Transcranial Magnetic Stimulation (TMS): rTMS, iTBS, and Deep TMS — Protocols, Neurobiological Mechanisms, Evidence Base, and Clinical Outcomes

Transcranial magnetic stimulation (TMS) has emerged as one of the most significant non-pharmacological treatment modalities in psychiatry over the past two decades. First approved by the U.S. Food and Drug Administration (FDA) for treatment-resistant depression (TRD) in 2008, TMS now encompasses a family of protocols — including repetitive TMS (rTMS), intermittent theta-burst stimulation (iTBS), and deep TMS (dTMS) — with expanding indications in obsessive-compulsive disorder (OCD), smoking cessation, anxious depression, and post-traumatic stress disorder (PTSD).

TMS operates on the principle of electromagnetic induction: a rapidly changing magnetic field generated by a coil placed on the scalp induces focal electrical currents in underlying cortical tissue, depolarizing neurons at the targeted site. Unlike electroconvulsive therapy (ECT), TMS does not require anesthesia, does not induce seizures intentionally, and carries a substantially different side-effect profile. Unlike pharmacotherapy, it bypasses systemic metabolism, avoiding hepatic first-pass effects and gastrointestinal absorption variability.

The clinical relevance of TMS is magnified by the substantial prevalence of treatment-resistant depression. Data from the landmark STAR*D trial demonstrated that after two adequate antidepressant trials, cumulative remission rates plateau near 50–55%, meaning roughly one in three patients with major depressive disorder (MDD) will fail to achieve remission with sequential pharmacotherapy. For these patients — and a growing range of neuropsychiatric conditions — TMS offers a mechanistically distinct therapeutic pathway. This article provides a detailed review of TMS protocols, their neurobiological underpinnings, clinical evidence, comparative effectiveness, prognostic predictors, and current research frontiers.

The therapeutic effects of TMS are mediated through several converging neurobiological mechanisms that operate at the synaptic, circuit, and network levels.

Synaptic Plasticity: LTP and LTD

The foundational mechanism of TMS is the induction of long-term potentiation (LTP) or long-term depression (LTD) at excitatory synapses, depending on stimulation frequency and pattern. High-frequency rTMS (≥5 Hz) applied to the left dorsolateral prefrontal cortex (L-DLPFC) induces LTP-like plasticity, increasing cortical excitability and glutamatergic neurotransmission in the stimulated region. Low-frequency rTMS (≤1 Hz), typically applied to the right DLPFC, induces LTD-like effects that reduce cortical excitability. These plasticity changes are NMDA receptor–dependent — blockade with ketamine or memantine attenuates TMS-induced aftereffects — and involve downstream modulation of AMPA receptor trafficking and brain-derived neurotrophic factor (BDNF) signaling.

Neurotransmitter Systems

TMS exerts measurable effects across multiple monoaminergic and amino acid neurotransmitter systems:

• Dopamine: PET studies demonstrate that rTMS to the L-DLPFC increases dopamine release in the ipsilateral caudate nucleus and putamen, likely through corticostriatal glutamatergic projections. This finding has implications for treating anhedonia and motivational deficits.
• Serotonin: Animal studies show that repeated rTMS sessions increase serotonin turnover in the hippocampus and frontal cortex, with changes in 5-HT1A and 5-HT2A receptor sensitivity analogous to those seen with antidepressant medications.
• GABA and Glutamate: Magnetic resonance spectroscopy (MRS) studies demonstrate that TMS modulates cortical GABA and glutamate concentrations in the DLPFC, with normalization of the GABA/glutamate ratio correlating with clinical response.
• BDNF: Serum BDNF levels increase over the course of rTMS treatment, paralleling patterns observed with antidepressant pharmacotherapy and ECT, suggesting shared downstream neurotrophic pathways.

