Gastrointestinal (GI) bleeding accounts for more than 450,000 hospitalisations annually in the US and is a frequent indication for red-cell transfusion. Blood transfusions are given to 43% of patients hospitalised with upper GI bleeding in the UK and to 21% of patients hospitalised with lower GI bleeding in the US. Transfusion practices for patients with GI bleeding have fluctuated over the past 100 years. Avoidance of transfusions early in the 20th century, owing to concern that increased blood pressure would induce rebleeding, gave way to more liberal use of transfusions and a haemoglobin threshold for transfusion of 10g per decilitre was recommended up to the early 2000s.

On the basis of more recent data, current guidelines for the management of GI bleeding have returned to a restrictive transfusion strategy, recommending a haemoglobin threshold of 7g per decilitre. Meta-analyses of randomised trials of restrictive transfusion thresholds as compared with liberal transfusion thresholds show no significant differences in 30-day mortality, length of hospital stay or rates of adverse events, and largely exclude the possibility of a clinical benefit with a liberal transfusion strategy. However, only 0-1% of the patients in these analyses had acute GI bleeding, which raises concerns about the generalisability of these results to patients with GI bleeding.

A recent study by Villanueva C et al in the New England Journal of Medicine1 provides evidence to guide practice and justify current recommendations for the management of upper gastrointestinal bleeding. A haemoglobin threshold for transfusion of 7g per decilitre, as compared with a threshold of 9g per decilitre, was associated with a significant 45% relative-risk reduction in 45-day mortality. 

On the basis of the results of this study, 25 patients would have to be treated according to a restrictive transfusion strategy rather than a liberal transfusion strategy to avert one additional death at 45 days. The decrease in mortality was accounted for primarily by fewer deaths from bleeding that could not be successfully controlled. Significant reductions with the restrictive strategy were also seen in the rates of further bleeding, transfusion reactions and cardiac events and in the length of hospital stay.

Although these results apply to a broad group of patients with upper gastrointestinal bleeding, modification of the transfusion threshold may be considered in specific subpopulations, such as patients with hypotension due to severe bleeding and patients with cardiovascular disease. Haemoglobin values early in the course of acute bleeding are minimally decreased and, in patients with substantial intravascular volume depletion, markedly overestimate the ‘true’ haemoglobin level that will be seen after fluid resuscitation and equilibration. 

In conclusion, the study by Villanueva C et al, provides important evidence to guide clinical practice. Most patients with upper GI bleeding, with or without portal hypertension, should have blood transfusions withheld until the haemoglobin level drops below 7g per decilitre.