Mind the evidence gap: The need for trials in the prevention and management of perioperative anemia in noncardiac surgery

Red blood cell (RBC) transfusions are the most widely administered inpatient therapeutic intervention in medicine.¹ Transfused RBCs increase oxygen delivery to tissues in situations such as anemia from perioperative bleeding. Each year, 108 million units of blood are transfused globally, with major surgical procedures accounting for approximately 20% of the utilization based on an estimate from the United States.^{2, 3} Internationally, surgical procedures have been estimated to account for a quarter to nearly a half of all RBC transfusions among hospitalized patients.^{4, 5} Although often lifesaving because major bleeding is the leading cause of perioperative death,⁶ blood products have been associated with adverse outcomes, and their use has a significant impact on our health system. Based primarily on observational studies, patients who receive transfusions have longer lengths of hospital stay and higher infection rates, spend more time in intensive care, and have a higher risk of acute respiratory distress.^7-9 Thus, clinicians attempting to balance the perceived benefits and risks of RBC transfusion is likely a factor that has produced considerable practice variation within and across institutions.^{10, 11}

Very little evidence exists in the noncardiac perioperative setting as to the most effective blood management strategies to reduce inappropriate transfusions and improve patient outcomes. Anemia is well tolerated by a majority of patients undergoing surgical procedures; thus the benefits of potentially corrective treatments such as RBC transfusions or pharmacologics need to be weighed against their risks. For noncardiac surgical procedures, recommendations to guide perioperative RBC transfusion practice are limited and nonspecific. For instance, clinical guidelines from AABB recommend a restrictive RBC transfusion threshold of a hemoglobin (Hb) level of 8 g/dL in patients undergoing orthopedic and cardiac procedures but do not define its applicability to other surgical procedures.¹² Thus, surgeons and anesthetists are left to generalize findings from critical care, orthopedic, and cardiac surgery populations demonstrating that restrictive transfusion thresholds are as safe and effective as liberal thresholds. In the absence of robust trial evidence in other noncardiac surgical populations, extrapolation and application of a restrictive strategy seems prudent. Yet, it is not ideal as surgery-specific evidence is necessary to reduce practice variation resulting from a physicianʼs uncertainty or unwillingness to accept evidence from other populations. A recent observational study of more than 6.6 million surgical patients captured in the American College of Surgeonsʼ National Surgical Quality Improvement Program identified 19 noncardiac procedures at very high risk of perioperative transfusion.¹³ Given the incidence of major noncardiac procedures and the risk of perioperative transfusion, thorough and contemporary evaluations of the clinical impact of perioperative anemia with population-based data are warranted.

In this issue of TRANSFUSION, Kerfeld and colleagues¹⁴ from the Mayo Clinic in Rochester, Minnesota, undertook a large retrospective cohort study of more than 3000 non–cardiac surgery patients residing within 120 miles of Rochester. Their objective was to evaluate the association between anemia at hospital discharge and important posthospitalization outcomes using routinely collected and well-validated perioperative, clinical, and administrative data. Based on a patientʼs discharge Hb, the authors categorized anemia as severe, moderate, mild, or none. As expected, the wide range of acceptable Hb levels at discharge reflected the current variation in clinical practice.

Contrary to their hypothesis, the authors found that hospital readmissions, strokes, myocardial infarctions, and mortality rates were not significantly associated with anemia. They did demonstrate higher posthospitalization transfusion rates in patients categorized as severely anemic (Hb <8 g/dL). While it is somewhat surprising that moderate to severe levels of anemia at hospital discharge did not translate into increased morbidity and mortality, the authorsʼ “real-world” observational data are in keeping with results from large pragmatic trials evaluating different levels of anemia tolerance. Contrary to widely held beliefs, the majority of large trials to date do not establish that increased transfusions via a liberal Hb threshold strategy result in higher morbidity and mortality.¹⁵ Importantly, some of these trials were designed and/or analyzed as noninferiority trials to test whether a restrictive threshold strategy is not unacceptably less effective than a liberal strategy; it is not surprising that such trials demonstrate that restrictive transfusion thresholds are no worse than liberal thresholds. Of course, on the basis of economics and preservation of a precious resource, restrictive transfusion strategies should be adopted when robust evidence supports restrictive transfusion strategies.

