Published ahead of print on July 12, 2007, doi:10.1164/rccm.200702-271OC
© 2007 American Thoracic Society doi: 10.1164/rccm.200702-271OC
Transfusion-related Acute Lung Injury in the Critically IllProspective Nested Case-Control Study1 Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, 2 Division of Transfusion Medicine, Department of Laboratory Medicine and Pathology, and 3 Division of Biostatistics, Department of Health Sciences Research, 4 Mayo Epidemiology and Translational Research in Intensive Care (METRIC), and 5 Department of Nursing, Mayo Clinic College of Medicine, Rochester, Minnesota Correspondence and requests for reprints should be addressed to Ognjen Gajic, M.D., Mayo Clinic, 200 First Street SW, Rochester, MN 55905. E-mail: gajic.ognjen{at}mayo.edu
Rationale: Acute lung injury (ALI) that develops 6 hours after transfusion (TRALI) is the leading cause of transfusion-related mortality. Several transfusion characteristics have been postulated as risk factors for TRALI, but the evidence is limited to retrospective studies. Objectives: To compare patient and transfusion risk factors between patients who do and do not develop ALI. Methods: In this prospective cohort study, consecutive transfused critically ill patients were closely observed for development of ALI. Donor samples were collected from the transfusion bags. Risk factors were compared between patients who developed ALI after transfusion and transfused control patients, matched by age, sex, and admission diagnosis. Measurements and Main Results: Seventy-four of 901 transfused patients developed ALI within 6 hours of transfusion (8%). Compared with transfused control subjects, patients with ALI were more likely to have sepsis (37 vs. 22%, P = 0.016) and a history of chronic alcohol abuse (37 vs. 18%, P = 0.006). When adjusted for patient characteristics, transfusion of plasma from female donors (odds ratio [OR], 5.09; 95% confidence interval [95% CI], 1.37–18.85) rather than male donors (OR, 1.60; 95% CI, 0.76 to 3.37), number of pregnancies among the donors (OR, 1.19; 95% CI, 1.05 to 1.34), number of donor units positive for anti-granulocyte antibodies (OR, 4.85; 95% CI, 1.32–17.86) and anti–HLA class II antibodies (OR, 3.08; 95% CI, 1.15–8.25), and concentration of lysophosphatidylcholine in the donor product (OR, 1.69; 95% CI, 1.10 to 2.59) were associated with the development of ALI. Conclusions: Both patient and transfusion risk factors determine the probability of ALI after transfusion. Transfusion factors represent attractive targets for the prevention of ALI.
Key Words: fresh-frozen plasma • platelet transfusion • pulmonary edema • female • blood donors
Since the original description (1), transfusion-related acute lung injury (TRALI) has emerged as the most important cause of morbidity and mortality resulting from blood transfusion (2–4), in part because of the decrease in morbidity and mortality from other adverse effects of transfusion. Although the exact etiology is not known, uncontrolled clinical studies and animal models have suggested an important role for donor-derived anti-leukocyte antibodies as well as biological response modifiers accumulated during blood storage (5–9). Anti-leukocyte antibodies (anti–HLA class I, anti–HLA class II, and anti-granulocyte) are known to be commonly present in the plasma of donors exposed to foreign tissue during previous pregnancy or transfusion and have been implicated in multiple case series and case reports (1, 3–5). Indeed, anti-HLA antibodies have been detected in up to 25% of multiparous female donors (10). TRALI is believed to be underdiagnosed and underreported (4, 5, 11), particularly in critically ill patients who often have multiple risk factors for acute lung injury (ALI). It is possible that current TRALI cases reported to blood banks represent the "tip of the iceberg" and that transfusion factors play a mechanistic role in many more patients with ALI, as one of "multiple hits" required for the full expression of this syndrome (12–14). Indeed, transfusion (massive) has long been recognized as one of the most important ALI risk factors (15). Submassive transfusion of all blood products, and particularly of high plasma volume products, has been identified as an independent risk factor for development of ALI in the critically ill (16, 17). In animal studies, the presence of additional ALI risk factors has been found to modulate both the development and severity of TRALI (multiple-hit hypothesis) (8, 9). The lack of a standardized definition, however, posed a significant limitation on previous clinical studies (3, 4). Two expert consensus panels have proposed a standardized TRALI definition (3, 4). Using the proposed definition, we reported retrospective data showing a much higher incidence of TRALI in critically ill patients than previously suggested (18, 19). In this prospective study we aimed to determine the incidence, risk factors, and outcome of ALI that develops 0–6 hours after the transfusion of blood products in a cohort of patients admitted to the medical intensive care unit (ICU). Using a matched case-control design, we compared types of blood products transfused by donor and storage characteristics, the prevalence of anti-leukocyte antibodies (anti-granulocyte, anti–HLA class I, and anti–HLA class II) (1, 5), and the amount of biological response modifiers (IL-8 and lysophosphatidylcholine [LysoPC]) (20) in the samples of donor blood units given to patients who did or did not develop ALI after transfusion. Some of the results of this study have been previously reported in the form of an abstract (21).
