Scientific standards of criticism: a reaction to "Scientific standards in epidemiologic studies of the menace of daily life," by A.R. Feinstein.

Feinstein's criticism of epidemiology are summarized and evaluated. His major premises appear to be that methodologic errors inevitably result in false positive associations, that epidemiologists are complacent about methodology, and that epidemiologists encourage overinterpretation of epidemiologic data by the public. We challenge these premises and discuss a number of technical errors that underlie Feinstein's writings: he generally ignores the tendency of nondifferential exposure misclassification to dilute associations; he inappropriately interchanges measure of disease occurrence, which leads him to erroneous assertions regarding differences among studies; and he asserts that absence of a dose-response gradient precludes causality, despite the fact that causal effects need not follow a monotonic dose-response curve over the entire range of exposure. We also challenge his assertions that contradictory results necessarily indicate poor study methods and that prior beliefs of the data collectors inevitably affect study results in ways not apparent in the study methods. Feinstein's approach to the use of epidemiology in policy formulation ignores the need to weigh all evidence and distinguish important from unimportant biases. We illustrate this point by documenting how the use of Feinstein's "scientific standards" in evaluating the evidence linking aspirin to Reye's syndrome would have led to costly errors in medical practice.