The relationship between blood transfusion, disease-free survival, and other potential prognostic factors was prospectively studied in 339 consecutive patients with colorectal cancer. Admission and discharge hematocrit, Dukes' stage, and blood loss were significantly related to both blood transfusion and disease-free survival. Using Cox proportional hazards model, however, the association of transfusion with disease-free survival was significant (p = 0.0196) after controlling for age, sex, blood loss, procedure, tumor differentiation, stage, admission hematocrit, duration of surgery, length of the specimen, and tumor size. Dukes' stage (p < 0.0001) and blood transfusion (p < 0.0001) were the only variables independently related to disease-free survival. Forty per cent (44) of the 110 patients who received transfusions developed cancer recurrence, compared with 22% (50) of the 229 patients who did not receive blood (p < 0.0001). Five-year disease-free survival of the transfused patients was 57%, compared with 77% for nontransfused patients. Patients who developed recurrence received an average of twice as much blood as patients without recurrence (1.26 versus 0.61 units, p = 0.0128). Perioperative blood transfusion is a significant independent prognostic factor for colorectal cancer.