Review of the clinical practice literature on allogeneic red blood cell transfusion
Paul C. Hébert,*† MD, FRCPC, MHSc; Irwin Schweitzer,*† MSc; Lisa Calder,* BScH; Morris Blajchman,‡ MD, FRCPC; Anthonio Giulivi,§ MD, FRCPC
CMAJ 1997;156(11 suppl):S9-26
From *the Department of Medicine, University of Ottawa, Ottawa, Ont.; †the Clinical Epidemiology Unit, University of Ottawa, Ottawa, Ont.; ‡the Department of Pathology, McMaster University, Hamilton, Ont.; and §the Canadian Red Cross Society Blood Services, Ottawa, Ont.
Correspondence to: Dr. Paul C. Hébert, Department of Medicine, LM-11, Ottawa General Hospital, 501 Smyth Rd., Ottawa ON K1H 8L6
Reprint requests to: Dr. Anne Carter, Director of Health Programs, Canadian Medical Association, 1867 Alta Vista Dr., Ottawa ON K1G 3Y6; fax 613 731-1779; cartea@cma.ca
© 1997 Canadian Medical Association (text and abstract)
Contents
Abstract
Objective: To review the evidence describing practice variation in the transfusion of allogeneic red blood cells as well as the risks, benefits, harms and costs associated with anemia and transfusion.
Literature search and selection: Searches of MEDLINE from January 1966 to December 1996 were combined with manual searches of bibliographies and references from experts. Two reviewers examined the abstracts of citations to identify those related to clinical practice involving red blood cell transfusions. Disagreement was resolved through consensus.
Literature synthesis: Selected articles were classified by study design and topic. Inferences were derived from the evidence.
Results: Of the 189 articles reviewed, 78 (41%) were interventional and 111 (59%) were observational studies. A number of observational studies reported a decrease in the number of transfusions since the mid 1980s, significant practice variation among physicians, institutions and various medical and surgical settings and rates of 4% to 66% of unnecessary transfusion. Of the 47 randomized clinical trials (RCTs) we found, 6 evaluated various "transfusion thresholds." Only 1 of the 6 RCTs in patients with sickle-cell disease was considered level I evidence. There was no consensus on a hemoglobin concentration that would act as a transfusion threshold. Two cohort studies suggested that adverse outcomes from anemia are greatest in patients with cardiac disease. In 8 studies evaluating the effect of hemoglobin concentration on health-related quality of life and symptoms such as dyspnea, fatigue and exercise capacity, no correlations or associations were noted.
Conclusion: The rate of transfusion has decreased since 1985; practice varies significantly as does the rate of unnecessary transfusion. Education programs and the use of algorithms may increase the appropriateness of RBC use. There was insufficient evidence to justify setting an optimal hemoglobin concentration as a transfusion threshold following acute or chronic anemia. RCTs should be conducted to determine best transfusion practice in a variety of clinical settings. Prospective cohort studies are also needed to describe transfusion practice.
[ Contents ]
Introduction
Allogeneic red blood cell (RBC) transfusions have been an important measure in the clinical care of many patients. They have been recommended for increasing oxygen delivery as well as for alleviating the symptoms of anemia.15 In the 1940s and early 1950s, allogeneic RBC transfusions were ordered readily as practitioners believed transfusion-related risks to be negligible. As early as the late 1950s, a more cautious approach was being suggested,6,7 but many practitioners continued to use transfusions with little overt concern about the risks.7 The appearance of viruses in the blood system, including the human immunodeficiency virus (HIV) and several types of viruses causing hepatitis, as well as the decreasing availability of blood, resulted in recommendations for a more conservative use of allogeneic RBCs.24,815
Despite increasing interest in all aspects of transfusion practice, we were unable to identify any systematic reviews of clinical practice in this area. To help health care practitioners make decisions about transfusion, we thought it important to describe transfusion practice by addressing the following questions: What are the indications for allogeneic RBC transfusions? Have there been changes in clinical transfusion practice and utilization of RBCs over time? Which patients are most frequently transfused with allogeneic RBCs? Is there published evidence of significant practice variation and unnecessary use of RBCs?
We also questioned the therapeutic aspects of this intervention: What are the benefits, risks, harms and costs associated with anemia and allogeneic RBC transfusions? What patient characteristics or diseases increase the risk of adverse outcomes following anemia? This systematic evaluation is an attempt to answer these questions and to provide a synthesis of a vast literature for physicians making transfusion-related clinical decisions.
[ Contents ]
Methods
Literature search and selection
A search of Medline from January 1966 to July 1996 was constructed using the following medical subject headings (MeSHs): blood transfusion, erythrocyte transfusion and blood component transfusion. This was combined with other searches using the MeSHs blood transfusion, adverse effects, postoperative complications, aged, immunosuppression and infection. The searches were designed to find these words in titles and abstracts of citations in all languages and study designs. In addition, manual searches of bibliographies were carried out. In this review, we excluded all laboratory studies and human studies focusing on physiologic mechanisms. Foreign language articles without a French or English abstract were also excluded.
