How Effective Is St. John's Wort? The Evidence Revisited

Background: St. John’s wort (Hypericum perforatum) has been identified as an effective treatment for depression in controlled studies and subsequent meta-analyses. However, 3 recently published large studies failed to demonstrate robust efficacy. Updated meta-analysis and assessment of publication bias may help determine the true effect of St. John’s wort.

Method: Meta-analysis to reevaluate the effectiveness of St. John’s wort as an antidepressant, funnel plot analysis, and meta-regression to assess the impact of publication bias, small-study effects, and variation in trial characteristics were performed. We conducted 2 analyses: a reproduction of a recent meta-analysis including 15 studies (Meta-15) and a meta-analysis extended by the 3 studies published since then (Meta-18). The studies in Meta-15 were identified through MEDLINE and EMBASE searches conducted in June 2000. The search terms used were St. John’s wort, hypericum, hypericin, depression, and antidepressant, and no language restrictions were applied. For both meta-analyses, we compared funnel plots, Begg’s rank correlation, Egger’s regression, trim and fill method, and meta-regression.

Results: In both analyses, effect sizes in recent studies were smaller than those reported in earlier studies; the addition of more recent studies into the analyses resulted in reduced effect size. In Meta-15, St. John’s wort was significantly more effective than placebo with a risk ratio (RR) of 1.97 (CI = 1.54 to 2.53). In Meta-18, the RR was reduced to 1.73 (CI = 1.40 to 2.14). On funnel plot analysis, the Meta-18 plot proved to be much more skewed than the Meta-15 plot. Meta-regression showed that increase in effect size was associated with smaller sample size only. The impact of baseline severity of depression could not be evaluated as the studies used different versions of the Hamilton Rating Scale for Depression.

Conclusion: St. John’s wort may be less effective in the treatment of depression than previously assumed and may finally be shown to be ineffective if future trials confirm this trend.