One hundred seventy-nine patients with advanced measurable colorectal cancer not previously treated with chemotherapy were entered into a prospective randomized clinical trial by the Mid-Atlantic Oncology Program (MAOP) to compare two schedules of delivery for single-agent fluorouracil (5-FU). The "standard" treatment was a schedule commonly employed in clinical practice, namely, a daily bolus dose administered intravenously (IV) for five consecutive days and repeated at 5-week intervals. The investigational treatment was a continuous infusion of 5-FU administered 24 hours a day for a protracted time (10 weeks or more). Both treatments were continued until the development of disease progression or unless interrupted for toxicity. Using stringent objective criteria requiring independent confirmation of x-ray or scan-documented response, the tumor response rate reached 7% (six of 87) for the bolus arm and 30% (26 of 87) for the infusion arms (P less than .001). Toxicity was substantially different for the two arms with major leukopenia observed only on the bolus arm, 22% developing grade 3 (severe) or grade 4 (life-threatening) leukopenia with four sepsis-related deaths. Hand-foot syndrome was observed only in the infusional arm, requiring treatment interruptions and dose reductions in 24% of patients, but with little impact on quality of life. In spite of the major difference in objective response rate, overall survival for the two groups was comparable. Administration of 5-FU as a continuous infusion for protracted periods clearly improves the therapeutic index for this agent in patients with advanced colon cancer with respect to response rate and reduced toxicity. This schedule appears workable in the community setting and yields response rates similar to those reported for 5-FU with high-dose leucovorin, but without the gastroin testinal toxicity profile of the latter combination.