A systematic review of transcranial magnetic stimulation for generalized anxiety disorder, obsessive-compulsive disorder, and posttraumatic stress disorder

Abstract

Background

The evidence base addressing transcranial magnetic stimulation (TMS) for behavioral health conditions other than treatment-resistant depression is not well established. This systematic review evaluated the efficacy and safety of TMS for persons with generalized anxiety disorder (GAD), obsessive-compulsive disorder (OCD), and posttraumatic stress disorder (PTSD).

Methods

PubMed, the Cochrane Library, and PsycINFO were searched for English language studies published from inception through July 7, 2025. Eligibility criteria included randomized, sham-controlled trials (RCTs) enrolling participants with GAD, OCD, and PTSD and measuring outcomes included remission, response, change in disease severity, and adverse events (AEs). One reviewer extracted data, and a senior reviewer checked for accuracy. Two reviewers completed independent risk-of-bias and strength-of-evidence (SOE) assessments. We conducted quantitative syntheses when appropriate.

Results

We identified 20 RCTs of TMS evaluating efficacy and 18 RCTs evaluating safety. Few studies reported follow-up beyond 12 weeks and several trials had high risk-of-bias. For remission, we found no evidence of benefit for TMS compared with sham for any condition. However, compared to sham, active TMS produced greater clinical response for OCD (Pooled RR, 1.74 [95% CI 1.06 to 2.84; 11 RCTs]; moderate SOE) and improvements in symptom severity for all three conditions (low SOE). There was no difference between active and sham TMS for serious AEs (low SOE) or total AEs (moderate SOE). Few studies reported follow-up beyond 12 weeks and many trials had high risk of bias and small sample sizes.

Conclusions

TMS probably improves clinical response in OCD and may improve symptom severity in GAD, OCD, and PTSD. Further research should determine optimal brain targets and TMS treatment parameters and assess the durability of outcomes at longer term follow-up times, to better support clinical decision making and coverage determinations.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-026-07930-4.

Background

Current treatment approaches for behavioral health disorders begin with psychotherapy, pharmacotherapy, or both. Transcranial magnetic stimulation (TMS), an outpatient noninvasive neuromodulation procedure, is an option for individuals who do not achieve adequate clinical responses or who have adverse effects from psychotherapy, pharmacotherapy, or other procedures (e.g., electroconvulsive therapy). During the TMS procedure, a coil placed against the scalp generates a focal magnetic field that passes through the skull and causes neurons on the cortical surface of the brain to depolarize, which can trigger activity in larger neural circuits and functional neural networks in the brain [1, 2]. TMS may be delivered in multiple pulses, most commonly referred to as repetitive TMS (rTMS). Variants of rTMS include deep TMS (dTMS), which uses specialized coils to stimulate deeper brain regions, and theta-burst stimulation (TBS), a high-intensity treatment delivered in pulses that mimic theta brain waves [3]. TMS has substantial evidence supporting its use for treatment-resistant depression; however, the evidence base for other mental health conditions is less conclusive, with studies reporting conflicting results or using non-comparative study designs [4–7]. The rationale for TMS is that it can be efficacious with fewer and more tolerable side effects than other therapies which has led to wider use of TMS for common non-depression conditions that often do not respond to first-line therapies [8, 9].

Generalized anxiety disorder (GAD), obsessive-compulsive disorder (OCD), and post-traumatic stress disorder (PTSD) have overlapping anxiety-related symptoms, comorbidities, and treatment approaches, and TMS may be considered in cases of limited response to psychotherapy and pharmacotherapy. In this review, we evaluate the efficacy and harms associated with TMS for the treatment of GAD, OCD, and PTSD compared to sham treatment to support clinical- and policy-related decisions regarding the use of TMS for these common mental health conditions.

Methods

Search strategy and study selection

We searched PubMed, the Cochrane Library, and PsycINFO for studies from database inception to July 7, 2025, and reviewed the reference lists of relevant studies and systematic reviews (Detailed search strategy in Appendix A1). Two reviewers independently screened titles and abstracts and full-text articles based on the study selection criteria detailed in Appendix A2. In brief, we selected studies conducted among individuals with a clinical diagnosis of GAD, OCD, or PTSD where rTMS, dTMS or TBS, was delivered over more than one session. Randomized controlled trials (RCTs) or crossover trials with a sham TMS comparator were included. For crossover trials, we extracted outcome data collected prior to crossover only. Eligible outcomes included clinical response, remission, change in severity of illness, measured by validated instruments or clinical evaluation, and adverse events. Clinical response and remission were defined according to study-specific thresholds and are summarized in the Appendix B efficacy outcome tables.

We excluded studies published in languages other than English or conducted in countries that were not very highly developed based on the 2020 United Nations Human Development Index (HDI) [10]. This criterion ensured applicability of findings to settings with similar standards for mental health care. We excluded head-to-head comparisons between alternative TMS protocols and comparisons between TMS and medication. Studies with multiple intervention arms were included if a sham control group was also included, as were studies including pharmacotherapy or psychotherapy co-interventions (per protocol or per usual care). Lastly, studies with fewer than 10 participants in eligible study arms were excluded due to low likelihood of being adequately powered.

Data abstraction and study risk-of-bias assessment

One investigator extracted study data that a senior investigator checked for accuracy. Two investigators conducted independent risk-of-bias (RoB) assessments using the Cochrane Risk of Bias tool (RoB 2) [11]. Disagreements were resolved through discussion or a third reviewer.

