BLOOD-TRANSFUSION AND INFECTION AFTER COLORECTAL-CANCER SURGERY

Background: Many observational studies have described an association b etween perioperative transfusion and postoperative infection, Detectio n of such a relationship may depend on which variables are considered as potential confounders of the association under study. However, most reports have not considered risk factors for postoperative infection at specific sites as possible explanations for the observed relationsh ip. Study Design and Methods: The records of 492 patients undergoing e lective colorectal cancer resection at the Massachusetts General Hospi tal between January 1992 and December 1994 were reviewed. The probabil ity of infection in association with transfusion was calculated with a nd without adjustment for the effects of chronic systemic illness, num ber of days with urinary catheter, endotracheal intubation, impaired c onsciousness, and specific risk factors for wound infection. Postopera tive infection at any site and infections at specific sites were analy zed as separate outcomes. Results: After adjustment for the effects of the variables listed above, allogeneic transfusion was not associated with postoperative infection at any site (p = 0.407). Only a specific association of transfusion with wound infection could be detected. Ho wever, in an analysis that adjusted for the effects of only the 18 con founders considered by previous authors, transfusion was the most sign ificant predictor of infection. In that analysis, the risk of postoper ative infection Increased by 14 percent per unit of red cells transfus ed (p<0.001). Conclusion: The overall adverse relationship between tra nsfusion and infection reported by previous observational studies may have been dire to an incomplete consideration of the variables that co nfound that association, This finding may help explain the disagreemen t between the conclusions of recent large, randomized, controlled tria ls (which failed to detect a deleterious transfusion effect) and the e arlier observational studies.