For decades, physicians believed in the benefit of prompt transfusion of blood to keep the hemoglobin level at arbitrary, optimum levels, ie, close to normal values, especially in the critically ill, the elderly, and those with coronary syndromes, stroke, or renal failure.
However, the evidence supporting arbitrary hemoglobin values as an indication for transfusion was weak or nonexistent. Also, blood transfusion can have complications and adverse effects, and blood is costly and scarce. These considerations prompted research into when blood transfusion should be considered, and recommendations that it should be used more sparingly than in the past.
This review offers a perspective on the evidence supporting restrictive blood use. First, we focus on hemodilution studies that demonstrated that humans can tolerate anemia. Then, we look at studies that compared a restrictive transfusion strategy with a liberal one in patients with critical illness and active bleeding. We conclude with current recommendations for blood transfusion.
EVIDENCE FROM HEMODILUTION STUDIES
Hemoglobin is essential for tissue oxygenation, but the serum hemoglobin concentration is just one of several factors involved.1–5 In anemia, the body can adapt not only by increasing production of red blood cells, but also by:
- Increasing cardiac output
- Increasing synthesis of 2,3-diphosphoglycerate (2,3-DPG), with a consequent shift in the oxyhemoglobin dissociation curve to the right, allowing enhanced release of oxygen at the tissue level
- Moving more carbon dioxide into the blood (the Bohr effect), which decreases pH and also shifts the dissociation curve to the right.
Just 20 years ago, physicians were using arbitrary cutoffs such as hemoglobin 10 g/dL or hematocrit 30% as indications for blood transfusion, without reasonable evidence to support these values. Not until acute normovolemic hemodilution studies were performed were we able to progressively appraise how well patients could tolerate lower levels of hemoglobin without significant adverse outcomes.
Acute normovolemic hemodilution involves withdrawing blood and replacing it with crystalloid or colloid solution to maintain the volume.6
Initial studies were done in animals and focused on the safety of acute anemia regarding splanchnic perfusion. Subsequently, studies proved that healthy, elderly, and stable cardiac patients can tolerate acute anemia with normal cardiovascular response. The targets in these studies were modest at first, but researchers aimed progressively for more aggressive hemodilution with lower hemoglobin targets and demonstrated that the body can tolerate and adapt to more severe anemia.6–8
Studies in healthy patients
Weiskopf et al9 assessed the effect of severe anemia in 32 conscious healthy patients (11 presurgical patients and 21 volunteers not undergoing surgery) by performing acute normovolemic hemodilution with 5% human albumin, autologous plasma, or both, with a target hemoglobin level of 5 g/dL. The process was done gradually, obtaining aliquots of blood of 500 to 900 mL. Cardiac index increased, along with a mild increase in oxygen consumption with no increase in plasma lactate levels, suggesting that in conscious healthy patients, tissue oxygenation remains adequate even in severe anemia.
Leung et al10 addressed the electrocardiographic changes that occur with severe anemia (hemoglobin 5 g/dL) in 55 healthy volunteers. Three developed transient, reversible ST-segment depression, which was associated with a higher heart rate than in the volunteers with no electrocardiographic changes; however, the changes were reversible and asymptomatic, and thus were considered physiologic and benign.
Hemodilution in healthy elderly patients
Spahn et al11 performed 6 and 12 mL/kg isovolemic exchange of blood for 6% hydroxyethyl starch in 20 patients older than 65 years (mean age 76, range 65–88) without underlying coronary disease.
The patients’ mean hemoglobin level decreased from 11.6 g/dL to 8.8 g/dL. Their cardiac index and oxygen extraction values increased adequately, with stable oxygen consumption during hemodilution. There were no electrocardiographic signs of ischemia.
Hemodilution in coronary artery disease
Spahn et al12 performed hemodilution studies in 60 patients (ages 35–81) with coronary artery disease managed chronically with beta-blockers who were scheduled for coronary artery bypass graft surgery. Hemodilution was performed with 6- and 12-mL/kg isovolemic exchange of blood for 6% hydroxyethyl starch maintaining normovolemia and stable filling pressures. Hemoglobin levels decreased from 12.6 g/dL to 9.9 g/dL. The hemodilution process was done before the revascularization. The authors monitored hemodynamic variables, ST-segment deviation, and oxygen consumption before and after each hemodilution.
There was a compensatory increase in cardiac index and oxygen extraction with consequent stable oxygen consumption. These changes were independent of patient age or left ventricular function. In addition, there were no electrocardiographic signs of ischemia.