Data synthesis and strength of evidence grading

We synthesized study characteristics and efficacy results by condition in tabular and narrative formats. Because we did not suspect that harms of the procedure would vary by condition, we evaluated harms across conditions. We conducted quantitative synthesis when appropriate using established guidance [12, 13] with random-effects models using the inverse variance method of DerSimionian and Laird to generate pooled mean differences (MD) or standardized mean differences (SMD) for continuous outcomes and pooled risk ratios (RRs) for categorical outcomes [14]. We performed a sensitivity analysis excluding high RoB studies from pooled analyses.

We graded the strength of evidence (SOE) for each clinical condition and category of outcome using the Agency for Healthcare Research and Quality Evidence-based Practice Center SOE approach [15], which is derived from the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach [16]. We combined multiple outcome measures within the same outcome domain and graded SOE for remission, response, disease-specific continuous outcomes, and safety. SOE was graded as insufficient, low, moderate, or high and reflected our level of confidence in the findings. Bodies of evidence from RCTs began with a high rating and were downgraded based on evaluation of consistency, precision, directness, and study limitations (i.e., risk of bias). We assessed consistency based on the direction and magnitude of effects; single-study bodies of evidence were downgraded one level. For precision, we used the confidence interval (CI) width for pooled estimates, overall sample size relative to optimal information size, or presence of complete data including variance. All included outcomes and comparisons were considered direct, and reporting bias was assessed as part of study limitations.

Results

We screened 2,422 unique citations and identified 20 RCTs that evaluated the efficacy (20 studies) and 18 studies evaluating harms of TMS (Fig. 1). Individual study characteristics, including population (e.g., comorbidities), and intervention characteristics (e.g., TMS parameters) and findings are provided in Appendix B. Studies excluded at the full-text stage are provided in Appendix C. Note that articles may have been excluded for more than one reason, but we report only one reason. We report individual study RoB assessments in Appendix D and provide a description of validated measures reported by studies in Appendix E.

Fig. 1.

Study flow diagram for systematic review on TMS for treatment of selected behavioral health disorders*. *Original search included additional conditions for MDD, SUD, and TUD for a health technology assessment with a larger scope. PRISMA adjusted to classify these conditions as wrong population. ECT = electroconvulsive therapy; MH = mental health; MDD = major depressive disorder; RCT = randomized controlled trial; SUD = substance use disorder; TMS = transcranial magnetic stimulation; TUD = tobacco use disorder

Clinical remission and response were commonly reported efficacy outcomes. Study authors required a score below a specific threshold on a condition specific assessment to define remission and response, though thresholds varied across studies. Authors reported changes in symptom severity using validated condition-specific instruments. Adverse event reporting was inconsistent; studies reported on any adverse events (AEs), serious AEs, or both, but may not have used similar definitions of these harm categories. Additionally, some studies only reported specific AEs, such as headache or scalp pain.

Generalized anxiety disorder

Study and population characteristics

Two RCTs compared rTMS to sham stimulation for the treatment of GAD (Appendix B, Tables B-1, B-2, and B-3). One trial had some concerns for RoB for differences in baseline severity of anxiety between groups that did not appear to be adjusted for in the analysis [17], and 1 trial was high RoB due to overall and differential attrition, and lack of transparency in reporting of patient flow [18].

Both studies recruited patients with moderate to severe anxiety; mean baseline Hamilton Anxiety Rating Scale (HARS) score ranged from 20 to 25 in one study [17] and 29 to 32 in the other study [18]. Neither study specified if participants were treatment-resistant.

In both RCTs, rTMS was provided over 6 weeks (2 to 5 sessions per week), targeted the right dorsolateral prefrontal cortex (DLPFC), and reported follow-up at 12 weeks. One study [18] used high frequency (HF) rTMS and the other used low frequency (LF) rTMS [17]. Studies included co-interventions of medications [17] or medications and psychotherapy [18].

Findings

A summary of findings and the SOE are in Table 1 (details in Appendix B, Tables B-4 and B-5). We assessed SOE as insufficient to evaluate the impact of active TMS on remission. We rated SOE as low for response to treatment and a larger improvement in clinician-rated symptoms (HARS score) for active versus sham TMS.

Table 1.

Summary of findings and SOE for TMS compared to sham stimulation for GAD

No. studies/no. participants TMS type	Summary of effect	Consistency	Precision	Directness	Study limitations	SOE/ direction of effect
Remission of GAD symptoms at post-treatment (6 weeks) and last follow-up (12 weeks post-treatment)
2 RCTs/76 participants • rTMS [17, 18]	One study reported remission measures favoring TMS immediately after treatment (6 weeks, 30.8% vs. 8.3%, p = NS) and at 12-weeks (53.8% vs. 0%, p = 0.003). One study reported remission in 80% of the intervention group at posttreatment and 100% at 12-week follow-up, but no remission data for the control group was reported.	Consistency NA*	Imprecise	Direct	High (1 SC, 1 high RoB)	Insufficienta, b NA
Response of GAD symptoms at post-treatment (6 weeks) and last follow-up (12 weeks post-treatment)
2 RCTs/76 participants • rTMS [17, 18]	Both studies reported greater response to treatment for participants allocated to active TMS compared to sham TMS immediately post-treatment (61.5% vs. 16.7%, p = 0.022 in 1 study; 100% vs. 12%, calculated p < 0.001). Similar findings at 12-week follow-up.	Consistent	Imprecise	Direct	High (1 SC, 1 high RoB)	Lowb, c Favor TMS
Symptom severity, at post-treatment (6 weeks) or last follow-up (12 weeks post-treatment), as measured by HARS score
2 RCTs/76 participants • rTMS [17, 18]	Larger change in HARS score from baseline immediately post treatment in 1 study (p < 0.001); lower mean HARS score immediately post-treatment and at 12 weeks in other study (p < 0.001).	Consistent	Imprecise	Direct	High (1 SC, 1 high RoB)	Lowb, c Favor TMS

