HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to Personal Folders Download to citation manager Author home page(s): Colleen Koch Tomislav Mihaljevic Lars Svensson Eugene H. Blackstone Permission Requests Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Koch, C. Articles by Blackstone, E. H. Search for Related Content PubMed PubMed Citation Articles by Koch, C. Articles by Blackstone, E. H. Related Collections Cardiac - other

---

Original Articles: Adult Cardiac

Transfusion and Pulmonary Morbidity After Cardiac Surgery

^a Department of Cardiothoracic Anesthesia, Cleveland Clinic, Cleveland, Ohio
^b Department of Quantitative Health Sciences, Cleveland Clinic, Cleveland, Ohio
^c Department of Laboratory Medicine and Clinical Pathology, Cleveland Clinic, Cleveland, Ohio
^d Department of Thoracic and Cardiovascular Surgery, Cleveland Clinic, Cleveland, Ohio

Accepted for publication July 10, 2009.

^_* Address correspondence to Dr Koch, Department of Cardiothoracic Anesthesia, J4-245, Cleveland Clinic, 9500 Euclid Ave, Cleveland, OH 44195 (Email: kochc{at}ccf.org).

	Abstract

Top Abstract Introduction Patients and Methods Results Comment Appendix 1 Appendix 2 Acknowledgments References

 
Background: True lung injury is among the leading causes of transfusion-related mortality. Pulmonary morbidity after cardiac surgery has been related to damaging effects of cardiopulmonary bypass and transfusion, but is confounded by cardiac-related events that may not reflect true lung injury. Thus, cardiac surgery poses unique challenges to criteria-specific diagnosis of transfusion-related acute lung injury (TRALI). Our objective was to determine the prevalence of pulmonary morbidity related to transfusion and whether TRALI consensus-criteria are applicable to cardiac surgery.

Methods: A total of 16,847 patients underwent on-pump, coronary artery bypass grafting (CABG), valve, or CABG-valve surgery from September 1998 to February 1, 2006. We performed four propensity-score-matching analyses with logistic regression on probability of receiving a transfusion: total hospital red blood cell (RBC) and fresh frozen plasma (FFP) transfusion and intraoperative RBC and FFP transfusion. Outcomes included traditional cardiac-surgery-defined pulmonary morbidity and ratio of arterial partial pressure of oxygen to fractional inspired oxygen concentration (PaO ₂/FiO ₂), a criterion for TRALI.

Results: Patients receiving RBC transfusion had more risk-adjusted pulmonary complications: respiratory distress 4.8% vs 1.5%, p < 0.001; respiratory failure 2.2% vs 0.39%, p < 0.0001; longer intubation times, 9.9 hours vs 7.5 hours, p < 0.0001; acute respiratory distress syndrome, 0.64% vs 0.21%, p = 0.015; and reintubation, 5.6% vs 1.3%, p < 0.0001. The FFP was similarly related to more pulmonary complications after surgery. By TRALI criteria, the majority manifested "lung injury" (PaO ₂/FiO ₂ ratio < 300) but unrelated to transfusion (65% vs 64%).

Conclusions: Transfusion is associated with many measures of postoperative pulmonary morbidity. Yet the PaO ₂/FiO ₂ ratio as important criterion of TRALI is unrelated to transfusion. Thus, due to the nature of cardiac surgery, application of consensus guided diagnosis of TRALI is problematic.

	Introduction

Top Abstract Introduction Patients and Methods Results Comment Appendix 1 Appendix 2 Acknowledgments References

 
Various measures of pulmonary morbidity after cardiac surgery have been related to cardiopulmonary bypass [1] and blood product administration [2–4]. True lung injury as it relates to transfusion of blood products is termed transfusion-related acute lung injury (TRALI), and has been characterized as adult respiratory distress syndrome (ARDS) due to transfusion [4, 5]. Patients undergoing cardiac surgery pose a unique challenge when attempting to apply standard definitions of TRALI, in part because measures of pulmonary morbidity may not reflect lung injury but cardiac events that prolong intubation support [6]. Current criteria for diagnosis of TRALI [7, 8] characterize it as onset of acute respiratory distress, evidence of hypoxemia, as reflected in arterial partial pressure of oxygen to fractional inspired oxygen concentration (PaO ₂/FiO ₂) ratio less than 300, presence of bilateral lung infiltrates on chest radiograph, and all temporally related to transfusion within a 6 hour time period. Other manifestations of TRALI include fever and mild to moderate hypotension [3, 4, 9].

In cardiac surgery, approximately half of transfusions occur in the operating room and the remainder in the postoperative period, thus making the timing of temporal relationships of transfusion to lung injury difficult. A majority of cardiac surgical patients arrive to the intensive care unit (ICU) on ventilatory support; consequently, clinical symptoms of dyspnea and acute respiratory distress are absent. In addition, patients arrive to the ICU mildly hypothermic; hence, increases in patient temperature are also not evident. Therefore, some suggest that, for a definitive diagnosis of TRALI, specific lung injury risk factors such as use of cardiopulmonary bypass [8] should be absent.

