Measurement error: effects and remedies in nutritional epidemiology.

Nutritional epidemiology is concerned to elucidate the relationship between intakes of specific foods and nutrients, and specified health outcomes. Usually the outcome of interest is the incidence of a disease. Typically epidemiological evidence for such a relationship exists at two levels: (1) the macro level, in which each data-point refers to an aggregation of subjects for example, a country, town or small area; (2) the micro level, in which the relationship is observed at the level of the individual subject. The ultimate challenge is the resolution of these two levels of evidence so that the observed differences in disease patterns between different communities can be fully explained in terms of relationships demonstrated at the level of individual subjects. That this is a difficult task is due in no small measure to the problem of measurement error; we are unable to obtain perfectly accurate assessments of dietary intakes either for individuals or for communities.