Some Observations on “Observational” Research

Evidence-based medicine (EBM) ranks different medical research methods on a hierarchy, at the top of which are randomized controlled trials (RCTs) and systematic reviews or meta-analyses of RCTs. Any study that does not randomly assign patients to a treatment or a control group is automatically placed at a lower level on the hierarchy. This article argues that what matters is whether the treatment and control groups are similar with respect to potential confounding factors, not whether they got that way through randomization. Moreover, nonrandomized studies tend to have other characteristics that make them useful sources of evidence, in that they tend to last longer and to enroll more patients than do randomized trials. Replacing the sharp dichotomy between randomized and nonrandomized studies with a continuum from “clean” studies (which have high internal validity but whose results do not readily generalize to clinical practice) to pragmatic studies (which are designed to more closely reflect clinical practice) would also make a place for outcomes research and research using clinical databases, which are not included in the current hierarchy of evidence but which can provide important information about the safety and efficacy of treatments.