Pancreatic adenocarcinoma is thought to arise from a noninvasive neoplastic precursor, the pancreatic intraductal lesion (PIL). Mutations of the K-ras gene are known to occur in PILs, but their high prevalence among PILs within the general population probably limit the use of K-ras as a marker of eventual clinical risk. In search of genetic constellations that might indicate the progression of some PILs toward an invasive phenotype, mutations at both the K-ras and p16 genes were sought within PILs of 10 pancreata resected for adenocarcinoma. K-ras mutations were present in most PILs and in nearly all PILs having nuclear atypia. In half of the patients, two or more unique K-ras mutations were identified among distinct PILs, which is evidence for the separate clonal evolution of multiple pancreatic neoplasms within individual patients. p16 alterations (one homozygous deletion and three point mutations) were found in 4 of the 10 carcinomas; these four pancreata harbored p16 alterations in three of nine PILs, of which one was a "histologically early" lesion. Two patients had p16 alterations in PILs matching those of the associated carcinomas. p16 mutations were not found in PILs of pancreata having wild-type p16 in the carcinoma, nor were they found in ducts having normal histology. It is suggested that alterations of the p16 gene affect a subset of PILs that contain mutations of the K-ras gene and that these mutations might identify high-risk precursors of the invasive malignancy.