Screening trials are even more difficult than we thought they were.

In this issue of the Journal, Black et al. (1) present an important analysis of methodologic pitfalls associated with randomized studies of screening interventions. The authors compare disease-specific and all-cause mortality from the 12 published randomized trials of cancer screening for which these end points were available. In seven of the 12 studies, major inconsistencies were detected in the direction or magnitude of these two outcomes. Black et al. propose that the use of diseasespecific mortality as the primary end point renders screening trials subject to at least two forms of serious bias: