Efficacy and tolerability of repetitive transcranial magnetic stimulation for the treatment of obsessive-compulsive disorder in adults: a systematic review and network meta-analysis

Obsessive-compulsive disorder (OCD) is a common, disabling disorder with lifetime prevalence of 2.5–3%, characterized by distressing obsessions and compulsions. First-line treatments (CBT, SSRIs, or their combination) leave about half of patients insufficiently responsive, motivating evaluation of alternative treatments such as repetitive transcranial magnetic stimulation (rTMS), which is FDA-approved for OCD. rTMS modulates cortical activity noninvasively and can target nodes of cortico-striato-thalamo-cortical (CSTC) circuits implicated in OCD. Frequency matters: low-frequency (≤1 Hz) is generally inhibitory, while high-frequency (≥5 Hz) tends to be excitatory, so target and frequency selection may influence outcomes. The study’s objective was to compare and rank the efficacy and tolerability of different rTMS strategies (targets and frequencies) for adult OCD using pairwise and network meta-analysis, addressing ongoing uncertainty about optimal parameters.

The authors note varied rTMS approaches for OCD without consensus: LF or HF stimulation over the dorsolateral prefrontal cortex (DLPFC), LF over the orbitofrontal cortex (OFC), and HF over the anterior cingulate cortex/medial prefrontal cortex (ACC/mPFC), typically compared with sham. They highlight frequency-dependent effects on cortical excitability and the theoretical rationale tied to CSTC circuitry abnormalities in OCD. Prior evidence has been heterogeneous, with mixed findings across targets and frequencies, prompting a comprehensive synthesis and ranking of strategies.

Design: Systematic review with pairwise and Bayesian network meta-analyses of randomized controlled trials (parallel-group or crossover; only first period of crossover used). Search: Cochrane Central, PubMed, Web of Science, Embase, PsycInfo from inception to March 25, 2020; no language restrictions. Search terms combined OCD-related terms with magnetic stimulation terms. Two independent reviewers conducted searches. Eligibility: Adults (≥18 years) diagnosed with OCD by standard criteria (RDC, DSM, or ICD). Required Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) assessment. Comparators included at least two of: LF-rTMS over DLPFC; HF-rTMS over DLPFC; LF-rTMS over OFC; HF-rTMS over ACC/mPFC; sham rTMS. LF defined as ≤1 Hz, HF as ≥5 Hz. Excluded non-RCTs, unclear rTMS definitions, or unclear randomization methods. Data extraction: Standardized form captured participant characteristics (mean age, female proportion, sample size), rTMS parameters (target location, frequency), treatment characteristics (number of sessions, treatment duration, follow-up), outcomes (means and SDs of baseline and post-treatment Y-BOCS, and drop-outs due to adverse events). Authors were contacted for missing data. Risk of bias: Cochrane risk of bias tool; studies categorized as low, unclear, or high risk. Percent contributions of studies with differing risk levels were computed for each comparison. Evidence grading: GRADE framework applied to network estimates, rating confidence as high, moderate, low, or very low, considering study limitations, imprecision, heterogeneity/inconsistency, indirectness, and publication bias. Statistical analysis: - Pairwise meta-analysis: Random-effects model (STATA metan v14.0). Reported mean difference (MD) in Y-BOCS change and odds ratio (OR) for drop-out, with 95% confidence intervals (CIs). Heterogeneity assessed. - Network meta-analysis: Random-effects Bayesian framework (R packages gemtc and rjags; R v3.6.3). Two MCMC chains with different initial values; convergence checked via Brooks–Gelman–Rubin diagnostics and trace plots. Assumed common heterogeneity across comparisons; global heterogeneity summarized by I². Reported MDs for Y-BOCS change and ORs for drop-out with 95% credible intervals (CrIs). Treatment rankings via SUCRA and cumulative ranking plots. Consistency assessed globally (design-by-treatment interaction test) and locally (node-splitting). - Small-study effects: Comparison-adjusted funnel plots for publication bias. - Subgroup analyses: Mean age, percent female, number of sessions, treatment resistance, baseline severity (Y-BOCS), lateralization (left/right/bilateral), and OCD subtype. - Sensitivity analyses: Excluding studies at high overall risk of bias; excluding studies with long follow-up; excluding studies missing drop-out data.