Circuit-Level Effects

Depression is increasingly understood as a disorder of distributed neural networks rather than any single brain region. The DLPFC serves as a critical hub in the frontoparietal control network (also called the central executive network), which is hypoactive in depression. Simultaneously, the default mode network (DMN) — centered on the medial prefrontal cortex and posterior cingulate — is hyperactive, driving rumination and self-referential negative processing. The subgenual anterior cingulate cortex (sgACC, Brodmann area 25) — a key node in affective regulation — shows hypermetabolism in depression.

TMS to the L-DLPFC normalizes the anticorrelation between the frontoparietal network and the DMN, and reduces sgACC hyperactivity through monosynaptic and polysynaptic prefrontal-cingulate connections. A landmark 2021 study by Siddiqi and colleagues in Nature Human Behaviour demonstrated that the specific DLPFC subregion most functionally anticorrelated with the sgACC predicts optimal TMS targeting — a finding that has catalyzed the move toward functional connectivity–guided, personalized TMS protocols.

Genetic Factors

Emerging pharmacogenomic research suggests genetic variation modulates TMS response. The BDNF Val66Met polymorphism — carried by approximately 30% of the population — affects activity-dependent BDNF secretion and synaptic plasticity. Some studies report that Met allele carriers show reduced cortical plasticity in response to TMS protocols, though clinical response data are mixed and this remains an active area of investigation. Polymorphisms in COMT (Val158Met), which affects prefrontal dopamine catabolism, and SLC6A4 (5-HTTLPR), the serotonin transporter promoter, have also been explored as potential predictors with preliminary but inconsistent results.

Several distinct TMS protocols are in clinical use, each with different stimulation parameters, coil geometries, treatment durations, and evidence bases.

Standard High-Frequency Repetitive TMS (HF-rTMS)

The most extensively studied protocol involves 10 Hz rTMS to the left DLPFC at 120% of the resting motor threshold (RMT). A standard course consists of 3,000 pulses per session, delivered over approximately 37.5 minutes, for 20–30 sessions (typically 5 days per week for 4–6 weeks). This was the protocol used in the pivotal trial by O'Reardon and colleagues (2007) that led to FDA clearance. Variations include 20 Hz protocols and right-sided low-frequency (1 Hz) stimulation, the latter targeting right DLPFC hyperactivity.

Intermittent Theta-Burst Stimulation (iTBS)

iTBS delivers triplet bursts at 50 Hz, repeated at 5 Hz (the theta rhythm), in an intermittent pattern — 2 seconds on, 8 seconds off — for a total of 600 pulses in approximately 3 minutes and 9 seconds. This dramatic reduction in treatment time (from ~37 minutes to ~3 minutes per session) has transformative implications for clinical throughput and patient access. iTBS was hypothesized to more efficiently engage LTP-like plasticity mechanisms due to its biomimicry of endogenous hippocampal theta rhythms involved in learning and memory consolidation.

The landmark THREE-D trial (Blumberger et al., 2018, published in The Lancet) was a multisite, randomized, noninferiority trial comparing iTBS to standard 10 Hz rTMS in 414 patients with treatment-resistant depression. The trial demonstrated noninferiority of iTBS, with response rates of 49% for iTBS versus 47% for 10 Hz rTMS, and remission rates of 32% for iTBS versus 27% for 10 Hz rTMS — with no significant difference between groups. This trial established iTBS as a time-efficient, clinically equivalent alternative.

Stanford Neuromodulation Therapy (SNT / SAINT Protocol)

The Stanford Accelerated Intelligent Neuromodulation Therapy (SAINT) protocol, developed by Nolan Williams and colleagues, represents a paradigm shift: it delivers 10 sessions per day of connectivity-guided iTBS over 5 consecutive days (50 sessions total, 90,000 pulses). Targeting uses individualized fMRI-based identification of the L-DLPFC subregion maximally anticorrelated with the sgACC. In the initial open-label study (Cole et al., 2020, American Journal of Psychiatry), 90% of participants with TRD achieved remission. The subsequent double-blind, sham-controlled trial (Cole et al., 2022, also in American Journal of Psychiatry) reported a remission rate of 78.6% for active treatment versus 14.3% for sham at one month. The FDA cleared this protocol in 2022 under the name SNT. While these results are striking, they come from relatively small samples (n = 29 in the RCT), and larger replication studies are ongoing.

