. Least squares means (LSM) estimates were adjusted for baseline CDRS-R total, age, sex at birth, and number of previously failed medications (measure of treatment resistance). Overall = timed-average LSM across weeks 1–6. A) CDRS-R total score over the 6-week study period by TMS treatment (1 Hz vs. 10 Hz). B) CDRS-R total score over the 6-week study period for 1 Hz vs. 10 Hz TMS treatment, stratified by low vs. high baseline ICF.

Table 2.

Effect of TMS on depression severity over the 6-week trial

	CDRS-R Total Score
Treatment Group	Week 1	Week 2	Week 3	Week 4	Week 5	Week 6	Overall Timed-Average (Weeks 1–6)		Overall Treatment Group Main Effect (Weeks 1–6)
	LSM (SE)	LSM (SE)	LSM (SE)	LSM (SE)	LSM (SE)	LSM (SE)	LSM (SE)	95% CI	Group Difference		F statistic	p	Cohen’s d
									LSM (SE)	95% CI
TMS group									−2.772 (1.321)	−5.377 to −0.166	F1,200 = 4.40	.0372	0.657
1 Hz group (n = 22)	51.268 (2.009)	47.725 (2.072)	40.972 (2.146)	40.805 (2.151)	36.622 (2.184)	33.218 (2.179)	41.768 (0.879)	40.034 to 43.503
10 Hz group (n = 19)	53.766 (2.236)	51.617 (2.237)	45.070 (2.349)	40.104 (2.385)	39.180 (2.391)	37.506 (2.386)	44.541 (0.967)	42.633 to 46.448

Note: LSM = least squares mean estimate, adjusted for baseline CDRS-R total score, age, sex at birth, and number of previously failed medications (measure of treatment resistance); SE = standard error; Difference = difference of TMS group LSM (1 Hz vs. 10 Hz); 95% CI = 95% confidence interval for the group difference of LSM estimate. p-value associated with the test (F-statistic) of the overall timed-average difference of the LSM estimate between the TMS groups (1 Hz vs. 10 Hz). Cohen’s d was calculated to estimate effect sizes for the overall between-participants TMS group difference of LSM estimates. Higher CDRS-R total score corresponds to greater depression severity.

Abbreviations: CDRS-R = Children’s Depression Rating Scale, Revised; TMS = transcranial magnetic stimulation.

Biomarker Stratification.

There was no difference in overall timed-average depressive symptom severity between the high baseline ICF and low baseline ICF groups [42.91 (SE = 0.84) vs. 43.39 (SE = 1.09), p = .7418; d = 0.104]. The mixed model repeated measures analysis revealed a significant TMS treatment × baseline ICF interaction effect (F_1,200 = 4.62, p = .0328) as well as significant main effects of TMS treatment (F_1,200 = 4.40, p = .0372) and time (F_5,200 = 18.75, p < .0001). However, there were no significant TMS treatment × baseline ICF × time interaction (p = .4338), baseline ICF × time interaction (p = .9532) or TMS treatment × time interaction effects (p = .8852), or baseline ICF main effect (p = .7418).

The simple TMS treatment group effects at individual timepoints revealed that week 6 symptom severity (adjusted CDRS-R total score) was lower for the group with low baseline ICF treated with 1 Hz TMS than the group with low baseline ICF treated with 10 Hz TMS [29.57 (SE = 3.35) vs. 42.19 (SE = 3.86), p = .0140, p_FDR = .0840; d = 1.223, Table 3, Figure 2B]. There were no such simple TMS group differences observed at the other individual weeks (weeks 1–5). Moreover, the pattern of the overall least squares TMS treatment group means over the 6-week trial showed that for participants within the low baseline ICF stratification group, depressive symptom severity was lower for those who received 1 Hz TMS than those who received 10 Hz TMS [40.61 (SE = 1.39) vs. 46.18 (SE = 1.61), p = .0080; p_FDR = .0160; d = 1.321; Table 3, Figure 2B]. However, for participants in the high baseline ICF stratification group, there was no difference in symptom severity between those who received 1 Hz TMS and those who received 10 Hz TMS [42.92 (SE = 1.15) vs. 42.89 (SE = 1.17), p = .9853; p_FDR = .9853; d = 0.009; Figure 2B, Table 3].

Table 3.

