Usefulness of population controls in retrospective studies of alcohol consumption. Experience from a case--control study of esophageal cancer in Ille-et-Vilaine, France.