The clinical records of patients identified by a prospective database as having undergone curative gastric resections for adenocarcinoma not involving the gastroesophageal junction were reviewed in order to examine transfusional practices and to determine if perioperative transfusion had an adverse effect on outcome. Between January 1985 and January 1992, 232 patients received such curative resections. The median follow-up for these patients was 19.0 months, whereas median survival for nonsurvivors was 12.3 months. Fifty-eight percent of the patients received transfusion of blood products. Fifty-four percent of these transfusions amounted to less than 2 units of blood products. By chi 2 analysis, advanced stage of disease (p = .03), advanced T-stage of primary tumor (p = .004), and total gastrectomy (p = .04) were associated with greater likelihood of transfusion. By univariate analysis, male sex (p = .004), total gastrectomy (p = .01), advanced stage of disease (p = .000006), high histologic grade of tumor (p = .03), and blood transfusion (p = .006) were predictors of poor outcome. By multivariate analysis using the proportional hazards model with stage, tumor grade, gender, extent of resection, and transfusion as covariates, blood transfusion was an independent predictor of poor outcome (p = .029, hazard 1.74). These results encourage prospective studies of transfusion on cancer recurrence and studies of alternatives to allogeneic blood transfusions in restoration of oxygen-carrying capacity during surgery in patients with gastric cancer.