Effect of TMS group (1 Hz vs. 10 Hz) by baseline ICF status (high vs. low) on depression severity over the 6-week trial

	CDRS-R Total Score
Treatment Group by Baseline ICF Status	Week 1	Week 2	Week 3	Week 4	Week 5	Week 6	Overall Timed-Average (Weeks 1–6)		Overall Treatment Group Main Effect (Weeks 1–6)
	LSM (SE)	LSM (SE)	LSM (SE)	LSM (SE)	LSM (SE)	LSM (SE)	LSM (SE)	95% CI	Group Difference		F statistic	p	p FDR	Cohen’s d
									LSM (SE)	95% CI
Low Baseline ICF (< 1.5)
TMS group									−5.573 (2.079)	−9.673 to −1.473	F1,200 = 7.19	.008	.0160	1.321
1 Hz group (n = 10)	49.954 (2.979)	47.286 (3.154)	40.946 (3.352)	40.696 (3.353)	35.196 (3.351)	29.571 (3.350)	40.609 (1.396)	37.855 to 43.363
10 Hz group (n = 7)	53.007 (3.580)	52.293 (3.582)	45.198 (3.863)	42.032 (3.864)	42.365 (3.862)	42.198 (3.863)	46.182 (1.614)	42.998 to 49.366
High Baseline ICF (> 1.5)
TMS group									0.029 (1.602)	−3.131 to 3.189	F1,200 = 0.01	.9853	.9853	0.009
1 Hz group (n = 12)	52.581 (2.717)	48.164 (2.718)	40.998 (2.727)	40.914 (2.719)	38.047 (2.836)	36.865 (2.837)	42.928 (1.151)	40.658 to 45.199
10 Hz group (n = 12)	54.524 (2.724)	50.941 (2.725)	44.941 (2.724)	38.177 (2.838)	35.995 (2.839)	32.814 (2.849)	42.899 (1.169)	40.592 to 45.205

Note: LSM = least squares mean estimate, adjusted for baseline CDRS-R total score, age, sex at birth, and number of previously failed medications (measure of treatment resistance); SE = standard error; Difference = difference of TMS group LSM (1 Hz vs. 10 Hz) by ICF condition; 95% CI = 95% confidence interval for the group difference of LSM estimate. p-value associated with the test (F-statistic) of the overall timed-average difference of the LSM estimate between the TMS groups (1 Hz vs. 10 Hz) by ICF condition; p_FDR = p-value corrected according to the False Discovery Rate procedure. Cohen’s d was calculated to estimate effect sizes for the overall between-participants TMS group difference of LSM estimates. Low ICF = baseline ICF≤1.5; high ICF = baseline ICF>1.5. Higher CDRS-R total score corresponds to greater depression severity.

Abbreviations: CDRS-R = Children’s Depression Rating Scale, Revised; ICF = intracortical facilitation (with 15-ms interstimulus interval); TMS = transcranial magnetic stimulation.

ICF

The mixed model repeated measures analysis revealed that there was no significant TMS treatment group × time interaction effect (p = .0936), as well as no significant main effect of TMS treatment group (p = .6706) or time (p = .1473), on ICF. Moreover, for the within-group simple effects, the pattern of the adjusted ICF least squares means revealed no significant change in ICF across the 6-week trial for either the 10 Hz TMS group (p = .0727; p_FDR = .1435) or the 1 Hz TMS group (p = .1435; p_FDR = .1435). The simple TMS treatment group effects at individual timepoints also revealed no significant TMS group differences on ICF.

We also examined the relationship between ICF (as the outcome) and depressive symptoms (CDRS-R total as a time-varying covariate) over 6 weeks of TMS treatment. The mixed model did not reveal a significant relationship between the change in CDRS-R total (over 6 weeks) and the change in adjusted ICF (over 6 weeks) for the 10 Hz TMS group ($\widehat{b}=-0.0117$, SE = 0.0075, p = .1179; p_FDR = .2358) or for the 1 Hz TMS group ($\widehat{b}=-0.0045$, SE = 0.0069, p = .5103; p_FDR = .5103). The interaction effect also revealed that the effect of change in CDRS-R total depressive symptoms (over 6 weeks) on change in adjusted ICF (over 6 weeks) was not significantly different between the 10 Hz TMS group and the 1 Hz TMS group ($\widehat{b}=-0.0072$, SE = 0.0087, p = .4141). In other words, TMS stimulus frequency (10 Hz vs. 1 Hz) did not moderate the relationship between the change in depressive symptoms and the change in ICF.

Blinding Assessment

At the end of treatment, the clinical rater guessed the correct treatment assignment in 36.9% of participants. Among adolescents and parents, 41.3% correctly guessed treatment assignment.