Deep TMS (dTMS)

Deep TMS uses the H-coil (developed by Brainsway), which generates a broader and deeper magnetic field than standard figure-of-eight coils, reaching structures 3–4 cm below the scalp compared to 1.5–2 cm with conventional coils. The standard dTMS protocol for depression uses the H1 coil targeting the L-DLPFC at 18 Hz, delivering 1,980 pulses per session over ~20 minutes. A pivotal multicenter trial by Levkovitz and colleagues (2015) demonstrated a response rate of 38.4% vs. 21.4% for sham and remission rate of 32.6% vs. 14.6% for sham.

Deep TMS has received separate FDA clearances for OCD (2018, targeting the medial prefrontal cortex/anterior cingulate with the H7 coil) and smoking cessation (2020, targeting the bilateral insula and prefrontal cortex with the H4 coil). The OCD clearance was based on a study by Carmi and colleagues (2019) showing a response rate of 38.1% vs. 11.1% for sham using symptom provocation prior to stimulation — a methodological innovation integrating cognitive-behavioral principles with neuromodulation.

Bilateral and Sequential Protocols

Bilateral stimulation — combining high-frequency left and low-frequency right DLPFC stimulation in a single session — has theoretical appeal for addressing both left hypoactivation and right hyperactivation in depression. Meta-analytic data suggest bilateral stimulation may offer modestly superior efficacy to unilateral stimulation, though head-to-head trials remain limited in number and statistical power.

TMS has FDA-cleared and evidence-supported indications across several neuropsychiatric conditions, each with its own epidemiological context.

Major Depressive Disorder (MDD)

MDD is the primary indication for TMS. The NIMH estimates the 12-month prevalence of MDD in U.S. adults at 8.3% (approximately 21 million adults) as of 2021 data. Lifetime prevalence is approximately 16–17%. Among these individuals, an estimated 30–40% meet criteria for treatment-resistant depression, defined variably but most commonly as failure of ≥2 adequate antidepressant trials in the current episode. This translates to approximately 6–8 million treatment-resistant individuals in the U.S. alone — the primary target population for TMS.

Obsessive-Compulsive Disorder

OCD has a lifetime prevalence of approximately 2–3% (DSM-5-TR). Approximately 40–60% of OCD patients are partial or non-responders to first-line treatments (SSRIs + exposure and response prevention). Deep TMS targeting the medial prefrontal cortex/dACC is FDA-cleared for OCD, and standard rTMS to the supplementary motor area (SMA) at 1 Hz has also shown efficacy in multiple trials.

Smoking Cessation

Tobacco use disorder affects approximately 11.5% of U.S. adults (28.3 million) per CDC 2020 data. Deep TMS targeting bilateral prefrontal cortex and insula received FDA clearance for smoking cessation based on the multicenter trial by Zangen and colleagues (2021), which demonstrated significantly higher continuous quit rates with active versus sham dTMS.

Emerging and Off-Label Indications

Active clinical research is investigating TMS for generalized anxiety disorder, PTSD, substance use disorders (particularly alcohol and cocaine), chronic pain, tinnitus, and cognitive symptoms of schizophrenia. The VA/DoD has funded large-scale trials of TMS for PTSD, and preliminary results suggest efficacy when targeting the right DLPFC at 1 Hz or using iTBS protocols. For negative symptoms of schizophrenia, rTMS to the L-DLPFC shows modest effect sizes in meta-analyses (Hedges' g ≈ 0.34), though clinical significance remains debated.

The evidence base for TMS has matured considerably, now encompassing dozens of randomized controlled trials, several large meta-analyses, and extensive real-world effectiveness data.

