Placebo-controlled randomized trials ideally produce unbiased estimates of the treatment effect after accounting for nonpharmacological effects (regression to the mean, Hawthorne, and placebo effects). Recognizing that the magnitude of these effects may help understand why investigators need to control for them, we sought to measure this magnitude. We reviewed published meta-analyses of randomized, placebo-controlled add-on trials of antiepileptic medications versus placebo, included in the Cochrane Library. In randomized trials of antiepileptic agents for epilepsy, 9.3-16.6% of patients in the placebo arm had a >50% reduction in seizure frequency. This effect represents 20-50% of the effect observed with active agents. Because patients with epilepsy in the placebo arm of randomized trials experience a large clinical benefit due to non pharmacological effects, randomized controlled trials are necessary to gauge the true magnitude of the treatment effect of new antiepileptic agents. © 2002 Elsevier Science (USA). All rights reserved.