- Included studies: 22 RCTs; total N=698 (active rTMS n≈365; sham n=333). Mean age 34.1 years; 48.0% women. Mean number of sessions 16.5 (range 10–30). - Pairwise meta-analysis (MD in Y-BOCS change vs sham; 95% CI): • LF-rTMS over DLPFC: MD 6.34 (2.81–9.87), significantly better than sham. • LF-rTMS over SMA: MD 4.33 (0.39–8.27), significantly better than sham. • HF-rTMS over DLPFC: MD 3.77 (1.43–6.11), significantly better than sham. • HF-rTMS over ACC/mPFC: MD 4.25 (1.31–7.18), significantly better than sham. • LF-rTMS over OFC: MD 4.19 (−0.35 to 8.73), not significantly different from sham. • Tolerability: No significant differences in drop-out rates between any active strategy and sham. - Network meta-analysis (MD vs sham; 95% CrI): • LF-rTMS over DLPFC: MD 6.34 (2.12–10.42), significantly better. • LF-rTMS over SMA: MD 4.18 (0.83–7.62), significantly better. • HF-rTMS over DLPFC: MD 3.75 (1.04–6.81), significantly better. • No significant differences among strategies for drop-out rates (tolerability). - Ranking (SUCRA): • Efficacy: LF-DLPFC ranked highest (SUCRA 71.78%), SMA second (54.91%). • Tolerability: LF-OFC most tolerated (SUCRA 60.02%), followed by LF-DLPFC (56.93%). - Heterogeneity and consistency: • Global I²: 73.5% for efficacy (high), 0.0% for tolerability. • High heterogeneity noted for HF-DLPFC vs sham (66.4%), LF-SMA vs sham (84.1%), and HF-DLPFC vs sham (63.0%). • No significant global inconsistency (efficacy P=0.442; tolerability P=0.987). Node-splitting showed no local inconsistency for LF-DLPFC vs HF-DLPFC (efficacy P=0.396; tolerability P=0.819). - Small-study effects: Funnel plots suggested possible asymmetry for efficacy (notably HF-DLPFC vs sham); no indication of publication bias for tolerability. - Subgroups: LF-DLPFC superior to sham particularly in patients <35 years, with lower proportion of females, in non–treatment-resistant OCD, and when targeting the right hemisphere, potentially explaining heterogeneity. - Evidence quality: GRADE rated efficacy evidence as very low; tolerability comparisons showed no differences.

The analyses address the central question of which rTMS parameters (target and frequency) are most effective and well tolerated for adult OCD. Both pairwise and network meta-analyses indicate that LF-rTMS over the DLPFC shows the most consistent and highest magnitude improvement in Y-BOCS scores compared with sham, with HF-DLPFC and LF-SMA also demonstrating benefit. Tolerability appears comparable across all active strategies and sham, suggesting rTMS is generally well tolerated. The ranking results reinforce LF-DLPFC as the leading approach for efficacy. However, substantial heterogeneity, potential small-study effects, and very low GRADE certainty temper confidence. Subgroup findings imply that patient characteristics (age, sex distribution, treatment resistance) and lateralization may moderate response, underscoring the need for personalized parameter selection and more homogeneous study designs. Overall, the findings support rTMS, particularly LF-DLPFC, as a potentially beneficial adjunctive therapy for OCD, while highlighting the need for higher-quality evidence to establish definitive recommendations.

Among rTMS strategies for adult OCD, low-frequency stimulation over the DLPFC appears to offer the greatest efficacy with tolerability comparable to sham, with HF-DLPFC and LF-SMA also showing benefit. Treatment rankings favor LF-DLPFC, but the certainty of evidence is very low due to heterogeneity and possible publication bias. Future research should prioritize large, high-quality, multicenter RCTs with rigorous blinding, standardized protocols (targets, frequencies, dosing, session number), stratification by clinical subgroups (age, sex, treatment resistance, lateralization), and direct head-to-head comparisons of leading strategies (e.g., LF-DLPFC vs HF-DLPFC vs SMA vs ACC/mPFC) to refine parameter selection and confirm efficacy and tolerability.

- High heterogeneity in efficacy outcomes (global I² 73.5%) and substantial heterogeneity in key comparisons (e.g., LF-SMA vs sham 84.1%, HF-DLPFC vs sham ~63–66%). - Very low GRADE certainty for efficacy estimates, limiting confidence in effect sizes and rankings. - Possible small-study effects/publication bias suggested for efficacy (funnel plot asymmetry), particularly for HF-DLPFC. - Majority of studies had unclear risk of bias (72.7%), with only 22.7% low risk and 4.6% high risk, raising concerns about selection and performance biases. - Variability across trials in patient characteristics (age, sex distribution, treatment resistance), stimulation parameters (targets, frequencies, lateralization), session numbers (10–30), and follow-up durations, potentially contributing to heterogeneity. - Some subgroup effects (e.g., benefit of LF-DLPFC limited to <35 years, lower female proportion, non–treatment-resistant, right-sided targeting) limit generalizability across all patient populations. - Incomplete reporting in some trials (e.g., drop-out data) required sensitivity analyses and may affect precision.