The aim of this prospective study was to evaluate if there was a statistical correlation between allogeneic blood transfusion and postoperative infections, and if this could have a dose-dependent pattern. The evaluation was based on multiple logistic and receiver operating characteristics (ROC) curve analyses. On hospital admission the following parameters were determined in 267 consecutive patients with colorectal cancer: hemoglobin, serum albumin, serum cholinesterase activity, total iron binding capacity and weight loss. Duration of operation, operative blood loss, amount of transfused blood, Dukes' cancer stage and occurrence of postoperative infections were also recorded. One hundred and thirty-two patients (49.4%) were given perioperatively allogeneic blood. Postoperative infections developed in 47 (17.6%) patients. Multivariate analysis identified allogeneic blood transfusion as the only variable related to the occurrence of postoperative infections (p < 0.05). ROC curve analysis showed that the risk for the occurrence of infection was significantly higher in patients transfused one unit of blood (p < 0.01). Moreover, a significant trend between increasing number of transfused blood units and susceptibility to infection was found (p < 0.00019).