Why is the effect size so small?
We may need to be more circumspect about the prescription of antidepressants than Turner and Rosenthal imply.1 There is agreement that the effect size is not large. However, merely eliminating outcome reporting bias by obtaining the full data set from drug companies does not mean that other sources of bias in clinical trials have also been removed. The true effect size in clinical practice could still be zero if the difference between drug and placebo is due to placebo amplification.2
In particular, methodological difficulties in clinical trials are introduced through unblinding.3 Blindness is relative and both raters and patients can significantly distinguish antidepressant and placebo treatment. Raters' expectations and patients' suggestibility could therefore entirely explain the small effect size found.4
- Turner EH and Robert Rosenthal R. Efficacy of antidepresants. BMJ 2008 336: 516-517. [Full Text]
- Thomson R. Side effects and placebo amplification. Br J Psychiat 1982 140: 64-68
- Even C, Siobud-Dorocant E, Dardennes RM. (2000) Critical approach to antidepressant trials. Blindness protection is necessary, feasible and measurable. Brit J Psychiat 2000177: 47-51
- Fisher S and Greenberg RP (eds) From placebo to panacea. Putting psychiatric drugs to the test. Chichester: John Wiley, 1997
Competing interests: No competing interests