RÉSUMÉ
OBJECTIVES: We re-analysed data from published meta-analyses testing the effects of Transcranial Magnetic Stimulation (TMS) on Major Depressive Disorder (MDD) in adults. We applied up-to-date meta-analytic techniques for handling heterogeneity including the random-effects Hartung-Knapp-Sidik-Jonkman method and estimated 95% prediction intervals. Heterogeneity practices in published meta-analyses were assessed as a secondary aim. STUDY DESIGN AND SETTING: We performed systematic searches of systematic reviews with meta-analyses that included randomised controlled trials assessing the efficacy, tolerability, and side effects of TMS on MDD. We performed risk of bias assessment using A MeaSurement Tool to Assess Reviews (AMSTAR) 2 and re-analysed meta-analyses involving 10 or more primary studies. RESULTS: We included 29 systematic reviews and re-analysed 15 meta-analyses. Authors of all meta-analyses interpreted findings to suggest TMS is safe and effective for MDD. Our re-analysis showed that in 14 out of 15 meta-analyses, the 95% prediction intervals included the null and captured values in the opposite effect direction. We also detected presence of small-study effects in some meta-analyses and 24 out of 25 systematic reviews received an AMSTAR 2 rating classed as critically low. CONCLUSION: Authors of all included meta-analyses interpreted findings to suggest TMS is safe and effective for MDD despite lack of comprehensive investigation of heterogeneity. Our re-analysis revealed the direction and magnitude of treatment effects vary widely across different settings. We also found high risk of bias in the majority of included systematic reviews and presence of small-study effects in some meta-analyses. Because of these reasons, we argue TMS for MDD may not be as effective and potentially less tolerated in some populations than current evidence suggests.