Supplementary materials

Transcription

1 Supplementary materials Kedzior, K.K., Gellersen, H., Brachetti, A., Berlim, M.T., Deep transcranial magnetic stimulation (DTMS) in the treatment of major depression: an exploratory systematic review and meta-analysis. Journal of Affective Disorders (2015), Contents Effect sizes in open-label studies in the current meta-analysis... 2 Computation of Cohen s d for paired means... 2 Computation of Hedges g for paired means... 4 Computation of response, remission, and dropout rates in one group of patients per study... 4 Figure S1. Standardised HDRS change scores, Hedges g (baseline after acute DTMS) in open-label studies... 7 Figure S2. Cumulative analysis: standardised HDRS change scores in open-label studies... 8 Figure S3. One-study removed analysis: standardised HDRS change scores in open-label studies Figure S4. Subgroup analyses: standardised HDRS change scores in open-label studies Figure S5. Funnel plot: standardised HDRS change scores in open-label studies Figure S6. Cumulative analysis: response rates in open-label studies Figure S7. One-study removed analysis: response rates in open-label studies Figure S8. Subgroup analyses: response rates in open-label studies Figure S9. Funnel plot: response rates in open-label studies Figure S10. Cumulative analysis: remission rates in open-label studies Figure S11. One-study removed analysis: remission rates in open-label studies Figure S12. Subgroup analyses: remission rates in open-label studies Figure S13. Funnel plot: remission rates in open-label studies Figure S14. Cumulative analysis: dropout rates in open-label studies Figure S15. One-study removed analysis: dropout rates in open-label studies Figure S16. Subgroup analyses: dropout rates in open-label studies Figure S17. Funnel plot: dropout rates in open-label studies PRISMA checklist (Moher et al., 2009) References

2 Effect sizes in open-label studies in the current meta-analysis There were nine open-label studies (Berlim et al., 2014a; Harel et al., 2014; Harel et al., 2011; Isserles et al., 2011; Levkovitz et al., 2009; Rapinesi et al., 2015a; Rapinesi et al., 2015b; Rosenberg et al., 2010a; Rosenberg et al., 2010b) regarding antidepressant effects of deep transcranial magnetic stimulation (DTMS) included in the quantitative meta-analysis and one additional double-blind, parallel-design, randomised-controlled trial (RCT) with an active DTMS group and an inactive sham DTMS group (Levkovitz et al., 2015) included in the qualitative synthesis. The mathematical approach to meta-analysis in this document is based on Borenstein et al. (Borenstein et al., 2009) and is utilised by the Comprehensive Meta-Analysis (CMA) software version 2.2 (CMA; Biostat, USA). Computation of Cohen s d for paired means Depression severity was measured using the Hamilton Depression Rating Scale (HDRS) (Hamilton, 1960) in all studies used in the current meta-analysis. The HDRS scores were collected at baseline and after DTMS from the same patients. Thus, the following formulae were utilised to compute a standardised difference in paired means (Cohen s d) in each study based on the following approach (Borenstein et al., 2009): d M diff pre post = =, S within M S M within where M pre is the mean HDRS score at baseline and M post is the mean HDRS score after DTMS treatment. The overall standard deviation (S within ) of the difference score (S diff ) was computed as follows (Borenstein et al., 2009): S within = S diff 2(1 r) diff 2 pre 2 post S = S + S 2rS S, pre post 2

