Repetitive Transcranial Magnetic Stimulation as an Augmentative Strategy for Treatment-resistant Depression

A Meta-analysis of Randomized, Double-blind and Sham-Controlled Study

Bangshan Liu; Yan Zhang; Li Zhang; Lingjiang Li


BMC Psychiatry. 2014;14(342) 

In This Article


Search Strategy

We identified articles for inclusion in this meta-analysis by:

  1. Searching MEDLINE and Cochrane Central Register of Controlled Trials (CENTRAL) from 1 January 1995 to 30 November 2013, using the key words "transcranial magnetic stimulation", "depress*", "augment*", "combin*", "adjunctive", "resistant" and "refractory". We restricted the article type to "randomized controlled trial" and the language to "English".

  2. Searching the references of all the previous relevant meta-analyses[11–26] focused on the efficacy of rTMS for MDD published earlier than 30 November 2013, as well as of all included RCTs.

The search procedures are described in details in Additional file 1

Inclusion Criteria and Exclusion Criteria

Studies included in this meta-analysis should satisfy the following criteria:

  1. Study validity: random allocation; double-blind (i.e. both patients and outcome raters were blind to the allocation); sham-controlled; rTMS was used as augmentation to antidepressants;

  2. Sample characteristics: subjects should be 18–75 years old with a diagnosis of MDD according to DSM-IV or ICD-10, comorbidity of psychotic symptoms was excluded;

  3. Efficacy evaluation and outcome reporting: efficacy should be rated by HAMD (17- or 21-items) or MADRS, and data are reported in a continuous (means and standard deviations (SDs) of pre- and post- treatment HAMD or MADRS scores) or dichotomous (response, remission and dropout rates) form able to be synthesized in this meta-analysis.

  4. Articles should be published in English.

Studies were excluded if they were:

  1. Non-RCT design, such as open trials;

  2. Subjects were limited to a specific type of MDD patients, such as postpartum MDD, old MDD or secondary depression (i.e., vascular depression);

  3. Sample size smaller than 5 in either rTMS or sham group;

Data Extraction

  1. Sample characteristics: mean age, gender, diagnosis criteria and definition of treatment resistance;

  2. rTMS-related: frequency, intensity, location, treatment strategy (number of sessions, duration of each stimulation and duration of each interval) and total pulses;

  3. Drug-related: drug strategy (standardized or non-standardized), washout, types and dosages of each type;

  4. Primary outcome measure: number of responders based on the RCTs' primary efficacy measure (defined as ≥50% reduction in post-treatment on the HAMD or MADRS scores) at the end of blinded treatment;

  5. Secondary outcome measure: number of remitters based on the RCTs' primary efficacy measure (e.g. 17- or 21-item HAMD scores ≤7 or ≤8, respectively, or MADRS scores ≤10[30]) at the end of blinded treatment, or the means and SDs of change of 17 or 21-item HAMD or MADRS scores;

  6. Acceptability of treatment: number of dropouts in rTMS and sham groups in each RCT.

Data Synthesis and Analysis

This meta-analysis was conducted according to the Cochrane Handbook for Systematic Reviews of Intervention,[31] all statistic work were performed by Review Manager 5.2 and Excel 2007.

We used a random-effect model because the effcicacy of rTMS between different RCTs was assumed to be varied considerably. This model endows small-sample studies with higher weight and leads to a relatively conservative result.[32] For dichotomous data, if available, an intention-to treat analysis was selected. In other words, we included all dropouts after randomization, because this is closer to clinical practice. When dropouts were excluded for efficacy assessment in any individual RCT (e.g. subjects who never returned for assessment after randomization), they were considered as non-responders. We calculated the pooled OR of response rate and associated NNT. As reported in other studies, an NNT ≤ 10 was considered as clinically meaningful because such a treatment difference would be regularly encountered in clinical practice.[33] For continuous data, we calculated standardized mean difference (SMD) of the baseline HAMD or MADRS scores and change from baseline of HAMD or MADRS scores after the blinded treatment between active rTMS and sham groups. When a study had more than 2 groups, the data of different active rTMS groups were combined together as one group (the data were combined only when the active groups did not show significant difference, if they did, the RCT were excluded).

Heterogeneity was assessed by chi-square and I-square statistics,[34] which is considered to be an indicator of study heterogeneity when p value for χ2 was lower than 0.1 or when I2 was higher than 35%. Sensitivity and subgroup analysis were conducted to determine the potential factors, such as sessions (≤10 or >10), intensity (≤100% or >100%) or total pulses (≤10000 or >10000), which may influence the antidepressant efficacy of rTMS. Nevertheless, because the number of included RCTs was relatively small, the heterogeneity of studies in subgroup analysis was considerably high, which lowered the reliability of the results. Finally, we used funnel plot and visual inspection to examine the publication bias.