Adverse Events

Nine participants reported 207 mild or moderate adverse events, with treatment site pain (46%) and headaches (34%) being the most common. Five participants (10 Hz, n = 4; 1 Hz, n = 1) exited the study prior to completing week 1 assessments due to discomfort with TMS treatments. The 10 Hz treatment arm had a higher overall rate of adverse events compared to the 1 Hz arm (χ² = 8.88, df = 1, p = .003). Six participants had serious adverse events during treatment. One participant reported transient blurred vision after treatment sessions, which resolved after five days and was classified as an unanticipated problem involving risks to subjects or others, and subsequently informed consent forms were modified to include blurred vision as a possible side effect. Other serious adverse events included suicidal ideation resulting in hospitalization; severe scalp pain and inability to tolerate treatments; left ear bleeding; chest pain; and groin pain. The participants reporting ear bleeding, chest pain, and groin pain were evaluated in the emergency department; these symptoms were determined not to be related to TMS, and all symptoms resolved. Performance on cognitive measures (NIH Toolbox Cognitive Battery) did not worsen in either 10 Hz or 1 Hz groups. There were no changes in vital signs, except for weight, which increased significantly in the 10 Hz group across the 6-week treatment course. Additional information is available in Supplement 1 and Tables S1 and S2, available online.

DISCUSSION

This study sought to compare the clinical effects of 1 Hz and 10 Hz TMS and to develop a biomarker-guided treatment approach for MDD in adolescents. Adolescent participants with low ICF at baseline who were treated with 1 Hz TMS had greater improvement in depressive symptom severity compared to adolescent participants with low baseline ICF who were treated with 10 Hz TMS. This superiority was particularly striking after 6 weeks of treatment. The present findings are noteworthy as 1 Hz TMS has not been extensively studied in adolescents with MDD, and as a prior 6-week trial of 10 Hz TMS did not demonstrate superiority over sham TMS.⁸ The present study design and results also challenge prior conventions, as 1 Hz TMS treatments were delivered to the left DLPFC in our study.^29,39,40

The current protocol did not employ a sham treatment control condition. In the context of high placebo response rates in studies of adolescents with MDD,^11,13,14 as well as ongoing concerns that sham response rates for neuromodulation treatments are increasing over time,¹⁵ this may have been an advantage of our study design. In addition to efforts to improve rigor of sham methodology in TMS, the study of active interventions with uncertain clinical effects as comparison groups has both methodological and ethical benefits in youth.⁹ This clinical trial also contributes to existing literature demonstrating that therapeutic TMS treatment is feasible, safe, and generally well-tolerated in adolescents with MDD.^8,9,16

Our findings also contribute to ongoing work suggesting that ICF has utility as a biomarker for adolescents with MDD.¹⁸ Prior work indicates that ICF is a general marker of cortical NMDA receptor (NMDAR)-mediated glutamatergic neurotransmission.^18,23,24 Collection of ICF measures is not as technically complex, expensive, or time-consuming as neuroimaging or electroencephalographic markers. The assessment of ICF is similar to motor threshold procedures that are currently an integral aspect of dosing TMS delivery in standard clinical practice.^5,6,29 Thus, determining TMS treatment parameters using ICF measurement may be more scalable and translatable to the bedside than many other biomarkers.

The present study suggests that 1 Hz TMS delivered to the left DLPFC may be superior to 10 Hz TMS for MDD in adolescents with low baseline ICF. Contrary to our hypothesis, weekly ICF markers were stable across treatment, suggesting that this biomarker is a trait-like measure of cortical NMDAR-mediated glutamatergic neurotransmission, or that ICF indexes tonic NMDAR activity. ICF stability across treatment also suggests that cortical excitability mechanisms may not account for changes in depressive symptoms directly, but instead may represent the relative receptiveness to induction of functional network changes via TMS, especially with 1 Hz stimulation. The antidepressant effects of 1 Hz TMS in adolescents with MDD may involve enhancement of long-term depression (LTD)-like effects on synaptic plasticity.^{20,26,41–43} Long-term potentiation (LTP) and LTD represent opposing mechanisms of neuroplasticity. Both LTP and LTD play critical roles in learning, memory, and neural circuit refinement. Synaptic connections are strengthened through LTP and weakened by LTD. One theory suggests that low-frequency forms of neuromodulation such as 1 Hz TMS induce neural effects that are similar to LTD. Optimal potentiation of NMDARs with ensuing increases in postsynaptic calcium is most likely necessary to induce LTD-like effects.^20,21,41 Previous work has found TMS to have LTD-like plasticity effects that correspond to changes in DLPFC oscillations measured by quantitative electroencephalography.⁴⁴ Other studies have demonstrated that repetitive TMS can entrain oscillatory synchrony,^45,46 and that TMS-induced oscillatory effects are specific to pulse frequency and may vary across brain regions.⁴⁷ The low ICF biomarker status in the present study may reflect an adolescent brain phenotype with optimal NMDAR activity for LTD-like neuroplasticity, which enhances antidepressant effects of 1 Hz TMS. As a result, adolescent patients with MDD and low ICF may have favorable clinical effects with 1 Hz TMS.^20,41,42,48 Further studies are needed to replicate our findings and to examine how TMS frequency affects other measures of network functioning, such as oscillatory synchrony and network functional connectivity, as potential mechanisms of antidepressant response to TMS. Such multimodal approaches will further explicate how diverse neurophysiologic phenotypes of MDD in youth underlie responsiveness to different TMS stimulation parameters. This mechanistic knowledge ultimately will enable more precise prediction of youth likely to respond to TMS, and guide more individualized TMS treatment interventions.

