The relevance of searching for effects under a clinical-trial lamppost: a key issue

Abstract

In economic evaluations of new medical technologies, analysts often need to use data from randomized controlled trials. Trials are designed to achieve high internal validity; however, their selected populations and often highly artificial environments may imply low external validity. Thus, the use of trial data in an economic evaluation may bias the results, since economic evaluation is concerned not with theoretical capability in a trial but with likely performance in the practice environment. This paper indicates both the probable bias of one aspect of artificiality in the trial environment--selected populations--and a method of adjusting the analysis so that results will be more likely to reflect actual practice. The judicious use of extra-trial information can be used to correct the biases of clinical trials.