Cost-effectiveness analysis: discount the placebo at your peril.

INTRODUCTION: The authors consider alternative mechanisms that might explain placebo responses and their implications for cost-effectiveness modeling. Three alternative placebo mechanisms are examined: a ''regression to the mean'' effect arising from natural variation and the preferential selection of patients with acutely severe disease into clinical trials, a patient expectancy effect specific to the clinical trial setting (Hawthorne effect), and a patient expectancy effect generalizable to routine clinical practice (true placebo effect). METHODS: To estimate cost-effectiveness, the authors needed to generalize from trial data to estimate responses to treatment that they would see in routine clinical practice. They use an example analysis of the cost-effectiveness of adjunct epilepsy treatments to illustrate the potential effects of these different placebo mechanisms on this generalization and subsequent cost-effectiveness estimates and adoption decisions. RESULTS: If an acceptable willingness-to-pay threshold of 30,000 per quality-adjusted life year (QALY) is assumed, then each of the placebo effect scenarios identifies a different treatment alternative as being optimum. DISCUSSION: Estimated cost-effectiveness ratios and associated policy decisions may be sensitive to assumptions regarding the mechanism underlying placebo responses. These assumptions should, if possible, be investigated through analysis of trial or observational data and, in the absence of other evidence, sensitivity analysis.