Résumé : Background. In phase II trials, irinotecan is active in patients with advanced colorectal cancer, but the survival and clinical benefit of irinotecan compared with second-line fluorouracil by continuous infusion is not known. Methods. 267 patients who had failed to respond to first-line fluorouracil, or whose disease had progressed after treatment with first-line fluorouracil were randomly allocated irinotecan 300-350 mg/m2 infused once every 3 weeks or fluorouracil by continuous infusion. Treatment was given until disease progression, unacceptable toxic effects, or the patient refused to continue treatment. The endpoint was survival, while progression-free response rate, symptom-free survival, adverse events, and quality of life (QoL) were secondary endpoints. Findings. 133 patients were randomly allocated irinotecan and 134 were allocated fluorouracil by continuous infusion. Patients treated with irinotecan lived for significantly longer than patients on fluorouracil (p = 0.035). Survival at 1 year was increased from 32% in the fluorouracil group to 45% in the irinotecan group. Median survival was 10.8 months in the irinotecan group and 8.5 months in the fluorouracil group. Median progression-free survival was longer with irinotecan (4.2 vs 2.9 months for irinotecan vs fluorouracil, respectively; p = 0.030). The median pain-free survival was 10.3 months and 8.5 months (p = 0.06) for irinotecan and fluorouracil, respectively. Both treatments were equally well tolerated. QoL was similar in both groups. Interpretation. Compared with fluorouracil by continuous infusion second-line irinotecan significantly improved survival in patients with advanced colorectal cancer.