The objective of this study was to clarify the clinical features of long-time survivors with unresectable pancreatic cancer treated by gemcitabine (GEM) alone and to predict survival time after the first course of treatment. Eighteen consecutive patients (median age 65.3 years, range 49-77 years; 12 males, 6 females) with unresectable pancreatic cancer and a baseline Karnofsky's performance state = or >60 were treated with GEM in a dose of 1,000 mg/m(2) weekly x 3 followed by 1 week of rest until progression. The overall response rate was 0% (CR 0, PR 0, SD 11 cases, and PD 5 cases), and the median survival time (MST) was 268 days. We observed a statistically significant difference in the patients with or without liver metastasis at the start of treatment (131.6 vs 324.9 days; p=0.0045). We also evaluated the usefulness of serial CA 19-9 measurements as a biochemical response marker and an outcome prognostic parameter in patients with unresectable pancreatic cancer receiving GEM alone. We classified two subgroups into responders (patients with a decrease of = or >10% of the baseline CA 19-9 level after 4 weeks of chemotherapy) and non-responder (patients with a increase of the baseline CA 19-9 level or with a decrease of <10% of the baseline CA 19-9 level after 4 weeks of chemotherapy). Responder had a significantly better median survival than non-responders (416.6 vs 138.3 days; p=0.009). In conclusion, CA 19-9 serum concentration serves as an early indicator of response to chemotherapy with GEM alone in unresectable pancreatic cancer.