Association of RBC Transfusion Thresholds and Outcomes in Medical Patients With Acute Respiratory Failure Supported With Extracorporeal Membrane Oxygenation: A Single-Center Retrospective Cohort Study

Abstract

The hemoglobin value to trigger RBC transfusion for patients receiving venovenous extracorporeal membrane oxygenation (ECMO) is controversial. Previous guidelines recommended transfusing to a normal hemoglobin level, but recent studies suggest that more RBC transfusions are associated with increased adverse outcomes. Is implementation of different institutional RBC transfusion thresholds for patients receiving venovenous ECMO associated with changes in RBC use and patient outcomes? This single-center retrospective study of patients receiving venovenous ECMO used segmented regression to test associations between implementation of institutional transfusion thresholds and trends in RBC use. Associations with secondary outcomes, including in-hospital survival, also were assessed. The study included 229 patients: 91 in the no threshold cohort, 48 in the hemoglobin< 8 g/dL cohort, and 90 in the hemoglobin< 7 g/dL cohort. Despite a decrease in number of RBC units transfused per day of ECMO support after implementation of different thresholds (mean ± SD: 0.6 ± 1.0 in the no threshold cohort, 0.3 ± 0.8 in the hemoglobin< 8 g/dL cohort, and 0.3 ± 1.1 in the hemoglobin< 7 g/dL cohort; P< .001), segmented regression showed no association between implementation of transfusion thresholds and changes in trends in number of RBC units per day of ECMO. We observed an increased hazard of death in the no threshold cohort compared with the hemoglobin< 8 g/dL cohort (adjusted hazard ratio [aHR], 2.08; 95%CI, 1.12-3.88) and in the hemoglobin< 7 g/dL cohort compared with the hemoglobin< 8 g/dL cohort (aHR, 1.93; 95%CI, 1.02-3.62). No difference was found in the hazard of death between the no threshold and hemoglobin< 7 g/dL cohorts (aHR, 1.08; 95%CI, 0.69-1.69). We observed a decrease in number of RBC units per day of ECMO over time, but changes were not associated temporally with implementation of transfusion thresholds. A transfusion threshold of hemoglobin< 8 g/dL was associated with a lower hazard of death, but these findings are limited by study methodology. Further research is needed to investigate optimal RBC transfusion practices for patients supported with venovenous ECMO.