^* Only one study reported posttreatment or follow-up data, so consistency was unable to be determined

^a Downgrade 2 levels for imprecision

^b Downgrade 1 level for study limitations

^c Downgrade 1 level for imprecision

CGI-S = clinical global impression scale-severity; GAD = generalized anxiety disorder; HARS = Hamilton Anxiety Rating Scale; NA = not applicable; NR = not reported; RCT = randomized controlled trial; RoB = risk of bias; SC= some concerns; SOE = strength of evidence; TMS = transcranial magnetic stimulation

Remission of GAD symptoms

Both RCTs [17, 18] reported on remission, defined as a HARS score < 10 in one study [18] and HARS score < 8 and a Clinical Global Impression-Improvement (CGI-I) score of 1 or 2 in the other study [17]. One study [17] reported increased incidence of remission in the TMS group immediately post-treatment (6 weeks, 30.8% vs. 8.3%, p = 0.161), although it did not achieve statistical significance. However, this result remained durable and became statistically significant at the 12-week follow-up (p = 0.003). In the other study [18] remission was reported in 80% of participants at post-treatment and in 100% of participants at the 12-week follow-up for the active TMS group. Remission was not reported for the control group and statistical significance was not reported for either time point.

Response to treatment

Both RCTs [17, 18] defined response as ≥ 50% improvement on HARS. One RCT [17] found that participants in the TMS group were more likely to have a response to treatment (61.5%) than those in the sham group immediately post-treatment (61.5% vs.16.7%; p = 0.022) and at 12-week follow-up (61.5% vs. 0%; p = 0.001). In the other RCT [18], authors reported a 100% response in for active TMS compared with 12% in the sham group at post-treatment (6 weeks) and at 12 weeks follow-up (calculated p < 0.001 at both time points).

Symptom severity

Both RCTs [17, 18] reported larger improvements in HARS scores for active TMS compared with sham. In the one study [17], authors reported a lower mean HARS score for the TMS group at post-treatment (12.1 vs. 14.4) and 12-week follow-up (10.4 vs. 18.0) compared to the sham group (p < 0.001 at both time points). The other study [18] reported a statistically significant larger improvement in the HARS score from baseline to post-treatment for active TMS compared to sham TMS (p < 0.001), but the actual change in scores was not reported.

Obsessive-compulsive disorder

Study population and characteristics

We identified 14 RCTs [19–32] that evaluated TMS compared to sham among individuals diagnosed with OCD (Appendix B, Tables B-6 to B-8). We assessed 11 studies as having low or some RoB [19, 21–25, 27, 29–32] and 3 as high RoB [20, 26, 28] due to no description of allocation concealment [23, 25–29, 32], failure to include all eligible randomized participants in the analyses [19, 20, 24, 26, 28], and no evidence that the analyses were preplanned [21, 22, 26–28].

Eight studies enrolled patients with minimum threshold Yale-Brown Obsessive Compulsive Scale (Y-BOCS) scores ranging from 15 to 20 (moderate OCD) [19–23, 27, 31, 32]. The remaining 6 studies did not have a threshold Y-BOCS requirement for enrollment, though baseline scores indicated moderate to severe OCD [24–26, 28–30]. All but 3 studies [25, 30, 32] enrolled participants who were defined as treatment-resistant, in which participants failed at least 1 to 2 medication trials. Two studies also required failed trials of cognitive behavioral therapy [19, 20]. Comorbid major depression ranged from 12% to 81% of participants in 5 studies [21, 22, 24, 27, 31]. Participants continued usual medication and/or psychotherapy treatment in 12 studies [19–24, 26–30, 32] or used a maximum dose of medications maintained at a stable regimen throughout treatment in 1 study [25]. Medication use was not reported in 1 study [31].

The active intervention was rTMS in 8 studies [22–28, 31], dTMS in 3 studies [19, 20, 29], continuous TBS in 2 studies [21, 32], and deep intermittent TBS in 1 study [30]. In 5 studies, the targeted area included the supplementary motor area (SMA) or pre-SMA [21–24, 31], the right or left DLPFC in 4 studies [24, 26, 27, 31]; the targeted areas for the remaining studies were right orbitofrontal cortex [28], and the anterior cingulate cortex (ACC) and the medial prefrontal cortex [19, 20, 25, 29, 30, 32]. Treatment duration ranged from 10 to 30 days with 3 studies administering TMS over two sessions per day [25, 29, 30]. Four studies incorporated OCD provocation before treatment [19, 20, 29, 30].