In this 2-year prospective cohort study, all patients transfused in a medical ICU of a tertiary care medical center were closely observed for 24 hours after transfusion. (For detailed methods, see the online supplement.) The institutional review board approved the study protocol. Patients who refused research authorization were excluded. Monitoring logs reflecting cardiopulmonary function, chest radiographs, and arterial blood gases were reviewed and monitored until 24 hours after each transfusion, allowing for identification of patients who had any worsening of respiratory status 0–24 hours after the transfusion (see the online supplement). Expert intensivists, blinded to specific transfusion factors, subsequently reviewed all clinical data and assigned the diagnosis of ALI within 6 hours of transfusion (suspected or possible TRALI) on the basis of the standard clinical definition (4, 22, 23). Interobserver agreement was measured by  statistics. Two of the reviewers (J.L.M. and O.B.R.) re-reviewed the discordant cases together and resolved any disagreements. In the nested case-control part of the study, the patient and transfusion factors were compared between patients who developed ALI after transfusion and transfused control patients matched by age ± 10 years, sex, and admission diagnostic group (19). Predictor variables were grouped as follows:
Statistical Analysis
We prospectively observed 901 critically ill patients who were transfused in the medical ICU over the 2-year study period (see outline of the study in Figure 1). Among 6,588 blood product units transfused during the ICU stay 3,383 (51%) were packed red blood cells, 2,728 (41%) were FFP, 306 (5%) were platelets and 171 (3%) were cryoprecipitates. Waste plasma samples were stored from 3,813 (57%) transfusions for subsequent laboratory studies. Seventy-four (8%) patients developed ALI within 6 hours of transfusion (Figure 1). Interobserver agreement was moderately good ( 0.6).
Table 1 describes clinical and transfusion characteristics of patients who developed ALI after transfusion and matched control subjects. Gastrointestinal bleeding was the most common reason for ICU admission (Table 1). Compared with patients who had no respiratory worsening after transfusion, patients who developed ALI were more likely to have sepsis, liver disease, and a history of chronic alcohol abuse (Table 1).
ALI case subjects were more likely to have received plasma-rich blood products (FFP or platelets). They were also more likely to have received blood products from female donors and larger volumes of plasma from female donors (Table 1). Five case subjects (7%) and four control subjects (5%) received massive transfusion (at least 10 units) during the exposure period. Donors to patients who developed ALI had a higher number of pregnancies and tested positive for anti-leukocyte antibodies (Tables 1 and 2) more often. There was no difference in IL-8 or storage age of red blood cell products between patients who developed ALI and matched control subjects. The concentration of LysoPC was significantly higher in blood products given to ALI case subjects than to control subjects (Table 2). At least one of the associated units was tested for anti-leukocyte antibodies for 82% of case subjects and 80% of control subjects. LysoPC and IL-8 were tested in at least one associated unit for 73% of case subjects and 78% of control subjects (Table 2).
