| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
* From the Departments of Anesthesia, Health Policy, Management, and
Evaluation,
Cardiac Surgery, and
Haematology, University Health Network, University of Toronto, Toronto, Ontario, Canada.
Address correspondence to: Dr. Keyvan Karkouti, University Health Network, Department of Anesthesia, Toronto General Hospital, EN 3-402, 200 Elizabeth Street, Toronto, Ontario M5G 2C4, Canada. Phone: 416-340-5164; Fax: 416-340-3698; E-mail: keyvan.karkouti{at}uhn.on.ca
 |
Abstract |
|---|
 TOP Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Methods: In this observational study, detailed baseline and perioperative data were prospectively collected on consecutive patients who underwent cardiac surgery at a single institution from 1999 to 2004. The independent associations of platelet transfusion with clinical outcomes (low output syndrome, myocardial infarction, stroke, renal failure, sepsis, and death) were determined by multivariable logistic regression analysis and propensity score case-control analysis.
Results: Of the 11,459 patients analyzed, 2,174 (19%) received (leukoreduced) platelets – 1,408 received 5 U, 471 received 10 U, 140 received 15 U, and 155 received 20 or more units. Although all measured adverse event rates were higher in those who received platelets, in neither the logistic regression analyses nor the propensity score analyses was there any association between platelet transfusion and any of the adverse events.
Conclusions: Transfusion of leukoreduced platelets in cardiac surgery is not associated with adverse clinical outcomes when adjustments are made for important confounders.
|
Introduction |
|---|
|
TOPAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Unmeasured confounders are an issue in observational studies comparing patients who received RBCs to those who did not because the decision to transfuse RBCs is, at least in part, based on clinical ‘gestalt’ rather than easily measured clinical variables. All else being equal, patients who are perceived to be sicker are more likely to be transfused RBCs, and they are also more likely to suffer adverse events. Platelet transfusion decisions, on the other hand, are generally not based on patients’ underlying medical status; rather, they are based on routinely measured criteria, such as platelet count and cardiopulmonary bypass (CPB) duration in cardiac surgery, that can be controlled by multivariable analysis techniques.
Observational studies, therefore, may be able to better control for the effects of confounders if they base their comparisons on platelet transfusions rather than RBC transfusions. To date, however, only two observational studies have directly assessed the association of platelet transfusions with adverse clinical outcomes. These studies had contradictory findings and were limited by small sample sizes or incomplete adjustment for confounding variables.3,4 We carried out this retrospective study in a large cohort of cardiac surgery patients to determine the independent association of platelet transfusions with several adverse postoperative events by using multivariable logistic regression analyses and propensity score-based case-control analyses to control for the effects of important confounders.
|
Methods |
|---|
|
TOPAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
RED BLOOD CELL TRANSFUSION
Hematocrit concentrations were measured frequently during the perioperative period: before CPB, every 15 min during CPB, every 30 min after CPB in the operating room, upon arrival and every four hours thereafter in the intensive care unit for the first 24 hr after surgery (more frequently in bleeding or unstable patients). Turnaround time for hematocrit concentration results was five to ten minutes. Leukoreduced RBC concentrate was transfused to maintain the hematocrit concentration above 21 to 24% in stable patients, and 
27% in unstable or bleeding patients during the perioperative period. During CPB, hematocrit concentration was maintained at greater than 17–20%.
PLATELET TRANSFUSION
Platelet counts were routinely measured before surgery and upon arrival to the intensive care unit after surgery. During surgery, platelet counts were measured at the discretion of the clinical team in patients with prolonged CPB times or microvascular hemorrhage following termination of CPB and neutralization of heparin. Turnaround time for platelet count results were 20 to 45 min. Indications for platelet transfusion included a platelet count of < 50 x 109·L–1, ongoing hemorrhage after complete reversal of heparin and a platelet count < 80 x 109·L–1, or ongoing hemorrhage after prolonged CPB (> two hours was considered prolonged by most clinicians) irrespective of plate-let counts. In cases where the risk of developing a qualitative platelet disorder after CPB was deemed to be very high (e.g., recent use of long-acting platelet glycoprotein IIb/IIIa receptor antagonist therapy or CPB > three hours), platelets were often transfused prophylactically upon termination of CPB irrespective of platelet counts or bleeding status. All platelets (random donor or single donor) were leukoreduced by the Canadian Blood Services.
OTHER BLOOD PRODUCTS
Institutional guidelines recommended that fresh frozen plasma (FFP) be transfused to patients with non-surgical bleeding if their international normalized ratio was > 1.5 after complete neutralization of heparin with protamine, and cryoprecipitate be transfused if the fibrinogen level was < 1 g·L–1. Turnaround time for these coagulation tests were 30 to 60 min. Given the long turnaround time, these products were at times administered empirically in massively bleeding patients.