Several limitations warrant consideration in the interpretation of our findings. First, although this was a large study in the context of related work with adolescents and TMS, the sample size was relatively small as much of the study recruitment occurred during the peak of the COVID-19 pandemic.^6,8 A detailed explanation of the sample size calculation is available in Supplement 1, available online. Second, it was not possible to provide blinding for the TMS operators, and blinding participants was challenging in the context of the study design. However, our results suggest that most adolescents could not accurately identify which frequency of TMS was provided. Third, the Beam F3 method was used for treatment coil localization rather than structural or functional targeting with magnetic resonance imaging. Our method closely mirrors common clinical practice localization methods, and prior work suggests the Beam F3 method provides comparable results to neuro-navigated treatment with TMS.²⁹ Nevertheless, other recent work has found that MRI-guided and computationally-optimized TMS targeting produce higher electric fields, which correlates with greater changes in depressive symptoms.⁴⁹ Fourth, the study design intentionally included patients naïve to treatment with antidepressant medications. This may have elevated placebo effects in both treatment arms of the study, but it is critical to advance knowledge and experience with this patient population.^9,15 Fifth, TMS treatments were provided as monotherapy, with no concurrent psychotropic medications. This provided additional rigor to the study design but reduces external validity, since in clinical practice TMS is commonly provided in addition to one or more psychotropic medications.^4,29 Notably, the adjunctive use of TMS was reflected in recent FDA clearance of TMS for treating MDD in adolescents aged 15 and older, which specified its use in addition to antidepressant treatment.⁷ Finally, as with other clinical trials of adolescents with MDD, the sample was heterogeneous, and inclusion was characterized based on clinical measures.^9,11 Inclusion criteria based on biomarkers, such as specific ICF or other physiologic thresholds, could more clearly test hypotheses about mechanisms of TMS.

In summary, in participants with low baseline ICF, acute treatment with 1 Hz TMS provided greater improvement in depressive symptom severity than 10 Hz TMS. Low ICF may reflect optimal conditions for LTD-like synaptic plasticity effects and clinical response to 1 Hz TMS. Further, weekly ICF measures were stable, suggesting that this measure potentially assesses trait-like or tonic NMDAR-mediated neurotransmission. This study provides additional practical experience with safety and tolerability of TMS in youth, and preliminary findings with low-frequency stimulation which will inform future innovations for the use of TMS treatments in adolescents with MDD.

Diversity & Inclusion Statement:

We worked to ensure sex and gender balance in the recruitment of human participants. We worked to ensure race, ethnic, and/or other types of diversity in the recruitment of human participants. We worked to ensure that the study questionnaires were prepared in an inclusive way. One or more of the authors of this paper self-identifies as a member of one or more historically underrepresented racial and/or ethnic groups in science. We actively worked to promote sex and gender balance in our author group. We actively worked to promote inclusion of historically underrepresented racial and/or ethnic groups in science in our author group. While citing references scientifically relevant for this work, we also actively worked to promote sex and gender balance in our reference list. While citing references scientifically relevant for this work, we also actively worked to promote inclusion of historically underrepresented racial and/or ethnic groups in science in our reference list.

Data Sharing:

Deidentified participant data, Data dictionary, Study Protocol, and Statistical Analysis Plan available to researchers (upon reasonable request) with publication for secondary analyses (upon reasonable request) through NIMH https://nda.nih.gov/ or data sharing agreement with PI.