Selection Bias, Phase II Trials, and the FDA Accelerated Approval Process

Selection Bias, Phase II Trials, and the FDA Accelerated Approval Process

Abstract

EDITORIALS Stephen L. George Several observational studies have found that the distance traveled from a patient’s home to the treatment center is an important prognostic variable. In this issue of the Journal, Lamont et al. (1) investigated this phenomenon for clinical trials at a specialized cancer treatment center and reported that patients with head and neck cancer enrolled on phase II clinical trials experienced better survival the further they lived from the treatment center. This finding could not be explained away by adjustment for other observable demographic, medical, or socioeconomic factors (e.g., age, stage of disease, performance status, income level) that are known to influence outcomes. We are left with the conclusion that distance traveled for care—something rarely reported in the analysis of trials—is a potentially important prognostic variable, probably reflecting the effect of other more fundamental but unmeasured variables. This conclusion is unsettling because it suggests that other unknown or unmeasured variables have important prognostic impact. The results of Lamont et al. also provide an important reminder of the degree to which selection factors can influence the outcomes of trials. If the studies they considered had been restricted to distant patients, the overall results would have been