PREOPERATIVE anemia and perioperative allogeneic erythrocyte (RBC) transfusion have both been shown to be independently associated with adverse outcome, such as increased mortality, postoperative infections, lung and kidney injury, myocardial infarction, stroke, and increased length of hospital stay.1–6The relative contribution of each of these conditions, and their interactions, is difficult to specify. The study by Glance et al . in this issue of Anesthesiology is important, because the authors studied more than 10,000 patients with preoperative anemia who underwent elective noncardiac surgery and received up to 2 units of RBCs intraoperatively. The study used records from the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) database.7
In an attempt to reduce confounders, the investigators included only patients within a preoperative hemoglobin range of 6–10 g/dl. In this hemoglobin range, according to American Society of Anesthesiologists guidelines, RBC transfusions may be administered on a clinical basis.8Thus, these RBC transfusions had likely occurred after limited blood loss. Observed outcomes thus may be attributed more clearly to RBC transfusion than if patients had started with normal preoperative hemoglobin and, as a consequence, had to have lost large amounts of blood during ill-controlled surgery before being transfused. In the latter case, it would be difficult to say whether the adverse outcomes were due to hemodynamic instability and other effects of acute blood loss or due to the RBC transfusion itself. Since in the study by Glance et al .7all patients were anemic preoperatively and may not have experienced massive blood loss before the index transfusion, the observed outcomes may be attributed quite clearly to the RBC transfusion administered. Also, the negative impact on outcome was substantial: Transfusion of only 1 or 2 units of RBCs resulted in more septic (odds ratio [OR]: 1.43), pulmonary (OR: 1.79), thromboembolic (OR: 1.77), and wound (OR: 1.87) complications and increased mortality (OR: 1.29).
The key strengths of this study lie in the number of the patients and the detailed data collection; the shortcomings are shared with many other transfusion cohort studies. Despite various efforts, it is always difficult to call transfusion as the cause of an adverse outcome, as opposed to some other associated patient comorbidity. A brief look at table 1 of the Glance et al . article indicates that the transfused cohort was generally in worse shape at baseline, compared with the not-transfused group (lower hematocrit, older age, more dependent, more comorbidities, etc.). Statistical methods are used to remedy this commonly seen disparity. Multivariate analysis can be used to adjust for confounders (the method used in the current study), or transfused patients can be matched against nontransfused patients (with varying degrees of complexity). Although these approaches often do marvels, they are not “cures,” because they are ultimately limited by the available variables and their data. Do we have all the potential confounders captured in our dataset? How certain are we of the primary reasons behind giving transfusion to every single case? What was the tipping point that made the physician decide to give blood to this patient, but not the other patient? These questions are often impossible to answer accurately in cohort studies, and they undermine the validity of the results.
Although the multivariate adjustment used in this study is a valid method, more data on the models would have been helpful (e .g ., strength of association as indicated by pseudo-R-squared or the reasoning behind creating the specific hematocrit categories). In addition, use of a matching approach may have allowed for more transparent interpretation of the results (at expense of reduced sample size). Given the large sample size, it would have been interesting to see whether the transfused and not-transfused patients could have been effectively matched to each other based on the risk factors for receiving transfusions (and, possibly, known risk factors of each outcome), thus allowing a more direct comparison.
In addition, large database analyses do have limitations of their own. Problems with miscoding of procedures and diagnoses are well recognized.9This study by Glance et al ., however, relies on the NSQIP database, which is more accurate than other state or federal administrative datasets. Data are collected by trained surgical chart reviewers, and participating hospitals undergo interrater reliability audits.10∥ Still, for any particular study, a generalized database such as NSQIP will not provide all the information that one might wish for.
In the case of the present study, we do not know the extent of pre- or postoperative transfusion in the study patients; because of NSQIP data definitions, patients could have received up to 4 units preoperatively and 4 units postoperatively. Transfusion before or after surgery, but not during, might decrease the detected effect of intraoperative transfusion; alternatively, those patients requiring pre- or postoperative transfusion might be more likely to receive blood intraoperatively. NSQIP does not record intraoperative nadir hematocrit or immediate postoperative hematocrit. Thus, we do not know if lower intraoperative hematocrit was associated with worse outcome. Because NSQIP data do not include a hospital identifier, the final results may have been confounded by hospital-based differences in practice, including differences in local transfusion triggers.10
Results of the present study are in line with previous studies in general,3cardiac,1orthopedic,4and trauma surgery.5,6Also, in a recent cohort data analysis in 239,286 elderly Veterans Administration patients with and without preoperative anemia RBC transfusions resulted in increased 30-day mortality (adjusted OR: 1.37 [1.27–1.48]).11Only in a 1-min subgroup of patients with a preoperative hematocrit of less than or equal to 24% (1.9% of the study population) did RBC transfusion result in a reduced mortality, whereas in patients with a preoperative hematocrit of greater than 30% (more than 80% of the study population), RBC transfusions actually increased mortality significantly (ORs: 1.37–1.59). Interestingly, in patients who eventually died, late postoperative complications, such as failure to wean from the ventilator and unplanned intubation indicative of lung injury, renal failure, and infectious complications, were very similar to the ones found in the article by Glance et al .,7and they were more frequent than in nontransfused patients.11
Another aspect of the study by Glance et al .7deserves mentioning, namely the fact that nonelective cases were excluded. Therefore, it should have been possible to treat preexisting anemia before surgery. This is important, because preoperative anemia is frequent,2,11–13easy to detect, and—in many situations—treatable at relatively low costs.14At least, in orthopedic surgery, there is strong evidence that preoperative treatment of anemia results not only in reduced RBC transfusion, but also in improved outcome.15
Where do we go from here? The call for prospective, randomized studies on the benefit or detriment of allogeneic RBC transfusions is the usual conclusion after studies of the likes of Glance et al . are published. This conclusion is certainly politically correct; however, if Glance et al. and the slew of others are correct—which is a valid possibility—can we take the responsibility to continue today's widespread transfusion practice, an approach that is putting patients at risk?
Contrarians to the above still overly rely on banked blood products, despite mounting evidence that patient blood management is achievable, avoids exposure to allogeneic blood products, improves patient outcome, and saves resources.15,16The ease of ignoring preoperative anemia as well as the simplicity of ordering and transfusing RBCs and thereby completely ignoring the burgeoning evidence of adverse transfusion outcome are not in the best interest of patients. Maintaining the clinical status quo under such circumstances would not be accepted in any other field of medicine in the context of current safety and quality standards.
Working on studies regarding benefit or detriment of allogeneic RBC transfusions is important, but more so is acting already today according to patient blood management principles; it is time for a change toward better patient care.
* Institute of Anesthesiology and Head Medical, Section Anesthesiology, Intensive Care Medicine and OR-Management, University Hospital Zurich, Zurich, Switzerland. donat.spahn@usz.ch. † Department of Anesthesiology and Critical Care and Hyperbaric Medicine and Institute for Patient Blood Management and Bloodless Medicine and Surgery, Englewood Hospital and Medical Center, Englewood, New Jersey. ‡ Medical Society for Blood Management, Laxenburg, Austria, and Center for Population Health Research, Curtin University, Western Australia. § Department of Anesthesiology, Columbia University College of Physicians and Surgeons, New York, New York.
