Sixteen trials were included in the review (n= 1,196; range 20 to 259). All were described as double blind, but only six described adequate allocation concealment. Six studies scored 5 on the Jadad scale, six scored 4, two scored 3, and two scored 2. There was a small but significant effect of magnets when results from all 16 trials are pooled regardless of scale (standardised mean difference was 0.23mm, 95% CI: 0.04 to 0.42, p=0.02), however, statistically significant heterogeneity was observed (p=0.009, 51.2%). There was no significant reduction in pain with magnet therapy compared to placebo when measured on a 100mm visual analogue scale when studies of patients with musculoskeletal pain who received an intervention for two to four months or when data from longer-term trials were analysed separately. Three out of four trials reported positive effects of magnets for the treatment of pain due to osteoarthritis; no consistent benefits were reported in trials for other subgroups presented. The test for publication bias was considered inconclusive.