Findings

A summary of findings and the SOE are in Table 2 (details in Appendix B, Tables B-9 and B-10). We rated SOE as insufficient for remission, moderate favoring TMS for response outcomes, and low favoring TMS for change in Y-BOCS scores.

Table 2.

Summary of findings and SOE for TMS compared to sham for OCD

No. studies/no. participants TMS type	Summary of effect	Consistency	Precision	Directness	Study limitations	Overall SOE/ direction
Clinical remission (Y-BOCS score ≤ 12) post-treatment (2 to 3 weeks) and last follow-up (2 weeks post-treatment)
2 RCTs/66 participants • dTMS [29] • d-iTBS [30]	Few remission events reported in either TMS (n = 1 to 2) or sham groups (n = 0 to 1).	Consistency unknown	Imprecise	Direct	Medium (2 SC)	Insufficienta
Clinical response (Decrease in Y-BOCS score of 25% to 35% or more) post-treatment (2 to 6 weeks) and last follow-up (2 to 12 weeks post-treatment)
11 RCTs/409 participants • rTMS [22–24, 27, 28, 31] • dTMS [19, 20, 29] • cTBS [21] • d-iTBS [30]	Pooled RR 1.74 (95% CI 1.06 to 2.84). ARD per 1,000 was 127 more (95% CI, 10 to 316 more). In the 3 of 6 studies with post-treatment follow-up, durability of response up to 4 weeks after the end of treatment was observed.	Consistent (I2 =37.2%)	Imprecise	Direct	Medium (1 low, 8 SC, 2 high RoB)	Moderateb Favor TMS
Symptom severity as measured by Y-BOCS score post-treatment (2 to 6 weeks) and last follow-up (2 to 24 weeks post-treatment)
14 RCTs/517 participants • rTMS [22–28, 31] • dTMS [19, 20, 29] • cTBS [21, 32] • d-iTBS [30]	At post-treatment, 8 of 14 studies reported symptom improvements for TMS compared to sham. There were no statistically significant results in the other 5 studies.	Consistent	Imprecise	Direct	High (1 low, 10 SC, 3 high RoB)	Lowb Favor TMS

^a Downgraded 3 levels for imprecision

^b Downgraded 1 level for imprecision

^c Downgraded 1 level for study limitations: lack of required data to adequately evaluate magnitude and direction of effect

CI = confidence interval; cTBS = controlled theta-burst stimulation; d-iTBS = deep intermittent theta burst stimulation; dTMS = deep transcranial magnetic stimulation; OCD = obsessive-compulsive disorder; RCT = randomized controlled trial; RoB= risk of bias; RR = risk ratio SC = some concerns; SOE = strength of evidence; TMS = transcranial magnetic stimulation; Y-BOCS = Yale-Brown Obsessive-Compulsive Scale

Remission

Two studies reported on remission, defined as posttreatment Y-BOCS score of 12 or lower [29, 30]. Few events occurred in either study; two participants in the active TMS achieved remission, while none achieved remission in the sham group at the end of treatment (3 weeks) in one study [29] and 1 participant in each study group achieved remission in the other study [30].

Response

Eleven studies reported clinical response [19–24, 27–31]. Most studies defined clinical response as ≥ 25% decrease in Y-BOCS score, 3 studies used a threshold of 30% [19, 20, 29] and 3 used a threshold of 35% [28, 30, 31]. In a pooled analysis, more participants receiving active TMS achieved a response compared to sham at 2 to 6 weeks post-treatment (pooled RR, 1.74; 95% confidence interval [CI], 1.06 to 2.84; I² = 35.2%; 409 participants. (Fig. 2). This is equivalent to an absolute risk difference of 127 more clinical responses per 1,000 participants (95% CI, from 10 more to 316 more). Studies of dTMS appeared to have a greater magnitude of effect compared with other types of TMS. In a sensitivity analysis removing the two high risk-of-bias studies [20, 28], the pooled RR was 1.64 (95% CI 0.97 to 2.76; I ² = 43.3%, Appendix F, Figure F1).

Fig. 2.

TMS vs. Sham for outcome of clinical response for OCD. ARD = absolute risk difference; CI = confidence interval; cTBS = continuous theta-burst stimulation; d-iTBS = deep intermittent theta-burst stimulation; DL = DerSimonian & Laird estimator for pooling estimates; dTMS = deep transcranial magnetic stimulation; LF-rTMS = low frequency repetitive transcranial magnetic stimulation; N = number; OCD = obsessive-compulsive disorder; TMS = transcranial magnetic stimulation; vs. = versus; Y-BOCS = Yale-Brown Obsessive-Compulsive Scale

Among the 6 studies [19–21, 23, 30, 31] reporting follow-up past the end of treatment, the improvement in response persisted in at follow-up time points ranging from 2 to 6 weeks post treatment for 3 studies [19, 20, 23].

Symptom severity

All 14 studies used the Y-BOCS to assess change in OCD symptoms, which was the primary endpoint in all but one study [25]. Six studies did not provide data (i.e., variance or follow-up Y-BOCS values) necessary to pool results.

