The significance of "nonsignificance" in randomized controlled studies: a discussion inspired by a double-blinded study on St. John's Wort (Hypericum perforatum L.) for premenstrual symptoms
Hicks, S.M., Walker, A.F., Gallagher, J., Middleton, R.W. and Wright, J. (2004) The significance of "nonsignificance" in randomized controlled studies: a discussion inspired by a double-blinded study on St. John's Wort (Hypericum perforatum L.) for premenstrual symptoms. Journal of Alternative and Complementary Medicine, 10 (6). pp. 925-932. ISSN 1075-5535
Full text not archived in this repository.
Official URL: http://www.liebertpub.com
Objectives: This study aimed to investigate the efficacy of St. John's wort extract (SJW) as a treatment for premenstrual symptoms. Design: The study was a randomized, double-blinded, placebo-controlled trial, with two parallel treatment groups. After a no-treatment baseline cycle, volunteers were randomized to either SJW or placebo for a further two menstrual cycles. Settings/location: A postal trial conducted from The University of Reading, Berkshire, England. Subjects: One hundred and sixty-nine (169) normally menstruating women who experienced recurrent premenstrual symptoms were recruited onto the study. One hundred and twenty-five (125) completed the protocol and were included in the analysis. Interventions: Six hundred milligrams (600) mg of SJW (standardized to contain 1800 mug of hypericin) or placebo (containing lactose and cellulose). Outcome measure: A menstrual diary was used to assess changes in premenstrual symptoms. The anxiety-related subgroup of symptoms of this instrument was used as the primary outcome measure. Results: After averaging the effects of treatment over both treatment cycles it was found that there was a trend for SJW to be superior to placebo. However, this finding was not statistically significant. Conclusion: The possibility that this nonsignificant finding resulted from insufficient statistical power in the study, rather than a lack of efficacy of SJW, is discussed. Following this discussion the recommendation is made that, in future, similar studies should be powered to detect a minimum clinically relevant difference between treatments.