3 where S pre is the standard deviation of HDRS scores at baseline, S post is the standard deviation of HDRS scores after DTMS and r is the Pearson correlation coefficient between HRDS scores at baseline and after DTMS. Since r was not reported in individual studies, we have used r=.5 as the most objective value for estimating the variance of d and, thus, the weight in each study (r close to 1.0 assumes that HDRS scores at both points in time are nearly perfectly correlated and this approach underestimates the variance of d, V d ; r close to 0.0 assumes that HDRS scores at both points in time are nearly independent-not correlated- and this approach overestimates V d ). V d was estimated using the following formula (Borenstein et al., 2009): V d 2 1 d = + 2(1 r), N 2N where d is Cohen s d, N is the sample size in the study (baseline N was used in case any patients dropped out from the study), and r is the Pearson correlation coefficient between HRDS scores at baseline and after DTMS. Some studies reported a raw mean difference score (M diff ) between the mean HDRS scores at baseline (M pre ) and after DTMS treatment (M post ) in all patients and its standard error of the mean (SE diff ) according to the following formulae: M diff = M pre M post These values were converted into Cohen s d and V d using the following formulae (Borenstein et al., 2011): d = M S diff diff = SE M diff diff N V d = 1 N 2 d + 2(1 r) 2N 3

4 As explained above, we have used r=.5 as the most objective value for estimating the variance of d. The formulae for the random-effects meta-analysis of Cohen s d are shown elsewhere (Borenstein et al., 2009). Computation of Hedges g for paired means Although commonly used, one limitation of Cohen s d is that it could be inflated in studies with small sample sizes (Borenstein et al., 2009), such as the primary studies used in the current meta-analysis. Correcting Cohen s d with a so-called J correction-factor produces another effect size, Hedges g, which is assumed to be an unbiased version of Cohen s d. Hedges g and its variance (V g ) were computed in addition to Cohen s d in the current metaanalysis according to the following formulae (Borenstein et al., 2009): g = d J 3 = d 1 4df 1 V = J 2, g V d where d is Cohen s d, J is the correction factor, and df are degrees of freedom for paired samples (sample size N-1). Computation of response, remission, and dropout rates in one group of patients per study Nominal (binary) outcomes (such as presence or absence of response or remission after treatment or dropout rates) can be analysed using the odds ratio if there are two treatment groups (such as an active DTMS group and an inactive sham group) in a study. The nominal outcome (response, remission, or dropout rate) from one group only (all patients receiving open-label DTMS) can be assessed using the response, remission, or dropout rates (Borenstein et al., 2009) referred to as event rates by the CMA software. The software applies the following approach using the symbols shown in the table below (Borenstein et al., 2009): 4

5 Response or remission No response or no remission Depression patients A (no. of patients) B (no. of patients) 1. Event rate (RR) is computed as follows based on the symbols used in the table above: RR = A A + B 2. Event rate is expressed as odds of the event as follows: RR odds = 1 RR 3. To maintain symmetry in the analysis, the odds of event are converted into a logarithmic scale using the natural logarithm (LN odds). The variance (V LNodds ), standard error (SE LNodds ), and the 95% confidence interval (95%CI) of LN odds (lower 95%CI LNodds and upper 95%CI LNodds ) are computed as follows: V 1 1 = + LNodds A B SE LNodds = V LNodds lowerci LNodds = LNodds ( 1.96 SELNodds ) upperci LNodds = LNodds + ( 1.96 SELNodds ) 4. All computed values are antilogged back to odds and 95%CI of odds. 5. Odds and 95%CI of odds are converted back to event rate and 95%CI of event rate (and subsequently shown on the forest plot for each study) as follows: odds RR = 1+ odds CI RR CI = 1+ CI odds odds 5

6 Random-effects meta-analysis of event rates is computed on natural logarithms of event rates in each study and the final result (overall mean weighted event rate) is antilogged and shown as an absolute event rate per study and for all studies on the forest plots. The formulae for the random-effects meta-analysis of event rates are shown elsewhere (Borenstein et al., 2009). 6