Eight of the 14 studies reported greater symptom improvements using the Y-BOCS score among the TMS-treated group compared to the sham group that were statistically significant or not reported [19, 20, 23, 25, 27, 29, 30, 32]. In the largest of the included studies for OCD patients (n = 100), the intention-to-treat and modified intention-to-treat (mITT) analyses favored the intervention group with an effect size of 0.48 (p = 0.09) and 0.69 (p = 0.01), respectively, at the end of the 6-week treatment. This finding persisted and was statistically significant at 4 weeks post-treatment (effect size = 0.62, p = 0.03, mITT population) [19]. A study of rTMS treatment for 3 weeks also reported significantly greater improvements in symptom severity for those receiving active treatment compared to those receiving sham (mean change − 10.7 [standard deviation (SD) 8.2] vs. -3.7 [SD 3.7], p = 0.005) at the end of treatment [27]. In another study, 6-week treatment with rTMS was associated with significant symptom improvement compared to sham (p < 0.001) up to 4 weeks post-treatment [23]. A study of rTMS treatment for 3 weeks also reported significantly greater improvements in symptom severity for those receiving active treatment compared to those receiving sham (mean change − 10.7 [standard deviation (SD) 8.2] vs. -3.7 [SD 3.7], p = 0.005) at the end of treatment [27].

Three studies reported persistent symptom reduction for TMS compared with sham at follow-up points after the end of treatment, ranging from 4 weeks to 6 months. In a study with a deep intermittent theta-burst stimulation (d-iTBS) intervention, mean YBOC scores decreased significantly over 4 weeks (p = 0.04) and at 2 weeks post-treatment [30]. Symptom severity improved at the end of treatment and persisted at 6 months post-treatment in a study of continuous theta-burst stimulation (cTBS) [32]. And in the largest study conducted in participants with OCD, reduction in symptom severity persisted 6 weeks after the end of treatment, as described above [19]. The remaining 3 studies, which used interventions of rTMS [25] and dTMS [20, 29] reported no statistically significant differences between groups for change in symptom severity at follow-up times ranging from 1 week to 3 months.

Seven studies reported no statistically significant differences in symptom severity as measured by the Y-BOCS for TMS compared to sham. There was no significant effect post-treatment or at 12 weeks following treatment in one study of rTMS [31]. The absolute values of follow-up scores or change in scores were not provided to determine which treatment had a greater effect in 3 studies [21, 24, 26], while 1 study favored the sham group [22]. In the one crossover trial, the Y-BOCS scores reported at 2 weeks (before crossover) were not significantly different between the rTMS and sham (27.1 [SD 2.6] vs. 26.1 [SD 2.5], p = 0.309); the change from baseline was not reported [28].

Posttraumatic stress disorder

Study and population characteristics

We identified 4 RCTs that focused on TMS [33–35] or TBS [36] compared to sham for the indication of PTSD. We assessed 1 of these trials as low RoB [36]; 2 as having some concerns for RoB due to lack of blinding of some participants, moderate attrition, and incomplete outcome data [33, 34]; and 1 study as having high RoB arising from the randomization process and selective outcome reporting [35]. Details of included TMS trials are in Appendix B, Tables B-11, B-12, and B-13. Two studies enrolled participants with Clinician Administered PTSD Scale (CAPS) scores of at least 25 [33] or 50 [35] the remaining 2 studies did not require minimum CAPS scores, although baseline scores corresponded to moderate [36] and severe [34] disease. One study [35] included participants defined as treatment-resistant, 1 study [36] included both treatment-naive and treatment-resistant participants, and 2 studies [33, 34] did not specify treatment history of the study populations. Three studies included only military veterans [34–36].

The active intervention in 2 RCTs was LF rTMS of the right DLPFC [34, 35]. One study [36] used intermittent TBS of the right DLPFC, and 1 study used dTMS to bilaterally stimulate the medial prefrontal cortex (MPC) and ACC [33]. Two studies [35, 36] provided active or sham treatment over 2 weeks with five sessions per week; 1 study [33] provided 4 weeks of treatment with 3 sessions per week and a booster treatment at weeks 5 and 9, while 1 study provided 1 session per week for 12 weeks [34]. Follow-up ranged from 2 weeks to 6 months post-treatment. One study [34] included cognitive processing therapy to all participants as a co-intervention, and 1 study included exposure therapy aimed to amplify the participants’ symptoms before each TMS session [33].

Findings

A summary of findings and the SOE are provided in Table 3 (details in Appendix B, Tables B-14 and B-15). We assessed the SOE as insufficient for remission and response. We assessed the SOE for change in symptom severity based on the CAPS or PTSD Checklist (PCL), the most commonly used measures for disease severity, as low for favoring TMS.

Table 3.