In order to better understand whether we could reasonably apply standard TRALI criteria or components of consensus criteria in the context of the cardiac surgical population we performed the following: (1) examined pulmonary morbidity related to transfusion both intraoperatively and postoperatively in a large cohort of patients undergoing cardiac surgery with use of cardiopulmonary bypass; (2) determined whether transfusions of red blood cell (RBC) and fresh frozen plasma (FFP) were associated with "lung injury" using a surrogate measurement, PaO ₂/FiO ₂; and (3) because of current consensus in England regarding female plasma donors, we investigated whether donor gender was associated with recipient pulmonary morbidity or lung injury.

	Patients and Methods

Top Abstract Introduction Patients and Methods Results Comment Appendix 1 Appendix 2 Acknowledgments References

 
Patients
Between September 1998 and February 2006, 16,847 patients underwent coronary artery bypass grafting (CABG), a valve procedure, or combination of CABG and valve surgery at Cleveland Clinic. Baseline demographics and perioperative variables were obtained from the Department of Cardiothoracic Anesthesia Registry. Registry information was prospectively collected concurrent with patient care by trained database personnel. The Cardiovascular Information Registry, a similar prospective registry, was also accessed for additional variables. Data from these registries are approved for use in research by the Institutional Review Board with patient consent waived. To examine the impact of donor gender on lung injury we received information from American Red Cross Biomedical Services on donor gender for products received from the American Red Cross supplier to Cleveland Clinic Blood Bank.

Pulmonary Outcomes
Pulmonary morbidity outcomes analyzed included the following: prolonged postoperative ventilation beyond 72 hours; respiratory failure defined documented need for reintubation secondary to respiratory failure; respiratory distress defined requirement for FiO ₂ greater than 80% for more than 24 hours or use of continuous positive airway pressure mask support; ARDS, defined as a clinical syndrome of PaO ₂ less than 60 mm Hg and significant infiltrate on chest radiograph, decreased lung compliance less than 0.5 mL/cm H₂O, and accurately documented by the infectious disease and the ICU teams; readmission to the ICU for respiratory failure; and total intubation time (initial intubation and reintubation duration) and total length of ICU stay (primary admission and readmission to ICU). The outcome variable reflective of lung injury for operating room transfusion only was PaO ₂/FiO ₂ ratio measured immediately after surgery on admission to the intensive case unit (ICU). (PaO ₂/FiO ₂ ratio of 300 mm Hg or less was considered a measure of hypoxemia [8]). A ratio less than 300 mm Hg was considered to be reflective of lung injury secondary to intraoperative events in our patient setting.

Statistical Methods
Primary statistical analyses of this study included four propensity score matching analyses [10, 30]. We propensity-matched patients on baseline and perioperative variables [Figs 1–4]. Using this technique we compared pulmonary outcomes with the following: (1) patients with and without RBC transfusion during the hospital stay; (2) patients with and without FFP transfusion during the hospital stay; (3) patients with and without RBC transfusion restricted to operating room only; and (4) patients with and without FFP transfusion also limited to the operating room only.

View larger version (22K): [in this window] [in a new window]   Fig 1. (A) Standardized differences of variables between patients who received red blood cell transfusions during the hospital stay and those who did not. Red crosses symbolize differences before propensity matching and blue open circles symbolize differences after propensity matching. Propensity matching effectively reduced heterogeneity among variables between the two patient populations in comparison. (* log transformed prior to the analysis). (B) Mirror histogram of propensity scores for patients who received red blood cell transfusion during the hospital stay (below the horizontal line at zero) and those who did not (above the horizontal line at zero). Matched patients are a subset of the original patient population and their volumes are highlighted in color. Matched groups have similar propensity score distributions. (BMI = body mass index; CABG = coronary artery bypass grafting; COPD = chronic obstructive pulmonary disease; ITA = internal thoracic artery; LVF = left ventricular function; NYHA = New York Heart Association.)

View larger version (18K): [in this window] [in a new window]   Fig 2. (A) Standardized differences of variables between patients who received fresh frozen plasma transfusion during the hospital stay and those who did not. Red crosses symbolize differences before propensity matching and blue open circles symbolize differences after propensity matching. Propensity matching effectively reduced heterogeneity among variables between the two patient populations in comparison. (* log transformed prior to the analysis.) (B) Mirror histogram of propensity scores for patients who received fresh frozen plasma transfusion during the hospital stay (below the horizontal line at zero) and those who did not (above the horizontal line at zero). Matched patients are a subset of the original patient population and their volumes are highlighted in color. Matched groups have similar propensity score distributions. (BMI = body mass index; CABG = coronary artery bypass grafting; COPD = chronic obstructive pulmonary disease; ITA = internal thoracic artery; LVF = left ventricular function; MI = myocardial infarction; NYHA = New York Heart Association.)

