PURPOSE To determine the predictive value of carbohydrate antigen (CA) 19-9 in pancreatic cancer treated with radiochemotherapy. METHODS AND MATERIALS Ninety-five patients with locally advanced unresectable adenocarcinoma of the pancreas were treated with hyperfractionated accelerated radiotherapy to a total dose of 44.8 Gy combined with 5-fluorouracil and folinic acid. CA 19-9 was measured before therapy, each week during therapy, and every 4 weeks during the follow-up period. RESULTS The median CA 19-9 before treatment was 420 U/mL; in the responder group it was 117 U/mL, and in the nonresponder group it was 806 U/mL. Patients with a pretreatment CA 19-9 less than the median had not only a significantly better tumor response (45.8%) but also a better survival prognosis (median survival 12.3 months) than those with a level higher than the median (tumor response 12.8%; median survival 7.1 months). The posttreatment median CA 19-9 for all patients also exhibited prognostic significance. The median survival of patients with a CA 19-9 level lower than the posttreatment median of 293 U/mL was 13.5 months, compared with 7.2 months for those with a CA 19-9 level greater than the median. To detect recurrent disease during follow-up, the sensitivity of CA 19-9 was 100% and the specificity 88%. CONCLUSION Our results indicate that CA 19-9 is of predictive value for prognosis, response, and detecting recurrence of pancreatic cancer in patients undergoing combined radiochemotherapy. Therefore, we recommend the routine implementation of CA 19-9 observation during the clinical course of treatment for patients with pancreatic cancer undergoing radiochemotherapy.