Summary of findings and SOE for TMS compared to sham for PTSD

No. studies/no. participants TMS type	Summary of effect	Consistency	Precision	Directness	Study limitations	Overall SOE/ direction
Remission (not defined) of PTSD symptoms post-treatment (specific time point NR)
1 RCT /134 participants • dTMS [33]	Study did not report number of events and non-events for TMS and sham groups, though did report no difference between groups.	Unknown (Single study)	Imprecise	Direct	Medium (1 SC RoB)	Insufficienta, b
Clinical response (decrease in symptoms of at least 50% reported in CAPS score) post-treatment (specific time point NR) and 4 weeks post-treatment
1 RCT /134 participants • dTMS [33]	Participants in the sham group were more likely to have a response to treatment at both time points, although results were not statistically significant, and number of events and nonevents were not reported.	Unknown (Single study)	Imprecise	Direct	Medium (1 SC RoB)	Insufficienta, b
Symptom Severity at post-treatment (2 to 5 weeks) and last follow-up (4 to 12 weeks), as measured by CAPS score
4 RCTs/307 participants • dTMS [33] • rTMS [34, 35] • iTBS [36]	Two studies using LF-rTMS showed improvement in CAPS scores for rTMS vs. sham, statistically significant in the larger study (n = 103). One study of iTBS showed no difference from sham, and one study favored sham over dTMS, though the results were not statistically significant.	Inconsistent	Imprecise	Direct	Medium (1 low, 2 SC, 1 high RoB)	Lowc, d Favor TMS
Symptom Severity at post-treatment (2 to 5 weeks) and last follow-up (12 weeks), as measured by PCL score
3 RCTs/173 participants • rTMS [34, 35] • iTBS [36]	One study (n = 20) with high RoB observed statistically significant larger improvements for TMS compared to sham, while a second study (n = 50) with low RoB showed both the iTBS and sham groups improved, but the difference in improvement between the two groups was not statistically significant. Another study (n = 103) with some RoB concerns measured change in score from baseline to last follow-up at 6 months post-treatment and found the rTMS group showed statistically significant improvement in PCL score compared to sham.	Consistent	Imprecise	Direct	Medium (1 low, 1 SC, 1 high RoB)	Lowb Favor TMS

^a Downgraded 1 level for inconsistency—single study

^b Downgraded 2 levels for imprecision

^c Downgraded 1 level for inconsistency

^d Downgraded 1 level for imprecision

CAPS = Clinician Administered PTSD Scale; CPT = cognitive processing therapy; dTMS = deep transcranial magnetic stimulation; iTBS = intermittent theta-burst transcranial magnetic stimulation; LF = low frequency; PCL = PTSD Checklist; PTSD = posttraumatic stress disorder; RCT = randomized controlled trial; RoB = risk of bias; rTMS = repetitive transcranial magnetic stimulation; SC = some concerns; SOE = strength of evidence; TMS = transcranial magnetic stimulation

Remission

One RCT reported that the incidence of remission was very low and did not statistically differ between the TMS and sham groups; however, the authors neither defined remission nor presented numerical findings [33].

Response to treatment

One RCT reported on response to treatment, defined as ≥ 50% decrease from baseline in the CAPS-5 score [33]. Authors found that participants in the sham group were more likely to have a response to treatment (55%) than those in the dTMS group immediately post-treatment (43%), results that were not statistically significant. Four weeks following the end of treatment, response to treatment improved across both groups but was also not statistically significant (63% sham vs. 54% active).

Symptom severity

Clinician administered PTSD scale

All four RCTs [33–36] reported on the change in CAPS score from baseline: 3 studies [33, 34, 36] used the CAPS-5, and one study [35] used the CAPS-1 scale (Fig. 3). Both LF-rTMS studies [34, 35] showed favorable impacts for rTMS, although findings were only statistically significant in one of the studies [34]. The study using intermittent theta-burst stimulation (iTBS) [36] did not show a statistically significant difference between treatments, and the study of dTMS [33] reported the sham group had a greater effect than the dTMS group, although not statistically significant.

Fig. 3.

TMS vs. Sham for Outcome of PTSD Symptom Severity. CAPS = Clinician Administered PTSD Scale; CI = confidence interval; dTMS = deep transcranial magnetic stimulation; iTBS = intermittent theta-burst stimulation; LF-rTMS = low frequency repetitive transcranial magnetic stimulation; N = number; PTSD = posttraumatic stress disorder; SMD = standardized mean difference; TMS = transcranial magnetic stimulation; w = week. Note: Symptom severity as measured by the CAPS-5 or CAPS 1 outcome

PTSD checklist

Three RCTs [34–36] reported on a change in PCL score from baseline to posttreatment. One study (n = 20) with high RoB reported a statistically significant change favoring the rTMS group compared to sham after 2 weeks of treatment [35]. The other study [36] (n = 50) with low RoB found that although the iTBS group improved more than the sham group, between-group difference was not statistically significant (p = 0.31) at 2 weeks posttreatment. One study [34] reported on the change from baseline to 6 months post-treatment and found the rTMS group showed statistically significant improvement in PCL score compared to sham (values not reported, p < 0.05).

Mississippi scale for combat related PTSD

One RCT [34], which was conducted among military veterans, reported on the change in the Mississippi Scale for Combat Related PTSD (M-PTSD) score from baseline to 6 months post-treatment and found the rTMS group showed statistically significant improvement in M-PTSD score compared to sham (values not reported, p < 0.05).

Modified PTSD symptom scale

One RCT [33] reported on a change in the Modified PTSD Symptom Scale (MPSS) score from baseline to end of treatment (MD, 4.6; 95% CI, 1.7 to 7.5) and to 4 weeks post-treatment (MD, 5.65; 95% CI, 2.1 to 9.2). At both time points that the sham group showed a statistically significant improvement in MPSS score compared to the dTMS group (p < 0.05).