The MEDLINE citation lists were scanned by 2 reviewers (PH,LC). Preliminary selection consisted of determining if a citation involved transfusion practice in humans. The abstracts of selected citations were then reviewed by the same 2 people to ensure that the following criteria defining clinical RBC transfusion practice were met: original data were used (i.e., primary studies); studies were in humans; clinical aspects of allogeneic RBC transfusions were examined. Selected review articles were used to highlight points made, but not to draw inferences. A few important studies reported before 1966 and between July 1996 and January 1997 were also added to the search results. We excluded studies that evaluated the collection, processing, storage, testing and other laboratory concerns related to RBCs. Disagreement was resolved through consensus.
Data synthesis
The selected articles were categorized according to topic and study design as defined by Meinert16 (Table 1). Studies were also assigned a level of evidence as proposed by Cook and colleagues17; i.e., inferences from clinical studies evaluating therapeutic interventions should be considered very weak if they are derived from case series (level V) and strong if derived from a large randomized, controlled clinical trial (RCT; level I). Evidence-based inferences were then formulated and graded according to the scale proposed by Wilson and co-workers,18 who described a 6-point scale in which recommendations are based on the type of design, the degree of similarity between study results and whether confidence intervals (CIs) overlap a treatment threshold. A grade A1 recommendation is based on more than 1 RCT with similar outcomes, all indicating that an intervention either does or does not result in clinically important benefits. A grade C2 recommendation, the weakest rating, is based on observational studies that do not have comparable outcomes.
[ Contents ]
Results
There were 189 primary studies identified for this review. We also incorporated non-peer-reviewed data from annual reports of the Canadian Red Cross Society, Blood Services, between 1981 and 1994.19 We excluded technical and laboratory studies (n = 9), reviews (n = 42), guidelines (n = 16) and commentaries (n = 9).
Of the 189 articles, 78 (41%) were interventional studies: 48 were RCTs, including 23 studies of hemodilution; 8 studies evaluating quality of life and symptoms such as dyspnea, fatigue and exercise tolerance;2027 8 studies of immunosuppressive complications from RBC transfusions (i.e., nosocomial infections and cancer recurrence);2835 6 on transfusion strategies;3641 2 on educational interventions;42,43 and 1 on a blood conservation device.44
Interventional studies also included 14 (9%) beforeafter trials primarily examining RBC utilization. Investigators evaluated education programs,4548 a monitoring program,49 clinical recommendations,50 statewide informed consent legislation,51 regulatory policies52,53 and an autologous blood program.54 There were also 4 non-randomized or concurrent control trials that examined the determinants of RBC blood requirement.5558
Under interventional studies, we also included non-randomized/concurrent control trials (n = 8) and interventional studies with historical controls (n = 8) that covered a broad range of clinical transfusion issues.
Among the 111 observational studies (59% of total), 60 (32%) were cross-sectional surveys, 27 (14%) were cohort studies (7 prospective and 20 retrospective), 18 (10%) were casecontrol studies and 6 (3%) were case series. The cross-sectional studies employed various designs including audits (n = 35), self-administered physician surveys (n = 8) and secondary analyses of administrative databases (n = 17). Most of these studies addressed issues related to RBC utilization. Physician surveys generally focused on transfusion triggers and determinants of transfusion practice.5966 Of the prospective cohort studies, 2 examined indications for RBC transfusion,67,68 3 addressed the effectiveness of educational strategies6971 and 2 focused on algorithms to control RBC utilization.72,73 Finally, 20 retrospective cohort studies addressed RBC utilization,7478 cost of transfusions7983 and the effects of educational tools84 and control measures85,86 on transfusion practice; the 7 remaining studies examined various issues related to RBC utilization using casecontrol and case series designs.
From the systematic evaluation, 19 inferences were drawn and graded according to the strength of supporting evidence (Table 2). Among these, 2 were not supported by any report and 7 were drawn only from observational or analytical studies (grade C). The inference supported by the strongest evidence (grade A) involved educational outreach programs and the use of intra-operative algorithms.
[ Contents ]
Discussion
What are the indications for allogeneic RBC transfusions?
The Canadian Red Cross Society's Clinical Guide to Transfusion10 states that the chief indication for RBC transfusion is anemia, mainly to increase O2 delivery, not solely to expand intravascular volume without evidence of decreased [Hb].
Other major medical organizations have also developed guidelines or position statements with similar indications.2,3,8,13,87 Many of the guidelines provide more specific criteria for the use of allogeneic RBCs. The American College of Physicians added that an empiric transfusion threshold should be avoided, that RBCs should be administered to relieve symptoms and on a unit-by-unit basis. They also explicitly stated that RBCs are contraindicated as a means to enhance well-being and promote wound healing.
Many guidelines and position papers3,4,8,8789 have also suggested a specific [Hb] or range of values to guide the transfusion decision: the National Institutes of Health4 suggest [Hb] 70 g/L; two separate publications of the American Association of Blood Banks (both developed as audit criteria not clinical practice guidelines)8,87 propose 80 g/L; and the American Society of Anesthesiologists3 states that RBC transfusion will provide the greatest benefit when [Hb] is 60 to 100 g/L.