7 Figure S1. Standardised HDRS change scores, Hedges g (baseline after acute DTMS) in open-label studies Statistics for each study Hedges's g and 95% CI Hedges's Lower Upper g limit limit Total Levkovitz et al Rosenberg et al. 2010a Rosenberg et al. 2010b Harel et al Isserles et al Berlim et al. 2014a Harel et al Rapinesi et al. 2015a Rapinesi et al. 2015b DTMS ineffective DTMS effective Note. Forest plot of a random-effects meta-analysis with Hedges g as effect size (standardised paired difference in means corrected for sample size in each study). A large antidepressant effect of DTMS occurred in nine open-label studies with 150 patients (overall mean weighted g=1.91, 95%CI: ). The individual study effect sizes based on HDRS change scores varied from 1.25 to 5.12 suggesting that DTMS had a large antidepressant effect compared to baseline in nine studies in our analysis. This moderate variability in effect sizes (Q(df 8)=24.67, p two-tailed =.002; I 2 =68%) was attributable to one study (Rapinesi et al., 2015a) and was reduced to 0% once the latter was removed from the analysis (the effect sizes in remaining eight studies varied from 1.25 to 2.46). Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation; HDRS, Hamilton Depression Rating Scale. 7

8 Figure S2. Cumulative analysis: standardised HDRS change scores in open-label studies A. Effect size: Cohen s d Lower Upper Point limit limit Total Levkovitz et al Rosenberg et al. 2010a Rosenberg et al. 2010b Harel et al Isserles et al Berlim et al. 2014a Harel et al Rapinesi et al. 2015a Rapinesi et al. 2015b Cumulative std paired difference (95% CI) DTMS ineffective DTMS effective B. Effect size: Hedges g Lower Upper Point limit limit Total Levkovitz et al Rosenberg et al. 2010a Rosenberg et al. 2010b Harel et al Isserles et al Berlim et al. 2014a Harel et al Rapinesi et al. 2015a Rapinesi et al. 2015b Cumulative hedges's g (95% CI) DTMS ineffective DTMS effective 8

9 Note. Forest plots of a random-effects, cumulative meta-analysis. Point refers to the overall mean weighted effect size of the study in each row and all prior studies. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation; HDRS, Hamilton Depression Rating Scale; Std paired difference, standardised paired difference in means (Cohen s d; effect size). 9

10 Figure S3. One-study removed analysis: standardised HDRS change scores in openlabel studies A. Effect size: Cohen s d Std Paired Difference (95% Lower Upper CI) with study removed Point limit limit Levkovitz et al Rosenberg et al. 2010a Rosenberg et al. 2010b Harel et al Isserles et al Berlim et al. 2014a Harel et al Rapinesi et al. 2015a Rapinesi et al. 2015b DTMS ineffective DTMS effective 10

11 B. Effect size: Hedges g Lower Upper Point limit limit Levkovitz et al Rosenberg et al. 2010a Rosenberg et al. 2010b Harel et al Isserles et al Berlim et al. 2014a Harel et al Rapinesi et al. 2015a Rapinesi et al. 2015b Hedges's g (95% CI) with study removed DTMS ineffective DTMS effective Note. Forest plots of a random-effects, one-study removed meta-analysis. Point refers to the overall mean weighted effect size of all studies except the study in each row. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation; HDRS, Hamilton Depression Rating Scale; Std paired difference, standardised paired difference in means (Cohen s d; effect size). 11

12 Figure S4. Subgroup analyses: standardised HDRS change scores in open-label studies A. All patients on concurrent antidepressants (yes vs. no) Group by Antidepressants Statistics for each study Std Paired Difference and 95% CI Std Paired Lower Upper Difference limit limit Total no Levkovitz et al no Rosenberg et al. 2010a no yes Harel et al yes Isserles et al yes Berlimet al. 2014a yes Rapinesi et al. 2015b yes DTMS ineffective DTMS effective B. All patients bipolar (yes vs. no) Group by Bipolar Statistics for each study Std Paired Difference and 95% CI Std Paired Lower Upper Difference limit limit Total no Levkovitz et al no Isserles et al no Berlim et al. 2014a no Harel et al no Rapinesi et al. 2015b no yes Harel et al yes DTMS ineffective DTMS effective 12

