Analyzing the relationship of exercise and health: methods, assumptions, and limitations

The papers in this supplement to Medicine and Science in Sports and Exercise seek to summarize our knowledge of the dose-response relationship of exercise with a variety of health outcomes. This type of rigorous evidence summation has become quite popular in the past 10 years as health care providers and policymakers seek to distinguish interventions that are helpful--the expected benefits justify human and economic costs--from those of no or uncertain benefit (7). Proponents of "evidence-based medicine" argue that in contrast to old-style literature reviews in which the writer, typically an "expert," develops a thesis and cites only those publications that support his point of view, an evidence-based approach finds all of the evidence and objectively weighs it to determine the "truth" (20,22). Although such methods offer the promise of increased objectivity, they are not immune to bias, and readers should understand that the declaration that something is "evidence-based" is no guarantee of its veracity. In fact, because the literature virtually never includes studies that perfectly match the setting and personal circumstance of the patient under treatment, "evidence-informed medicine" would be a more honest title for care grounded in systematic reviews of the literature (23). This paper begins with a discussion of potential theoretical problems with systematic reviews and concludes by considering specific problems that members of the panel were likely to encounter when evaluating the literature on exercise and health. These concepts were presented in the initial lecture of the conference so that the 24 lecturers and 6 members of the consensus panel would be forewarned of the dangers that lay ahead.