Twenty-seven studies were included in the review. Nineteen studies were randomised controlled trials (RCTs) (n=1,005); all other studies were uncontrolled, but exact design was unclear (n also unclear).
Between-group analysis: The weighted overall mean effect size was 0.51 (95% CI 0.46 to 0.63; 19 RCTs), which indicated a medium efficacy. Statistically significant heterogeneity was detected (p<0.05). Pooled effect sizes ranged from 0.43 (multi-modality relaxation technique; three RCTs) to 0.95 (applied relaxation; one RCT). Meditation also showed a large pooled effect size (0.86). Pooled effect sizes were larger for volunteers than for patient groups (p<0.001). There was a positive correlation between length of training and effect size (p<0.05) and between the use of homework and effect size (p<0.001). The failsafe N was 79 to reduce the mean d to an effect size of 0.1.
Within-group analysis of 25 studies (n=748) showed a weighted overall mean effect size of 0.57 (95% CI 0.52 to 0.68; 25 studies), which indicated a medium efficacy. Statistically significant heterogeneity was detected (p<0.01).