The CRIT Study: Anemia and Blood Transfusion in the Critically Ill - Current Clinical Practice in the United States

Howard L. Corwin, MD; Andrew Gettinger, MD; Ronald G. Pearl, MD, PhD; Mitchell P. Fink, MD; Mitchell M. Levy, MD; Edward Abraham, MD; Neil R. MacIntyre, MD; M. Michael Shabot, MD; Mei-Sheng Duh, MPH, ScD; Marc J. Shapiro, MD


Crit Care Med. 2004;32(1) 

In This Article


Anemia is common in critically ill patients and is seen early in their ICU course. By 48 hrs after ICU admission, almost 70% of patients admitted to the ICU had a baseline hemoglobin level of <12 g/dL, and half of these had a level of <10 g/dL. The anemia in these critically ill patients persisted throughout the duration of their ICU and hospital stay, with or without RBC transfusion. The extent of anemia observed is consistent with previous studies.[6]

Transfusion practice in response to this anemia has changed little over the last decade. Based on our survey, almost 50% of patients admitted to ICUs across the United States today receive transfusions. Although the initial RBC transfusion tended to occur early in the ICU stay (<3 days), there were ongoing RBC transfusions in these patients throughout their ICU stay. These observations are virtually identical to those made in earlier studies.[6,7,8] Similarly, the mean pretransfusion hemoglobin observed (i.e., the transfusion trigger) was 8.6 ± 1.7 g/dL in the present study, a value that is comparable with that described in earlier reports.[6,7] RBC transfusions were not restricted to the ICU. Thirteen percent of patients discharged from the ICU received an average of 2.7 ± 2.8 units after ICU discharge. Post-ICU RBC transfusions increased the total number of RBC units transfused by almost 15% and therefore had a significant impact on the total number of RBC units consumed by critically ill patients.

The ICUs and institutions that participated in the study are a representative cross-section of ICUs in the United States, including large and small hospitals and teaching and nonteaching hospitals. Therefore, the results of the study likely reflect general transfusion practice patterns in ICUs in the United States today. A minority of ICUs (<20%) participating in the current study had an institutional or ICU transfusion protocol in place at the time of the study; however, the existence of such a protocol did not seem to affect transfusion-related practices. Although transfusion practice was reasonably consistent across institutions and ICUs, surgical patients tended to receive transfusions more frequently. This observation is consistent with the study by Groeger et al.,[8] which reported that as many as 25% of patients in surgical ICUs receive transfusions on any given day, as compared with 14% in the overall population.

Recently, a similar observational study of transfusion practice in ICUs across Western Europe was performed.[9] Data were collected on 3,534 patients admitted to ICUs during a 2-wk period in late 1999. The similarity of the results of this study with our study is striking and suggests a remarkable consistency in current transfusion practice within the critical care community. A total of 37% of patients received transfusions of a mean of 4.8 RBC units while in the ICU, and 12.7% of patients received transfusions in the post-ICU period. In total, 42% of patients received transfusions during the 28-day study period. The mean pretransfusion hemoglobin level was 8.4 g/dL.

The magnitude of RBC transfusion in the critically ill today is surprising, given the scrutiny to which transfusion practice has been subjected during the last decade. In a prospective randomized study of critically ill patients, Hebert et al.[10] demonstrated that maintaining hemoglobin levels in the 7-9 g/dL range is at least equivalent, and in some patients (APACHE II of ≤20 or age of <55 yrs) superior, to maintaining hemoglobin levels of >10 g/dL with RBC transfusion. This finding also seemed to apply to patients with underlying cardiac disease, although other data suggest that patients with active ischemic cardiac disease may require a higher hemoglobin level.[11] The studies by Hebert et al.[10,11] and Dietrich et al.[12] have raised questions regarding the validity of the historic assumption that RBC transfusion is beneficial for all critically ill patients with anemia. Recent recommendations have advocated that empirical automatic transfusion thresholds be abandoned in favor of a practice of RBC transfusion only for defined physiologic need.[2,3] However, the suggestion for a more conservative approach to RBC transfusion does not as yet seem to have resulted in any major alteration in practice patterns.

Decisions about transfusing RBCs are often made without a complete understanding of the risks and benefits of transfusion.[13] Although a much clearer understanding of the risks of RBC transfusion has developed since the 1980s, the risks of anemia and the benefit of RBC transfusion are much less well characterized. For more than five decades, a hemoglobin level of 10 g/dL and a hematocrit of 30% were generally accepted minimum levels, particularly in the surgical setting. First proposed in 1942,[14] the "10/30" rule has become more a matter of faith than data. Although it is clear that transfusions at hemoglobin levels in the 10 g/dL range are much less common today, we observed that only about 25% of RBC transfusions occur in a range consistent with the findings reported by Hebert et al.[10]

This transfusion behavior is consistent with previous studies, which noted that transfusion decisions tend to be driven by individual transfusion triggers rather than specific physiologic indications.[6] In these studies, pretransfusion hematocrit was the same, regardless of transfusion indication. In the present study, there was little evidence that either age or co-morbidities significantly influenced transfusion practice. On the other hand, a low baseline hemoglobin level was associated with more RBC transfusions. The time to first transfusion was significantly longer in those patients who presented with a high baseline hemoglobin level (1.8 ± 1.7 days with baseline hemoglobin of ≤8 g/dL vs. 6.3 ± 6.2 days with baseline hemoglobin of >12 g/dL, p < .05). These results support the hypothesis that RBC transfusion in many critically ill patients is driven by arbitrary transfusion triggers rather than physiologic findings.[6] The fact that a low hemoglobin level was noted as one of the transfusion indications in 90% of transfusions is consistent with this hypothesis. The similarity between the apparent transfusion thresholds in the ICU and after ICU discharge also supports the view that hemoglobin level rather than clinical or physiologic factors drives transfusion decisions.

