We thank Dr Lin for his comments and the opportunity to emphasise important aspects of our recently published analysis.[2, 3] In fact, we indeed performed thorough multivariable risk adjustment for all of factors listed by Dr Lin. We adjusted for blood on rectal examination or in the nasogastric tube aspirate, ASA score, co-morbidities, haemodynamic instability, initial haemoglobin, use of fresh frozen plasma, red blood cell transfusion, presence of high risk endoscopic stigmata, performance of endoscopic therapy and proton pump inhibitor use.
We included all parameters found in the Blatchford scale except for urea, as the data used in this analysis were also used to validate a modified Blatchford scale that excludes urea. Our results, and those of a large UK audit, suggest that transfusion practice, and more specifically blood administration in the first 24 h, may independently predict poorer outcomes.
This observation was recently validated in a large randomized trial of patients with upper gastrointestinal bleeding, and more specifically amongst class A and B cirrhotic patients. Unfortunately, issues of generalisability in the aforementioned trial prevent definitive conclusions about the impact of transfusion policy on the outcomes of patients with nonvariceal bleeding; importantly, additional information will soon be available from a cluster randomized trial in the United Kingdom.
In the interim, based on all currently available evidence, clinicians should continue to follow consensus recommendations for patients with nonvariceal bleeding. We would also like to correct Dr Lin regarding his comments about current recommendations. The 2010 International Consensus Group in fact suggested a restrictive transfusion practice, with blood transfusions administered to patients with a haemoglobin level of 7 g/dL or less, pending further data. Furthermore, threshold haemoglobin levels of 6–10 g/dL may warrant transfusion in patients with underlying cardiac disease.