One hundred and twenty-five previously untreated patients bearing metastatic or advanced recurrent (inoperable) colorectal carcinoma and measurable disease were prospectively randomized. Those in arm A received 5-fluorouracil (5-FU), 1,200 mg/m2 i.v. infusion over 2 h, while those in arm B received methotrexate (MTX), 200 mg/m2 i.v. (push injection), followed 20 h later by 5-FU, 1,200 mg/m2 i.v. infusion over 2 h, plus calcium leucovorin (LV), 25 mg i.m. every 6 h for eight doses beginning 24 h after MTX administration. Cycles were repeated every 15 days. All patients receiving treatment were evaluable for toxicity and survival, and 118 patients were evaluable for response. The objective regression rate (complete plus partial response) was 12% (7 of 58) in arm A and 28% (17 of 60) in arm B (p = 0.049). No change was observed in 24% (14 of 58) in arm A and in 35% (21 of 60) in arm B (p = 0.28), while progressive disease was registered in 64% (37 of 58) and 37% (22 of 60) in arms A and B, respectively (p = 0.006). Median duration of response was 3 months in arm A and 5 months in arm B (p = 0.39). The median survival was 8.3 months in arm A and 11.2 months in arm B (p = 0.25). No statistically significant differences were found when objective regression and survival were related to site of primary tumor, performance status, and number of involved organs. There were two drug-related deaths in arm B due to severe myelosuppression followed by mucositis and sepsis. Of nonhematologic toxicities, diarrhea was more frequently observed in arm B, as were mucositis and infectious complications. Our results indicate that the sequential schedule MTX-5-FU-LV with 20-h intervals between MTX and 5-FU is superior in terms of objective regression to 5-FU alone given at the dose and schedule used in the present study. However, MTX-5-FU-LV did not have a significant impact on survival.