Blood Transfusions in UGI Bleeding



Transfusion thresholds in medicine have been steadily falling, as more and more literature demonstrates either no improvement in outcomes or associated harms.  


Today’s mini-JC focuses on a topic extremely relevant to our practice as emergency physicians: transfusion thresholds of PRBC’s in patients with acute upper gastrointestinal bleeding (UGIB). If the patient is exsanguinating, we all know what to do, but what about in hemodynamically stable patients with an upper GI bleed?


1.  Why is this topic important?


Several non-randomized trials, and animals studies have evaluated a restrictive approach to blood transfusion in patients with bleeding from portal hypertension, as increases in effective circulating volume may lead to a rebound in portal pressure, which is associated with a risk of re-bleeding.


In addition, trauma surgery has propagated the idea of “don’t pop the clot” in patients who have penetrating trauma in order to prevent rebleeding and coagulopathy after surgery. These same mechanisms have been suggested to be beneficial in patients with arterial hemorrhage from a perforated peptic ulcer.



2.  What does this study attempt to show?


This study attempts to demonstrate that in patients with UGIB without massive exsanguination, a restrictive transfusion threshold of 7 is non-inferior to a transfusion threshold of 9. Patients were eligible if they had hematemesis, (or bloody nasogastric aspirate), melena or both as confirmed by hospital staff.


They attempted to exclude critically ill patients as well as completely stable patients as defined below:


-Patients were excluded if they had a lower GI bleed, active acute coronary syndrome (new EKG findings or positive biomarkers), massive exsanguination (hypotension, with active bleeding requiring emergent transfusion), or acute CVA.


-Patients with rockall score of 0, or a hemoglobin of 12 or greater were also excluded as these patients represented the lowest risk group.


Patients were randomized to a restrictive (transfuse if hgb <7, goal hgb 7-9) or liberal (transfuse if hgb <9, goal 9-11) transfusion groups. Randomization also was stratified for presence or absence of cirrhosis in blocks of four.


Treatment was otherwise kept the same between the two groups (protonix bolus and drip, emergency EGD within 6 hours with banding or sclerotherapy as necessary, octreotide and antibiotics if the bleed was deemed to be variceal.)


3.  What are the essential findings?


A total of 921 patients underwent randomization and 32 were withdrawn or withdrawn by investigators. 444 patients were left in the restrictive-strategy group and 445 in the liberal-transfusion group. Baseline characteristics were similar in both groups, as described in table 1.


225 (51%) of patients in the liberal group got transfusions, as compared with 61 (14%) in the restrictive group.


Mortality at 45 days was significantly lower in the restrictive-strategy group than in the liberal group 5% (23 patients) as compared with 9% (41 patients) hazard ratio 0.55 (0.33-0.92) respectively, p < 0.05.  


Death was due to unsuccessfully controlled bleeding in 3 patients in the restrictive strategy group (0.7%), and 14 patients in the liberal strategy group (3.1%).


Overall the liberal strategy group got more blood, had more rebleeding and thus required more salvage procedures especially in patients with variceal bleeding (TIPS and balloon tamponade). There were more medical associated deaths (sepsis, acute MI, acute CVA etc.) in patients who received the liberal transfusion strategies.  



4.  How is patient care impacted?


Among patients with severe acute upper gastrointestinal hemorrhage a restrictive strategy of blood transfusion improved mortality at 45 days. This survival benefit was likely conferred due to better control of factors contributing to death, such as further bleeding, rescue therapy and adverse events.  


5.  Is this an area of controversy?


Although this is an area of controversy, the evidence that blood transfusions, especially older blood in which the oxygen carrying capacity has diminished, may result in increased complications has been suggested previously. The TRICC trial, the Houston study, and several other large randomized controlled trials have all shown that patients have improved or equivalent outcomes when using a restrictive transfusion strategy.  The mechanisms for this difference in mortality are certainly different, as we know that transfusions may propagate further bleeding in patients at risk of hemorrhage. In addition, it is known that blood transfusions can result in significant immunomodulation and cytokine release resulting in acute renal failure, pulmonary edema, ARDS, and higher infectious complications in the critically ill medical patient.


This must be balanced in the situation in which acute hemorrhage is life threatening and rapid transfusion is required to restore the effective circulating blood volume.  


6.  Major Limitations of the study?


The study is a well completed randomized controlled trial. However, several limitations still exist. First, there is no direct statement as to what was defined as a life-threatening hemorrhage that excluded patients from the trial, and therefore knowing how to apply this to our patients is somewhat confusing. In addition, 32 patients were withdrawn from the study because the investigator did not believe that there was clinical equipoise. They do not further specify this in the intention to treat analysis, and how many patients from each group were removed (if all 32 patients from the restrictive strategy group were removed because the attending physicians believed that they required blood transfusions this could have a significant effect on the statistically significant mortality benefit shown in this paper.)


However, despite some limitations in the study there are several important concepts:

1.    Transfusing patients based on hemoglobin levels only may be short sighted. Looking at perfusion parameters (lactate, oxygen delivery, end organ malperfusion) may be a more appropriate approach rather than transfusing to a number (MUST KEEP KGB > 9).  

2.    Giving a blood product should be viewed as giving a potentially dangerous medication (not dissimilar to heparin, Plavix etc.) and each time you order it you should consider the risks and the benefits. It has been demonstrated time and time again that blood products have associated harms (certainly I am not arguing against transfusing the actively bleeding patient with obvious signs of malperfusion), especially in the elderly patients with multiple comorbidities or patients with cancer.




Villanueva et al. Transfusions Strategies for Acute Upper gastrointestinal Bleeding. NEJM. 2013; 368:11-21.