Licker et al13 studied the hemodynamic effect of preoperative hemodilution in 50 patients with coronary artery disease undergoing coronary artery bypass graft surgery, performing transesophageal echocardiography before and after hemodilution. The patients underwent isovolemic exchange with iso-oncotic starch to target a hematocrit of 28%.
Acute normovolemic hemodilution triggered an increase in cardiac stroke volume, which had a direct correlation with an increase in the central venous pressure and the left ventricular end-diastolic area. No signs of ischemia were seen in these patients on electrocardiography or echocardiography (eg, left ventricular wall-motion abnormalities).
Hemodilution in mitral regurgitation
Spahn et al14 performed acute isovolemic hemodilution with 6% hydroxyethyl starch in 20 patients with mitral regurgitation. The cardiac filling pressures were stable before and after hemodilution; the mean hemoglobin value decreased from 13 to 10.3 g/dL. The cardiac index and oxygen extraction increased proportionally, with stable oxygen consumption; these findings were the same regardless of whether the patient was in normal sinus rhythm or atrial fibrillation.
Effect of hemodilution on cognition
Weiskopf et al15 assessed the effect of anemia on executive and memory function by inducing progressive acute isovolemic anemia in 90 healthy volunteers (age 29 ± 5), reducing their hemoglobin values to 7, 6, and 5 g/dL and performing repetitive neuropsychological and memory testing before and after the hemodilution, as well as after autologous blood transfusion to return their hemoglobin level to 7 g/dL.
There were no changes in reaction time or error rate at a hemoglobin concentration of 7 g/dL compared with the performance at a baseline hemoglobin concentration of 14 g/dL. The volunteers got slower on a mathematics test at hemoglobin levels of 6 g/dL and 5 g/dL, but their error rate did not increase. Immediate and delayed memory were significantly impaired at hemoglobin of 5 g/dL but not at 6 g/dL. All tests normalized with blood transfusion once the hemoglobin level reached 7 g/dL.15
Weiskopf et al16 subsequently investigated whether giving supplemental oxygen to raise the arterial partial pressure of oxygen (Pao2) to 350 mm Hg or greater would overcome the neurocognitive effects of severe acute anemia. They followed a protocol similar to the one in the earlier study15 and induced anemia in 31 healthy volunteers, age 28 ± 4 years, with a mean baseline hemoglobin concentration of 12.7 g/dL.
When the volunteers reached a hemoglobin concentration of 5.7 ± 0.3 g/dL, they were significantly slower on the mathematics test, and their delayed memory was significantly impaired. Then, in a double-blind fashion, they were given either room air or oxygen. Oxygen increased the Pao2 to 406 mm Hg and normalized neurocognitive performance.
Hemodilution studies in surgical patients
A 2015 meta-analysis17 of 63 studies involving 3,819 surgical patients compared the risk of perioperative allogeneic blood transfusion as well as the overall volume of transfused blood in patients undergoing preoperative acute normovolemic hemodilution vs a control group. Though the overall data showed that the patients who underwent acute normovolemic hemodilution needed fewer transfusions and less blood (relative risk [RR] 0.74, 95% confidence interval [CI] 0.63–0.88, P = .0006), the authors noted significant heterogeneity and publication bias.
However, the hemodilution studies paved the way for justifying a more conservative and restrictive transfusion strategy, with a hemoglobin cutoff value of 7 g/dL, and in acute anemia, using oxygen to overcome acute neurocognitive effects while searching for and correcting the cause of the anemia.
STUDIES OF RESTRICTIVE VS LIBERAL TRANSFUSION STRATEGIES
Studies in critical care and high-risk patients
Hébert et al18 randomized 418 critical care patients to a restrictive transfusion approach (in which they were given red blood cells if their hemoglobin concentration dropped below 7.0 g/dL) and 420 patients to a liberal strategy (given red blood cells if their hemoglobin concentration dropped below 10.0 g/dL). Mortality rates (restrictive vs liberal strategy) were as follows:
- Overall at 30 days 18.7% vs 23.3%, P = .11
- In the subgroup with less-severe disease (Acute Physiology and Chronic Health Evaluation II [APACHE II] score < 20), 8.7% vs 16.1%, P = .03
- In the subgroup under age 55, 5.7% vs 13%, P = .02
- In the subgroup with clinically significant cardiac disease, 20.5% vs 22.9%, P = .69
- In the hospital, 22.2% vs 28.1%; P = .05.