Harms

A total of 18 RCTs reported on harms: 2 GAD studies [17, 18], 13 OCD studies [19–22, 24–32], and three PTSD studies [33, 34, 36]. Four studies reported no differences between groups (p > 0.1) for any AEs [19, 21, 29, 33] and of the 16 studies reporting SAEs, most studies reported zero events [21, 22, 24–28, 30, 32]. Specific SAEs included seizure (1 event in the TMS group [18]), suicide ideation (3 events in the TMS group across 2 studies [19, 33] and 2 events across 2 studies in the sham group [32, 36]), moderate to severe anxiety (2 events in the sham group for 1 study [33]), and chest pain [17]. Treatment site discomfort and headache were the most common AEs; other specific harms identified were fatigue and insomnia. We assessed the SOE for any AEs and SAEs as moderate and low for no difference between active and sham TMS, respectively (Table 4).

Table 4.

Summary of findings and SOE ratings for safety of TMS

No. studies/no. participants TMS type	Summary of effect	Consistency	Precision	Directness	Study limitations	Overall SOE/ direction
Safety (any AEs) post-treatment (2 to 9 weeks or NR) and at last follow up (2 to 14 weeks post-treatment or NR)
4 RCTs/288 participants • dTMS [19, 29, 33] • cTBS [21]	Four studies reported on any AEs, with no differences found between groups (p > 0.1). In 2 of these studies, participants had an adverse event, in both the intervention (range: 73%-77%) and sham (range: 63%-69%) groups.	Consistent	Imprecise	Direct	Medium (3 SC, 1 high RoB)	Moderatea No difference
Safety (SAEs) post-treatment (2 to 9 weeks or NR) and at last follow up (2 to 14 weeks post-treatment or NR)
16 RCTs/755 participants • rTMS [17, 18, 22, 24–28] • rTMS + CPT [34] • dTMS [19, 20, 33] • d-iTBS [30] • cTBS [21, 32] • iTBS [36]	Ten studies reported zero events in both groups. Five studies reported SAEs among one to five participants in the intervention group, including suicidal ideation, moderate to severe anxiety, seizure, and chest pain. Three studies reported SAE among two to four participants in the control group, including suicide ideation and moderate to severe anxiety.	Consistent	Imprecise	Direct	Medium (2 low, 10 SC, 4 high RoB)	Lowb No difference

^a Downgrade 1 level for imprecision

^b Downgrade 2 levels for imprecision

AE = adverse event; CPT = cognitive processing therapy; cTBS = continuous theta-burst stimulation; dTMS = deep transcranial magnetic stimulation; iTBS = intermittent theta-burst stimulation; NR = not reported; RCT = randomized controlled trial; RoB = risk of bias; SAE = serious adverse event; SC = some concerns; SOE = strength of evidence; rTMS = repetitive transcranial magnetic stimulation; TMS = transcranial magnetic stimulation

Discussion

Relative to treatment-resistant depression, fewer studies have evaluated the efficacy and harms of TMS for GAD, OCD, and PTSD. The majority of studies identified in this review were conducted in study populations with OCD, whereas the evidence base for GAD and PTSD was more limited. We rated the SOE as insufficient to assess the impact on remission outcomes for all three conditions. Across conditions, we rated the SOE as low (SAEs) or moderate (AEs) for no difference in harms between active and sham TMS.

We rated the improved response to treatment for participants with OCD from TMS as moderate SOE, while improvement in symptom severity was low SOE, based on all or nearly 80% of identified studies, respectively. The results highlight a growing evidence base supporting the use of TMS for individuals with OCD. Future research that addresses limitations of the current evidence base, such as high RoB and heterogeneity of protocols and co-treatments, may provide more insight into and confidence in the role of TMS as a therapeutic option for the treatment of OCD.

Findings for GAD and PTSD were drawn from a small number of heterogenous studies with difference in TMS type, TMS target regions, and treatment protocols, and it remains unclear if TMS improves outcomes for individuals with these conditions. We rated the SOE for response to treatment as insufficient for PTSD and low for GAD. We found an improvement in symptom severity for both conditions with active TMS compared to sham (low SOE); however, data was limited to evaluate whether these improvements were clinically meaningful. The low or insufficient SOE for all GAD and PTSD efficacy outcomes reflects the small number of trials, imprecision, and study limitations, which limit confidence in the whether observed changes represent meaningful clinical benefit.

We identified variation with respect to TMS protocol used, including brain location targets, treatment duration, TMS type, and co-treatments, which likely accounts for the heterogeneity of effects that we observed, including the heterogeneity reported in the meta-analysis of OCD response.

Identifying individuals who would benefit most from TMS is a crucial step in clinical decision making as well as policy-level decisions such as coverage determinations. TMS is generally considered when individuals do not achieve adequate response to current first-line psychotherapy or pharmacotherapy, which is often termed treatment-resistant [37] and indicates greater severity of disease. Most OCD studies in this review enrolled treatment-resistant populations, while GAD studies did not specify treatment resistance and PTSD studies varied with respect to prior treatment response. Studies of TMS in populations with different levels of severity of disease could elucidate which patients would benefit most from TMS; for example, whether TMS would provide greater benefit at a less severe stage of illness or if TMS is more appropriate to consider in the current step-wise approach. The results of this review suggest that TMS may be of benefit for individuals with higher severity of OCD. More information about disease severity data in future studies would better support decisions about when to consider TMS in the treatment pathway.

