The use of attributable fraction in the design and interpretation of epidemiologic studies.