View larger version (19K): [in this window] [in a new window]   Fig 3. (A) Standardized differences of variables between patients who received red blood cell transfusion during the operation and those who did not. Red crosses symbolize differences before propensity matching and blue open circles symbolize differences after propensity matching. Propensity matching effectively reduced heterogeneity among variables between two comparison groups. (* log transformed prior to the analysis.) (B) Mirror histogram of propensity scores for patients who received red blood cell transfusion during the operation (below horizontal line at zero) and those who did not (above horizontal line at zero). Matched patients are a subset of the original patient population and their volumes are highlighted in color. Matched groups have similar propensity score distributions. (BMI = body mass index; CABG = coronary artery bypass grafting; COPD = chronic obstructive pulmonary disease; ITA = internal thoracic artery; LVF = left ventricular function; MI = myocardial infarction; NYHA = New York Heart Association; OR = operating room.)

View larger version (19K): [in this window] [in a new window]   Fig 4. (A) Standardized differences of variables between patients who received fresh frozen plasma transfusion during the operation and those who did not. Red crosses symbolize differences before propensity matching and blue open circles symbolize differences after propensity matching. Propensity matching effectively reduced heterogeneity among variables between groups. (* log transformed prior to the analysis.) (B) Mirror histogram of propensity scores for patients who received fresh frozen plasma transfusion during the operation (below the horizontal line at zero) and those who did not (above the horizontal line at zero). Matched patients are a subset of the original patient population and their volumes are highlighted in color. Matched groups have similar propensity score distributions. (BMI = body mass index; CABG = coronary artery bypass grafting; COPD = chronic obstructive pulmonary disease; ITA = internal thoracic artery; LVF = left ventricular function; MI = myocardial infarction; NYHA = New York Heart Association; OR = operating room.)

We developed a linear prognostic model for the PaO ₂/FiO ₂ ratio using the predictors considered in the propensity score analysis. Splines were used to account for nonlinearity in some continuous predictors. Because almost all predictors were statistically significant, no model selection was done.

We studied the effect of donor gender being female on the outcomes in an exploratory analysis. Details of that analysis are included in Appendix 1.

Significance level of all tests was set at 0.05. Analyses and plotting were carried out using SAS 9.1 (SAS Institute Inc, Cary, NC) and R 2.6.2 (http://www.r-project.org). Further details of overall statistical methodology can be found in Appendix 2.

	Results

Top Abstract Introduction Patients and Methods Results Comment Appendix 1 Appendix 2 Acknowledgments References

 
Pulmonary Morbidity and Transfusion
A total of 79,530 red blood cell and component blood product units were transfused. Among propensity-matched patients, those who received RBC transfusion had more pulmonary morbidity postoperatively (Table 1). They had longer intubation time, higher prevalence of respiratory distress and failure, and more frequently required reintubation for pulmonary-related reasons. Prevalence of ARDS was higher. They also had more readmissions to the ICU, and longer overall ICU length of stay.

View larger version (32K): [in this window] [in a new window]   Fig 5. Frequency histograms of PaO 2/FiO 2 ratio of those patients receiving red blood cell (RBC) transfusions in the operating room (OR) (PaO 2/FiO 2 < 250 [49.9%], PaO 2/FiO 2 < 300 [65%]), and those that did not (PaO 2/FiO 2 < 250 [49%], PaO 2/FiO 2 < 300 [64%]). (PaO 2/FiO 2 = ratio of arterial partial pressure of oxygen to fractional inspired oxygen concentration.)

Emergency surgery, higher New York Heart Association (NYHA) functional class, longer cardiopulmonary bypass duration, and operating room FFP transfusion were among a number of perioperative variables related to significantly lower the PaO ₂/FiO ₂ ratio (see Table 3).

	Comment

Top Abstract Introduction Patients and Methods Results Comment Appendix 1 Appendix 2 Acknowledgments References

 
We report that transfusion was associated with a higher risk-adjusted prevalence of respiratory distress and failure, ARDS, reintubation for pulmonary-related reasons, longer total intubation time, more readmissions to the ICU for pulmonary-related reasons, and longer ICU length of stay. Similarly, FFP transfusion was associated with significantly more pulmonary complications in the postoperative period. Surprisingly, a substantial proportion of patients manifested (PaO ₂/FiO ₂ < 300) for both transfused and nontransfused groups, which is considered to be a marker of lung injury. Important determinants for a lower PaO ₂/FiO ₂ included duration of cardiopulmonary bypass, emergency surgery, clinical risk factors, and intraoperative FFP administration.

Cherry and colleagues [11] summarized current criteria for diagnosis of TRALI from the American-European Consensus Conference Definition of ALI [12], the National Heart, Lung and Blood Institute Definition of TRALI [8], and the European Haemovigilance Network Definition of TRALI [7]. Common among published criteria is the temporal relationship between transfusion and onset of clinical signs and symptoms. Acute onset of clinical symptoms commonly occurs within 1 to 2 hours of transfusion [4] with 100% of patients' clinical symptoms presenting within 6 hours of transfusion [5, 13]. Evidence of hypoxemia, bilateral pulmonary infiltrates, and absence of fluid overload are among TRALI criteria. In large case series, 72% necessitate ventilatory support and a majority have symptom resolution within 96 hours of presentation [13, 14]. Transfusion-related acute lung injury has been further differentiated as "possible TRALI" if the patient has an alternative risk factor for acute lung injury [8].