13 C. Stimuli per session: 1680, 1980, or 3000 Group by Stimuli per session Statistics for each study Std Paired Difference and 95% CI Std Paired Lower Upper Difference limit limit Total Levkovitz et al Rosenberg et al. 2010a Rosenberg et al. 2010b Harel et al Isserles et al Harel et al Rapinesi et al. 2015a Rapinesi et al. 2015b Berlimet al. 2014a DTMS ineffective DTMS effective Note. Forest plots of random-effects subgroup meta-analyses. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation; HDRS, Hamilton Depression Rating Scale; Std paired difference, standardised paired difference in means (Cohen s d; effect size). 13

14 Figure S5. Funnel plot: standardised HDRS change scores in open-label studies A. Effect size: Cohen s d B. Effect size: Hedges g Note. Random-effects funnel plots of standardised effect sizes (X-axes) and their estimated precision (Y-axes; 1/standard error of the mean) in each study. There are two studies with higher than average standardised HDRS change scores theoretically missing from the current analysis. If the two missing studies were included in the current meta-analysis then the antidepressant effect of DTMS would have increased compared to the observed effect 14

15 according to the trim-and-fill analysis: the overall mean weighted d would have increased from 2.04 to 2.26 (95%CI: ), and the overall mean weighted g would have increased from 1.91 to 2.12 (95%CI: ). Therefore, there is little evidence for publication bias in this analysis. Abbreviations: HDRS, Hamilton Depression Rating Scale; Std Err, standard error of the mean; Std paired difference, standardised paired difference in means (Cohen s d; effect size). 15

16 Figure S6. Cumulative analysis: response rates in open-label studies Lower Upper Point limit limit Total Levkovitz et al / 19 Rosenberg et al. 2010a / 26 Rosenberg et al. 2010b / 32 Harel et al / 51 Isserles et al / 71 Berlim et al. 2014a / 88 Harel et al / 114 Rapinesi et al. 2015a / 138 Rapinesi et al. 2015b / Cumulative event rate (95% CI) DTMS ineffective DTMS effective Note. Point refers to the overall mean weighted effect size (response rate referred to as event rate) of the study in each row and all prior studies. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation. 16

17 Figure S7. One-study removed analysis: response rates in open-label studies Lower Upper Point limit limit Levkovitz et al Rosenberg et al. 2010a Rosenberg et al. 2010b Harel et al Isserles et al Berlim et al. 2014a Harel et al Rapinesi et al. 2015a Rapinesi et al. 2015b Event rate (95% CI) with study removed DTMS ineffective DTMS effective Note. Point refers to the overall mean weighted effect size (response rate referred to as event rate) of all studies except the study in each row. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation. 17

18 Figure S8. Subgroup analyses: response rates in open-label studies A. All patients on concurrent antidepressants (yes vs. no) Group by Antidepressants Event Lower Upper rate limit limit Total no Levkovitz et al / 19 no Rosenberg et al. 2010a / 7 no yes Harel et al / 19 yes Isserles et al / 20 yes Berlim et al. 2014a / 17 yes Rapinesi et al. 2015b / 12 yes Event rate and 95% CI DTMS ineffective DTMS effective B. All patients bipolar (yes vs. no) Group by Bipolar Event Lower Upper rate limit limit Total no Levkovitz et al / 19 no Isserles et al / 20 no Berlim et al. 2014a / 17 no Harel et al / 26 no Rapinesi et al. 2015b / 12 no yes Harel et al / 19 yes Event rate and 95%CI DTMS ineffective DTMS effective 18

19 C. Stimuli per session: 1680, 1980, or 3000 Group by Stimuli Event Lower Upper rate limit limit Total Levkovitz et al / Rosenberg et al. 2010a / Rosenberg et al. 2010b / Harel et al / Isserles et al / Harel et al / Rapinesi et al. 2015a / Rapinesi et al. 2015b / Berlim et al. 2014a / Event rate and 95%CI DTMS ineffective DTMS effective Note. Forest plots of random-effects subgroup meta-analyses. Response rate is referred to as event rate. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation. 19

20 Figure S9. Funnel plot: response rates in open-label studies Note. There are no studies theoretically missing from the current analysis. Therefore, there is little evidence for publication bias in this analysis. Abbreviation: Std Err, standard error of the mean. 20