Evidence was limited with respect to longer term follow-up of outcomes for the TMS group compared to sham. Only one study in participants with GAD and another study in participants with PTSD had follow-up of 3 months or longer [17, 34]; results were durable at 3 to 6 months with respect to remission, response, or reduction in symptom severity. For the 9 studies in participants with OCD reporting on response to TMS and 6 studies reporting on change in severity after the end of treatment, results were mixed on whether improvements in these measures persisted after the end of treatment. Thus, the important clinical question of whether TMS provides durable benefits remains, a relevant consideration given the most commonly studied protocols involve multiple sessions over several weeks. While there is a growing evidence base for accelerated protocols [5], data on durability of results help providers and patients decide where TMS sits within the context of longer-term treatment planning.

The strengths of our review included using narrower study selection criteria compared with existing systematic reviews of TMS for the treatment of GAD, OCD, and PTSD, which often included uncontrolled studies, studies using small sample sizes (< 10 participants per study arm), studies conducted in non-very high HDI countries; or studies enrolling individuals with conditions other than GAD, OCD, or PTSD. Our systematic review limited inclusion criteria to controlled trial designs, validated clinical outcome measures to better compare the effects of TMS compared to sham, and a minimal sample size (> 10 participants per study arm) to focus on studies with a higher likelihood of being adequately powered for relevant clinical outcomes. Our findings of improvement in GAD symptoms are consistent with some existing systematic reviews and others with no effect [4–6, 38–42]; however, these reviews included uncontrolled studies, which likely accounts for the larger magnitude of effect sizes reported, or included study populations with multiple anxiety disorders or combined other non-invasive brain stimulation techniques, such as transcranial direct current stimulation, with TMS. Several systematic reviews reported reduction in OCD symptoms and improved response rate, similar to our results [43–46]. However, in our review, study populations primarily included moderate to severe OCD, who are more likely to receive TMS. Other reviews included patients with less severe disease [43, 47]. Existing reviews of TMS for PTSD [7, 48–52] reported reduction of symptoms [48, 50], or no difference compared with sham [7, 49, 52].

This review has several limitations. We included only English-language studies and studies conducted in countries with very high HDI due to resource constraints. However, the prevalence of TMS in lower income countries may be limited by several factors, including the high cost of the equipment and lack of trained personnel; the lower adoption of this treatment modality outside of very high HDI countries makes it less likely we missed relevant studies. We did not evaluate the comparative effectiveness of alternative TMS protocols and excluded medications as a comparator. However, in current clinical practice, TMS is most often used after failure of multiple medications, making the comparison of TMS to pharmacotherapy less relevant.

The evidence base has several limitations. Several RCTs had high RoB. Removing high RoB studies from the meta-analysis of OCD response for TMS compared with sham therapy, we found a slightly attenuated risk ratio with confidence intervals that no longer excluded a null effect largely attributed to the exclusion of 1 RCT with a very large effect size. Studies inconsistently reported harms limiting our ability to evaluate harms overall and by TMS types. To mitigate this limitation, we combined harms outcomes across indications for TMS. As discussed above, minimal reporting of long-term results and inconsistent or absent reporting of treatment resistance, prevented evaluation of the durability of TMS treatment, as well as optimal target populations for TMS treatment. We identified fewer studies of TMS used for GAD and PTSD, which may be the result of diagnostic complexity; for example, in a review of national and international consensus definitions of treatment resistance for common psychiatric diagnoses, guidelines defining treatment resistance for PTSD or GAD were not identified [53]. Additionally, there may be challenges in standardizing treatment and limited funding as treatment for PTSD and GAD often rely of psychotherapeutic interventions.

Further research needs include identifying the optimal brain target and treatment parameters to evaluate in larger effectiveness trials. Trials should be adequately powered and executed to minimize RoB from attrition, outcome assessment, and unplanned post hoc analyses. Additionally, including a measure of disease severity and treatment resistance of the enrolled population would support clinical decision-making of when to use TMS compared to another therapy, such as additional medication trial or neuromodulation option. For all conditions, longer term follow-up would elucidate the durability of treatment effects and harms, supporting shared decision making between individuals and providers.

Conclusions

TMS probably improves response in OCD and may improve symptom severity in GAD, OCD, and PTSD. Further research should determine optimal brain targets and TMS treatment parameters, standardize the disease severity of the enrolled populations, and assess the durability of outcomes at longer term follow-up times, all of which would better support clinical decision making and coverage determinations.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

SR was the principal investigator and contributed to study design, systematic search, risk of bias assessment, and statistical analysis. LK contributed to study design, risk of bias assessment and statistical analysis. CR, SK, VN, and OV contributed to systematic search, screening, and risk of bias assessment. BG contributed to study design and analysis. All authors contribute to drafting and revising the manuscript and have approved the final article.

Funding

This manuscript is derived from a health technology assessment (HTA) (https://www.hca.wa.gov/assets/program/TMS-final-report.pdf) that our team conducted for the State of Washington Health Care Authority (HCA) that was presented at the State’s Health Technology Clinical Committee meeting (https://www.hca.wa.gov/about-hca/programs-and-initiatives/health-technology-assessment/clinical-committee-meetings-and-materials) on March 17, 2023. The sponsor gave input on inclusion and exclusion criteria for this HTA, but the final study selection criteria were determined by our team. The sponsor played no role in study selection or the collection, analysis, and interpretation of data, in the writing of the report; or in the decision to submit this manuscript. Additional support was provided by the RTI Fellows Program (Dr. Kahwati).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Human ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.