Patients who were transfused in our investigation had more pulmonary morbidity; however, lack of ability to apply clear-cut criteria for TRALI complicated our ability to diagnose true lung injury. Identification of TRALI in our investigation was complicated by lack of formal surveillance; inability to detect TRALI-specific symptoms in addition to mortality as a competing risk factor.

Differentiating TRALI from transfusion-associated circulatory overload (TACO), is also problematic, but probably pertinent to the cardiac surgery setting. Although TRALI can be considered "permeability edema" versus "hydrostatic edema" from TACO, clinical and radiographic manifestations are similar and both may coexist [15]. Patients come to surgery with cardiovascular disease, multiple comorbidities, in heart failure, and with poor left ventricular function and are precisely the high-risk patients receiving blood products. In addition, use of cardiopulmonary bypass exposes these patients to risk for acute lung injury. Others have had difficulty differentiating TRALI from TACO in the clinical practice setting [16]. Brain natriuretic peptide (BNP) and N-terminal pro-brain natriuretic (NT-pro-BNP) have been used in an attempt to differentiate these two in critically ill patients; however, neither test reliably distinguished these [17]. Popovsky [18] noted the overlap of hypoxemia and signs and symptoms between TACO and TRALI which complicated differentiation. However, the author highlighted problematic clinical consequences of TACO: need for ICU care and longer hospital length of stay [18, 19].

Furthermore, it is difficult to differentiate lung injury related to transfusion with PaO ₂/FiO ₂ TRALI criteria measures. Over 60% of our patients with and without transfusion manifested a ratio less than 300 on ICU admission and almost 50% of patients in the transfused and nontransfused groups manifested PaO ₂/FiO ₂ less than 250. Pulmonary dysfunction after cardiac surgery is well recognized. Contributing factors are use of cardiopulmonary bypass, general anesthesia, breach of pleura, and cessation of ventilation while on cardiopulmonary bypass [20, 21]. However, others have attributed lung injury to RBC transfusion rather than cardiopulmonary bypass duration as measured by a rise in pulmonary vascular permeability as measured with ⁶⁷Ga-transferrin pulmonary leak index [22].

Characterizing TRALI
All plasma-containing blood components have been implicated in TRALI [2–4]. In its most severe presentation, it appears clinically similar to ARDS; however, it has lower mortality and shorter time to clinical resolution [4, 23]. Silliman and colleagues [24] reported prevalence of TRALI as 1 in 1,120 cellular components, suggesting TRALI may be more common than previously recognized. In an investigation of recipients of blood components from a donor with a granulocyte antibody, Kopko and colleagues [14] suggested that TRALI may present with mild signs and symptoms such as dyspnea and oxygen desaturation without pulmonary edema. These authors noted lack of recognition and spectrum of clinical presentation may result in inappropriate treatment.

While the causative pathogenesis of lung injury in TRALI is unknown, the end result is increased vascular permeability and pulmonary edema [23]. One theory is that TRALI is immune-mediated with antibodies from donor plasma acting against recipient leukocyte antigens, resulting in recipient leukocyte activation [4, 23]. Another proposed theory is the "two-hit model" characterized by a "first hit" such as cardiac surgery which activates leukocytes; the "second hit" occurs with transfusion which initiates release of leukocyte cytotoxic contents [5, 25]. The immune-mediated and two-hit model of TRALI are not mutually exclusive and both are supported by clinical and experimental observations [26]. Other mechanisms include immune complex formation with complement activation and cytokine network activation [23].

Role of Gender and HLA antigens
In a review of 58 TRALI-related transfusion fatalities reported to the Center for Biologics Evaluation and Research from 1997 to 2002, Holness and colleagues [2] reported that FFP was the most common blood component implicated in the development of TRALI; followed by RBC, platelets, and cryoprecipitate-reduced plasma. Forty-eight of the 63 donors implicated in these TRALI deaths were tested for antihuman leukocyte antigen (HLA) and (or) antigranulocyte antibodies; 83% were found to be positive. Parity, particularly multiparity, was a risk factor for antibody development and a frequent finding among implicated female donors. Insunza and colleagues [27] reported that HLA antibodies were present in 18% of previously pregnant apheresis donors. Rates of HLA alloimmunization for parous women range from 17% to 40% [23, 28]. Current recommendations for TRALI risk reduction include reducing the use of plasma products from female donors and screening female and other at risk donors for anti-HLA antibodies, including HLA class II and antigranulocyte antibodies [2, 29]. Although there has been an observation that the majority of blood donors implicated in these reactions are females who have been sensitized to fetal antigens by multiple pregnancies, there are many blood products from multiparous donors that contain such antibodies and do not cause clinically apparent TRALI reactions [14, 23, 24]. We reported similar outcomes among gender groups; however, our investigation of donor gender was limited by the small number of morbid outcomes.

