The article by Bower is yet another observational study assessing the relationship between transfusion and serious morbidity and/or mortality in patients undergoing non-cardiac surgery. A single-center academic hospital experience of patients undergoing colorectal, head and neck, liver, thyroid, and vascular surgeries from 2006 to 2009 is examined to demonstrate an independent negative effect of transfusion on length hospital stay, morbidity, wound infection, postoperative complications, and mortality. The methods and analyses of this article are either inadequate or not transparent. For instance, definitions of exposure (blood transfusion) and outcome variables are not provided nor the timing of the events (before or after transfusion). The number of matched pairs that were used in the propensity analyses along with their baseline characteristics is not provided and they are essential to understanding the results. Weak or non-transparent methods and results aside, Table 1 tells the most important story. Namely, transfused patients are very different from non-transfused patients. Of course, this is intuitive as the sicker you are, the greater the probability of requiring a transfusion. This is why we need large randomized trials to assess the independent effect of transfusions (i.e. transfusion triggers) on clinically important outcomes. The study by Bower and colleagues is of limited clinical importance due to its serious flaws in methods and analysis and overinterpretation of results.
– Dean A. Fergusson