Beware using secular trends in deaths to judge effectiveness of breast screening