Limitations
This was an observational cohort investigation and the presence of unknown confounders could certainly influence the observed study results. Lack of formal TRALI surveillance was problematic; however, should a surveillance system been in place, the ability to apply TRALI consensus-specific criteria would have been challenging due to the nature of our patient population. Timing of transfusion as it relates to outcome is difficult because exact timing of transfusion is not a measured variable. Lung injury as expressed by the PaO ₂/FiO ₂ ratio was measured only on ICU admission; we were unable to determine the impact of transfusion on PaO ₂/FiO ₂ further into the postoperative period. However, the PaO ₂/FiO ₂ ratio as one of the metrics of TRALI is insensitive in the setting of cardiac surgery. Furthermore, our ability to examine the impact of donor gender was hindered by insufficient donor gender information and lack of information on parity of female donors.

Conclusion
Cardiac surgical patients with poor cardiopulmonary reserve, who received transfusion, had more risk-adjusted pulmonary morbidity after surgery. Whether increased pulmonary morbidity was related to TRALI, TACO, or both is uncertain in part because current consensus-criteria for TRALI are insufficient for the cardiac surgical population. Almost 80,000 units of blood products were transfused in our subset of cardiac surgical patients in a period of less than 10 years. According to current prevalence estimates, TRALI could have been identifiable in approximately 72 patients. While lack of formal surveillance was problematic, current consensus criteria is insensitive to the diagnosis of TRALI.

	Appendix 1

Top Abstract Introduction Patients and Methods Results Comment Appendix 1 Appendix 2 Acknowledgments References

 
Among the 16,847 patients in the study sample, 4,020 of them (23.86%) did not receive transfusion of any blood products (RBC, FFP, cryoprecipitate, and platelets). The 12,827 (76.14%) patients that had transfusions received a total of 79,530 units of blood products, among which 38,613 (48.55%) units had missing donor gender, 18,556 (23.33%) units came from female donors, and 22,361 (28.12%) units came from male donors. Among the 12,827 patients that received a transfusion, 1,365 (10.64%) had complete information on donor gender of all the units they received; 5,698 (44.42%) did not have any information on donor gender of all the units they received.

Among 1,365 patients with complete donor gender information of all the transfused units, 310 received exclusively female units, and 542 patients received exclusively male units. Because the donor gender is independent of a patient's condition, these two groups are similar in preoperative and intraoperative variables (this is confirmed in a two-group comparison of these variables, results not shown). Therefore, any difference in the outcomes can be attributed to the effect of donor gender. However, we did not find any difference between female donor versus male donor comparison in arterial partial pressure of oxygen to fractional inspired oxygen concentration (PaO ₂/FiO ₂) (median [25th quintile, 75th quintile]: 257 (197, 328), 250 (187,332), p = 0.56, respectively; ICU length of stay (median [25th quintile, 75th quintile]: 28.8 (24.4, 51.9), 30.0 (24.5, 55.3), p = 0.68, respectively; and intubation time (median [25th quintile, 75th quintile]: 8.25 (5.67, 13.1), 8.58 (5.72, 14.7), p = 0.38, respectively. Wilcoxon rank sum tests were used here for the first imputed data set. Because no statistical significance was found, we expected that the multiple imputation adjusted p values were not significant either. We did not find any difference in the binary morbid event because the incident rates of those events were very low and did not show up in these two small groups.

	Appendix 2

Top Abstract Introduction Patients and Methods Results Comment Appendix 1 Appendix 2 Acknowledgments References

 
In this Appendix, we provide a more detailed account of the statistical analysis plan used in this study. Primary analyses of this study included four propensity score matching analyses. To conduct the propensity score analysis, we first performed a logistic regression to calculate the predictive probability of a patient receiving a transfusion, given that patient's set of preoperative and intraoperative variables; this probability is the propensity score for that patient. It has been shown [Rosenbaum P, Rubin D. The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70:41–55] that patients with similar propensity scores have similarly distributed pre- and intraoperative variables, regardless of whether they received a transfusion or not. Therefore, assuming that all relevant confounders are considered, identification of a subset of patients from the transfusion and nontransfusion groups with similar propensity scores allows differences in outcomes between the two subsets to be attributed to transfusion itself.

A greedy match [Parsons L. Reducing bias in a propensity score matched-pair sample using greedy matching techniques. Proceedings of the 26th Annual SAS Users Group International Conference. Cary NC: SAS Institute 2001:214–26] algorithm was used to perform 1:1 matching between the transfusion and nontransfusion groups. In order to evaluate the quality of matching for each confounding variable, we calculated the standardized differences between the transfusion and nontransfusion groups, before and after matching. Good matching removes or reduces standardized differences in confounders between comparison groups. Standardized difference was defined to be the difference in means between the groups divided by the average standard deviation.