Meta-Analytic Evidence for rTMS in Depression

A comprehensive meta-analysis by Berlim and colleagues (2014) pooling data from sham-controlled trials found that HF-rTMS to the L-DLPFC produced a response rate of approximately 29% versus 10% for sham (NNT ≈ 5–6) and a remission rate of approximately 18% versus 5% for sham (NNT ≈ 8). More recent network meta-analyses have confirmed similar effect sizes, with pooled odds ratios of approximately 3.0–3.7 for response relative to sham.

A 2024 network meta-analysis by Mutz and colleagues, one of the most comprehensive to date, ranked bilateral rTMS and iTBS among the protocols with the strongest evidence, while confirming that most active TMS protocols significantly outperform sham. The standardized mean difference (SMD) for depression symptom reduction across all active TMS protocols versus sham typically ranges from 0.4–0.8, reflecting moderate to large effect sizes.

Real-World Effectiveness Data

Critically, outcomes from real-world clinical practice consistently exceed those from RCTs. The NeuroStar Outcomes Registry — the largest naturalistic TMS database — reported data on over 5,000 patients showing a response rate of 58% and remission rate of 37% in clinical practice, measured by the Patient Health Questionnaire-9 (PHQ-9). This discrepancy between RCT and real-world data likely reflects several factors: RCT populations often have higher treatment resistance; real-world protocols allow clinician flexibility in dosing and duration; and sham response rates inflate NNT in controlled trials.

TMS vs. Antidepressant Medications

Direct head-to-head comparisons between TMS and antidepressants are limited. However, indirect evidence is informative: the response rate for an individual SSRI or SNRI in treatment-resistant depression (after ≥1 prior failure) is approximately 25–30% based on STAR*D Level 2–4 data, which is roughly comparable to the RCT response rates for TMS. A key difference is the side-effect profile: TMS avoids the sexual dysfunction (~30–60% with SSRIs), weight gain, gastrointestinal effects, and discontinuation symptoms associated with antidepressants. The most common TMS side effect is application-site discomfort/pain (reported in 20–40% of patients, typically habituating over sessions), followed by headache (~15–25%). Seizure risk is approximately 1 in 30,000 sessions when safety guidelines are followed.

TMS vs. ECT

ECT remains the most effective acute treatment for severe depression, with remission rates of 50–65% in most trials. TMS does not match ECT's efficacy, particularly for severe, psychotic, or catatonic depression. However, a randomized noninferiority trial by Eranti and colleagues (2007) and subsequent studies suggest that for non-psychotic, moderate-severity TRD, TMS may approach ECT's effectiveness while avoiding general anesthesia, cognitive side effects (particularly retrograde amnesia), and the logistical burden of ECT. The choice between TMS and ECT should be individualized based on depression severity, psychotic features, urgency (suicidality), patient preference, and cognitive risk tolerance.

iTBS vs. Standard rTMS: The THREE-D Trial in Detail

The THREE-D trial merits additional discussion given its pivotal role. With 414 randomized participants across three Canadian sites, it remains the largest head-to-head TMS protocol comparison. Both arms received 20 sessions over 4 weeks followed by tapering. The primary outcome (change in HRSD-17 score) showed noninferiority of iTBS (mean change –10.3 vs. –9.9 for 10 Hz, within the 2-point noninferiority margin). The clinical implications are substantial: iTBS's 3-minute treatment time versus 37 minutes for standard rTMS dramatically increases clinic capacity and reduces patient burden.

Identifying reliable predictors of TMS response is a critical research priority, as approximately 40–50% of patients treated with standard protocols do not achieve clinically significant response. Known and emerging predictive factors span clinical, neuroimaging, neurophysiological, and genetic domains.

