Thirty-eight patients with histologically proven pancreatic adenocarcinoma were investigated to establish the utility of serum CA 19-9 as a prognostic indicator. CA 19-9 assays were performed serially during the course of the disease. In four patients with negative Lewis blood type, the CA 19-9 levels remained essentially normal throughout the disease course. In the remaining 34 patients, (1) CA 19-9 levels were significantly lower in patients with tumor size no larger than 5 cm in diameter, and in patients with resectable tumors than in those with tumor size larger than 5 cm or with unresectable tumors (p less than 0.01). 2) CA 19-9 levels dropped sharply after resection in all 11 resectable patients, whereas no significant change was found after laparotomy without resection. (3) The average survival time in seven patients whose CA 19-9 levels returned to normal after resection was significantly longer than in those four patients with postoperative CA 19-9 levels that decreased but did not return to normal (21.9 versus 8.7 months, p less than 0.05). (4) In 6 of 11 patients who underwent resection, recurrent elevation of CA 19-9 preceded changes detectable by computed tomography or clinical examination by 2 to 9 months. (5) In 23 patients who died of pancreatic carcinoma, 15 (65%) had an obvious rise in CA 19-9 level before death. There was a correlation between the doubling time of the CA 19-9 serum level and survival time (r = 0.5, p less than 0.05). Because it can be demonstrated that the reduction of tumor burden by resection lowers serum CA 19-9 levels, serum CA 19-9 levels may be a useful indicator of whether other forms of treatment such as radiation therapy or chemotherapy also reduce the tumor burden.