21 Figure S10. Cumulative analysis: remission rates in open-label studies Lower Upper Point limit limit Total Levkovitz et al / 19 Rosenberg et al. 2010a / 26 Rosenberg et al. 2010b / 32 Harel et al / 51 Isserles et al / 71 Berlim et al. 2014a / 88 Rapinesi et al. 2015a / 112 Rapinesi et al. 2015b / Cumulative event rate (95% CI) DTMS ineffective DTMS effective Note. Point refers to the overall mean weighted effect size (remission rate referred to as event rate) of the study in each row and all prior studies. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation. 21

22 Figure S11. One-study removed analysis: remission rates in open-label studies Lower Upper Point limit limit Levkovitz et al Rosenberg et al. 2010a Rosenberg et al. 2010b Harel et al Isserles et al Berlim et al. 2014a Rapinesi et al. 2015a Rapinesi et al. 2015b Event rate (95% CI) with study removed DTMS ineffective DTMS effective Note. Point refers to the overall mean weighted effect size (remission rate referred to as event rate) of all studies except the study in each row. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation. 22

23 Figure S12. Subgroup analyses: remission rates in open-label studies A. All patients on concurrent antidepressants (yes vs. no) Group by Antidepressants Event Lower Upper rate limit limit Total no Levkovitz et al / 19 no Rosenberg et al. 2010a / 7 no yes Harel et al / 19 yes Isserles et al / 20 yes Berlim et al. 2014a / 17 yes Rapinesi et al. 2015b / 12 yes Event rate and 95% CI DTMS ineffective DTMS effective B. All patients bipolar (yes vs. no) Group by Bipolar Event Lower Upper rate limit limit Total no Levkovitz et al / 19 no Isserles et al / 20 no Berlim et al. 2014a / 17 no Rapinesi et al. 2015b / 12 no yes Harel et al / 19 yes Event rate and 95%CI DTMS ineffective DTMS effective 23

24 C. Stimuli per session: 1680, 1980, or 3000 Group by Stimuli Event Lower Upper rate limit limit Total Levkovitz et al / Rosenberg et al. 2010a / Rosenberg et al. 2010b / Harel et al / Isserles et al / Rapinesi et al. 2015a / Rapinesi et al. 2015b / Berlim et al. 2014a / Event rate and 95%CI DTMS ineffective DTMS effective Note. Forest plot of random-effects subgroup meta-analyses. Remission rate is referred to as event rate. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation. 24

25 Figure S13. Funnel plot: remission rates in open-label studies Note. There are three studies with higher than average remission rates theoretically missing from the current analysis. If the three missing studies were included in the current metaanalysis then the remission rates would have increased compared to the observed remission rates according to the trim-and-fill analysis: the overall mean weighted remission rate would have increased from 29% to 39% (95%CI: 23%-57%). Therefore, there is little evidence for publication bias in this analysis. Abbreviation: Std Err, standard error of the mean. 25

26 Figure S14. Cumulative analysis: dropout rates in open-label studies Lower Upper Point limit limit Total Levkovitz et al / 23 Rosenberg et al. 2010a / 30 Rosenberg et al. 2010b / 36 Harel et al / 55 Isserles et al / 80 Berlim et al. 2014a / 97 Harel et al / 126 Rapinesi et al. 2015a / 150 Rapinesi et al. 2015b / Cumulative event rate (95% CI) DTMS ineffective DTMS effective Note. Point refers to the overall mean weighted effect size (dropout rate referred to as event rate) of the study in each row and all prior studies. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation. 26

27 Figure S15. One-study removed analysis: dropout rates in open-label studies Lower Upper Point limit limit Levkovitz et al Rosenberg et al. 2010a Rosenberg et al. 2010b Harel et al Isserles et al Berlim et al. 2014a Harel et al Rapinesi et al. 2015a Rapinesi et al. 2015b Event rate (95% CI) with study removed DTMS ineffective DTMS effective Note. Point refers to the overall mean weighted effect size (dropout rate referred to as event rate) of all studies except the study in each row. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation. 27

