Background: The indications for transfusion have never been evaluated in an adequately sized clinical trial. A pilot study was conducted to plan larger clinical trials.
Study design and methods: Hip fracture patients undergoing surgical repair who had postoperative hemoglobin levels less than 10 g per dL were randomly assigned to receive 1) symptomatic transfusion: that is, transfusion for symptoms of anemia or for a hemoglobin level that dropped below 8 g per dL or 2) threshold transfusion: that is, patients receive 1 unit of packed RBCs at the time of random assignment and as much blood as necessary to keep the hemoglobin level above 10 g per dL. Outcomes were 60-day mortality, morbidity, functional status, and place of residence.
Results: Among 84 eligible patients enrolled, mean (+/- SD) prerandomization hemoglobin was 9.1 (+/- 0.6) g/ dL. The median number of units transfused in the threshold transfusion group was 2 (interquartile range, = 1-2), and that in the symptomatic transfusion group was 0 (6; interquartile range, = 0-2) (p < 0.001). Mean hemoglobin levels were approximately 1 g per dL higher in the threshold group than in the symptomatic group: for example, on Day 2, 10.3 (+/- 0.9) g per dL versus 9.3 (+/- 1.2) g per dL, respectively (p < 0.001). At 60 days, death or inability to walk across the room without assistance occurred in 16 (39.0%) of the symptomatic transfusion group and 19 (45.2%) of the threshold transfusion group. Death occurred by 60 days in 5 (11.9%) of the symptomatic transfusion group and 2 (4.8%) in the threshold transfusion group (relative risk = 2.5; 95% CI, 0.5-12.2). Other outcomes were similar for the two groups.
Conclusions: Symptomatic transfusion may be an effective blood-sparing protocol associated with the transfusion of appreciably fewer units of RBCs and lower mean hemoglobin levels than are associated with the threshold transfusion policy. However, it is unknown whether these two clinical strategies have comparable mortality, morbidity, or functional status. A definitive trial is needed.