However, the indications from these many sources may not be useful to practitioners because of limitations in the evidence and their inability to determine the ideal rate of O2 delivery for a particular patient and to identify patients at increased risk of either anemia or transfusion.
Have there been changes in clinical transfusion practice and utilization of RBCs over time?
For decades, a [Hb] of 100 g/L or a hematocrit of 30% was advocated as the threshold or transfusion trigger at which most patients with acute anemia should be administered RBCs without any consideration of the patient's clinical course.1,9093 Zauder94 surmised that the popular 100 g/L threshold originated in a discussion of preoperative anemia in a 1941 publication by Adams and Lundy.95 Subsequently, Clark and colleagues9698 described a condition they labeled as chronic shock in chronically anemic patients. These authors went on to say that such patients should be transfused when their [Hb] decreased below 100 g/L.
Throughout the 1950s and 60s, most major anesthesia and surgical textbooks incorporated the notion of a 100 g/L transfusion trigger.95 In 1945, Brannon and colleagues99 documented incremental increases in cardiac output in anemic patients with [Hb] below 100 g/L and interpreted this observation as potentially detrimental. In a series of studies on hemodilution, Messmer and co-workers100105 concluded that a hematocrit of 30% (or [Hb] of 100 g/L) provides maximal O2 delivery. A 1970 study106 found an association between increased mortality and [Hb] below 100 g/L.
More recently, clinical practice guidelines from major medical organizations24,8,13,107,108 have refuted the concept of a single [Hb] as a transfusion threshold, emphasizing the need for clinical judgement in transfusion-related decisions. Recent reviews90,92,93,109 have also concluded that insufficient evidence exists to support a single [Hb] threshold.
Although experts have recommended moving away from the use of transfusion triggers or thresholds, practitioners appear to be slow in modifying their approach to allogeneic transfusions. Several self-administered surveys have examined the use of triggers.59,60,6264 In 1970, the first such survey59 reported that 88% of anesthesiologists required a [Hb] of at least 90 g/L prior to surgery and 44% required a concentration greater than 100 g/L. In 1987, a self-administered survey of American anesthesiologists60 revealed that 65% required a preoperative [Hb] of 100 g/L. A 1992 Canadian survey of critical care practitioners62 observed that 35% of respondents identified 90 g/L as minimum concentration and an additional 40% selected 100 g/L. The authors also noted significant difference in pretransfusion [Hb] among the 4 clinical scenarios (normovolemic ventilated patients with severe sepsis, trauma, gastrointestinal bleeding and postoperative vascular surgery) and a number of potential risk factors including age, disease severity, hypoxia acidosis and myocardial ischemia. In a recent survey of gynecologic oncologists,64 most respondents reported that they adopted very low transfusion thresholds which varied considerably according to the clinical setting.
From these studies, it appears that a proportion of physicians tolerated increasing levels of anemia in their patients. Because of differences in study populations, study designs and quality of the methods used, it is not possible to infer that physicians have modified their practice over time or have incorporated recommendations from published guidelines into their clinical practice.
Trends noted in clinical transfusion practice have been mirrored by significant changes in utilization rates. Several American studies,110119 1 Taiwanese120 and 1 older British study7 have documented trends in the administration of RBCs. The British study described a 13.5% increase in the number of units transfused from 1954 to 1958 in a section of London. Between 1948 and 1960 at the Cleveland Clinic, there was a 4-fold increase in the number of units transfused.110 In Connecticut,114 the use of allogeneic RBCs increased between 1966 and 1976. Three National Institutes of Health (NIH) surveys117119 indicated that the number of transfusions in the United States doubled from 1971 to 1980. From 1980 to 1986, transfusion rates increased only minimally and dropped significantly in 1987.121 Wallace and colleagues112,116 reported a 3.1% decrease from 1989 to 1992. Overall, surveys in the United States suggest that the rate of RBC transfusions increased until the mid-80s and subsequently declined. These estimates should be interpreted with caution given that the data used for examining longitudinal trends were gathered from cross-sectional surveys rather than through prospective studies.
Unfortunately, there are no published Canadian studies describing trends over time. Data from Canadian Red Cross Society annual reports19 show that the number of RBC units transfused increased between 1980 and 1986, stabilized until 1991, then declined steadily until 1995. In 1995, 856 267 RBC units were collected whereas 757 674 (88.5%) were transfused. There also appears to have been a decrease in the number of allogeneic RBC units administered from 3.01 to 2.82 units per patient between 1991/92 and 1994. These data, gathered by a single national organization in a similar fashion from year to year, confirm observations based on published estimates in the United States.
In summary, significant changes have occurred, both in the clinician's approach to RBC administration and in overall utilization in the past 3 decades. The more conservative approach to transfusion practice and the decline in RBC use in Canada since the mid-1980s have coincided with the advent of HIV detection in allogeneic RBCs.
Which patients are most frequently transfused with allogeneic RBCs?
