Moderator analysis for effect of Z drugs (treatment) or placebo on insomnia. Effect sizes are weighted standardised mean differences
| Moderator variable† | Effect size (95% CI) | β‡ |
|---|---|---|
| Polysomnographic sleep latency | ||
| Dose (22 comparisons): | ||
| 1 mg | −0.24(−0.38 to −0.11) | −0.22* |
| 20 mg | −0.50 (−0.64 to −0.35) | |
| Subjective sleep latency | ||
| Year of data collection (9 comparisons): | ||
| 1988 | −0.88 (−1.19 to −0.58) | 0.63*** |
| 2004 | −0.03 (−0.16 to 0.10) | |
| Age (9 comparisons) : | ||
| 38 years | −0.65 (−0.82 to −0.48) | 0.89*** |
| 72 years | 0.31 (0.13 to 0.50) | |
| Percentage of women (9 comparisons): | ||
| 55.9 % | 0.01 (−0.17 to 0.18) | −0.40** |
| 67.5 % | −0.67 (−0.99 to −0.34) | |
| Type of drug (11 comparisons): | ||
| Eszopiclone (3 comparisons) | 0.02 (−0.13 to 0.18) | −0.57** |
| Zolpidem (8 comparisons) | −0.47 (−0.61 to −0.33) | |
| Dose (11 comparisons): | ||
| 1 mg | 0.13 (−037 to 0.30) | −0.70*** |
| 20 mg | −1.01 (−1.31 to −0.70) | |
*P<0.05; **P<0.01; ***P<0 .001.
†Effect size of each outcome was entered as dependent variable into separate weighted least squares regressions under fixed effects assumptions for each moderator variable independently; negative ds imply lower outcome at final available measures; estimates appear for observed extremes of continuous features.
‡Standardised regression coefficient.