Clinical Predictors

• Degree of treatment resistance: This is the most consistently identified negative predictor. Patients who have failed fewer prior antidepressant trials respond better to TMS. One analysis found that failure of ≤1 adequate trial predicted a response rate of ~55%, while failure of ≥4 trials predicted ~35%.
• Episode duration: Shorter current episode duration is associated with better outcomes. Episodes lasting >2 years show diminished response rates.
• Age: Some studies report modest negative correlations between age and TMS response, potentially related to age-related cortical atrophy increasing the scalp-to-cortex distance and reducing the delivered dose to the DLPFC. However, this finding is inconsistent, and TMS is effective across the adult age range.
• Comorbid anxiety: Mixed findings. While severe comorbid anxiety was historically considered a negative predictor, more recent data (including from the NeuroStar registry) suggest that comorbid anxious depression may respond comparably or even favorably to TMS.
• Psychotic features: Depression with psychotic features responds poorly to TMS and is generally considered a relative contraindication; ECT is preferred.
• Concurrent antidepressant medication: Meta-analytic data suggest TMS combined with concurrent antidepressant pharmacotherapy produces better outcomes than TMS alone, though this is difficult to disentangle from selection effects.

Neuroimaging and Neurophysiological Predictors

• Baseline sgACC hyperactivity: Greater pretreatment sgACC metabolic activity (measured by PET or functional connectivity) predicts better TMS response, mirroring findings in deep brain stimulation and pharmacotherapy research.
• DLPFC-sgACC functional connectivity: Patients with stronger negative (anticorrelated) functional connectivity between the stimulation site on the DLPFC and the sgACC show better outcomes, supporting the rationale for connectivity-guided targeting.
• Prefrontal theta cordance on EEG: Some studies report that early changes in frontal theta cordance (within the first week of treatment) predict eventual response, raising the possibility of EEG-based early response monitoring.
• Cortical thickness and scalp-to-cortex distance: Increased distance from scalp to cortex at the stimulation site (due to atrophy or individual anatomy) attenuates the delivered electromagnetic field. Neuronavigated protocols that account for individual anatomy may mitigate this effect.

Genetic and Molecular Predictors

As noted earlier, the BDNF Val66Met polymorphism has been studied, with Val/Val carriers potentially showing enhanced plasticity responses. However, no single genetic biomarker has sufficient predictive accuracy for clinical use. Polygenic risk scores for depression response are under development but remain preliminary.

Depression rarely presents in isolation, and the impact of psychiatric comorbidities on TMS outcomes is a clinically important consideration.

Anxiety Disorders

Comorbid anxiety disorders — including generalized anxiety disorder, social anxiety disorder, and panic disorder — co-occur with MDD in approximately 50–60% of cases. The DSM-5-TR "anxious distress" specifier applies to roughly 50–75% of depressed patients. Data from large registries suggest that patients with anxious depression respond to TMS at rates comparable to those without comorbid anxiety, and some studies report incidental anxiolytic effects, with significant reductions in GAD-7 and Hamilton Anxiety Rating Scale scores alongside depression improvement.

Post-Traumatic Stress Disorder

PTSD co-occurs with MDD in approximately 48% of cases. TMS protocols targeting the right DLPFC at 1 Hz or bilateral prefrontal stimulation have shown promise for PTSD symptoms, though evidence is still emerging. The presence of PTSD does not appear to substantially reduce TMS efficacy for comorbid depression.

Substance Use Disorders

Co-occurring substance use disorders are present in approximately 20–25% of patients with MDD. Active substance use disorder can complicate TMS treatment through irregular attendance, altered seizure thresholds (particularly with alcohol withdrawal or stimulant use), and confounding of outcome measurement. However, preclinical and early clinical data suggest TMS may directly reduce craving through modulation of prefrontal-striatal circuits involved in reward processing and impulse control. Studies targeting the DLPFC and insula for alcohol, cocaine, and methamphetamine use disorders have shown encouraging reductions in craving, though long-term abstinence data remain limited.

