Worsening of depressive symptoms prior to randomization in clinical trials: a possible screen for placebo responders?

A common practice in depression trials is to exclude patients whose depressive symptoms improve between Screen and Baseline evaluations under the assumption that they are more likely to respond to placebo. The present study investigated this contention by examining the relationship between pre-randomization changes in Hamilton depression rating scale (HAMD) scores to post-randomization placebo response and drug-placebo separation. Four randomized, double-blind, placebo-controlled trials (active medication=fluoxetine or paroxetine) were conducted in outpatients with Major Depressive Disorder using a novel design in which a depressive severity inclusion criterion (HAMD >/= 22) was utilized only at Screen. Patients with no change or minimal (1 point) improvement on the HAMD between Screen and Baseline had the lowest placebo response and the best drug-placebo separation. Patients with pre-randomization improvement of 2 points or greater had moderately higher placebo response and poorer drug-placebo separation. Patients who worsened between Screen and Baseline showed the highest placebo response and the poorest drug-placebo separation. There were no obvious differences in demographic variables between the groups which could account for the findings. In our original analyses 3/4 studies failed to show significant drug-placebo separation. When only patients with no change or pre-randomization improvement of 1 point were used in the analyses, 3/4 studies showed significant drug-placebo separation while the other study approached significance, p < 0.07. These results suggest that pre-randomization changes in HAMD scores may predict post-randomization placebo response and drug-placebo separation. Further, pre-randomization increases in HAMD scores (i.e., worsening) may be the best predictor of heightened placebo responding and poor drug placebo separation.