The relation between selected indicator foods, alcohol and coffee intake, and the risk of pancreatic cancer was evaluated in a case-control study conducted between 1983 and 1992 in northern Italy on 362 patients with histologically confirmed, incident cancers of the pancreas, and 1,552 controls in hospital for acute, non-neoplastic diseases. Odds ratios (ORs) for subsequent tertiles of intake were computed after allowance for sociodemographic factors and tobacco smoking. Pancreatic cancer risk was directly associated with consumption of meat (OR for the highest frequency tertile = 1.43), liver (OR = 1.43) and ham and sausages (OR = 1.64), and inversely with consumption of fresh fruit (OR = 0.59), fish (OR = 0.65) and olive oil (OR = 0.58). No appreciable association was found with coffee (OR = 1.21) and alcohol consumption (OR = 1.20). A summary score was derived by summing the six related food items; compared to the lowest level, the OR was 2.7 for the highest quintile, and the population attributable risk was 36% (95% confidence interval, 15-57%), indicating the scope of diet for the prevention of this common neoplasm in the Italian population.