The analysis involved a summation of all RBC and FFP therapy transfused for the hospital stay and a number of outcomes reflective of lung injury in the postoperative period. This analysis included an examination of the following.

(1) Patients with and without RBC transfusion during the hospital stay; and

(2) patients with and without FFP transfusion during the hospital stay.

Outcomes reflective of lung injury in the second analysis included the following: pulmonary morbidity as defined by prolonged postoperative ventilation beyond 72 hours; respiratory failure defined by patients who were extubated after the surgical procedure with documented need for reintubation secondary to respiratory failure; respiratory distress was defined as a patient requiring an FiO ₂ of greater than 80% for more than 24 hours, or use of continuous positive airway pressure mask support; acute respiratory distress syndrome (ARDS), was defined as acute onset with clinical diagnosis of PaO ₂ less than 60 mm Hg, significant infiltrate on chest radiograph, decreased lung compliance less than 0.5 mL/cm water and documented by the infectious disease and the ICU teams; readmission to the ICU for respiratory failure; and total intubation time (initial intubation and reintubation duration) and total length of ICU stay (primary admission and readmission to ICU).

Propensity matching between patients with (n = 6,459) and without (n = 10,388) RBC transfusion during the hospital stay resulted in 4,388 matched-pairs. Propensity matching resulted in a similar distribution of perioperative variables between the two groups (Figs 1A and 1B) Propensity matching between patients with (n = 1,238) and without (n = 15, 609) FFP transfusion during the hospital stay resulted in 964 matched pairs. Propensity matching resulted in a similar distribution of baseline variables between the two groups. (Figs 2A and 2B).

(3) Patients with and without RBC transfusion restricted to operating room only.

(4) Patients with and without FFP transfusion also limited to the operating room only.

The outcome variable was the ratio of arterial partial pressure of oxygen to fractional inspired oxygen concentration (PaO ₂/FiO ₂) measured immediately after surgery on admission to the intensive care unit (ICU). (PaO ₂/FiO ₂ ratio of 300 mm Hg was considered a measure of hypoxemia [Kleinman S, Caulfield T, Chan P, et al. Toward an understanding of transfusion-related acute lung injury: statement of a consensus panel. Transfusion 2004;44:1774–89].). A ratio less than 300 mm Hg was considered to be reflective of lung injury secondary to intraoperative events in our patient setting.

Propensity matching between patients with (n = 5,237) and without (n = 11,610) intraoperative RBC transfusion during the hospital stay resulted in 2,941 matched-pairs. Propensity matching effectively reduced the heterogeneity among perioperative variables between the two groups. (Fig 3A; B) Propensity matching between patients with (n = 728) and without (n = 16,119) an intraoperative FFP transfusion resulted in 625 matched pairs. Following propensity matching baseline variables were comparable between the two groups (Fig 4A; B).

Descriptive statistics were calculated and compared between matched groups. Effect of transfusion was tested in a stratified regression model for the outcome, with each matched-pair as a stratum. For continuous outcomes, we used general linear model with cluster-specific random effects. Proper transformations on the continuous outcomes were used to meet the model assumptions. For binary outcomes, we used conditional logistic regression.

There were missing values in preoperative creatinine (260), preoperative hematocrit (2,770), body mass index (9), hypertension (150), diabetes (285), smoking (274), and PaO ₂/FiO ₂ ratio (902). We imputed the missing values five times using the multiple imputation method (SAS PROC MI). All analyses described above, including the development of the propensity score, greedy matching, and post-matching analysis, were repeated for each of five imputed data sets and results were combined by multiple imputation formula for proper inference and p values (SAS PROC MIANALYZE). The p values from the combined analysis were in general a little larger than the p values obtained from the analysis of a single imputed data set, which reflects the proper adjustment for the uncertainty in the missing data. Unless otherwise stated, all p values were adjusted for multiple imputations. For parsimony of presentation, descriptive statistics were reported only for the first imputed data set when there were no substantial differences between the five imputed data sets.

We developed a linear prognostic model for PaO ₂/FiO ₂ ratio using the predictors considered in the propensity score analysis. The PaO ₂/FiO ₂ ratio was square root transformed to make the residual assumptions of the model satisfied. The model building process started with a semiparametric generalized additive model in which all the continuous predictors were assumed to have a smooth but otherwise unrestricted relationship with the mean outcome variable. If the effect of the predictor appeared to be nonlinear, we took a monotone transformation of that predictor or approximated it with a low rank regression spline to reflect such nonlinear relationship in broad strokes. Hence, the final model was a parametric linear model with predictors and their spline terms. The estimates, standard errors and p values in the final model were adjusted appropriately to account for the uncertainty of multiple imputations. Since almost all predictors are significant, no model selection was done.

Significance level of all tests was set at 0.05. No p value adjustment for multiple statistical tests was made; however, all the endpoints were predefined during the planning stage of this study. The analyses and plotting were carried out using SAS 9.1 and R 2.6.2 (http://www.r-project.org).