A number of studies describe the overall patient population receiving RBC transfusions,85,122126 and a significant number of studies described RBC utilization in selected patient populations.47,48,50,51,53,55,65,76,77,127139 In a 1992 survey conducted in 45 Toronto area hospitals, 65% of the allogeneic RBCs used were administered to patients undergoing operative procedures categorized as digestive and abdominal, cardiovascular and musculoskeletal.122 Brien and co-workers123 reported that 56% of all RBC units were administered to surgical patients whereas Ghali and colleagues85 determined that 69% of RBC units are transfused into surgical patients.
In general, cardiac surgical procedures, orthopedic procedures (e.g., total hip and knee replacement) and selected gynecologic (e.g., radical hysterectomy) and urologic procedures (e.g., radical prostatectomy) were noted to have a high proportion of patients requiring RBCs.54,122,125,140 The proportion of patients receiving RBCs ranges from 50% to 80% for aortic aneurysms and coronary revascularization to as few as 2%6% for cholecystectomy.54,122,130 In nonoperative settings, allogeneic RBCs were most frequently administered to patients with malignancies.122 Despite the lack of optimal study designs describing trends over time and utilization in various patient populations, published studies appear to be remarkably consistent.
Is there published evidence of significant practice variation in the use of RBCs?
Several clinical studies have commented on the appropriateness45,56,84,85,123,125,141145 and practice variation114,124,128,129,138,146148 in RBC use. Several investigators identified practice variation as being an interinstitutional phenomenon. A secondary analysis of a large database conducted in 1978114 found striking variation among hospitals in Connecticut: large hospitals used more blood and plasma per discharge than smaller ones. The authors inferred that physician habits and personal preferences determined institutional variation in blood utilization. However, others criticized the study for failing to control for the effects of case mix.1 Subsequently, other studies have documented significant practice variation within specific disease categories,146,147 clinical settings149 and surgical procedures128,129 including hip130,132,150,151 and knee132,150,151 arthroplasty and coronary revascularization.138,139,148
Controlling for population differences, blood loss and pump time, a prospective audit of patients undergoing coronary artery bypass grafting148 identified transfusion factors, such as the nadir and discharge hematocrits, that accounted for significant variation in blood use among 18 tertiary care hospitals. In similar patients, Surgenor and colleagues138 found that there were significant differences between hospitals in the percentage of patients transfused. Hébert and co-workers149 found a significant variation in transfusion practice (in terms of lowest [Hb]) among 6 Canadian intensive care units after controlling for the effects of disease severity, diagnosis, age and sex.
Retrospective chart reviews and self-administered surveys have also been carried out to determine whether physicians account for significant variations. In the Sanguis study,129 transfusion rates were found to depend more on physicians than the patient population, type of procedure or hospital. Wide variation was found among 43 hospitals in 10 European countries128 and between hospitals within the same country. Some factors found to influence this variation were age, sex, preoperative hematocrit and blood loss. In a survey of anesthesiologists, Stehling and colleagues60 observed a wide variation in the use of a transfusion trigger of [Hb] 100 g/L. The variations depended more on the institution and physician than on patient characteristics.
There is substantial evidence that transfusion practice varies. Many authors have concluded that such differences suggest inappropriate use by physicians. However, there are few, if any, studies that explore the reasons for these observations. It is possible, for example, that the limited number of large RCTs as well as competing risks elaborated in the existing evidence may be a significant source of variation.
Is there evidence of unnecessary use of RBCs?
Despite significant differences in both criteria and reported rates, studies consistently show that a proportion of transfusions are unnecessary (Table 3). Criteria for appropriateness included selected guidelines,45,85,141,152 clinical indicators,144,153 specific [Hb],69,127,144,154 algorithms,84,123 some combination of these142 or other criteria.56
The rates of unnecessary or inappropriate RBC use range from 4% to 66%.1 In a Canadian teaching hospital, in which 55% of 170 allogeneic RBC transfusions were deemed inappropriate,85 most unnecessary transfusions were in normovolemic, hemodynamically stable patients with anemia and multiple-unit transfusions. Brien and colleagues123 determined that 67% of family medicine patients to 95% of obstetricsgynecology patients were transfused according to appropriate indications. In several studies, there were differences in rates of appropriateness between men and women.56,132,138,139,156,157 Because they had a lower baseline [Hb], a greater proportion of women were transfused as [Hb] fell below an arbitrary transfusion trigger.
In explaining these observations, 1 study61 suggested that more junior staff were often coerced to transfuse patients unnecessarily by attending physicians. The authors also noted that attending physicians exhibited more widespread deficiencies in their knowledge of transfusion risks and indications. Differences in study designs (audits versus secondary analysis of databases), sample size, study population (diagnostic category or procedure, age and sex differences and disease severity) and appropriateness criteria may all account for the variation in the rates of unnecessary RBC use. Indeed, Hasley and co-workers158 and Goodnough and colleagues56 noted a relation between the use of restrictive criteria and the increased number of inappropriate transfusions. Difficulty with missing or incomplete data, preconceived biases, and any number of measurement biases weaken any inference drawn from these retrospective studies attempting to evaluate bedside decisions.
