Blood transfusion harmful in some patients with GI bleeds (RCT)

Blood Transfusion: Deadly for GI Bleeds?

You read the headline right: in a randomized trial published in the New England Journal of Medicine, liberal blood transfusions (to a hemoglobin of 9 g/dL) seemed to harm people with acute upper gastrointestinal bleeding, as compared to transfusing when hemoglobin fell below 7 g/dL. People with mild to moderate cirrhosis with variceal bleeding were at increased risk of death by blood transfusion, while the risks or benefits of transfusion for people with bleeding gastric ulcers or other novariceal bleeds was uncertain.

What They Did

Investigators randomized 921 patients with acute GI bleeding (as evidenced by hematemesis and/or melena) to undergo either a restrictive blood transfusion strategy (to hemoglobin >7 g/dL) or a liberal strategy (to Hb > 9 g/dL). This was done at a single center, the Hospital de la Santa Creu i Sant Pau in Barcelona, Spain.

Patients were stratified at randomization according to whether or not they had cirrhosis -- important for their post hoc subgroup analysis (see below).

Excluded were patients who had undergone recent surgery or who were exsanguinating; acute coronary syndrome or stroke; recent blood transfusions, or concomitant lower GI bleeding.

• All patients got an upper endoscopy within 6 hours with standard endoscopic treatment of varices and ulcers; prophylactic antibiotics and somatostatin were given to people with varices.

• Units of packed red blood cells were transfused one at a time, with hemoglobin checks in between.

• The packed red blood cells were an average of 15 days old in both groups of patients; read further for why this might matter.

The primary outcome was death at 45 days from any cause.

What They Found

At 45 days, 23 people (5%) with acute GI bleeding in the restrictive blood transfusion group had died, and 41 people (9%) had died in the liberal blood transfusion group. Those in the restrictive group -- who received blood transfusions only for hemoglobin of 6.9 and below -- had a 45% relative reduction, and a 4% absolute risk reduction for death (5% vs. 9%, p=0.02).

The risk reduction was highly concentrated in the subgroup of patients with Childs Class A or B cirrhosis (a relative risk of 0.30, statistically significant); those with nonvariceal bleeding had a non-stat.significant relative risk of 0.70, and those with Class C (advanced) cirrhosis had a RR of 1.04.

Those in the restricted-blood transfusion group also had less re-bleeding; this signal was also concentrated in the Childs Class A and B cirrhosis subgroup, with equivocal signals in the other subgroups.

Half of the patients in the restrictive-transfusion strategy did not receive any blood at all during their clinical course for acute GI bleeding; the average number of units transfused was 1.5 vs. 3.7.

What It Means

Blood transfusion practice has always been guided above all by tradition, emotion, and whim of the treating physician. By proving that transfusing less blood could be safe, and even beneficial for critically ill patients, Hebert et al's landmark 1999 TRICC trial began a slow sea change toward a more rational and judicious approach to blood transfusions by intensivists. In 2011, the American Association of Blood Blanks recommended that a restrictive approach be extended to less-ill patients on the hospital ward, as well. (Isn't "restrictive" inappropriately named? If less blood is better, nothing is being restricted; rather, excessive blood transfusions are being avoided).

For acute variceal bleeding, gastroenterologists have long been advising a conservative transfusion practice, and this study supports that. The harmful effects of blood here may have been hydrostatic: the most-affected group (Childs A and B cirrhosis) had a high risk of both rebleeding and mortality when transfused liberally, and had measurable excessive increases in their portal pressures.

There are important caveats to generalizing these findings to people with nonvariceal upper GI bleeding. This was a single center trial; single-center trials consistently show larger effects than multicenter trials. Any benefit among the nonvariceal subgroup was unclear, just a trend with a wide confidence interval.

Whether the age of stored donated blood prior to transfusion is responsible for the apparent harm from some blood transfusions is an area of ongoing research. The blood here was 15 days old. Damage to stored red blood cells become apparent by 14 days, but federal regulations allow blood to be stored for up to 42 days. The average age of transfused blood in the U.S. is about 17 days, but it varies widely between institutions according to how often they transfuse blood. High-transfusing institutions (like with trauma centers) get the oldest blood from the Red Cross, which may seem paradoxical, but it's because they're most likely to use it and therefore the least likely to waste it. The INFORM trial did not show a difference in outcome according to the age of transfused blood.

Clinical Takeaway: Patients with moderate cirrhosis and variceal bleeding might be hurt by aggressive blood transfusion. An increasing body of data suggests that rational blood transfusions to low thresholds (7 g/dL) might be best for the large majority of patients, despite the anxiety this practice provokes in medical care teams. If larger trials agree with this one, that less-blood-is-better category could expand to include patients with active upper GI bleeding from any source. At which time a new motto for physicians might be: "Save a Life: Don't Give Blood."

Read more:

Càndid Villanueva, et al. Transfusion Strategies for Acute Upper Gastrointestinal Bleeding. NEJM 2013; 368:11-21.