Eleven studies (n=230) were included in the review; all but one were randomised controlled trials (n=212).
All but one of the included studies were double-blind (i.e. neither the patient nor the evaluator was aware of the treatment condition) and sham-controlled. All studies except one (for which no information was available) were described as randomised.
The mean standardised gain effect size for change in hallucination severity after rTMS was 0.76 (95% CI: 0.36, 1.17, p=0.0001; 10 studies) compared with sham treatment. There was significant heterogeneity between the studies (p=0.01) which disappeared after excluding a study inserting multiples pauses during stimulation sessions. After the exclusion of that study, the mean standardised gain effect size was 0.88 (95% CI: 0.52, 1.23, p<0.0001). The results in studies with less than five stimulation sessions versus more than five sessions were similar. Studies with a parallel-group design had a mean weighted effect size of 0.63 (95% CI: 0.30, 0.97, p=0.0001), while those with a crossover design had a mean weighted effect size of 0.93 (95% CI: 0.17, 1.69, p=0.009). The authors did not report whether this difference was statistically significant.
There was no significant improvement in the severity of all positive symptoms (PANSS positive subscale or SAPS); the effect size was 0.21 (95% CI: -0.29, 0.72, p=0.20; 6 studies).