Secondary analysis of clinical trials--a cautionary note

There is concern in published reports and reviews that patients are being harmed or denied effective treatment by the use of questionable results from secondary analyses of data from clinical trials. A well-reported secondary analysis must make clear to the reader the uncertainty of the result--so clear, in fact, that it should be an obvious part of the conclusions that implementation should await confirmation as the primary outcome in an adequately powered trial. Those who write, review and publish these reports have a responsibility to ensure that reports accurately describe the sources of uncertainty, explain complex methods and their weaknesses with clarity, and convince readers to require better evidence before changing their practice.