Case-control studies of pancreatic cancer were conducted in 5 populations with moderate to high rates and differing dietary practices, using a common protocol and questionnaire. Comprehensive diet histories were completed for a total of 802 cases and 1669 controls identified in Adelaide (Australia), Montreal and Toronto (Canada), Utrecht (The Netherlands) and Opole (Poland). Positive associations were observed with intake of carbohydrates and cholesterol, and inverse associations with dietary fiber and vitamin C. These relationships were generally consistent among the 5 studies, and showed statistically significant and generally monotonic dose-response relationships. The relative risks for highest vs. lowest quintile of intake were estimated for carbohydrates to be 2.57 (95% confidence interval 1.64-4.03), cholesterol 2.68 (1.72-4.17), dietary fiber 0.45 (0.30-0.63), and vitamin C 0.53 (0.38-0.76). The consistency, strength, and specificity of these associations provides evidence for the hypothesis that some or all of these dietary factors may alter the risk of pancreatic cancer.