Use of Restrictive Transfusion Threshold May Reduce Mortality Compared to Liberal Threshold in Patients with Upper GI Bleeding
EBM Focus - Volume 8, Issue 3
Blood transfusions are common in patients hospitalized with acute upper gastrointestinal bleeding, but the optimal hemoglobin threshold for administering blood products remains unclear. For critically ill patients in general, use of lower, more restrictive transfusion thresholds is associated with reduced transfusion rates and decreased in-hospital mortality (but not longer-term mortality) compared to more liberal thresholds (Cochrane Database Syst Rev 2012 Apr 18;(4):CD002042). Current guidelines recommend the use of a restrictive threshold (7 g/dL) in patients with upper GI bleeding (Am J Gastroenterol 2012 Mar;107(3):345, Ann Intern Med 2010 Jan 19;152(2):101), but little of the evidence supporting these guidelines comes from studies in this specific population.
A recent randomized trial compared restrictive vs. liberal transfusion protocols in 921 adults with acute upper GI bleeding (49% with peptic ulcer, 21% with esophageal varices). Patients were randomized to have blood transfusion when hemoglobin reached < 7 g/dL with a target range 7-9 g/dL (restrictive strategy) or when hemoglobin reached < 9 g/dL with target range 9-11 g/dL (liberal strategy). Patients reaching the threshold initially received 1 unit of red blood cells and had a second unit if hemoglobin fell below the target range. Exclusion criteria included exsanguinating bleeding and low risk of rebleeding.
A total of 889 patients (96.5%) were included in the analysis following exclusions for ineligibility, major protocol violations, or withdrawal. The rates of transfusion were 49% with the restrictive strategy and 85% with the liberal strategy (p < 0.001). All-cause mortality at 45 days was significantly reduced in the restrictive group (5% vs. 9%, p = 0.02, NNT 25) (level 2 [mid-level] evidence). The restrictive threshold was also associated with a lower rate of death due to uncontrolled bleeding (0.7% vs. 3.1%, p = 0.01, NNT 42) and of severe adverse events (12% vs. 18%, p = 0.01, NNT 17) (N Engl J Med 2013 Jan 3;368(1):11).