In general, more severely ill patients, as measured by either APACHE II or SOFA score, had a low baseline hemoglobin level and received more RBC transfusions. However, after correcting for baseline hemoglobin level and severity of illness, more RBC transfusions were independently associated with worse clinical outcomes. This is similar to the finding by Vincent et al.[9] On the other hand, although both studies found that baseline hemoglobin level was not associated with mortality, we did find that a nadir hemoglobin level of <9 g/dL was associated with a higher mortality. Given the observational design of these studies, these findings should be interpreted with caution. However, the transfusion results are consistent with recent data suggesting that a liberal RBC transfusion policy may be deleterious for some critically ill patients.[10,15]

Why RBC transfusions are associated with worse clinical outcomes is unclear. A substantial amount of literature has developed since the early 1980s suggesting that exposure to allogeneic leukocytes in transfusions may trigger an immune system response in the recipient leading to increased risk of infection, earlier recurrence of malignancy, and increased likelihood of mortality.[16] A significant association between the number of RBC transfusions and risk of subsequent infection has been reported in patients after trauma, burns, and a variety of surgical procedures, both elective and emergency.[17,18,19] In the critically ill, Taylor et al.[20] demonstrated an association between RBC transfusion and nosocomial infection and mortality in the critically ill. These data have in turn led to the hypothesis that giving patients transfusions with leuko-reduced blood should result in reduced morbidity and mortality compared with patients receiving transfusions with non-leuko-reduced blood. However, most of the studies bearing on these questions have been flawed by retrospective design and inadequate consideration of the effects of co-morbidities, whereas the few prospective studies in specific patient populations have reached contradictory conclusions. Meta-analyses of this substantial literature have failed to identify a statistically significant effect of leuko-reduction.[16,21,22] A recent study evaluating clinical outcomes after the institution of a universal leuko-reduction program in Canada noted a reduction in hospital mortality after introduction of this program.[23] On the other hand, a randomized prospective study comparing outcomes in patients receiving either leuko-reduced or non-leuko-reduced RBCs failed to demonstrate any beneficial effect of leuko-reduction on clinical outcome.[24] The question therefore still remains as to whether there are in fact clinical benefits associated with leuko-reduction of transfused RBCs.[25] We do not have data from the current study that would allow us to answer this question.

Recent data have also raised the issue that transfusion of "old" blood may be associated with worse outcomes.[26,27] The current study provides robust data regarding the age of RBCs critically ill patients receive. The mean age of RBCs transfused was 3 wks, and >25% of transfused RBCs were >1 month old. This is somewhat older than the mean 16 ± 6.7 days for RBCs transfused in Western Europe.[9] There were no differences in the age of RBCs transfused across institutions or ICUs. Although there was a trend toward worse clinical outcome among patients receiving transfusions with relatively old blood, this relationship was weak and did not achieve statistical significance. The clinical significance of the age of blood remains controversial and will require further study.

Why critically ill patients are anemic is multifactorial. Phlebotomy and blood loss undoubtedly play a role.[6,9] Nevertheless, a number of studies suggest that an underproduction of erythrocytes similar to that observed in chronic inflammatory diseases significantly contributes to the anemia in critical illness.[28] More than 90% of ICU patients have low serum iron, total iron binding capacity, and iron-total iron binding capacity ratio.[5,29] In addition, these patients typically have an elevated serum ferritin level.[5,19] At a time when the iron studies are abnormal, serum erythropoietin levels are only mildly elevated, with little evidence of a reticulocyte response to endogenous erythropoietin.[5] Therefore, the anemia of critical illness is a distinct clinical entity characterized by a blunted erythropoietin production and abnormalities in iron metabolism. This is reflected in the fall in hemoglobin level observed during the course of a patient's critical illness.

The data from this study should be interpreted recognizing that this is an observational study. Although the analysis attempted to control for confounding factors, it was limited to only the factors recorded. Given the complexity of critical illness, all of the factors influencing outcome may not have been included. For example, although there are considerable baseline clinical data available, fewer data are available regarding a patient's clinical status at the time of a RBC transfusion. The inferences drawn between variables can only indicate association not causation.

In conclusion, anemia is common in the critically ill patient, and persists throughout the ICU and hospital stay. Despite the scrutiny of transfusion practice during recent years, practice in the United States in 2000-2001 is little changed as compared with the preceding decade.[6,7,8] Transfusion practice in the United States is also very similar to transfusion practice as recently reported in Western Europe.[9] Current data regarding RBC transfusion thresholds and risks of RBC transfusion have not as yet significantly altered practice patterns.[10] RBC transfusions seem to be associated with worse clinical outcomes. Clearly, approaches to reduce RBC transfusion would be desirable.[10,30] However, further study is required to more fully explore the risk of anemia, optimal hemoglobin level, and the risk and efficacy of RBC transfusion in the critically ill.