Personality Disorders

Comorbid borderline personality disorder (BPD) is present in approximately 10–15% of TRD patients presenting for TMS. Historically, personality pathology has been considered a negative prognostic factor for all depression treatments. Limited data specific to TMS suggest modestly lower response rates in patients with significant personality disorder comorbidity, though clinically meaningful improvement is still achievable. Emotional dysregulation — a core BPD feature — may partially respond to prefrontal stimulation given the DLPFC's role in emotion regulation.

Medical Comorbidity

TMS is generally safe in the context of medical comorbidity, with fewer medical contraindications than ECT. The primary absolute contraindication is ferromagnetic material near the stimulation site (e.g., cochlear implants, metallic cranial plates, or implanted stimulators). Cardiac pacemakers were historically considered a contraindication, though newer evidence suggests standard TMS coils do not interfere with cardiac devices at typical treatment distances, and this is no longer an absolute contraindication in most guidelines. Epilepsy is a relative contraindication given the seizure risk, though some protocols for depression in epilepsy patients have been safely conducted under careful monitoring.

The evolution of TMS targeting methods reflects the field's maturation from crude anatomical approximations to precision-medicine approaches.

The 5-cm Rule

The original targeting method involved identifying the motor cortex (the site producing a visible thumb twitch) and then moving the coil 5 cm anteriorly along the scalp. This approach, used in many early trials including the pivotal O'Reardon (2007) study, is now recognized as imprecise: it often targets the premotor cortex rather than the DLPFC, and the optimal target varies substantially across individuals due to head size, cortical folding patterns, and functional neuroanatomy differences.

Beam F3 Method

The Beam F3 method uses scalp measurements based on the 10-20 EEG system to approximate the F3 electrode position, which overlies the L-DLPFC. This approach is more anatomically consistent than the 5-cm rule, requires no neuroimaging, and is widely used in clinical practice. A study by Mir-Moghtadaei and colleagues (2015) demonstrated that Beam F3 targeting produced results comparable to MRI-guided neuronavigation in a clinical sample.

MRI-Guided Neuronavigation

Structural MRI-guided neuronavigation uses individual anatomical imaging to identify the DLPFC target (often Brodmann areas 9 or 46) and a frameless stereotactic system to guide coil placement in real time. This ensures consistent targeting across sessions and compensates for individual anatomical variability.

Functional Connectivity–Guided Targeting

The frontier of TMS targeting involves using individual resting-state fMRI to identify the specific DLPFC subregion most functionally connected (anticorrelated) to the sgACC, as described in the SAINT/SNT protocol. Retrospective analyses by Weigand and colleagues (2018) demonstrated that optimal responders to standard rTMS had been inadvertently stimulated at sites with stronger DLPFC–sgACC anticorrelation. Prospective validation in larger cohorts is ongoing, but this approach represents the clearest embodiment of precision psychiatry applied to neuromodulation.

Dosing Considerations

Standard intensity is set at 120% of the resting motor threshold (RMT), which is determined individually by finding the minimum stimulator output that produces a visible motor evoked potential (MEP) in the contralateral hand ~50% of the time. This normalization to individual motor threshold accounts for variability in skull thickness, cortical excitability, and other factors. However, the DLPFC and motor cortex differ in depth and excitability, meaning motor threshold may not perfectly calibrate prefrontal dosing. Newer approaches using electric field modeling to estimate the actual dose delivered to the DLPFC are under investigation.

TMS is among the safest brain stimulation modalities available, with a well-characterized adverse effect profile.

Common Side Effects

• Scalp pain/discomfort at the stimulation site: The most common side effect, reported by 20–40% of patients, particularly during initial sessions. It typically habituates over the first week and can be managed with pre-treatment analgesics (acetaminophen or ibuprofen) or minor adjustments to coil position.
• Headache: Reported by approximately 15–25% of patients, typically mild and transient, resolving within hours of treatment.
• Facial twitching: Involuntary contraction of scalp or facial muscles during stimulation is common and benign.

