Association between physician beliefs regarding assigned treatment and clinical response: re-analysis of data from the Hypericum Depression Trial Study Group.

BACKGROUND: We have previously shown an association between patient belief and treatment response in the Hypericum Depression Trial Study Group's 2002 study. We re-examined these data to determine whether clinical improvement was associated with physician belief about assigned therapy. METHODS: Three hundred and forty adults with major depression and baseline scores ≥20 on the 17-item Hamilton Depression Scale (HDRS-17) were randomized to Hypericum 900-1500mg/day, sertraline 50-100mg/day, or placebo for 8 weeks. At week 8, physicians guessed their patients' treatment. We analyzed 277 subjects with at least one post-baseline visit and physician guess data. We examined association between guess and improvement in HDRS-17 and whether treatment assignment moderated the effect of belief on remission (final HDRS-17 score <8). RESULTS: Patient and doctor guesses agreed at 53% for sertraline, 68% for Hypericum, and 52% for placebo (kappa=0.37). Doctors guessed placebo correctly (38%) more than sertraline (18%) or Hypericum (19%) (p=0.001). Adverse event scores were significantly greater among subjects for which the clinicians guessed Hypericum (p<0.001) or sertraline (p=0.005) compared to placebo. Significant improvements in HDRS-17 score were found when comparing the Hypericum-guess (p<0.001) or the sertraline-guess group (p<0.001) against the placebo-guess group. Remission rates were significantly greater for subjects whose clinicians guessed sertraline (p<0.001) or Hypericum (p<0.001) versus placebo. CONCLUSION: Doctors tended to guess placebo more easily than Hypericum or sertraline, and their guesses tended to favor active therapies when improvement was more robust. Results show association but not causation, and merit more careful investigation.