PRACTICE CHANGER
Do not order transfusions of red blood cells for patients with acute upper gastrointestinal bleeding unless their hemoglobin level <7 g/dL.
Villanueva C, Colomo A, Bosch A, et al. Transfusion strategies for acute upper gastrointestinal bleeding. N Engl J Med. 2013;368:11-21.1
A: Based on a single randomized controlled trial (RCT) consistent with other RCTs on recommendations for transfusion.
ILLUSTRATED CASE
An 82-year-old patient presents to the emergency department with several episodes of melena over the past week and one episode of hematemesis this morning. He denies any shortness of breath, dizziness, lightheadedness, or fatigue. He is tachycardic but normotensive. Lab results note a hemoglobin level of 8.3 g/dL. Should you order a transfusion of red blood cells?
Acute upper gastrointestinal bleeding (UGIB) commonly requires hospital admission, with approximately 61 cases per 100,000 population in the United States in 2009.2 Gastroduodenal peptic ulcer disease accounts for the majority of these cases.3 Although trends indicate an overall decrease in cases requiring hospitalization, UGIB remains a condition associated with a mortality rate of 2.5% and inpatient costs of $2 billion annually.2,3
Studies have been inconclusive—until now
An RCT published in 1999 showed a restrictive transfusion strategy (hemoglobin threshold of 7 g/dL) to be at least as effective as—and possibly superior to—a liberal strategy (threshold of 10 g/dL) in critically ill patients.4 In 2010, an RCT demonstrated that a liberal transfusion strategy (also defined as a transfusion threshold of 10 g/dL) did not reduce the rates of death or in-hospital morbidity in elderly patients after hip surgery.5 A recent Cochrane review of transfusion strategies for UGIB included only 3 small studies (N=93), so its authors could not draw any firm conclusions.6 The results of a new RCT, detailed below, are more conclusive.
STUDY SUMMARY: Restrictive transfusion policy lowers mortality risk
Villanueva et al conducted a nonblinded RCT comparing outcomes in patients admitted to the hospital with moderate-risk acute UGIB transfused on a liberal vs a restrictive strategy.1 The restrictive group used a transfusion hemoglobin threshold of 7 g/dL and a posttransfusion target of 7 to 9 g/dL; the liberal group used a threshold of 9 g/dL, with a posttransfusion target of 9 to 11 g/dL. Patients received one unit of red blood cells at a time until their hemoglobin was above the predetermined threshold.
Patients were excluded if they declined blood transfusion; had massive exsanguinating bleeding, acute coronary syndrome, symptomatic peripheral vasculopathy, stroke, lower GI bleeding, or a transient ischemic attack; had received a transfusion within the previous 90 days; or had a recent history of surgery or trauma. Patients at low risk of rebleeding (as defined by the Rockall risk scoring system) were also excluded. Randomization was stratified by the presence or absence of cirrhosis of the liver.
Participants (N=921) had confirmed hematemesis and/or melena on admission. All underwent emergency gastroscopy within 6 hours of admission, with subsequent interventions based on endoscopic findings. In addition to established hemoglobin levels, patients received a transfusion anytime they developed signs or symptoms related to anemia, massive bleeding, or the need for surgery. Staff monitored hemoglobin levels every 8 hours during the first 48 hours, then daily thereafter.
Both groups had similar baseline characteristics, including hemoglobin on admission and source of bleeding. The authors used intention-to-treat analysis to identify the primary outcome: death from any cause at 45 days. Secondary outcomes were further bleeding and in-hospital complications.
During hospitalization, 49% of patients in the restrictive group and 86% of those in the liberal group received a blood transfusion (P<.001). Thirty-two patients (17 from the restrictive group and 15 from the liberal group) withdrew from the study, leaving 889 patients for overall analysis.
At 45 days, overall mortality from any cause was 5% in the restrictive group and 9% in the liberal group (P=.02; number needed to treat [NNT]=25). Sub-group analysis revealed a lower risk of death in patients with cirrhosis and Child-Pugh class A or B disease assigned to the restrictive transfusion group vs the liberal group. The results showed a trend toward a lower risk of death in patients with bleeding from varices or peptic ulcers for the restrictive group, as well.
In addition, the restrictive transfusion group had a significantly lower rate of adverse events (40% vs 48% for the liberal transfusion group; P=.02, NNT=13), with a significant reduction in transfusion reactions (3% vs 9%; P=.001, NNT=17) and cardiac complications (11% vs 16%; P=.04, NNT=20). The restrictive group had a lower rate of further bleeding (10% vs 16% for the liberal transfusion group; P=.01, NNT=17), as well.