28 Figure S16. Subgroup analyses: dropout rates in open-label studies A. All patients on concurrent antidepressants (yes vs. no) Group by Antidepressants Event Lower Upper rate limit limit Total no Levkovitz et al / 23 no Rosenberg et al. 2010a / 7 no yes Harel et al / 19 yes Isserles et al / 25 yes Berlim et al. 2014a / 17 yes Rapinesi et al. 2015b / 12 yes Event rate and 95% CI DTMS ineffective DTMS effective B. All patients bipolar (yes vs. no) Group by Bipolar Event Lower Upper rate limit limit Total no Levkovitz et al / 23 no Isserles et al / 25 no Berlim et al. 2014a / 17 no Harel et al / 29 no Rapinesi et al. 2015b / 12 no yes Harel et al / 19 yes Event rate and 95%CI DTMS ineffective DTMS effective 28

29 C. Stimuli per session: 1680, 1980, or 3000 Group by Stimuli Event Lower Upper rate limit limit Total Levkovitz et al / Rosenberg et al. 2010a / Rosenberg et al. 2010b / Harel et al / Isserles et al / Harel et al / Rapinesi et al. 2015a / Rapinesi et al. 2015b / Berlim et al. 2014a / Event rate and 95%CI DTMS ineffective DTMS effective Note. Forest plot of random-effects subgroup meta-analyses. Dropout rate is referred to as event rate. Abbreviations: CI, confidence interval; DTMS, deep transcranial magnetic stimulation. 29

30 Figure S17. Funnel plot: dropout rates in open-label studies Note. There are two studies with higher than average dropout rates theoretically missing from the current analysis. If the two missing studies were included in the current meta-analysis then the dropout rates would have increased compared to the observed dropout rates according to the trim-and-fill analysis: the overall mean weighted dropout rate would have increased from 18% to 23% (95%CI: 12%-38%). Therefore, publication bias could not be ruled out in this analysis. Abbreviation: Std Err, standard error of the mean. 30

31 PRISMA checklist (Moher et al., 2009) Section/topic # Checklist item Reported on page # TITLE Title 1 Identify the report as a systematic review, meta-analysis, or both. 1 ABSTRACT Structured summary 2 Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, participants, and interventions; study appraisal and synthesis methods; results; limitations; conclusions and implications of key findings; systematic review registration number. INTRODUCTION Rationale 3 Describe the rationale for the review in the context of what is already known. 2 Objectives 4 Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design (PICOS). 1 2 METHODS Protocol and registration 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number. Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale. Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched. Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated. Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis). Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators. - 2 Table 1 Table 1 2; Figure

32 Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made. Risk of bias in individual studies 12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis. Summary measures 13 State the principal summary measures (e.g., risk ratio, difference in means). 3 Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I 2 ) for each meta-analysis. 3; Tables 2-4 3; 5-6; 10 3; 5 Section/topic # Checklist item Reported on page # Risk of bias across studies 15 Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting within studies). Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified. 3; 5-6; 10 6 RESULTS Study selection 17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram. Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations. Figure 1 Tables 2-4 Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12). Figures 2, S1-S17 Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot. Figure 2, Table 6 Synthesis of results 21 Present results of each meta-analysis done, including confidence intervals and measures of consistency. 6-8; Table 6 Risk of bias across studies 22 Present results of any assessment of risk of bias across studies (see Item 15). 6-8; Figures S1-S17 Additional analysis 23 Give results of additional analyses, if done (e.g., sensitivity or subgroup analyses, meta-regression [see Item 16]). DISCUSSION Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers). 6-8; Figures S1-S

33 Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias). Conclusions 26 Provide a general interpretation of the results in the context of other evidence, and implications for future research. FUNDING Funding 27 Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the systematic review

