A total of 139 treatment groups from 66 studies were assessed (n=1,538 and n=369 for experimental and no-treatment participants, respectively). The numbers of studies and the number of participants used to analyse the various outcomes are unclear.
Progressive muscular relaxation-based approaches were assessed from 40 groups. Other relaxation approaches, including meditation, desensitisation, imagery relaxation, hypnosis and autogenic training, were assessed from 16 groups.
Stimulus control approaches were assessed from 25 groups.
Paradoxical intention approaches were assessed from 14 groups.
Sleep restriction approaches were assessed from 4 groups.
Combination treatments were assessed from 23 groups.
Placebo effect sizes were assessed from 17 groups.
Results are given for all 4 outcomes and for all treatments investigated. Only summary results are quoted below.
SOL: overall weighted mean effect 0.87 (95% CI: 0.58, 1.16); the weighted mean effect ranged from 0.46 (placebo) to 1.16 (stimulus control).
Follow-up weighted mean effect: 1.01 (95% CI: 1.01, 1.01); the weighted mean effect ranged from 0.43 (placebo) to 2.04 (other relaxation).
TST: overall weighted mean effect 0.49 (95% CI: 0.49, 0.49); the weighted mean effect ranged from 0.1 (paradoxical intention and placebo) to 0.78 (combination).
Follow-up weighted mean effect: 0.54 (95% CI: 0.54, 0.54).
Number of nocturnal awakenings: overall weighted mean effect 0.63 (95% CI: 0.63, 0.63); the weighted mean effect ranged from 0.37 (other relaxation) to 1.00 (paradoxical intention).
Follow-up weighted mean effect: 0.78 (95% CI: -0.08, 1.64).
Sleep quality ratings: overall weighted mean effect 0.94 (95% CI: 0.28, 1.60); weighted mean effect ranged from 0.21 (placebo) to 1.30 (stimulus control).
Follow-up weighted mean effect: 1.30 (95% CI: 1.30, 1.30).
The following factors were found to be significant predictors of effect size: source of participant, i.e. clinically-referred compared to solicited volunteer, non-hypnotic drug user compared to hypnotic user, year of study and presence of home practice. These 4 variables accounted for 28.5% of the variance. After controlling for differences in patient characteristics, treatment settings and methodological features, the significant predictors of the effect size were the source of participants, non-use of drugs and placebo treatment. None of the design variables considered were found to influence effect size estimates.