Table 3 provides the unadjusted and adjusted ALI odds ratios for specific transfusion risk factors. When adjusted for baseline APACHE III scores, sepsis, and alcohol abuse in the conditional logistic regression analysis, several transfusion factors remained associated with the development of ALI (Table 3). The results were similar after additional adjustment for baseline liver disease (see Table E2 in the online supplement). The distribution of transfusion factors was similar in subgroups of patients who had significant additional ALI risk factors ("possible TRALI") and those who did not ("suspected" TRALI) (see Table E1).
Of 74 patients who developed ALI, 58 were treated with mechanical ventilation (48 invasive, 10 noninvasive) and 16 were treated with oxygen supplementation via face mask. The median duration of mechanical ventilation was 3.6 (1.6 to 7.1) days. Twenty-seven case subjects and 27 control subjects were mechanically ventilated at the time of transfusion. An additional 31 case subjects were started on mechanical ventilation (26 invasive, 5 noninvasive) because of the development of ALI. Hospital mortality was higher in patients who developed ALI (41%) than in matched control subjects (23%) (P < 0.01) (see also the online supplement).
The principal findings of our study are the following: (1) in a cohort of critically ill medical patients, the development of ALI shortly after transfusion was more common than usually appreciated and an order of magnitude more common than what is reported to blood banks; (2) both underlying patient characteristics ("first hit") and specific transfusion factors ("second hit") were associated with the development of ALI after blood transfusion; and (3) our results support both of the proposed TRALI mechanisms as well as the general multiple-hit model of TRALI and ALI. To our knowledge, this is the first time the recommended standardized clinical definition (3, 4, 22) was implemented in a prospective study of TRALI. In critically ill medical patients admitted to a tertiary care medical center, clinically defined TRALI appears to be more common than previously thought, supporting the general notion that TRALI is grossly underrecognized and underreported (3–5). Rather than concentrating on clinically suspected TRALI reactions reported to blood banks, we prospectively examined consecutive transfused critically ill medical patients. Patients who developed ALI after transfusion often had other important ALI risk factors not related to the transfusion itself, for example, sepsis and a history of chronic alcohol abuse. The fact that potentially modifiable transfusion risk factors (such as donor sex, parity, and alloimmunization) are commonly associated with the development of ALI may have important implications in the prevention and treatment of ALI in critically ill medical patients. Observational data already suggest a decrease in incidence of ALI in patients treated according to a less liberal transfusion policy (32, 33). Moreover, preliminary data from the United Kingdom suggest a significant decrease in postoperative ALI associated with a decreased use of FFP from female donor plasma (34). Our results to some extent support both of the main proposed mechanisms for TRALI (6, 7, 35, 36) as well as the overall multiple-hit paradigm in ALI development (13, 14). Both the presence of anti-leukocyte antibodies and the concentration of bioactive lipid factors in the donor product were higher in patients who developed ALI than in matched control subjects both with and without adjustment for baseline characteristics ("first hit"). Although the small sample size precludes wider inferences, a closer look at the distribution of antibodies and LysoPC across different blood products (see Tables E3 and E4) reveals a trend toward more antibody-positive units given to patients with ALI versus control subjects across each of the blood products (red blood cells, FFP, and platelets). On the other hand, a higher concentration of LysoPC seems merely to reflect the larger number of high plasma volume products (platelets and FFP) given to patients with ALI versus control subjects. Our study design did not allow us to determine the sensitivity and specificity of laboratory diagnosis of TRALI, or to assess the specific etiologic (as opposed to a biomarker) role of anti-leukocyte antibodies and LysoPC. The association between infusion of plasma from female donors and the subsequent development of lung injury is an intriguing finding of our study and has important implications regarding the etiology and prevention of TRALI. Our findings support the results of the small randomized trial by Palfi and colleagues, in which plasma transfusion from multiparous female donors led to worsening in oxygenation and increased inflammatory response (35). The U.K. blood system has nearly eliminated female donors from the production of high plasma volume products (FFP) and the first reports suggest a significant decrease in TRALI reactions reported to the national transfusion surveillance system (SHOT [serious hazards of transfusion]) (37). The AABB TRALI Working Group has recommended similar guidelines for North American blood centers, with a plan to eliminate transfusion of high plasma volume products from donors at