Groupe hospitalier Pitié-Salpêtrière Service de médecine nucléaire 83, bd de l'Hôpital 75651 Paris cedex 13
Epidemiology is an essential tool in public health decision-making and in clinical medicine, furnishing as it does data on health status that are essential for risk analysis and for the evaluation of health practices and policies. But epidemiology encounters problems when it seeks to provide clear answers about the level, or even the reality, of risks associated with low exposure, especially to environmental factors. Contradictory studies and false-positive findings are frequent in this domain. Among the causes of these inconsistencies we find the failure to consider either the uncertainty associated with exposure to the suspected risk factors or relevant confounding factors (particularly in the case-control studies, where they are assessed retrospectively), multiple statistical tests without appropriate correction, and unverifiable differential bias (especially related to history-taking). These difficulties can generally be controlled. A scale of validity for epidemiological studies should be developed that assesses their methodological reliability and the degree of proof they provide.