Prevention for multifactorial diseases.

This paper addresses the question of choosing between alternative preventive strategies for diseases with several risk factors. Estimates are obtained of the reduction in cases to be anticipated following a modification of the population exposure to one or several of these hazards. In practice, such estimates must often be made without knowledge of the joint distribution of exposure to risk, and so general conditions are established under which the factor attributable risks, calculated by ignoring all other factors, are unbiased effect measures. These concepts are illustrated by data from several epidemiologic studies.