OBJECTIVE: To investigate the efficacy of hypericum extract WS 5570 (St John's wort) compared with paroxetine in patients with moderate to severe major depression. DESIGN: Randomised double blind, double dummy, reference controlled, multicentre non-inferiority trial. SETTING: 21 psychiatric primary care practices in Germany. PARTICIPANTS: 251 adult outpatients with acute major depression with total score > or = 22 on the 17 item Hamilton depression scale. INTERVENTIONS: 900 mg/day hypericum extract WS 5570 three times a day or 20 mg paroxetine once a day for six weeks. In initial non-responders doses were increased to 1800 mg/day hypericum or 40 mg/day paroxetine after two weeks. MAIN OUTCOME MEASURES: Change in score on Hamilton depression scale from baseline to day 42 (primary outcome). Secondary measures were change in scores on Montgomery-Asberg depression rating scale, clinical global impressions, and Beck depression inventory. RESULTS: The Hamilton depression total score decreased by mean 14.4 (SD 8.8) points, corresponding to 56.6% (SD 34.3%) of the baseline value, in the hypericum group and by 11.4 (SD 8.6) points (44.8% (SD 33.5%) of baseline value) in the paroxetine group (intention to treat analysis; similar results were observed in the per protocol analysis). The intention to treat analysis (lower one sided 97.5% confidence limit 1.5 points for the difference hypericum minus paroxetine) and the per protocol analysis (lower confidence limit 0.7 points) showed non-inferiority of hypericum and statistical superiority over paroxetine. The lower limits in both cases exceeded the pre-specified non-inferiority margin of -2.5 points and the superiority margin of 0. The incidence of adverse events was 0.035 and 0.060 events per day of exposure for hypericum and paroxetine, respectively. CONCLUSIONS: In the treatment of moderate to severe major depression, hypericum extract WS 5570 is at least as effective as paroxetine and is better tolerated.