This study demonstrated that parsimonious blood use did not worsen clinical outcomes in critical care patients.
Carson et al19 evaluated 2,016 patients age 50 and older who had a history of or risk factors for cardiovascular disease and a baseline hemoglobin level below 10 g/dL who underwent surgery for hip fracture. Patients were randomized to two transfusion strategies based on threshold hemoglobin level: restrictive (< 8 g/dL) or liberal (< 10 g/dL). The primary outcome was death or inability to walk without assistance at 60-day follow-up. The median number of units of blood used was 2 in the liberal group and 0 in the restrictive group.
There was no significant difference in the rates of the primary outcome (odds ratio [OR] 1.01, 95% CI 0.84–1.22), infection, venous thromboembolism, or reoperation. This study demonstrated that a liberal transfusion strategy offered no benefit over a restrictive one.
Rao et al20 analyzed the impact of blood transfusion in 24,112 patients with acute coronary syndromes enrolled in three large trials. Ten percent of the patients received at least 1 blood transfusion during their hospitalization, and they were older and had more complex comorbidity.
At 30 days, the group that had received blood had higher rates of death (adjusted hazard ratio [HR] 3.94, 95% CI 3.26–4.75) and the combined outcome of death or myocardial infarction (HR 2.92, 95% CI 2.55–3.35). Transfusion in patients whose nadir hematocrit was higher than 25% was associated with worse outcomes.
This study suggests being cautious about routinely transfusing blood in stable patients with ischemic heart disease solely on the basis of arbitrary hematocrit levels.
Carson et al,21 however, in a later trial, found a trend toward worse outcomes with a restrictive strategy than with a liberal one. Here, 110 patients with acute coronary syndrome or stable angina undergoing cardiac catheterization were randomized to a target hemoglobin level of either at least 8 mg/dL or at least 10 g/dL. The primary outcome (a composite of death, myocardial infarction, or unscheduled revascularization 30 days after randomization) occurred in 14 patients (25.5%) in the restrictive group and 6 patients (10.9%) in the liberal group (P = .054), and 7 (13.0%) vs 1 (1.8%) of the patients died (P = .032).
Murphy et al22 similarly found trends toward worse outcomes with a restrictive strategy in cardiac patients. The investigators randomized 2,007 elective cardiac surgery patients with a postoperative hemoglobin level lower than 9 g/dL to a hemoglobin transfusion threshold of either 7.5 or 9 g/dL. Outcomes (restrictive vs liberal strategies):
- Transfusion rates 53.4% vs 92.2%
- Rates of the primary outcome (a serious infection [sepsis or wound infection] or ischemic event [stroke, myocardial infarction, mesenteric ischemia, or acute kidney injury] within 3 months):
35.1% vs 33.0%, OR 1.11, 95% CI 0.91–1.34, P = .30)
- Mortality rates 4.2% vs 2.6%, HR 1.64, 95% CI 1.00–2.67, P = .045
- Total costs did not differ significantly between the groups.
These studies21,22 suggest the need for more definitive trials in patients with active coronary disease and in cardiac surgery patients.
Holst et al23 randomized 998 intensive care patients in septic shock to hemoglobin thresholds for transfusion of 7 vs 9 g/dL. Mortality rates at 90 days (the primary outcome) were 43.0% vs 45.0%, RR 0.94, 95% CI 0.78–1.09, P = .44.
This study suggests that even in septic shock, a liberal transfusion strategy has no advantage over a parsimonious one.
Active bleeding, especially active gastrointestinal bleeding, poses a significant stress that may trigger empirical transfusion even without evidence of the real hemoglobin level.
Villanueva et al24 randomized 921 patients with severe acute upper-gastrointestinal bleeding to two groups, with hemoglobin transfusion triggers of 7 vs 9 g/dL. The findings were impressive:
- Freedom from transfusion 51% vs 14% (P < .001)
- Survival rates at 6 weeks 95% vs 91% (HR 0.55, 95% CI 0.33–0.92, P = .02)
- Rebleeding 10% vs 16% (P = .01).
Patients with peptic ulcer disease as well as those with cirrhosis stage Child-Pugh class A or B had higher survival rates with a restrictive transfusion strategy.
The RELIEVE trial25 compared the effect of a restrictive transfusion strategy in elderly patients on mechanical ventilation in 6 intensive care units in the United Kingdom. Transfusion triggers were hemoglobin 7 vs 9 g/dL, and the mortality rate at 180 days was 55% vs 37%, RR 0.68, 95% CI 0.44–1.05, P = .073.