Patient population and data collection
Following Institutional Ethics approval, data on consecutive adult patients (> 18-yr-old) undergoing cardiac surgery from June 1999 to June 2004 were obtained from two prospectively collected, validated, and accurate databases. These databases have been previously described.5 Full-time research personnel blinded to the details of this current study adjudicated all patient outcomes included in the databases.
For patients who underwent more than one cardiac operation during the study period, only data from their first operation was used. Missing values were completed from the medical records if possible; otherwise, patients with missing categorical variable values were excluded. For continuous variables, missing values were imputed based on the mean for the entire sample.
Statistical analyses
Statistical analyses were performed using SASTM version 8.2 (SAS Institute, Inc., Cary, NC, USA).
ADVERSE EVENTS
Measured adverse postoperative events were: low output syndrome (use of inotropes or mechanical devices for > 30 min to maintain blood pressure > 90 mmHg with a cardiac index < 2.2 L·min–1·m2 despite appropriate preload and afterload conditions), stroke (any new persistent postoperative neurological deficit), acute renal failure (new requirement for dialysis support), myocardial infarction (new Q wave on postoperative electrocardiogram; or MB isoenzyme of creatine kinase > 50 U·L–1, CK-MB/CK ratio > 5%, and new electrocardiogram changes), sepsis (requiring a positive blood culture), and in-hospital death. A composite outcome of any of the above was also analyzed.
INDEPENDENT VARIABLES
All measured covariates that could be related to plate-let transfusion, perioperative hemorrhage, or any of the dependent variables were assessed for inclusion in the multivariable analysis and propensity score case-control matching.
MULTIVARIABLE LOGISTIC REGRESSION
Multivariable logistic regression modelling was carried out on the entire sample to assess the independent relationships between platelet transfusion (both as a binomial variable – yes or no – and as a continuous variable – number of units transfused) and each of the adverse events. First, bivariate analysis (using the Chi-square statistic for categorical variables and the t test or Wilcoxon rank-sum test for continuous variables) was carried out to identify which independent variables were associated (P < 0.1) with platelet transfusion and adverse events for inclusion in the modelling. Modelling was by backward logistic regression, with P < 0.05 as the criteria for variable retention in the models. Various classifications (categorical and continuous) of important confounders (such as RBC and plasma transfusions) were included in the modelling process. Linearity and collinearity issues were managed according to standard methodolgies.6,7 A separate model was constructed for each of the measured adverse events to determine if platelet transfusion remained in any of the models.
PROPENSITY SCORE-BASED MATCHING
Patient matching based on propensity scores was employed to obtain a precise and approximately unbiased estimate of treatment effect by balancing measured covariates in transfused and untransfused patients.8 The propensity score derivation model for receiving a platelet transfusion was derived by multivariable logistic regression using all measured independent variables that could be related to platelet transfusion or perioperative hemorrhage (32 variables), as well as important two-way interaction terms. The model was then used to derive the predicted probability (or propensity score) of platelet transfusion for each patient.
Individual patients who received platelets were then matched 1:1 to patients who did not receive platelets on the basis of similar propensity scores. A 5 
1 computerized greedy matching technique was employed for this matching process whereby cases were first matched to controls that had a propensity score that was identical in all five digits. Those that did not match were then matched to controls on four digits of the propensity score. This continued down to a one–digit match on propensity score for those that remained unmatched.9 Measured covariates and adverse postoperative events in these matched pairs were compared using paired t test or Wilcoxon signed-rank test for continuous variables, and conditional matched-pair logistic regression for categorical variables.10,11
|
Results |
|---|
|
TOPAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Of the 11,459 patients, 2,174 (19.0%) received platelets, 6,657 (58.1%) received RBC transfusions, and 3,227 (28.2%) received FFP. Of those who received platelets, 1,408 (64.8%) received 5 U, 471 (21.7%) received 10 U, 140 (6.4%) received 15 U, and 155 (7.1%) received 20 U or more of platelets (median = 5 U; tenth and ninetieth percentiles = 5 and 15 U). (In our analyses, 1 U of apheresis single donor platelets was counted as 5 U of random donor platelets).
