2',2'-Difluorodeoxycytidine (gemcitabine) is a cytidine analogue with established antitumor activity against several experimental tumor types and against human ovarian and non-small-cell lung cancer. Both preclinical studies and most clinical trials involving patients with solid tumors have focused on short-term administration schedules; however, mechanistic studies indicate that a continuous-infusion schedule may be more effective. We determined the maximal tolerated dose (MTD) of gemcitabine in mice using various schedules. At these MTDs we observed considerably better antitumor activity of gemcitabine in two of three murine colon carcinoma lines using a prolonged administration as compared with a standard bolus protocol (i.p. 120 mg/kg q3d x 4). On the latter schedule, Colon 26-10 grown in BALB/c mice was the most sensitive tumor line, showing a growth-delay factor (GDF, number of doubling times gained by the treatment) of 6.7, whereas Colon 38 (grown in C57/B16 mice) was the least sensitive tumor, displaying a GDF of 0.9. Prolonged treatment (q3d x 6) of Colon 26-10 at a lower dose (100 mg/kg) enhanced the antitumor activity (GDF 9.6) while producing similar toxicity. A similar weight loss was found following the continuous infusion (c.i.) of gemcitabine using Alzet osmotic pumps s.c. for 3 or 7 days (2 mg/kg), but the GDF increased to 2.4 in Colon 38 (C57/B16) as compared with that provided by the bolus injections. Continuous infusion of gemcitabine at 15 mg/kg per 24 h q7d x 2 i.v. via the tail vein was more effective than bolus injection against Colon 26-10, with the GDF being > 17.7 and 73% of the tumors regressing completely. However, against Colon 38 tumors this schedule was not effective (GDF 0.4), even with a 25% higher dose. The plasma pharmacokinetics of gemcitabine was determined after one bolus dose (120 mg/kg). The peak concentration of gemcitabine was 225 microM and that of the deaminated catabolite 2',2'-difluorodeoxyuridine (dFdU) was 79 microM. The elimination of gemcitabine was much faster than that of dFdU, with the t1/2 beta values being 15 min and 8 h, respectively. For the c.i. schedules, plasma concentrations were below the detection limit of the assay (< 0.5 microM). Our results suggest that prolonged infusion of gemcitabine can give a better antitumor activity than bolus injections and shows promise of being active in clinical trials.