Study subgroups^{a}

Nstudies (%)^{b}

d (95% CI)

p
_{
twotailed
}

Metaregression predictors

Nstudies^{b}

Mode

Range

Regression line slopep
_{
twotailed
}


Treatment resistance

13
  
Patient characteristics
    
YES (all failed ≥2 AD trials)

10 (77%)

.56 (−.81, −.32)

<.001*

Female patients (%)

32

47, 50

2292

.003*

NO (all failed ≥1 AD trials)

3 (23%)

.58 (−.90, −.26)

<.001*

Mean age (years)

32

43, 44

3962

.194

Antidepressants

32
  
Bipolar patients (%)

24

0

0100

.211

YES (any % of patients)

25 (78%)

.51 (−.68, −.35)

<.001*

Psychotic patients (%)

14

0

08

.789

Stable dose

20 (80%)

.54 (−.74, −.34)

<.001*

rTMS parameters
    
Started with rTMS

5 (20%)

.43 (−.75, −.12)

.007*

Frequency (Hz)

32

10

520

.824

NO (all patients)

7 (22%)

.33 (−.53, −.14)

.001*

Motor threshold (%)

32

80, 90, 110

80120

.984

Bipolar depression

24
  
Total sessions

32

10

520

.813

YES (any % of patients)

16 (67%)

.39 (−.62, −.16)

.001*

Stimuli/session

20

1600

2503000

.021*^{c}

NO (all patients)

8 (33%)

.45 (−.62, −.28)

<.001*

Total stimuli/study

20

16000

125060000

.124

Psychotic depression

14
  
Trains/session

28

20

5100

.217

YES (any % of patients)

3 (21%)

.73 (−1.78, .33)

.177

Intertrain interval (s)

31

30

860

.680

NO (all patients)

11 (79%)

.59 (−.81, −.36)

<.001*
     
Coiltype

30
       
Figureofeight

28 (93%)

.50 (−.66, −.34)

<.001*
     
Circular

2 (7%)

.59 (−1.59, .42)

.254
     
Coil angle sham

32
       
0° (inactive coil)

2 (6%)

.33 (−.85, .19)

.217
     
0° (sham coil)

5 (16%)

.63 (−.98, −.28)

<.001*
     
45°

12 (38%)

.27 (−.50, −.04)

.022*
     
90°

13 (41%)

.61 (−.82, −.40)

<.001*
     
 Notes: ^{a}There were no statistically significant differences in effect sizes between subgroups. Subgroups were compared using the mixedeffect model (randomeffects model was used to compute the overall mean weighted d in each subgroup, overall mean weighted d of subgroups were compared using the fixedeffect model because the number of subgroups was fixed). ^{b}Not all studies reported the characteristics investigated in this table. ^{c}Unlike % female patients, the slope of regression line of stimuli/session on weighted d was driven by the largest study in the current analysis ( O’Reardon et al. [2007]). After removal of this study the pvalue of the slope of regression line was .061. Abbreviations: AD, antidepressant; d, weighted standardised mean difference (Cohen’s d); HFL, highfrequency left rTMS; rTMS, repetitive transcranial magnetic stimulation. *p < .05