high risk of leukocyte alloimunization (including previously pregnant female donors) by November of 2008 (28). Because this intervention may be associated with costs and/or shortage of blood products in some blood centers, further studies will need to assess whether the benefit of this intervention outweighs the costs. Relatively high mortality of our suspected or possible TRALI cases is comparable to a report from Wallis and colleagues (38) and our retrospective study of the critically ill (19), but higher than in some other epidemiological studies (6). It is important to emphasize the characteristics of the population studied: medical ICU patients in a tertiary care medical center, a group with an inherently high mortality rate. The observed mortality may or may not be attributable to TRALI as these critically ill patients had multiple additional risk factors for both ALI and poor outcome. Our study design includes some limitations. The study was conducted in a single medical ICU in a tertiary care center and, although internal validity may be high, external validity is limited. In addition, the observational nature of the study does not allow independent estimation of the cause-and-effect relationship between the predictors and outcome. Residual confounding is always an issue in observational studies but is unlikely to fully explain our observed differences in donor product characteristics (anti-leukocyte antibodies from previously pregnant female donors and the concentration of LysoPC). Because there is no way that clinicians could have specifically ordered "male" or "low LysoPC" product in our institution, indication bias should be less of an issue. Although the trend toward higher amounts of total plasma could explain some of the observed differences it is again unlikely to explain the association between the donor sex and pregnancy and the development of ALI. An important limitation of our study is that we did not test for corresponding leukocyte antigens and were not able to determine the specificities of donor antibodies. We did not perform additional testing or treatment of the sera (e.g., adsorption of HLA class I antibodies with platelets) to confirm whether our granulocyte immunofluorescence-positive samples were indeed directed against granulocyte-specific antigens. Because the fresh granulocyte panels used for the assay could not be selected to include all possible granulocyte antigens, the potential exists that antibodies directed toward uncommon granulocyte antigens would not be detected. Therefore, we could not distinguish between the etiologic, rather than a biomarker, role of specific antibodies. In a number of units given to both case and control subjects we were unable to extract an adequate sample for laboratory analysis. Although the proportion of units tested for anti-leukocyte antibodies was similar for case and control subjects, a somewhat higher proportion of control units was tested for IL-8 and LysoPC. This difference, if anything, would make it less likely that a significant difference would be detected by chance alone. Although the differences in pretransfusion risk factors may indicate a limitation of our matching procedure, we were able to distinguish important, but not as well appreciated, patient risk factors, that is, chronic alcohol abuse. Moss and coworkers previously reported a strong independent association between alcohol use and the development of ALI (27). Laboratory experiments suggest the depletion of antioxidant glutathione stores associated with chronic alcohol abuse as the most plausible mechanistic explanation (39). Because neutrophil-mediated oxidative injury is the hallmark of TRALI (7, 8, 36, 40), it seems plausible that antioxidant depletion may be associated with greatly increased risk of this type of ALI. In conclusion, by prospectively applying the standardized clinical definition, we have found a high incidence of suspected and possible TRALI among transfused critically ill medical patients. The association between specific donor and transfusion characteristics and subsequent development of ALI has important implications relative to both the etiology and prevention of this syndrome. Ongoing multicenter studies in transfused patients with and without underlying critical illness will allow us to further elucidate the incidence and mechanisms of TRALI and the effect of potential preventive strategies.
Supported by the NHLBI (1 K23 HL087843-01A1) and by a grant from the National Blood Foundation (NBF 2004-05). This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org Originally Published in Press as DOI: 10.1164/rccm.200702-271OC on July 12, 2007 Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript Received in original form February 16, 2007; accepted in final form July 12, 2007
Related articles in AJRCCM:
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
AT A GLANCE COMMENTARY
= 0.05 level test. Conditional logistic regression was used to compare specific transfusion factors after the adjustment for pretransfusion ALI risk factors. SAS statistical software was used for all analyses (SAS version 9; SAS Institute, Inc., Cary, NC). 
receptors are essential in a mouse model of transfusion-related acute lung injury. J Clin Invest 2006;116:1615–1623.
