Abstract

BACKGROUND: Erythrocyte blood transfusions are commonly given to intensive care unit (ICU) patients. Recent studies have suggested that transfusions may increase, have no effect on, or decrease short-term mortality. These disparate results may reflect different patient populations, diseases, and characteristics; different transfusion triggers and thresholds; or differences in controlling for confounding variables. The aim of this study was to determine the association of transfusion and 90-day mortality in patients with acute lung injury. METHODS: Retrospective analysis using binary logistic regression and then propensity analysis to control for variables on the day of the transfusion decision. One thousand patients initially enrolled in ARDSNET 05 Fluid and Catheter Treatment Trial. RESULTS: A total of 361 patients (36%) received transfusions during the 8-day course of the study and by univariable analysis were more likely to be dead by 90 days (131 of 361, 36%, vs. 153 of 639, 24%, P < .001). After multivariable correction using propensity analysis to control for factors present on the day the transfusion was given, transfusion was not associated with mortality. Instead, factors such as hypotension and laboratory values as measured by propensity scores that predicted transfusion on days 1, 2, and 5 were associated with increased mortality, and not transfusion itself. CONCLUSION: In patients suffering with acute lung injury, transfusion of erythrocytes in the ICU was not associated with 90-day mortality.