Several approaches have been used to improve transfusion practice. In an RCT,43 focused teaching sessions decreased noncompliance with transfusion guidelines by 40% among surgeons in the study group compared with a 9% increase in the control group. Despotis and colleagues42 evaluated an intra-operative transfusion algorithm in patients experiencing microvascular bleeding after cardiac surgery and found significant effects on transfusion practice. Patients treated according to the algorithm had fewer exposures to red blood cells and other blood components than patients treated according to standard policy. The algorithm had a significant impact on surgeons' transfusion practice, assisted in decision-making and served as an effective teaching tool.
Interventions that directly affect clinical decision-making, such as clinical practice guidelines, education programs, conferences, academic detailing and audits, may improve a physician's transfusion practice. A number of other interventions, involving various levels of the health care and blood system, may modify clinical decision-making (Table 4). However, in many instances, the impact of each method on transfusion practice, alone or in combination, has not been evaluated.
What are the relative benefits, risks, harms and costs associated with anemia and allogeneic RBC transfusions?
In the treatment of anemia, all clinically important potential benefits, risks and costs must be considered in decisions to adopt one approach over another. One should also consider whether alternative therapies such as preoperative autologous donations and pharmacologic interventions, including aprotinin and erythropoeitin, should replace or be incorporated into any transfusion strategy. Thus, the practitioner should weigh the risks of anemia against the benefits and risks of administering allogeneic RBCs (or alternatives). Ideally, rigorously conducted RCTs in patients with anemia should compare approaches and interventions to provide clinicians with the most accurate estimates of treatment benefit.16,17,159
To develop an optimal treatment approach in patients with anemia, RCTs should first compare outcomes such as mortality and myocardial infarction rates in a variety of clinical settings (i.e., perioperative anemia and anemia in the critically ill) and diseases (patients with and without cardiac disease). More subjective but equally important outcomes should also be examined. Comparisons should include health-related quality of life, activity levels and exercise tolerance as well as symptoms including fatigue, dyspnea and exercise tolerance. Other outcomes to be considered include the rates of viral transmission and infectious complications following various allogeneic RBC transfusion strategies.
We identified 6 RCTs contrasting 2 transfusion strategies in a total of 813 patients (Table 5).3641 Only 1 study, conducted in patients with sickle-cell disease, was large enough to rule out clinically important differences in its primary outcome, perioperative sickle-cell crises.40 In this study, an aggressive transfusion strategy, compared with a more conservative regimen, was unable to prevent sickle-cell crises.
In a second study, 50 consecutive patients with severe gastrointestinal bleeding were randomly chosen to receive at least 2 allogeneic RBC units immediately or no transfusions unless [Hb] fell below 80 g/L.41 In the immediate-transfusion cohort, 9 patients had a recurrence of gastrointestinal bleeding compared with 1 patient in the delayed-transfusion cohort (p < 0.001).
Patients undergoing coronary revascularization were examined in 2 studies.36,37 In 1,37 there was no difference in postoperative complication rates between patients treated with a liberal transfusion strategy compared with those subjected to a conservative strategy, although for the conservative group there was a significant decrease in total postoperative blood use. The other36 assessed day 5 exercise tolerance as well as hemodynamic and myocardial metabolic response following normovolemic hemodilution in 27 patients. One group of 13 patients received RBCs if their [Hb] fell below 120 g/L in addition to colloids; the other 14 patients received crystalloids and allogeneic RBCs only if [Hb] fell below 70 g/L. Although patients in the low [Hb] trigger group received significantly fewer RBCs than the other group, there were no differences in morbidity, mortality or exercise tolerance. In a small subset of 6 patients, there were differences in the rate of myocardial lactate recovery in the low [Hb] trigger group suggesting increased myocardial ischemia from anemia.
Hébert and colleagues39 randomly assigned 69 critically ill patients to a restrictive or liberal RBC transfusion strategy to evaluate the impact of the treatments on mortality rates, organ dysfunction scores and other markers of morbidity. Neither mortality nor the development of organ dysfunction were affected by the transfusion strategy. However, maintaining [Hb] between 70 and 90 g/L decreased the average number of units transfused from 4.8 to 2.5 (48% reduction, p < 0.001).
Finally, 25 critically ill trauma victims were randomly chosen to receive allogeneic RBC transfusions once hematocrit levels reached either 30% or 40%.38 The authors concluded that there were no discernable differences in O2 transport variables between the 2 groups.
In summary, 5 of the 6 studies enrolled too few patients to make significant inferences regarding important outcomes from RBC transfusions and the 1 large RCT reported only the effects of transfusions using a disease-specific outcome in sickle-cell disease.
A total of 23 RCTs evaluating perioperative hemodilution in patients undergoing the following surgical procedures were identified: cardiac surgical interventions (primarily coronary revascularization),160170 vascular procedures,171173 tumor resection,174,175 hip arthroplasty,176179 thoracic procedures180 and prostate resection.181,182 Of these 18 studies reported the number of units or volume of allogeneic RBCs used; and 12 reported allogeneic RBC exposure rates. The remaining studies focused on oxygen transport as well as the cardiac and coronary effects of this intervention (see Hébert and associates, Review of physiologic mechanisms in response to anemia, this issue).
