Stephen J.LurieMD, PhD, Senior EditorIndividualAuthor
In Reply: Most of the specific concerns expressed
in these letters can be addressed by emphasizing several fundamental principles.
First, we compared 3 drugs on many outcomes at 4 follow-up time points. The
problem of multiple hypothesis testing is exacerbated if one does pairwise
comparisons of drugs as suggested because, for each outcome, there are 3 potential
pairwise comparisons: paroxetine vs fluoxetine, paroxetine vs sertraline,
and fluoxetine vs sertraline. Focusing on the 25 outcomes in Tables 2 and
3 would result in 300 pairwise comparisons (25 outcomes × 3 pairwise
comparisons × 4 follow-up time points). Including other outcomes reported
in Tables 4 and 5 and the text could result in more than 450 pairwise comparisons.
Contrast this with a standard clinical trial that examines 1 primary outcome
at 1 end point, such as depression severity at 12 weeks in a comparative study
of 2 antidepressants.
Kroenke K, Eckert G, Swindle R, West SL. Is One Selective Serotonin Reuptake Inhibitor Better Than Another?—Reply. JAMA. 2002;287(15):1935-1938. doi:10.1001/jama.287.15.1935