Pancreatic ductal carcinoma is one of the malignant diseases with the poorest prognosis. To develop effective methods for better treatment of pancreatic cancer patients, we tried to analyze the course of multistep carcinogenesis of the pancreatic ductal cells. IPMT (intraductal papillary-mucinous tumor) is thought to be one of the premalignant lesions of the pancreas, which would transform into carcinomas. Loss of 18q at the SMAD4 locus is known to be an early genetic change in pancreatic ductal carcinomas. It is not clear, however, whether or not the target gene for inactivation is SMAD4. Using 18 IPMTs, we analyzed LOH at the SMAD4 locus and observed frequent LOH (7/14, 50%). No mutations were observed in any of the tumors. Moreover, the expression level of the SMAD4 protein did not show a reduction in IPMTs. These results suggested that (i) inactivating mutation of the SMAD4 gene is a rather late genetic change in pancreatic carcinogenesis, and (ii) there may be an unknown tumor suppressor gene in 18q, other than SMAD4, that is involved in pancreatic ductal carcinogenesis.