In the 12 studies reporting a statistically significant difference in allogeneic RBC transfusion volumes, the decrease was small and clinically unimportant (250500 mL of RBCs). Thus, the efficacy of perioperative hemodilution in limiting allogeneic RBC exposure has yet to be established. In addition, inferences about the safety of anemia based on these studies are limited because the technique is used in highly selected patients in a controlled setting; intra-operative transfusion threshold and protocols were not explicitly outlined, therefore the degree of anemia is not known; and the techniques (including degree of hemodilution, replacement fluids, storage and reinfusion) were not comparable from 1 study to another. The lack of significant differences in cardiac events or mortality rates in these small studies should not be interpreted as inferring that moderate degrees of intra-operative normovolemic anemia are safer than avoidance of anemia through RBC transfusion.
Several observational studies were also identified, including numerous reports of toleration of severe anemia in terms of mortality risk and adverse consequences such as myocardial infarction in surgical patients.91,106,183193 In a prospective study of 1769 anemic patients undergoing coronary revascularization, Bayer and co-workers183 found that a decreased hematocrit (27% to 30%) was well tolerated, with no reported increase in either morbidity or mortality compared with historical controls from the same institutions. Additional reports or case series91,190,194196 describe successful outcomes in patients with chronic anemia as a result of renal failure.
Finally, descriptive studies in patients refusing RBC transfusion184186,191,192 and in regions where blood supplies have been limited187,197 have demonstrated that patients can survive surgical interventions with [Hb] as low as 45 g/L. Two studies, 1 casecontrol185 and 1 case series,186 in the same cohort of Jehovah's Witness patients191 documented an association between preoperative [Hb], intra-operative estimated blood loss and postoperative mortality. No deaths were reported in more than 100 patients undergoing major elective surgery when preoperative [Hb] was greater than 80 g/L and estimated blood loss was less than 500 mL. In a single-centre series of 542 Jehovah's Witness patients undergoing a cardiac surgical procedure, the overall mortality was 10.7%; only 2.2% of the deaths were considered to be a direct consequence of anemia.
In summary, these observational studies suggest that moderate degrees of anemia are well tolerated in low-risk patients. However, such studies only provide weak evidence in support of a lowered transfusion threshold given the potential for selection and measurement biases as well as the possibility of significant confounding by clinical factors such as disease severity and comorbid diseases. Although a significant number of articles have examined the impact of various transfusion strategies on clinically important outcomes such as mortality and rates of myocardial infarction, few were considered to be level I or II RCTs. In addition, the clinical heterogeneity of patient populations and interventions would not permit the use of meta-analytic techniques to combine the results of the RCTs.
Determining the relative benefit of RBC transfusions should include not only an assessment of mortality, but also consideration of the impact of therapy on anemia-related symptoms such as dyspnea and fatigue and overall measures of health status such as quality of life. This is most relevant in patients with chronic anemia and patients at low risk of death from acute anemia. We were unable to identify any RCT comparing the effect of various [Hb] or transfusion strategies on symptoms, physical functioning or health-related quality of life. The most compelling evidence supporting an association between [Hb] and quality of life arises from studies evaluating erythropoeitin use in a number of clinical settings2025 (Table 6). Improvements in health-related quality of life were observed in patients on hemodialysis,2123,198 in those with chronic anemia as a result of human immunodeficiency virus (HIV) or HIV therapy24 and those with cancer-related anemia.25,199
The greatest benefit of increased [Hb] in erythropoeitin therapy appears to be in terms of increased energy and activity levels.21,24,198 Using a disease-specific quality of life instrument, hemodialysis patients in the Canadian Erythropoeitin Study22 reported significant improvements in their scores for fatigue and physical symptoms without significant changes in exercise capacity assessed using the 6-minute walk test. In contrast, an RCT26 evaluating iron therapy in anemic women did not report any improvement in fatigue and breathlessness despite significant increases in [Hb]. In another study,27 psychomotor function in anemic women was also found to be unaffected by iron therapy. Therefore, there are conflicting conclusions regarding the association between anemia and subjective outcomes that have arisen from well-controlled clinical trials evaluating interventions other than RBC transfusions.
When considering transfusion with allogeneic RBCs, the physician must weigh the consequences against the risks associated with ongoing anemia. Many of the risks associated with allogeneic RBC transfusions have been difficult to quantify because they are very small. However, the risk of transmitting viruses such as HIV and hepatitis has been uppermost in the minds of practitioners and the public in the past few years. Currently, the risk of contracting a viral infection from a unit of blood ranges from 1:63 000 for hepatitis B and 1:103 000 for hepatitis C to as low as 1:676 000 for HIV and 1:641 000 for human T-cell lymphotropic virus (types I and II) based on a recent American study200 and Canadian Red Cross Society data.19
There are no prospective cohort studies describing the rates of viral transmission and associated complications in recipients of blood products. In addition, the donor population is constantly changing and the screening process evolving. New diseases or mutations of older diseases are continually threatening the system. These risks are difficult to quantify and incorporate into decision-making. A number of other potential complications include hemolytic reactions -- acute (1 in 25 000 units transfused) and delayed (1 in 25009000) -- anaphylaxis (1 in 20 00050 000), transfusion-related lung injury, graft-versus-host disease, posttransfusion purpura, congestive heart failure (1 in 100) and iron overload (begins after more than 20 RBC units transfused).
