Eighteen RCTs were included in the review (n=1,317 participants). Sample sizes ranged from 5 to 434 participants. Only eight RCTs obtained the maximum score of 5 points; three RCTs scored 3 points and seven RCTs scored 2 points.
There were no significant differences between valerian and placebo for latency time (WMD 0.70 min, 95% CI -3.44 to 4.83; I2=57%; 10 RCTs), or for sleep-quality measured using visual analogical scales (SMD -0.02, 95% CI -0.35 to 0.31; I2=62%; seven RCTs). However, there was substantial heterogeneity for both analyses. Neither trial quality nor latency time measurement technique were significant in the regression analysis. Results did not change significantly for sensitivity analyses, although heterogeneity was reduced for some analyses.
There was a significant improvement in sleep quality with valerian compared to placebo (RR 1.37, 95% CI 1.05 to 1.78; six RCTs) when assessed dichotomously (self reported yes/no), but there was substantial heterogeneity for this analysis (I2=60%). When only trials of greater methodological rigour (4 or more points on Jadad scale) were included in the analysis, effectiveness was reduced (RR 1.31, 95% CI 1.00 to 1.17; I2=35%), but moderate heterogeneity remained. Results did not change significantly for other sensitivity and subgroup analyses.
Only one trial reported on adverse effects, with a greater frequency of diarrhoea in patients taking valerian compared with those taking a placebo (18% versus 8%; p=0.02).
There was no evidence of significant publication bias.