34 References Berlim, M.T., Van den Eynde, F., Tovar-Perdomo, S., Chachamovich, E., Zangen, A., Turecki, G., 2014a. Augmenting antidepressants with deep transcranial magnetic stimulation (DTMS) in treatment-resistant major depression. World J. Biol. Psychiatry 15, Borenstein, M., Hedges, L., Higgins, J., Rothstein, H., Introduction to meta-analysis. Wiley, Great Britain. Borenstein, M., Hedges, L., Higgins, J., Rothstein, H., Comprehensive Meta Analysis (CMA) Software 2.0 User Manual. Biostat, USA. Hamilton, M., A rating scale for depression. J. Neurol. Neurosurg. Psychiatry 23, Harel, E.V., Rabany, L., Deutsch, L., Bloch, Y., Zangen, A., Levkovitz, Y., H-coil repetitive transcranial magnetic stimulation for treatment resistant major depressive disorder: An 18-week continuation safety and feasibility study. World J. Biol. Psychiatry 15, Harel, E.V., Zangen, A., Roth, Y., Reti, I., Braw, Y., Levkovitz, Y., H-Coil repetitive transcranial magnetic stimulation for the treatment of bipolar depression: An add-on, safety and feasibility study. World J. Biol. Psychiatry 12, Isserles, M., Rosenberg, O., Dannon, P., Levkovitz, Y., Kotler, M., Deutsch, F., Lerer, B., Zangen, A., Cognitive-emotional reactivation during deep transcranial magnetic stimulation over the prefrontal cortex of depressive patients affects antidepressant outcome. J. Affect. Disord. 128, Levkovitz, Y., Harel, E.V., Roth, Y., Braw, Y., Most, D., Katz, L.N., Sheer, A., Gersner, R., Zangen, A., Deep transcranial magnetic stimulation over the prefrontal cortex: 34

35 evaluation of antidepressant and cognitive effects in depressive patients. Brain Stimul. 2, Levkovitz, Y., Isserles, M., Padberg, F., Lisanby, S.H., Bystritsky, A., Xia, G., Tendler, A., Daskalakis, Z.J., Winston, J.L., Dannon, P., Hafez, H.M., Reti, I.M., Morales, O.G., Schlaepfer, T.E., Hollander, E., Berman, J.A., Husain, M.M., Sofer, U., Stein, A., Adler, S., Deutsch, L., Deutsch, F., Roth, Y., George, M.S., Zangen, A., Efficacy and safety of deep transcranial magnetic stimulation for major depression: a prospective multicenter randomized controlled trial. World Psychiatry 14, Moher, D., Liberati, A., Tetzlaff, J., Altman, D., Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. B. M. J. 339, b2535. Rapinesi, C., Bersani, F., Kotzalidis, G., Imperatori, C., Del Casale, A., Di Pietro, S., Ferri, V., Serata, D., Raccah, R., Zangen, A., Angeletti, G., Girardi, P., 2015a. Maintenance deep transcranial magnetic stimulation sessions is associated with reduced depressive relapses in patients with unipolar or bipolar depression. Front. Neurol. 6, 16. Rapinesi, C., Curto, M., Kotzalidis, G.D., Del Casale, A., Serata, D., Ferri, V.R., Di Pietro, S., Scatena, P., Bersani, F.S., Raccah, R.N., Digiacomantonio, V., Ferracuti, S., Bersani, G., Zangen, A., Angeletti, G., Girardi, P., 2015b. Antidepressant effectiveness of deep Transcranial Magnetic Stimulation (dtms) in patients with Major Depressive Disorder (MDD) with or without Alcohol Use Disorders (AUDs): A 6-month, open label, followup study. J. Affect. Disord. 174, Rosenberg, O., Shoenfeld, N., Zangen, A., Kotler, M., Dannon, P.N., 2010a. Deep TMS in a resistant major depressive disorder: a brief report. Depress. Anxiety 27, Rosenberg, O., Zangen, A., Stryjer, R., Kotler, M., Dannon, P.N., 2010b. Response to deep TMS in depressive patients with previous electroconvulsive treatment. Brain Stimul. 3,