Many investigators have studied and commented on the immunosuppressive effects of allogeneic RBC transfusion.201220 Observational studies201,207,221,222 have suggested an association between the administration of allogeneic RBCs and the recurrence of cancer as well as postoperative infections. It has been hypothesized that a unit of allogeneic blood depresses immune function, thereby increasing a host's susceptibility to infections and promoting tumour growth.
We identified 8 RCTs evaluating the immune consequences of RBC transfusions, contrasting either rates of cancer recurrence (n = 2) or postoperative infections (n = 6). Investigators compared either leukocyte-depleted2833 or autologous34,35 transfusion with allogeneic RBC transfusion. Contradictory conclusions were drawn from the 6 RCTs examining postoperative infections (Table 7). Two studies28,34 did not find any significant difference in the rates of infection among patients who had undergone colorectal surgery. Houbiers and colleagues28 found a higher rate of postoperative infection in patients receiving leukocyte-depleted as opposed to allogeneic RBCs (42% versus 36%, p > 0.05). However, the 4 remaining studies29,32,33,35 reported clinically important decreases in postoperative infections in patients receiving leukocyte-depleted RBCs compared with standard allogeneic RBC products.
In a recent RCT, Jensen and colleagues29 demonstrated that the rates of wound infections and intra-abdominal abscesses were significantly lower in patients receiving allogeneic RBCs compared with untransfused groups (12% v. 1%, p < 0.0001). The frequency of pneumonia was also lower in patients receiving leukocyte-depleted RBCs (3%) or no transfusions (3%) compared with patients receiving allogeneic transfusions (23%, p < 0.001).
In summary, the 6 level-I studies arrived at divergent conclusions concerning the risks of postoperative infections attributed to allogeneic RBC transfusions. A meta-analysis using either aggregate or individual patient data might provide useful insights from these 6 conflicting RCTs.
The 2 studies evaluating cancer recurrence have not convincingly demonstrated that allogeneic RBCs truly affect the rates of tumour recurrence through immune modulation. Therefore, the clinical significance of the immunosuppressive effects of RBC transfusions have not been clearly established.
In Canada, the cost of administering allogeneic RBCs is not passed on to patients directly and, as a result, have no impact on bedside transfusion decisions. However, information regarding costs may be extremely relevant when comparing allogeneic cells to alternative strategies such as autologous blood programs, use of other O2 carriers or pharmaceutical interventions. We identified 7 studies79,80,223227 that attempted to establish the cost of allogeneic RBC transfusions. In a Canadian study,224 data from a 13-centre clinical trial evaluating erythropoeitin were used to determine the cost of allogeneic RBC units; $210/unit was estimated using a hospital perspective,80 data collected from 8 Canadian hospitals and 6 blood centres were used to establish the unit cost at $210 for in-patient allogeneic RBC transfusions and $280 for outpatient transfusions. In this same study, 59% ($124) of the cost was related to blood banking, personnel and hospital equipment; 31% ($64) was incurred in the collection process.
A multicentre study in the United States estimated comparable costs for an allogeneic RBC unit ($155 [US]). However, the average cost to society in the United States was estimated at $458 [US]. This dollar value included indirect costs such as lost productivity, decreased psychological well-being and travel expenses. A study comparing the cost-effectiveness of allogeneic RBCs with leukocyte-depleted products found the latter to be more cost-effective per patient treated225 because of decreased length of stay and associated costs ($7867 v. $12 347 [US], p < 0.01). Future studies may establish the cost-effectiveness of various approaches to the administration of RBCs as well as alternative interventions.
Finally, when making bedside transfusion decisions, one should also consider possible alternatives to RBC transfusions. All benefits, risks, harms and costs of new therapies should be compared to the best available therapy: allogeneic RBC transfusion. Although the use of autologous RBCs and erythropoeitin may decrease exposure to allogeneic RBCs, these alternatives have not been convincingly demonstrated to result in an overall benefit to patients. Future studies may help elucidate the optimal role of alternatives.
In summary, significant limitations are identified in the transfusion literature evaluating various strategies. Published clinical studies do not provide conclusive evidence supporting a specific approach to allogeneic RBC transfusions. Thus, clinical practice guidelines for the use of RBCs must still rely heavily on expert opinion.
What patient characteristics or diseases increase the risk of adverse outcomes following anemia?