Table I
contains the descriptive statistics for the measured covariates for patients who received platelets and those who did not. As expected, the two groups differed on many covariates. Table II contains the adverse event rates according to number of units of platelets transfused; as can be seen, patients who were transfused platelets had higher adverse event rates when no adjustments were made for confounders.
|
|
Adjusting for confounders by propensity score matching
The propensity score derivation model, which consisted of 22 variables (three of which were interaction terms), included, in order of importance: FFP transfused (three categories: 0 U, 1–4 U, and > 4 U), preoperative platelet count (three categories: > 150 x 109·L–1, between 100 x 109·L–1 – 150 x 109·L–1, and < 100 x 109·L–1), CPB duration (continuous variable), and massive perioperative RBC transfusion (binomial variable: 5 U within one day of surgery) – had excellent discrimination (c-index = 0.95) and was well calibrated (Hosmer-Lemeshow test Chi-square = 9.75; P = 0.28). On the basis of similar propensity scores, 924 (42.5%) patients who received platelets were matched to 924 untransfused, control patients. Patients were well matched for all measured covariates except for number of patients who received 1 U of RBC transfusion, which was higher in the platelet group (Table III). (The propensity score derivation model included two other variables that were related to perioperative blood loss: number of units of FFP transfused, and transfusion of 5 U of RBC within one day of surgery, and the patients were well matched on these variables.) Before matching, the mean (± standard deviation) propensity scores in the transfused and untransfused patients were 0.67 ± 0.29 and 0.08 ± 0.16, respectively. After matching, the mean scores were 0.44 ± 0.24 in both groups.
|
). The primary composite outcome occurred in 11.4% of the platelet group and 12.2% of the control group (P = 0.6). The study had 80% power to detect a 4% absolute increase in the rate of this composite outcome.
|
|
Discussion |
|---|
|
TOPAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
An unwanted effect of platelet transfusion is immune modulation, which is caused in part by plate-let-derived (as a result of platelet activation or lysis) and white blood cell (WBC)-derived cytokines that accumulate during storage.12 Platelet related immune modulation has long been implicated as the cause of febrile non-hemolytic transfusion reactions.12 More recently, it has been postulated that the immunomodulatory effects of platelets may have more general detrimental effects that increase overall morbidity and mortality.13 The present study’s findings are not in agreement with this hypothesis.
The association between platelet transfusions and adverse outcomes has been assessed in only two other studies.3,4 Vamvakas and Carven found no association between the volume of platelet supernatant transfused and respiratory dysfunction in their retrospective multivariable analysis of 416 coronary artery bypass grafting patients.3 Spiess et al., on the other hand, found a substantial increase in the risk of adverse postoperative events associated with platelet transfusion in their retrospective study on 1,720 cardiac surgery patients.4 There were several important differences between this study and ours. First, patients in our study received leukoreduced platelets whereas platelets were not leukoreduced in the study by Spiess et al. Leukoreduction is thought to reduce the immunomodulatory effects of blood products by attenuating the production of WBC-derived cytokines in stored blood products.14,15 To date, however, no clear link has been demonstrated between leukoreduction and reduced blood product transfusion related morbidity and mortality.2 Second, our study included a larger and potentially more generalizable cohort of cardiac surgery patients. And third, we adjusted for a larger set of confounders. One important confounding variable is massive perioperative RBC transfusion, which is strongly associated with both platelet transfusion and adverse postoperative events.5 This point is illustrated in the Figure, which plots the relationship between the composite adverse event rate and number of units of platelets transfused for the study sample as a whole, in the subgroup of patients that received less than 5 U of RBCs, and in the subgroup that received 5 or more units of RBCs. The dose-effect relationship between platelet transfusion and the composite adverse event rate is largely explained by this single confounding variable. Other important confounders that we adjusted for included difficult wean from CPB and patients’ baseline plate-let count, both of which are associated with platelet transfusions as well as with postoperative outcomes.16 When our data were analyzed without adjustment for these three confounders, platelet transfusions were associated with an increased incidence of low output syndrome (P = 0.0001), renal failure (P = 0.05), death (P = 0.03), as well as prolonged duration of mechanical ventilation (P = 0.0002), intensive care unit stay (P < 0.0001) and hospitalization (P = 0.001).
