Robustness and evidence of mechanisms in early experimental atherosclerosis research.

This article considers the evaluation of experimental evidence for a causal relation between cholesterol and atherosclerosis from the beginning of the 1900s until the late 1950s. It has been argued that the medical community failed to see the implications of this early research, and at first unjustifiably rejected a causal link between cholesterol and atherosclerosis. This article argues to the contrary that the medical community was justified to conclude based on the experimental evidence that cholesterol (dietary or blood) is probably not an effective target for preventive treatment. However, the evidence would have been sufficient to ascribe to cholesterol a contributing causal role in atherosclerotic heart disease. This view is argued for based on a rational reconstruction of the researchers' evaluation of evidence, specifically, the robustness of evidence for a manipulable dependence between cholesterol and atherosclerosis on the one hand, and the evidence for a mediating mechanism on the other. The case study is used to illustrate that robustness is a feasible methodological principle even when evidence is discordant, and evidence of mechanism should be evaluated on a par with evidence of statistical dependence in establishing causal claims.