ROCHESTER, N.Y. -- Filtering WBCs from donated blood may reduce postoperative infections by 10% among transfused surgical patients, according to a meta-analysis.
ROCHESTER, N.Y., April 9 -- Filtering WBCs from donated blood may reduce postoperative infections by 10% among transfused surgical patients, according to a meta-analysis.
The meta-analyses excluded data from patients who never received a transfusion, a confounder that clouded previous meta-analyses of the benefits of leukoreduction, said Neil Blumberg, M.D., of the University of Rochester Medical Center here, and colleagues.
Minus patients who never received transfusions, the current meta-analysis revealed about a 50% reduction in the odds of postoperative infection from leukoreduced blood versus standard transfusions, Dr. Blumberg and colleagues reported in the April issue of the journal Transfusion.
Leukoreduction may reduce the occurrence of infections carried in white blood cells, such as human T-cell leukemia virus, cytomegalovirus, herpesviruses, and Epstein-Barr virus, they said. They speculated that leukoreduced blood could also reduce impaired immunity that follows a transfusion, which could contribute to post-surgical infection rates.
Leukoreduction is done prior to storage for donated blood in Canadian blood banks, but is not recommended by the FDA nor widely adopted in the United States.
The meta-analysis included nine randomized trials and 11 meta-analyses comparing pre- or post-storage leukoreduced transfusions with transfusions of unmodified red cells, whole blood, buffy coat-depleted red blood cells (60% to 80% leukoreduced), or plasma-reduced red blood cells. All had post-operative infections as a primary or secondary outcome. Surgery was predominantly colorectal or cardiac.
More than half of the studies did intent-to-treat analyses, which resulted in 35% of the overall patients contributing infection data without having ever received a transfusion. Although excluding these untransfused patients would have diluted the statistical power of any single study, including them would have significantly reduced the power of the meta-analysis to detect a difference between treatments, the researchers said.
"Because it is unlikely that randomization assignment in the transfusion service affects the likelihood of transfusion in the operating room, excluding patients who do not receive a transfusion will have little or no effect on the integrity and robustness of the randomization process," they added.
So, randomized but untreated patients were excluded with nearly equal numbers excluded in control and leukoreduced arms. After adding unpublished data from the authors of the original trials, a total of 3,093 transfused patients were analyzed.
The findings were:
The difference in postoperative infection risk between leukoreduced to partial leukoreduced (buffy coat-removed red blood cells) transfusions was less than the difference between leukoreduced and unmodified red blood cells or whole blood transfusions (OR 0.668, 95% CI 0.416 to 1.074, versus 0.425, 95% CI 0.257 to 0.703).
This finding suggested a dose response between degree of leukoreduction and postoperative infection, Dr. Blumberg and colleagues said.
Heterogeneity between studies was most likely due to variation in postoperative infection rates at different centers, the two types of surgery included, and the inclusion of two multicenter studies amidst seven single- or two-center studies, they added.
All but one of the previous meta-analyses concluded that there was little or no evidence for a treatment benefit from leukoreduced transfusions in surgery. However, the researchers noted that they almost all used intent-to-treat data even when the original studies did not.
"These results demonstrate the importance of including only scientifically valid data in clinical trials and meta-analyses," they wrote. "The intention-to-treat principle should never lead to inclusion of data not actually derived from experimental results."
In clinical trials of blood transfusion, blinding treating and evaluating personnel to allocation is particularly challenging, they noted. Randomization at the time of transfusion is usually impractical due to logistics. "Innovative new approaches are needed to allow lower rates of protocol violations in transfusion studies, and randomization closer to the time transfusion is required," they concluded.
The studies included in the meta-analysis only included surgical patients. The infection risk may differ for cancer patients or those who receive transfusions in other settings.