|
In contrast to this study’s findings for platelet transfusion, RBC transfusion has almost uniformly been associated with increased morbidity or mortality in numerous observational studies.1 There are at least two potential reasons for the discrepancy between these findings and ours. One is that the mechanism of injury related to RBC transfusion may be distinct from that related to platelet transfusion. Red blood cell transfusion, which result in the release of many of the same immunomodulatory cytokines that accumulate in units of stored platelets and are thought to play a role in TRIM, may lead to organ injury by reducing microcirculatory oxygen delivery.18–22 As discussed earlier, a second reason for the discrepant findings may be that unmeasured confounders play a greater role in RBC transfusions vs platelet transfusions. This concept is best demonstrated by a hypothetical example: two identical coronary artery bypass grafting surgery patients (with respect to comorbidities, intraoperative course, and laboratory values) present with a hemoglobin concentration of 80 g·L–1 in the immediate post-CPB period. Patient-1 has an uneventful postoperative course but Patient-2 develops sustained ST-segment depression consistent with myocardial ischemia. Given that myocardial ischemia in the setting of low hemoglobin concentration is an indication for RBC transfusion,22 and post-CPB ST-segment depression is associated with adverse cardiac outcomes,23 Patient-2 is more likely to both receive a RBC blood transfusion and to suffer a myocardial infarction. In contrast, Patient-2 will probably not receive a platelet transfusion, and therefore no association between myocardial infarctions and platelet transfusions will be apparent. Thus, unless the incidence, duration, and temporal pattern of perioperative ST-segment depression are measured and properly adjusted for by multivariable analysis, an observational study comparing patients who did and did not receive RBC transfusions will find a spurious association between RBC transfusion and postoperative myocardial infarction. This (and other) confounding variables that affect RBC transfusion decisions, however, are not routinely measured.
In conclusion, in this large, retrospective study, no association was found between transfusion of leukoreduced platelets and postcardiac surgery morbidity or mortality. Although not a randomized controlled clinical trial, the results of this study should allay any concerns physicians may have about transfusing leukoreduced platelets to patients who would otherwise need them. Nevertheless, given the important role of confounders in observational studies examining the effects of blood product transfusion on morbidity and mortality, a definitive randomized controlled trial to determine the true relationship between the transfusion of different blood components and adverse events is needed.
|
Footnotes |
|---|
Accepted for publication September 1, 2005. Revision accepted September 15, 2005.
|
References |
|---|
|
TOPAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
2 Raghavan M, Marik PE. Anemia, allogenic blood transfusion, and immunomodulation in critically ill. Chest 2005; 127: 295–307.[Medline]
3 Vamvakas EC, Carven JH. Allogeneic blood transfusion and postoperative duration of mechanical ventilation: effects of red cell supernatant, platelet supernatant, plasma components and total transfused fluid. Vox Sang 2002; 82: 141–9.[Medline]
4 Spiess BD, Royston D, Levy JH, et al. Platelet transfusions during coronary artery bypass graft surgery are associated with serious adverse events. Transfusion 2004; 44: 1143–8.[Medline]
5 Karkouti K, Wijeysundera DN, Yau TM, et al. The independent association of massive blood loss with mortality in cardiac surgery. Transfusion 2004; 44: 1453–62.[Medline]
6 Katz MH. Multivariable Analysis: a Practical Guide for Clinicians, 1st ed. Cambridge: Cambridge University Press; 1999.
7 Feinstein AR. Multiple logistic regression. In: Feinstein AR (Ed.). Multivariable Analysis: an Introduction. New Haven: Yale University Press; 1996: 297–330.
8 D’Agostino RB Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med 1998; 17: 2265–81.[Medline]
9 Parson LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques. Proceedings of the Twenty-sixth Annual SAS® Users Group International Conference, Long Beach, CA, 2001.
10 Schlesselman JJ. Case-control studies: design, conduct, analysis. New York: Oxford University Press; 1982.
11 Allison PD. Logit analysis of longitudinal and other clustered data. In: Allison PD (Ed.). Logistic Regression Using SAS System: Theory and Application. Cary, NC: SAS Institute and Wiley; 1991: 179–216.
12 Boehlen F, Clemetson KJ. Platelet chemokines and their receptors: what is their relevance to platelet storage and transfusion practice? Transf Med 2001; 11: 403–17.[Medline]
13 Spiess BD. Risks of transfusion: outcome focus. Transfusion 2004; 44: 4S–14S.[Medline]
14 Corwin HL, AuBuchon JP. Is leukoreduction of blood components for everyone? JAMA 2003; 289: 1993–5.
15 Heddle NM. Universal leukoreduction and acute transfusion reactions: putting the puzzle together (Editorial). Transfusion 2004; 44: 1–4.[Medline]
16 Stephan F, Hollande J, Richard O, Cheffi A, Maier-Redelsperger M, Flahautl A. Thrombocytopenia in a surgical intensive care unit: incidence, risk factors and outcome. Chest 1999; 115: 1363–70.[Medline]
17 Itescu S, Tung TC, Burke EM, et al. Preformed IgG antibodies against major histocompatibility complex class II antigens are major risk factors for high-grade cellular rejection in recipients of heart transplantation. Circulation 1998; 98: 786–93.
