Literature Search and Screening
All the 7 RCTs[27,30.35–39] included in this meta-analysis were identified by electronic database searching, and the hand searching of bibliographies of previous meta-analyses did not result in additional studies available for data synthesis; two RCTs[40,41] on rTMS's augmentative effect for TRD obtained by hand searching were excluded, because the data described in these two reports are unable to synthesize in this study. The process of literature search and screening was shown in Figure 1 and a detailed description of the process was available in Additional file 1 https://www.biomedcentral.com/1471-244X/14/342/additional.
Included Studies: Main Characteristics
All included RCTs[27,30,35–39] used HAMD as a primary outcome measurement. They all mentioned randomization, while no one described the scheme of allocation concealment in details. Three studies[30,38,39] reported how they guaranteed the blindness of the patients and the efficacy raters, while the other 4 RCTs[27,36,37] just mentioned "the patients and raters did not know which group they were allocated". The risk of bias table was shown in Figure 2. All the studies had applied high frequency (≥5Hz) rTMS on the LDLPFC except for the study of Garcia-toro et al., which simultaneously applied high frequency rTMS on the LDLPFC and low frequency rTMS on the RDLPFC. The main characteristics of included RCTs were described in Table 1.
Risk of bias graph: review authors' judgments about each risk of bias item presented as percentages across all included studies.
Six RCTs had reported qualified data about response rates. As a whole, 68/146 (46.6%) and 15/84 (22.1%) subjects in the active or sham rTMS groups were classified as responders, respectively. The pooled OR was 5.12 (95% CI 2.11–12.45, z = 3.60, p = 0.0003), implying a significant difference favoring the active rTMS group (Figure 3). The risk difference translated into NNT was 3.4, namely, one patient would get clinical response in every 3.4 patients being treated.
Meta-analysis of active rTMS vs sham condition used as an augmentative strategy for antidepressants in treatment-resistant depression: response rates.
Heterogeneity between RCTs did not exceed that expected by chance (χ2 = 6.09, p = 0.30, I2 = 18%), meaning that the variance among the effect sizes was no greater than that expected by sampling error. The associated Funnel Plot (Figure 4) is roughly symmetric, please refer to the Additional file 1 https://www.biomedcentral.com/1471-244X/14/342/additional.
Change From Baseline of HAMD Scores
Data relating to change from baseline of HAMD scores were available from 6 RCTs, with a total of 126 and 87 subjects in the active rTMS and sham group, respectively. The pooled SMD was 0.86 (95% CI 0.57–1.15, z = 5.75, p < 0.00001), indicating the superiority of active rTMS in alleviating depression severity compared with sham rTMS (Figure 5).
Meta-analysis of active rTMS versus sham condition used as an augmentative strategy for antidepressants in treatment-resistant depression: change from baseline in HAMD scores.
Heterogeneity between RCTs did not exceed that expected by chance (χ2 = 2.67, p = 0.75, I2 = 0%), indicating that the data were reasonably appropriate for synthesis. The associated Funnel Plot was approximately symmetrical. For the associated funnel plot, please refer to the Additional file 1 https://www.biomedcentral.com/1471-244X/14/342/additional.
Acceptability of Treatment
All the 7 studies included in this meta-analysis reported the data of dropouts. In total, 9/171 (5.7%) and 5/108 (5.04%) dropped after the blinded treatment in the active rTMS and sham groups, respectively. The dropout rates in both groups were relatively low and had no significant difference [risk difference (RD) = 0.01, 95% CI −0.04–0.07, z = 0.48, p = 0.63] (because there were no dropouts in two RCTs,[30,37] to enroll the data of the two studies, we chose RD rather than OR), indicating the relatively low adverse effect and high acceptability of rTMS. Additionally, the side effects reported by patients were presented in Table 1. The most frequent reported side effects were mild headache and discomfort in stimulation location, the headache and discomfort were commonly transient and did not differ significantly between groups, implying the safety of rTMS. For the associated forest plot and funnel plot, please refer to the Additional file 1 https://www.biomedcentral.com/1471-244X/14/342/additional.
Only 2 RCTs[30,38] reported the number of remitters at the end of blinded rTMS treatment, both of them found a significant difference between the active and sham groups in achieving remission. For the small number or RCTs, data were not synthesized. For detailed description, please refer to the Additional file 1 https://www.biomedcentral.com/1471-244X/14/342/additional.
rTMS vs Sham Group: Baseline Depression Severity
No significant difference were observed in baseline severity between active and sham rTMS groups, the pooled SMD was −0.09 [95% CI −0.34–0.17, z = 0.66, p = 0.51], indicating that the two groups are comparable at baseline and baseline depression severity cannot be a confounding factor of the efficacy. For the associated forest plot and funnel plot, please refer to the Additional file 1 https://www.biomedcentral.com/1471-244X/14/342/additional.
Sensitivity and Subgroup Analyses
We tried to carry out sensitivity and subgroup analyses to explore the potential confounding factors, such as number of sessions, intensity and total pulses, and drug strategy (standardized or non-standardized). However, because the number of included studies is relatively low and the heterogeneity between RCTs in subgroups were very high, the preplanned subgroup analyses were finally abandoned to avoid the misleading results. For detailed description, please refer to the Additional file 1 https://www.biomedcentral.com/1471-244X/14/342/additional.
Two studies[36,37] implemented open intervention after the blinded treatment. In these follow-up studies, patients in sham groups who did not achieve response in blinded treatment were enrolled to receive active rTMS treatment, both studies reported additional reduction of HAMD scores reduction after the "subsequent rTMS treatment". The other three studies[27,35,38] reported non-interventional observation of the follow-up efficacy of rTMS., Interestingly, in the study of Chen et al., augmentative active rTMS didn't achieve significant difference compared to sham condition, while after 1 month's follow-up, the active group showed significant greater HAMD scores reduction than the sham group, indicating a lagging efficacy of rTMS treatment, while in the study of Bretlau et al., the results is inverse: rTMS group achieved significant greater HAMD scores reduction after treatment than the sham group, while after 9 weeks of follow-up, the difference is no longer significant between the groups. And the study of Rossini et al. demonstrated a figure that shows active rTMS group achieved greater HAMD scores reduction both after treatment and in the follow-up duration than sham group, and the follow-up data in this study is not available.
BMC Psychiatry. 2014;14(342) © 2014 BioMed Central, Ltd.
© 1999-2006 BioMed Central Ltd