Causation in the presence of weak associations.

摘要

This article discusses identification of cause-effect relationships in the presence of weak associations, with particular reference to cancer and diet. Methodological problems encountered by observational research are considered, which includes chance (random error), bias (systematic error) and confounding (bias caused by a third unmeasured factor). [Presented at a workshop entitled "Decision-making for recommendations and communication based on totality of food-related research" sponsored by the International Life Sciences Institute, held in Washington DC, USA, on Dec. 15, 2008.]