We have studied serum carbohydrate antigen 19-9 (CA 19-9) and carcinoembryonic antigen (CEA) in 221 persons to assess their usefulness in the diagnosis of pancreatic carcinoma. Although serum CA 19-9 and CEA in all healthy controls were within normal limits, the positive ratings of serum CA 19-9 and CEA in all benign disease were 9.8% and 18.1%, respectively. Sensitivity of serum CA 19-9 for pancreatic carcinoma was 70.5%, which was higher than that found in healthy controls, benign disease, and other malignant disease except biliary carcinoma; but sensitivity of serum CEA levels (67.7%) was not different from that seen in malignant disease. Three of 34 patients (8.8%) with pancreatic carcinoma who had a above-normal levels of serum CA 19-9 but not serum CEA were resectable. Although there was no correlation between serum CA 19-9 and CEA, advanced stages of pancreatic, gastric, and colorectal carcinoma tend to show high serum CA 19-9 and CEA, but no statistical differences were observed in relation to the stages of these carcinomas. Comparative studies of serum CA 19-9 and CEA for sensitivity and the predictive value of true positive and negative results for detecting pancreatic, gastric, and colorectal carcinoma showed that serum CA 19-9 has significantly higher sensitivity and predictive value of true positive results for pancreatic carcinoma than for gastric and colorectal carcinoma (P less than 0.05). However, serum CEA measurements did not show any difference between these carcinomas, and the highest predictive value of a true negative result for excluding pancreatic carcinoma was also observed in serum CA 19-9. These results indicate that although the CA 19-9 assay is not specific for pancreatic carcinoma, it is more useful adjunct method for diagnosing pancreatic carcinoma, possibly in resectable stages.