Résumé : Background: Ki-ras mutation analysis from material collected during ERCP has been claimed to improve the diagnosis of pancreatic and bile duct carcinomas as compared with conventional cytology. Our aim was to study the relative contribution of both Ki-ras analysis and brush cytology in patients with a significant stricture at ERCP. Methods: Brushings were collected in duplicate for both analyses in 142 patients in whom a definitive diagnosis was obtained by histology or a minimal follow-up of 6 months. Results: For pancreatic strictures, sensitivity, specificity, and accuracy of Ki-ras analysis vs. cytology in detecting malignancy were 81% vs. 66%, 72% vs. 100%, and 70% vs. 74%, respectively. For biliary strictures, they were 25% vs. 42%, 100% vs. 100%, and 35% vs. 43%, respectively. The combination of the two methods only marginally increased their sensitivity and accuracy in both types of strictures. Conclusion: Ki-ras analysis is a sensitive method for diagnosing pancreatic but not biliary carcinoma. However, its specificity is lowered by a high frequency of Ki-ras mutations in patients with chronic pancreatitis (25%) who did not manifest cancer development within a 6-month follow-up period. In pancreatic duct strictures, brush cytology appears to be more specific in detecting malignancy; specificity for Ki-ras and cytology are equivalent for the diagnosis of malignant bile duct strictures. Therefore, making a clinical decision on the sole basis of Ki-ras analysis is probably not justified in the majority of the cases.