Rare but Serious Adverse Events

• Seizure: The most serious acute risk. The estimated incidence is approximately 0.003% per session (approximately 1 in 30,000) when International Federation of Clinical Neurophysiology (IFCN) safety guidelines are followed. Risk factors include personal or family history of seizures, concurrent medications that lower seizure threshold (bupropion, clozapine, theophylline), alcohol withdrawal, and sleep deprivation. All established safety guidelines (Rossi et al., 2009; updated 2021) specify maximum stimulation parameters (frequency, intensity, train duration, inter-train interval) designed to minimize seizure risk.
• Mania/hypomania induction: Rare, reported primarily in patients with undiagnosed or known bipolar disorder. Estimated incidence is <1%, but screening for bipolar spectrum disorders is recommended before initiating TMS.
• Hearing changes: The TMS coil produces a loud clicking sound (~100+ dB). Earplugs are mandatory for all treatments. Transient hearing threshold shifts have been reported rarely; permanent hearing loss is essentially unreported with proper ear protection.

Cognitive Effects

In contrast to ECT, TMS has no negative effects on cognition. Neuropsychological testing consistently shows no decline in memory, attention, or executive function during or after TMS courses. Some studies report modest improvements in cognitive function, likely secondary to depression amelioration rather than direct cognitive enhancement, though prefrontal stimulation may independently enhance certain executive functions.

Absolute Contraindications

• Ferromagnetic implants in or near the head (excluding dental fillings)
• Implanted metallic particles near the coil (e.g., shrapnel)

Relative Contraindications

• Personal or family history of epilepsy
• Medications that substantially lower seizure threshold
• Active or recent alcohol/benzodiazepine withdrawal
• Intracranial lesions or raised intracranial pressure
• Cochlear implants (require case-by-case assessment)

A critical clinical question is how long TMS benefits persist after an acute treatment course and whether maintenance strategies can extend durability.

Durability of Acute Treatment Response

Naturalistic follow-up studies suggest that approximately 50–70% of initial responders maintain their clinical gains at 6 months and 40–50% at 12 months. However, relapse rates are substantial: approximately 30–40% of responders experience symptom recurrence within the first year, a rate comparable to relapse after antidepressant pharmacotherapy. The Carpenter and colleagues (2012) durability study following the original FDA pivotal trial reported that 63% of patients who had responded maintained response over 24 weeks with monthly taper and maintenance sessions.

Maintenance TMS

Maintenance TMS — typically 1–2 sessions per week or per month following acute response — has been studied in several open-label trials and a smaller number of controlled studies. While no universally standardized maintenance protocol exists, clinical practice often involves a gradual taper (e.g., 3 sessions/week for 2 weeks, then 2 sessions/week for 2 weeks, then 1 session/week, then biweekly or monthly) with re-intensification at early signs of relapse. Observational data suggest that maintenance TMS can extend remission periods, though large RCTs specifically comparing maintenance TMS to no maintenance in TMS responders are still needed.

Retreatment

Patients who relapse after a successful TMS course can be retreated. Data from the NeuroStar registry and other sources indicate that retreatment response rates are comparable to initial treatment, with approximately 80–85% of patients who responded to initial TMS responding to retreatment. This is an important clinical reassurance for patients and providers.

Despite significant progress, several major questions remain open in TMS research.

Accelerated Protocols

The SAINT/SNT protocol has demonstrated that multiple daily sessions can be safely administered with striking efficacy over just 5 days. However, the initial trials were small (n = 14–29), and larger replication trials are essential. Whether similar accelerated approaches can be successfully applied using standard figure-of-eight coils (rather than functional connectivity guidance and specialized infrastructure) is being investigated in pragmatic trials.

Precision Targeting

The shift toward fMRI-guided, individually optimized targeting is perhaps the most transformative research direction. However, barriers include the cost and availability of individual resting-state fMRI, reliability of functional connectivity maps across sessions, and the need for automated targeting pipelines. Companies and academic groups are developing scalable commercial solutions, but widespread clinical adoption is likely years away.

