(Can J Psychiatry 2007;52:96-97) Antidepressant drugs are claimed to have specific effects on depressive symptoms. It is assumed that they do this by acting on an abnormal brain state that gives rise to depression. In contrast, I suggest that there is no evidence for this position. The effects of antidepressants seen in depression trials can easily be accounted for by nonspecific pharmacologic and psychological actions. Short-Term Studies There are thousands of randomized controlled trials (RCTs) comparing antidepressants with placebo, using various measures of depression. Metaanalyses of these have generally come to the conclusion that, overall, they show a small advantage for antidepressants. However, this obscures enormous heterogeneity among published study results, with many trials finding no difference between medication and placebo. In addition, we know that negative trials are less likely to be published than positive ones and that positive outcomes are selectively reported within trials. A recent metaanalysis that included unpublished studies found that the difference between antidepressants and placebo amounted to less than 2 points on the Hamilton Depression Rating Scale (HDRS).1 There are 2 reasons why the small difference found in this and other recent metaanalyses does not necessarily imply that antidepressants have a truly antidepressant effect. First, all rating scales contain items-such as sleeping difficulties, anxiety, and agitation-that are not specific to depression and that are likely to respond to nonspecific sedative effects associated with many antidepressants. For example, the HDRS contains 6 such items; these can score a total of 16 points (a score of 19 to 20 points indicates severe depression as defined by the American Psychiatric Association). Therefore, improvements according to rating scales may simply reflect nonspecific effects and do not necessarily imply any change in mood, per se. Second, antidepressants are active drugs and, as such, produce a range of physiological effects when ingested. These effects may indicate to assessors or trial participants whether they are taking the or the placebo; thus, the double-blind design is often penetrated in trials of antidepressants and other psychotropic agents. This is especially likely to happen in contemporary trials in which subjects are forewarned in detail about the randomized design, the use of placebo, and the nature of likely side effects. Patients on antidepressants may therefore experience an amplified placebo effect as a consequence of suspecting that they are taking the active drug. Similarly, raters may inflate ratings for individuals they suspect to be taking the active drugs on the basis of reported side effects. This amplified placebo effect is difficult to demonstrate empirically, but several metaanalyses have found that trials with more rigorously blind conditions demonstrated lower medication effects, compared with placebo, than other trials.2,3 These metaanalyses have been criticized with some justification, partly owing to the low quality of the included trials. Nevertheless, although critics are right to point out that the effects of placebo amplification have not been conclusively demonstrated, neither have they been refuted. Therefore, the clinical significance of small differences in rating scale scores in RCTs comparing antidepressants with placebo is unclear. These differences could easily be accounted for by nonspecific pharmacologie effects, such as sedation, or by amplified placebo effects. Evidence from other sources does not suggest that antidepressants have a beneficial impact on the outcome of depression. Naturalistic studies indicate that people treated with antidepressants do less well than people who are not treated with them, even after controlling for the severity of the original condition.4 Despite the enormous increase in prescribing in the West over the last decade and a half, epidemiologic evidence suggests that the prevalence of depressive episodes is higher than ever. …