	Acknowledgments

Top Abstract Introduction Patients and Methods Results Comment Appendix 1 Appendix 2 Acknowledgments References

 
There was no external funding for this investigation. All support was obtained within the Cleveland Clinic. The authors had full control of the design of the study, methods used, and outcome parameters, analysis of data, and production of the written report.

	References

Top Abstract Introduction Patients and Methods Results Comment Appendix 1 Appendix 2 Acknowledgments References

 

1. Weiss YG, Merin G, Koganov E, et al. Postcardiopulmonary bypass hypoxemia: a prospective study on incidence, risk factors, and clinical significance J Cardiothorac Vasc Anesth 2000;14:506-513.[Medline]
2. Holness L, Knippen MA, Simmons L, Lachenbruch PA. Fatalities caused by TRALI Transfus Med Rev 2004;18:184-188.[Medline]
3. Kopko PM, Holland PV. Transfusion-related acute lung injury Br J Haematol 1999;105:322-329.[Medline]
4. Popovsky MA. Transfusion and lung injury Transfus Clin Biol 2001;8:272-277.[Medline]
5. Popovsky MA. Pulmonary consequences of transfusion: TRALI and TACO Transfus Apher Sci 2006;34:243-244.[Medline]
6. Murthy SC, Arroliga AC, Walts PA, et al. Ventilatory dependency after cardiovascular surgery J Thorac Cardiovasc Surg 2007;134:484-490.[Abstract/Free Full Text]
7. European Haemovigilance Network E. European Haemovigilance Network, (EHN) Definition of adverse transfusion eventshttp://www.ehn-org.net 2007.
8. Kleinman S, Caulfield T, Chan P, et al. Toward an understanding of transfusion-related acute lung injury: statement of a consensus panel Transfusion 2004;44:1774-1789.[Medline]
9. Silliman CC, Paterson AJ, Dickey WO, et al. The association of biologically active lipids with the development of transfusion-related acute lung injury: a retrospective study Transfusion 1997;37:719-726.[Medline]
10. Rosenbaum P, Rubin D. The central role of the propensity score in observational studies for causal effects Biometrika 1983;70:41-55.[Abstract/Free Full Text]
11. Cherry T, Steciuk M, Reddy VV, Marques MB. Transfusion-related acute lung injury: past, present, and future Am J Clin Pathol 2008;129:287-297.[Abstract/Free Full Text]
12. Bernard GR, Artigas A, Brigham KL, et al. The American-European Consensus Conference on ARDS. Definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994;149(3 pt 1):818-824.[Abstract/Free Full Text]
13. Popovsky MA, Moore SB. Diagnostic and pathogenetic considerations in transfusion-related acute lung injury Transfusion 1985;25:573-577.[Medline]
14. Kopko PM, Marshall CS, MacKenzie MR, Holland PV, Popovsky MA. Transfusion-related acute lung injury: report of a clinical look-back investigation JAMA 2002;287:1968-1971.[Abstract/Free Full Text]
15. Gajic O, Gropper MA, Hubmayr RD. Pulmonary edema after transfusion: how to differentiate transfusion-associated circulatory overload from transfusion-related acute lung injury Crit Care Med 2006;34(5 suppl):S109-S113.[Medline]
16. Hume H. TRALI: moving toward prevention Transfusion 2009;49:402-405.[Medline]
17. Li G, Daniels CE, Kojicic M, et al. The accuracy of natriuretic peptides (brain natriuretic peptide and N-terminal pro-brain natriuretic) in the differentiation between transfusion-related acute lung injury and transfusion-related circulatory overload in the critically ill Transfusion 2009;49:13-20.[Medline]
18. Popovsky MA. Transfusion-associated circulatory overload: the plot thickens Transfusion 2009;49:2-4.[Medline]
19. Popovsky MA, Audet AM, Andrzejewski Jr C. Transfusion-associated circulatory overload in orthopedic surgery patients: a multi-institutional study Immunohematology 1996;12:87-89.[Medline]
20. Ng CS, Wan S, Yim AP, Arifi AA. Pulmonary dysfunction after cardiac surgery Chest 2002;121:1269-1277.[Abstract/Free Full Text]
21. Wan S, LeClerc JL, Vincent JL. Inflammatory response to cardiopulmonary bypass: mechanisms involved and possible therapeutic strategies Chest 1997;112:676-692.[Abstract/Free Full Text]
22. Groeneveld A, Jansen E, Verheij J. Mechanisms of pulmonary dysfunction after on-pump and off-pump cardiac surgery: a prospective cohort study J Cardiothorac Surg 2007;2:111–7.[Medline]
23. Curtis BR, McFarland JG. Mechanisms of transfusion-related acute lung injury (TRALI): anti-leukocyte antibodies Crit Care Med 2006;34(5 suppl):S118-S123.[Medline]
24. Silliman CC, Boshkov LK, Mehdizadehkashi Z, et al. Transfusion-related acute lung injury: epidemiology and a prospective analysis of etiologic factors Blood 2003;101:454-462.[Abstract/Free Full Text]
25. Silliman CC. The two-event model of transfusion-related acute lung injury Crit Care Med 2006;34(5 suppl):S124-S131.[Medline]
26. Looney MR, Matthay MA. Animal models of transfusion-related acute lung injury Crit Care Med 2006;34(5 suppl):S132-S136.[Medline]
27. Insunza A, Romon I, Gonzalez-Ponte ML, et al. Implementation of a strategy to prevent TRALI in a regional blood centre Transfus Med 2004;14:157-164.[Medline]
28. Bux J. Transfusion-related acute lung injury (TRALI): a serious adverse event of blood transfusion Vox Sang 2005;89:1-10.[Medline]
29. American Association of Blood Banks. Transfusion-related acute lung injury. Association Bulletin November 3, 2006;#06-07(Bethesda, MD).
30. Parsons L. Reducing bias in a propensity score matched-pair sample using greedy matching techniquesProceedings of the 26th Annual SAS Users Group International ConferenceCary NC: SAS Institute; 2001214–226.

