In view of the increased incidence of pancreatic cancer and the possible aetiological role of certain dietary factors, a retrospective epidemiological study was undertaken to investigate the roles of tobacco, alcohol, fat, protein and carbohydrate intakes. Sixty-nine patients with pancreatic adenocarcinoma, and 199 normal subjects were interviewed. Data were obtained on life time drinking, smoking and dietary habits. Conditional logistic regression models were used to analyse the relative risk variations. It was shown that the relative risk of cancer of the pancreas increases with fat and alcohol intakes, does not vary with protein intake, and decreases with carbohydrate intake and duration of alcohol consumption. Alcohol may be not directly involved in the aetiology of cancer of the pancreas: its effect could be due to the contents of some alcoholic beverages.