Association of Postoperative Transfusions With Adverse Outcomes After Noncardiac Surgery
With great interest we read the recent article by Abdelsattar et al1 identifying the patient-level effects of blood transfusion on postoperative outcomes and estimating the effects of different transfusion practices on hospital-level risk-adjusted outcomes in patients undergoing noncardiac surgery. They showed that patients receiving a postoperative transfusion had significantly increased risks for 30-day mortality, any morbidity, and infectious complications, but were less at risk for postoperative myocardial infarction. By matching patients by propensity scores generated from more than 60 variables, the authors had attempted to control the effects of potential confounders on study endpoints. Furthermore, the authors openly discuss the limitations of their work. In our view, the 2 important issues in this study seemed not to be well addressed.
First, most patients had a history of preoperative cardiovascular diseases, such as hypertension, congestive heart failure, arrhythmias, and peripheral vascular disease. Did the dataset from which Abdelsattar et al retrieved information for analysis include perioperative cardiac medications of patients? If it does, perioperative cardiac medications should be used as adjusted confounders’ propensity score-matching model. It has been shown that perioperative aspirin, angiotensin-converting enzyme inhibitors, β-blockers, and statins are significantly associated with reduced short-term morbidity and mortality after noncardiac surgery.2–4 Furthermore, a combination of β-blockers with statins may offer additional prevention for perioperative death.5
Second, we were not provided with detail of anesthesia and intraoperative managements. Consequently, it is difficult to estimate the extent to which interventions by anesthesiologists might have influenced outcomes. The recent evidence shows that anesthetic induction with etomidate is associated with increased 30-day mortality and cardiovascular morbidity after noncardiac surgery,6 whereas intraoperative nitrous oxide administration is associated with decreased risk of 30-day mortality and decreased risks of inhospital mortality/morbidity.7 Actually, it has been shown that even short durations of an intraoperative MAP below 55 mm Hg can result in postoperative myocardial and kidney injury, with an independent graded relationship between duration of intraoperative hypotension and postoperative myocardial injury and kidney injury.8 Similarly, intraoperative hypoxemia, hypotension, and tachycardia have been associated independently with short-term morbidity and mortality after noncardiac surgery.9
It must be emphasized that although propensity score-matching model is very useful for controlling the patients’ baseline characteristic and selection biases in a retrospective study, it carries the significant assumption of no unmeasured confounders.10 That is, all important known factors that affect intervention selection and measured outcomes must be measured and taken into account within the propensity score-matching model. Thus, we believe that addressing the above factors would improve the transparency of this study.