Guidelines3,4,155 and reviews1,5,228 have indicated that anemia is less well tolerated in older patients, in the severely ill and in patients with coronary, cerebrovascular or respiratory disease. However, clinical evidence confirming that these factors are independently associated with an increased risk of adverse outcome is lacking. One small casecontrol study189 following high-risk vascular surgery suggests an increase in postoperative cardiac events with increasing severity of anemia. Two large cohort studies of perioperative229 and critically ill patients230 have reported increasing degrees of anemia associated with a disproportionate increase in mortality rate in the subgroup of patients with cardiac disease. In 1958, in Jehovah's witness patients229 adjusted odds of death increased from 2.3 (95% CI 1.4 to 4.0) to 12.3 (95% CI 2.5 to 62.1) as preoperative [Hb] declined from 100109 g/L to 6069 g/L in patients with cardiac disease. There was no significant increase in mortality in noncardiac patients with comparable levels of anemia. Critically ill patients230 with cardiac disease also tended to have higher mortality when [Hb] < 95 g/L (55% versus 42%, p = 0.09) compared with anemic patients with other diagnoses. Patients with anemia, a high APACHE II score (>20) and a cardiac diagnosis had a significantly lower mortality rate when given 13 or 46 units of allogeneic RBCs: 55% (no transfusion) v. 35% (13 units) or 32% (46 units), p = 0.01). Although both cohort studies were retrospective and may not have controlled for a number of important confounders, the evidence suggested that anemia increased the risk of death in patients with significant cardiac disease.
Severity of illness also appears to be a risk factor in the critically ill.185,230 Two retrospective studies report that degree of blood loss contributes to perioperative mortality.185,230 However, no studies have examined the independent contribution of age, cerebrovascular disease and respiratory disease to increased mortality risk in anemic patients. This relation may well be complex given that age and cerebrovascular disease are risk factors associated with coronary artery disease. Smoking-related respiratory diseases may have similar associations with cardiac disease. Therefore, the association between anemia and increased rates of adverse outcomes in these patients can best be described as speculative, at this time.
[ Contents ]
Conclusion
We were able to draw several inferences from the literature. A significant variation in allogeneic RBC transfusion practice has been reported in a number of studies and a variety of patient populations. Despite this, few studies attempt to explain or minimize excessive practice variation. Similarly, studies evaluating the appropriateness of RBC transfusions reveal that a proportion may be unnecessary. Only 2 randomized, controlled trials, 1 evaluating a teaching program and another the use of an intra-operative transfusion algorithm, demonstrated that specific interventions may be employed to maximize appropriateness.
One of the most important questions facing the practitioner is whether there is an optimal [Hb] at which to maintain most anemic patients or certain patient groups. Six RCTs evaluated various [Hb] transfusion thresholds. A single level-I study demonstrated that there were no differences in the frequency of sickle-cell crises in patients treated with a conservative transfusion strategy compared with more liberal use of allogeneic RBCs. The 5 other small RCTs did not provide conclusive evidence to support an optimal [Hb] or approach to the administration of RBCs. Therefore, clinical practice guidelines addressing optimal [Hb] at which to maintain patients or administer RBCs would not be based on well-controlled clinical trials but rather on weaker grades of evidence as well as expert opinion. We suggest that level-I RCTs comparing transfusion strategies in various patient populations be conducted to develop high-grade evidence-based recommendations.
Clinicians wish to know if certain patients are at increased risk of suffering adverse outcomes following the development of anemia. Two clinical studies suggested that complications from anemia are greatest in patients with cardiac disease. Associations between anemia and adverse outcomes, as well as modification in the degree of risk in patients with other potential risk factors such as increased age and disease severity, respiratory and cerebrovascular disease, have not been clearly established using rigorous study designs.
A number of RCTs have also evaluated the effect of [Hb] on health-related quality of life and symptoms such as dyspnea, fatigue and exercise capacity using erythropoeitin and iron as means of increasing [Hb]. Most erythropoeitin studies suggested improvements in many of these subjective outcomes whereas studies using iron therapy did not find significant differences. Unfortunately, we found no level-I studies comparing patients who were maintained at low [Hb] with patients transfused to higher [Hb]. Recently, well-conducted clinical trials failed to demonstrate that observed increased rates of postoperative infections were more frequent in patients administered standard allogeneic RBCs compared with untransfused patients or patients receiving leukocyte-depleted or autologous RBC products. Thus, there is still no consensus on whether early immunosuppressive effects of allogeneic RBCs may have clinically important consequences.
Despite the many deficiencies in the clinical transfusion literature, there was a substantial body of evidence of practice variation and unnecessary transfusion. The clinical studies identified did not indicate an optimal [Hb], but did suggest that patients with cardiac disease were at increased risk.
This systematic review was sponsored by the Canadian Medical Association. Financial support was provided by Health Canada, the Canadian Blood Agency and the Canadian Red Cross Society, Blood Services. Dr. Hébert is an Ontario Ministry of Health Career Scientist.
We are grateful to Patricia Chung, Eric Partington, Dr. Ling Qun Hu and the Expert Working Group for Guidelines on Transfusion of Red Blood Cells and Plasma in Adults and Children for their support and assistance in preparing this manuscript. We thank Jessica McGowan for her assistance in the computer searches, and Christine Niles for her help in the preparation of the manuscript.
[ Contents ]
References
