Between March 1984 and July 1988, 1,158 patients with Dukes' A, B, and C carcinoma of the colon were entered into National Surgical Adjuvant Breast and Bowel Project (NSABP) Protocol C-02. Patients were randomized to either no further treatment following curative resection or to postoperative fluorouracil (5-FU) and heparin administered via the portal vein. Therapy began on day of operation and consisted of constant infusion for 7 successive day. Average time on study was 41.8 months. A comparison between the two groups of patients indicated both an improvement in disease-free survival (74% v 64% at 4 years, overall P = .02) and a survival advantage (81% v 73% at 4 years, overall P = .07) in favor of the chemotherapy-treated group. When compared with the treated group, patients who received no further treatment had 1.26 times the risk of developing a treatment failure and 1.25 times the likelihood of dying after 4 years. Particularly significant was the failure to demonstrate an advantage from 5-FU in decreasing the incidence of hepatic metastases. The liver was the first site of treatment failure in 32.9% of 82 patients with documented recurrences in the control group and in 46.3% of 67 patients who received additional treatment. Therapy is administered via a regional route to affect the incidence of recurrence within the perfused anatomic boundary. Since, in this study, adjuvant portal-vein 5-FU infusion failed to reduce the incidence of hepatic metastases, it may be concluded that its use thus far is not justified. It may also be speculated that the disease-free survival and survival advantages (the latter of borderline significance) are a result of the systemic effects of 5-FU.