Falsification end points for observational studies

摘要

To the Editor: Observational studies on the association between exposures or treatments and health outcomes are potentially biased (eg, by confounding). Drs Prasad and Jena discussed falsification end points, a potential solution to validate findings from observational studies. The idea of falsification end points, or negative controls, is elegant. Think of an end point that is known to be unrelated to the exposure or treatment under study. If that association is indeed absent (and thus unbiased), it may be concluded that the association between the treatment and the actual end point is unbiased as well. Unfortunately, this will not necessarily be the case.