18 Kristiansson M, Soop M, Saraste L, Sundqvist KGl. Cytokines in stored red blood cell concentrates: promoters of systemic inflammation and simulators of acute transfusion reactions? Acta Anaesthesiol Scand 1996; 40: 496–501.[Medline]
19 Zallen G, Moore EE, Ciesla DJ, Brown M, Biffl WL, Silliman CC. Stored red blood cells selectively activate human neutrophils to release IL-8 and secretory PLA2. Shock 2000; 13: 29–33.[Medline]
20 Buttnerova I, Baumler H, Kern F, et al. Release of WBC-derived IL-1 receptor antagonist into supernatants of RBCs: influence of storage time and filtration. Transfusion 2001; 41: 67–73.[Medline]
21 Lin JS, Tzeng CH, Hao TC, et al. Cytokine release in febrile non-haemolytic red cell transfusion reactions. Vox Sang 2002; 82: 156–60.[Medline]
22 Spiess BD. Transfusion of blood products affects outcome in cardiac surgery. Semin Cardiothoracic Vasc Anest 2004; 8: 267–81.
23 Spahn DR. Strategies for transfusion therapy. Best Pract Res Clin Anaesthesiol 2004; 18: 661–73.[Medline]
24 Smith RC, Leung JM, Mangano DT. Postoperative myocardial ischemia in patients undergoing coronary artery bypass graft surgery. S.P.I. Research Group. Anesthesiology 1991; 74: 464–73.[Medline]
This article has been cited by other articles:
 |
|
 |
|
  T. McGrath, C. G. Koch, M. Xu, L. Li, T. Mihaljevic, P. Figueroa, and E. H. Blackstone Platelet transfusion in cardiac surgery does not confer increased risk for adverse morbid outcomes. Ann. Thorac. Surg., August 1, 2008; 86(2): 543 - 553. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Karkouti and S. A. McCluskey REPLY Can J Anesth, April 1, 2008; 55(4): 249 - 250. [Full Text] [PDF]
|
|
|
|
|
|
M. Ranucci, A. Pazzaglia, C. Bianchini, G. Bozzetti, and G. Isgro Body Size, Gender, and Transfusions as Determinants of Outcome After Coronary Operations Ann. Thorac. Surg., February 1, 2008; 85(2): 481 - 486. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. C. Harle Point-of-Care Platelet Function Testing Seminars in Cardiothoracic and Vascular Anesthesia, December 1, 2007; 11(4): 247 - 251. [Abstract] [PDF]
|
|
|
|
|
|
Y. Lin, N. Kanani, F. Naughton, J. Pendergrast, and K. Karkouti Case report: Transfusion-related acute lung injury (TRALI) A clear and present danger: [Presentation de cas : Lesion pulmonaire aigue associee a la transfusion (TRALI) Un danger net et present] Can J Anesth, December 1, 2007; 54(12): 1011 - 1016. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Karkouti REPLY Can J Anesth, December 1, 2007; 54(12): 1027 - 1028. [Full Text] [PDF]
|
|
|
|
|
|
The Society of Thoracic Surgeons Blood Conservatio, V. A. Ferraris, S. P. Ferraris, S. P. Saha, E. A. Hessel II, C. K. Haan, B. D. Royston, C. R. Bridges, R. S.D. Higgins, G. Despotis, et al. Perioperative Blood Transfusion and Blood Conservation in Cardiac Surgery: The Society of Thoracic Surgeons and The Society of Cardiovascular Anesthesiologists Clinical Practice Guideline Ann. Thorac. Surg., May 1, 2007; 83(5_Supplement): S27 - S86. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Karkouti, T. M. Yau, S. Riazi, K. M. Dattilo, M. Wasowicz, M. Meineri, S. A. McCluskey, D. N. Wijeysundera, A. van Rensburg, and W. S. Beattie Determinants of complications with recombinant factor VIIa for refractory blood loss in cardiac surgery: [Les determinants de complications avec le facteur VIIa recombinant utilise pour les pertes sanguines refractaires en chirurgie cardiaque]. Can J Anesth, August 1, 2006; 53(8): 802 - 809. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Karkouti, S. Shayan, D. N. Wijeysundera, S. A. McCluskey, M. Ghannam, and W. S. Beattie Prognostic value of postoperative proteinuria in cardiac surgery: a pilot study: [Valeur pronostique de la proteinurie postoperatoire en chirurgie cardiaque : une etude pilote]. Can J Anesth, May 1, 2006; 53(5): 500 - 506. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
