Placebos are essential when extent and variability of placebo response are unknown

EDITOR,—We need placebo controlled trials1 because responses to placebo are remarkably variable. Figure 1 shows the results of randomised double blind trials of anticonvulsants,2 antidepressants,3 and topical capsaicin4 in painful diabetic neuropathy. The proportion of patients in whom the active drug relieved at least 50% of the pain is plotted against the proportion in whom placebo relieved at least 50% of the pain. The active treatments achieved this degree of pain relief in 50–90% of the patients, while placebo did so in …