New pitch for placebo power.

Much fanfare has greeted drugs touted as balms for depression, such as fluoxetine (Prozac). These medications indeed elicit chemical changes in the brain linked to mood improvements, but the lion's share of their effectiveness stems from the placebo effect, according to a statistical analysis of 39 studies of treatments for depression.

"How people think about an antidepressant drug and its effect on how they feel may influence improvement more powerfully than the chemical substance itself," contends Guy Sapirstein of Westwood Lodge Hospital in Westwood, Mass.

Sapirstein and Irving Kirsch of the University of Connecticut in Storrs compiled data from published studies consisting of 3,252 moderately to severely depressed individuals who had randomly received antidepressants (including Prozac or any of several related drugs), placebo pills, psychotherapy, or no treatment. Trial periods lasted from 2 to 15 weeks.

The researchers then calculated the extent to which each treatment, on average, yielded brighter moods.

From a comparison of these estimates, Sapirstein and Kirsch assert that the placebo effect-a measure of the psychological impact of being given a medication, even if it contains no active substances-plays a dominant role in sparking favorable responses to antidepressants. The placebo effect is twice as strong as either the pharmacological effects of antidepressants or "nonspecific" factors, such as the passage of time, they maintain.

This conclusion gains force from some prior evidence (SN: 10/10/92, p. 231), but it remains tentative until issues such as differences between placebo and no treatment effects are better understood, the researchers hold. Still, they say that available data probably underestimate the placebo effect.

For instance, antidepressant researchers typically do a pretest to weed out volunteers who respond strongly to a single placebo pill. Moreover, an unknown number of studies in which antidepressants fail to outperform placebos are either not submitted or not accepted for publication, Sapirstein adds.