Bleeding is rarely good for you.

I n a careful statistical analysis of 651 775 patients undergoing elective noncardiac, non-neurological, noncarotid surgery in the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) database between 2005 and 2009, Kamel and colleagues 1 have reaffirmed that bleeding is bad, and that more bleeding is worse, at least in the intraoperative period. Patients who received Ͼ4 units of packed red blood cells or whole blood had an Ϸ2.5-fold increased risk of a subsequent stroke or Q-wave myocardial infarction (MI), independent of a multitude of other risk factors. The relationship between blood loss and postoperative MI or stroke was related to the severity of bleeding: For each unit of intraoperative transfusion with packed red blood cells or whole blood, the risk was reported to increase by Ϸ1.1-fold (95% confidence interval, 1.10 –1.13). In a sensitivity analysis reported in their Discussion section, these results were unchanged when " limited to large intraoperative transfusions in patients without baseline anemia. " Article see p 207 This study must be interpreted in the context of another study from the NSQIP dataset. Glance and colleagues 2 reported that a 1-to 2-unit intraoperative blood transfusion was not significantly associated with cardiac or neurological complications in an analysis limited to 10 100 patients who underwent general, vascular, or orthopedic surgery. The adjusted risk of cardiac complications (odds ratio, 1.31; 95% confidence interval, 0.88 –1.95) was not inconsistent with the findings of Kamel and colleagues. But, in this report, the risk of central nervous system complications was actually insignificantly lower (odds ratio, 0.68; 95% confidence interval, 0.34 –1.38) in patients who received 1 to 2 units of intraop-erative blood. These 2 studies, put together, indicate that a lot of bleeding is bad and a little bit is probably not good. The authors acknowledge some of the limitations of their study design. The database that they searched had no information regarding the perioperative use of antiplatelet medications. The database could not distinguish hemorrhagic as compared with ischemic strokes, although most perioperative strokes are thought to be ischemic. 3 More importantly, the NSQIP database also lacks information about the intraoperative courses of the patients in it. Even if bleeding is not good, it is unclear whether the 2.5-fold increase risk of complications related to major intraoperative bleeding is more a function of the associated hypotension than of a specific reduction in oxygen-carrying capacity. For example, data from decades …