Observational studies evaluating the impact of transfusion are highly susceptible to bias with a number of sources of selection and information bias at play. Of utmost concern is the issue of confounding by indication whereby sicker patients are more likely to receive a transfusion, and thus an association with a clinical outcome is likely confounded by patient prognosis, physician preference based on prognosis, and physician choices. Bias also arises from the potential for patients to receive multiple transfusions over time and the difficulty in assigning the timing of a related deleterious outcome. Further compounding the issue is the potential heterogeneity of each RBC product including different donors, different manufacturing processes, and different storage times, all of which may influence outcomes. No amount of elaborate analysis can account entirely for these factors. Altogether, these issues highlight the need for rigorous design and analysis in transfusion epidemiology.

Kerfeld and colleagues are to be commended for conducting a thoughtful observational study with a well-defined goal of assessing levels of anemia at discharge in transfused non–cardiac surgery patients. The authors carefully considered and addressed many of the potential sources of bias in their study. They purposely kept the design and respective analysis simple by selecting discharge Hb values in a transfused population as a marker of perioperative transfusion Hb targets. By choosing discharge Hb as their starting point and capturing important baseline risk factors at discharge, they then looked forward in time to capture deleterious outcomes. Despite the simplicity, we need to be conscious of potential biases when interpreting their results, especially in light of the proverbial “black box” of what occurs in the pre-, intra-, and postoperative periods before their chosen starting point of patient discharge. We must also be cognizant of the authorsʼ requirement of the need to receive RBCs to meet their eligibility criteria. Given that patient blood management consists of a number of perioperative interventions and approaches that influence transfusion administration such as pharmacologics, transfusion triggers, blood management algorithms, and other blood products, a focus on those receiving RBC transfusions provides a narrower picture on the clinical impact of anemia at discharge. It also ignores those patients in which good patient blood management practices led to an avoidance of RBC transfusions. Furthermore, their patient population is limited to patients receiving intraoperative RBC transfusions and does not consider those who received postoperative transfusions. Interestingly, even though we would intuitively expect greater baseline acuity in the severe anemia category, the baseline characteristics across the three levels of anemia defined at discharge are relatively well-balanced. One possible explanation is that this group may paradoxically include some of the healthiest patients within the cohort. A proportion of patients in the “severe anemia” category likely included those who could tolerate low intraoperative Hb thresholds but were administered a single RBC unit, appropriately or inappropriately, and went on to have a quick and uneventful recovery. The impact of their inclusion would dilute the effects of observed outcomes. This severe anemia group also included patients with greater acuity and poorer or more complicated prognoses requiring clinically appropriate multiple transfusions. The mixing of patient acuity within anemia level categories may have obscured the “true” clinical impact of intraoperative transfusion. Complicating the issue further is the relatively low incidence of some of the clinical outcomes such as mortality, stroke, and myocardial infarction, leading to wide confidence intervals.

Pragmatic population-based observational studies using robust data sources provide a needed perspective into real-world clinical problems. To this end, Kerfeld and colleagues highlight both the prevalence of postoperative anemia in a transfused non–cardiac surgery patient population and the fact that the level of anemia at discharge did not have an appreciable impact on many major postdischarge outcomes. Their study also underlines the power of integrating routinely collected electronic health and medical data sources rather than relying on the time-consuming process of manual data collecting. The efficiency of utilizing existing health informatics systems should be considered and adopted in clinical trials where appropriate. As the authors note, their findings support further research to evaluate approaches to prevent severe anemia after surgery. More broadly, we would suggest that their findings support the evaluation of acceptable levels of anemia tolerance in non–cardiac surgery populations. Given the inherent limitations of observational studies in transfusion medicine, randomized clinical trials are needed to properly answer such questions.

Despite a burden of more than 300 million surgical procedures and millions of blood transfusions administered each year, there remain significant evidence gaps as to when physicians should transfuse perioperatively.¹⁶ Kerfeld and colleagues have helped identify such a gap in non–cardiac surgery and underscore variability in perioperative transfusion practice even within a single center. Their study supports a call for randomized trials evaluating the clinical impact of perioperative blood management strategies including prophylactic agents and Hb thresholds. Doing so will improve patient outcomes and reduce potentially wasteful and inappropriate variation in perioperative transfusion practice and management.