Veenith and colleagues from Cambridge, UK, describe perioperative risk factors for all-cause mortality (primary outcome), length of stay in intensive care unit and total hospital stay (secondary outcomes) in 874 patients 80 years of age or older undergoing cardiac surgery. After controlling for baseline patient characteristics and intraoperative factors, transfusion of greater than 2 units of red blood cells was associated with 8.47 fold higher risk of death and the use of FFP, platelets or cryoprecipitate was associated with 23.5 fold higher risk of death.
There are many observational studies in the literature showing that blood transfusion is harmful. Most have controlled for differences in prognostic clinical factors using regression analysis. The issue is always the same; can regression analysis control for confounding when evaluating effect of transfusion? In my view the answer is no because blood transfusion itself is a marker for severity of illness (Carson JL et al. Crit Care Med 2008;36:2707-8). Transfusion is given to the sickest patients and will always appear harmful. Only a randomized clinical trial will provide an unbiased assessment.
– Jeffrey L. Carson