Biomarker-Guided Treatment Selection

Identifying prospective biomarkers that predict who will respond to TMS (versus medication, psychotherapy, or ECT) would transform treatment selection. Candidate biomarkers include baseline functional connectivity profiles, frontal theta power on EEG, inflammatory markers (CRP, IL-6), and machine learning–derived composite scores. The NIMH-funded EMBARC study and similar precision psychiatry initiatives are working toward clinically actionable prediction models, but none have yet achieved sufficient accuracy for individual-patient decision-making.

Novel Indications

Large-scale trials of TMS for anorexia nervosa (targeting the DLPFC or DMPFC), autism spectrum disorder (social cognition), and post-stroke depression and aphasia are in progress. The use of TMS as a cognitive neuroscience tool — for mapping causal brain-behavior relationships — continues to inform these clinical applications.

Limitations of the Evidence Base

• Sham credibility: Maintaining effective blinding in TMS trials is challenging because active TMS produces scalp sensation and muscle twitching that sham conditions may not fully replicate. While newer sham coils and active sham conditions (e.g., electrical scalp stimulation) improve blinding, some degree of unblinding in TMS trials is likely, potentially inflating effect sizes.
• Heterogeneity: Substantial heterogeneity in TMS protocols (coil type, frequency, intensity, number of sessions, targeting method), patient populations, and outcome measures across trials complicates meta-analytic synthesis and direct comparison.
• Representation: TMS trials have predominantly enrolled White, educated, insured patients in North America and Europe. Evidence from diverse populations, including racial/ethnic minorities, older adults, and patients in low- and middle-income countries, is lacking.
• Active comparator trials: Head-to-head trials comparing TMS to established pharmacotherapy or psychotherapy are rare. Most evidence compares TMS to sham, limiting clinicians' ability to position TMS within treatment algorithms based on direct comparative data.

Current clinical practice guidelines — including those from the American Psychiatric Association (APA), the Canadian Network for Mood and Anxiety Treatments (CANMAT), and the National Institute for Health and Care Excellence (NICE) — position TMS as a second- to third-line treatment for MDD, typically recommended after failure of at least one (and more commonly two) adequate pharmacotherapy trials.

The 2016 CANMAT guidelines explicitly recommend rTMS as a first-line option for patients who have not responded to at least one antidepressant in the current episode, and the 2019 update incorporated iTBS alongside standard 10 Hz rTMS. The 2023 APA Practice Guidelines similarly endorse TMS for pharmacoresistant depression, noting its favorable safety profile relative to ECT for patients who do not require urgent intervention.

In clinical practice, TMS is most commonly used as an augmentation strategy — added to ongoing pharmacotherapy and/or psychotherapy — rather than as monotherapy. The decision to pursue TMS versus other augmentation strategies (lithium augmentation, atypical antipsychotic augmentation, combination antidepressants, or ketamine/esketamine) should be individualized based on prior treatment history, patient preference, medical comorbidities, local availability, insurance coverage, and the relative urgency of the clinical presentation.

A pragmatic clinical algorithm might position TMS after failure of 1–2 adequate pharmacotherapy trials, in parallel with psychotherapy (particularly CBT or behavioral activation), for patients with moderate-severity nonpsychotic depression. ECT would be preferred for severe, psychotic, or imminently life-threatening presentations. Esketamine (Spravato) occupies a similar niche to TMS for treatment-resistant depression and may be appropriate when rapid onset (within days rather than weeks) is prioritized, though it requires monitoring for dissociation and has abuse potential concerns.

The practical barriers to TMS access — including cost (typically $6,000–$15,000 for a full course), time commitment (daily sessions for 4–6 weeks for standard protocols), geographic availability, and variable insurance coverage — remain significant. The adoption of iTBS (reducing per-session time dramatically) and accelerated protocols (reducing total treatment duration to days) addresses some but not all of these barriers.