Confounding and Effect Modification in Epidemiologic Studies