In retrospective studies, perioperative blood transfusions were associated with poor prognosis after surgery for cancer and were a major independent risk factor for postoperative bacterial infection. Leucocyte-depleted, in contrast to buffy-coat-depleted, blood has no immunosuppressive effects in transplantation and so might lack detrimental effects on cancer prognosis and postoperative infections. We studied this hypothesis in a controlled trial by randomly allocating patients to receive either leucocyte-depleted red cells or packed cells without buffy coat when blood was needed. Between 1987 and 1990, 871 eligible patients with colorectal cancer, including 697 patients operated upon with curative intent, were randomised in the 16 participating hospitals. Neither the eligible group nor the curative group showed significant differences between the two trial transfusions in survival, disease-free survival, cancer recurrence rates, or overall infection rates after an average follow-up of 36 months. Patients who had a curative resection and who received blood of any sort had a lower 3-year survival than non-transfused patients (69% vs 81%, p = 0.001) and a higher infection rate (39% vs 24%, p < 0.001). Colorectal cancer recurrence rates, however, were not influenced by blood transfusion (30% vs 26%, p = 0.22). These combined observations confirm the association between blood transfusion and poor patient survival but indicate that the relation is not due to promotion of cancer.