This article has been cited by other articles:

				G. Paone, R. Brewer, P. F. Theurer, G. F. Bell, C. M. Cogan, R. L. Prager, and Michigan Society of Thoracic and Cardiovascular Su Preoperative predicted risk does not fully explain the association between red blood cell transfusion and mortality in coronary artery bypass grafting J. Thorac. Cardiovasc. Surg., January 1, 2012; 143(1): 178 - 185. [Abstract] [Full Text] [PDF]

				A. Shander, M. Javidroozi, S. Ozawa, and G. M. T. Hare What is really dangerous: anaemia or transfusion? Br. J. Anaesth., December 1, 2011; 107(suppl_1): i41 - i59. [Abstract] [Full Text] [PDF]

				A. Alfirevic, M. Xu, D. Johnston, P. Figueroa, and C. G. Koch Transfusion Increases the Risk for Vasoplegia After Cardiac Operations Ann. Thorac. Surg., September 1, 2011; 92(3): 812 - 819. [Abstract] [Full Text] [PDF]

				A. Ballotta, H. Kandil, T. Generali, L. Menicanti, G. Pelissero, M. Ranucci, and Surgical and Clinical Outcome Research (SCORE) Gro Tracheostomy After Cardiac Operations: In-Hospital and Long-Term Survival Ann. Thorac. Surg., August 1, 2011; 92(2): 528 - 533. [Abstract] [Full Text] [PDF]

				S. A. Falk and A. T. Cheung Invited Commentary Ann. Thorac. Surg., August 1, 2011; 92(2): 533 - 534. [Full Text] [PDF]

				A. Vivacqua, C. G. Koch, A. M. Yousuf, E. R. Nowicki, P. L. Houghtaling, E. H. Blackstone, and J. F. Sabik III Morbidity of Bleeding After Cardiac Surgery: Is It Blood Transfusion, Reoperation for Bleeding, or Both? Ann. Thorac. Surg., June 1, 2011; 91(6): 1780 - 1790. [Abstract] [Full Text] [PDF]

				G. J. Fitzmaurice and H. D. Parissis Platelet Transfusion Associated With Acute Lung Injury After Coronary Artery Bypass Grafting Ann. Thorac. Surg., June 1, 2011; 91(6): 1977 - 1979. [Abstract] [Full Text] [PDF]

				A. P. J. Vlaar, J. J. Hofstra, R. M. Determann, D. P. Veelo, F. Paulus, W. Kulik, J. Korevaar, B. A. de Mol, M. M. W. Koopman, L. Porcelijn, et al. The incidence, risk factors, and outcome of transfusion-related acute lung injury in a cohort of cardiac surgery patients: a prospective nested case-control study Blood, April 21, 2011; 117(16): 4218 - 4225. [Abstract] [Full Text] [PDF]

				B. H. Shaz, S. R. Stowell, and C. D. Hillyer Transfusion-related acute lung injury: from bedside to bench and back Blood, February 3, 2011; 117(5): 1463 - 1471. [Abstract] [Full Text] [PDF]

				R. Hall and C. D. Mazer Review Article: Antiplatelet Drugs: A Review of Their Pharmacology and Management in the Perioperative Period Anesth. Analg., February 1, 2011; 112(2): 292 - 318. [Abstract] [Full Text] [PDF]

				D. M. Moskowitz, J. N. McCullough, A. Shander, J. J. Klein, C. A. Bodian, R. S. Goldweit, and M. A. Ergin The Impact of Blood Conservation on Outcomes in Cardiac Surgery: Is It Safe and Effective? Ann. Thorac. Surg., August 1, 2010; 90(2): 451 - 458. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to Personal Folders Download to citation manager Author home page(s): Colleen Koch Tomislav Mihaljevic Lars Svensson Eugene H. Blackstone Permission Requests Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Koch, C. Articles by Blackstone, E. H. Search for Related Content PubMed PubMed Citation Articles by Koch, C. Articles by Blackstone, E. H. Related Collections Cardiac - other