| Evidence-based medicine (EBM) has been defined as "the conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients" (Sackett et al., 1996). EBM has had great influence, particularly on the development of "secondary resources" that summarize the medical literature in order to make it more readily available to practicing clinicians. These secondary resources are based on the idea that there is a "hierarchy of evidence" so that the higher on this hierarchy a study ranks, the stronger the evidence it provides. At the top of the hierarchy is the randomized controlled trial (RCT). It is the central contention of this thesis that, while RCTs may in many cases be necessary tests of the efficacy and safety of a treatment, they are far from sufficient. Moreover, I argue that the hierarchy of evidence should be replaced by a view of evidence that (1) classifies population-level (or epidemiological) studies on a variety of methodological criteria, rather than solely on whether or not the study subjects were randomly assigned to their study groups and (2) is sensitive to the potential of variation within study groups to affect the results of a study, as well as to the need to identify demographic or physiological differences associated with these differences in clinical outcome. In making my case for the need for a richer view of medical evidence than that currently provided by EBM, I draw on research in philosophy of science. In particular, I use Nancy Cartwright's concept of a nomological machine to show that (and why) the results of RCTs cannot easily be extrapolated to a broader clinical context, and I build on recent work on the importance of mechanisms in science to illustrate the close relationship between epidemiological research (including RCTs) and laboratory studies. Based on these arguments, I sketch an alternative approach to the nature of medical research that I argue is superior to EBM's hierarchy of evidence for both epistemological and ethical reasons. Finally, I consider the implications of my proposed alternative for the way in which secondary resources should be compiled. |
