Objective: To reanalyse data from a 2002 study by the Hypericum Depression Trial Study Group to determine whether patients who believed they were receiving active therapy rather than placebo obtained greater improvement, independent of treatment.

Method: 340 adults with major depressive disorder (according to the Structured Clinical Interview for DSM-IV) and baseline scores of 20 or higher on the 17-item Hamilton Depression Rating Scale (HDRS-17) were randomised to Hypericum perforatum 900 to 1500 mg/d, sertraline 50 to 100 mg/d or placebo and were asked to guess their assigned treatment after 8 weeks.  This reanalysis of data was performed from 1 Oct 2009 to 15 Apr 2011.  The intention-to-treat sample included 207 subjects (mean age = 44 years) who had (1) at least 1 post-baseline visit; (2) adherence data based on serum levels of hyperforin, sertraline and desmethylsertraline; and (3) guess data.  Univariate factorial analysis of variance was used to determine whether treatment assignment affected clinical improvement according to HDRS-17 score and whether this effect was moderated by patient guess of sertraline, Hypericum or placebo.  Analysis of covariance was used to determine whether side effects mediated improvement in the context of patient guess and assigned treatment.  Chi2 analyses compared response rates (50% or greater decrease in HDRS-17 score) between the guess groups and between the treatment groups within each guess group.

Results: Assigned treatment had no significant effect on clinical improvement (P = 0.65), but patient guess was significantly associated with improvement (P less than 0.001), and treatment and guess interacted significantly (P = 0.005).  Among subjects who guessed placebo, clinical improvement was small and did not differ significantly across treatments.  Among subjects who guessed Hypericum, improvement was large and did not differ significantly across treatments.  Among subjects who guessed sertraline, those who received placebo or sertraline had large improvements, but those who received Hypericum had significantly less improvement (P less than 0.001).  Similar findings were obtained for response rates.

Conclusions: Patient beliefs regarding treatment may have a stronger association with clinical outcome than the actual medication received, and the strength of this association may depend upon the particular combination of treatment guessed and treatment received.