The risk of fallacious conclusions from autopsy data on the incidence of diseases with applications to heart disease.