How To Find Out If A Platelet Transfusion After A Coronary Artery Bypass Surgery Is Associated With Adverse Outcomes

Transcription

1 European Journal of Cardio-Thoracic Surgery Advance Access published September 2, 2015 European Journal of Cardio-Thoracic Surgery (2015) 1 8 doi: /ejcts/ezv297 ORIGINAL ARTICLE Cite this article as: Kremke M, Hansen MK, Christensen S, Tang MAndreasen JJ, Jakobsen C-J. The association between platelet transfusion and adverse outcomes after coronary artery bypass surgery. Eur J Cardiothorac Surg 2015; doi: /ejcts/ezv297. The association between platelet transfusion and adverse outcomes after coronary artery bypass surgery Michael Kremke a, *, Malene Kærslund Hansen b, Steffen Christensen a, Mariann Tang c, Jan Jesper Andreasen d and Carl-Johan Jakobsen a ADULT CARDIAC a Department of Anaesthesiology and Intensive Care, Aarhus University Hospital, Aarhus, Denmark b Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark c Department of Cardiothoracic and Vascular Surgery, Aarhus University Hospital, Aarhus, Denmark d Departments of Cardiothoracic Surgery and Clinical Medicine, Aalborg University Hospital, Aalborg, Denmark * Corresponding author. Department of Anaesthesiology and Intensive Care, Aarhus University Hospital, Skejby, Palle Juul-Jensens Boulevard 99, 8200 Aarhus N, Denmark. Tel: ; fax: ; michhinr@rm.dk. (M. Kremke). Received 6 February 2015; received in revised form 2 July 2015; accepted 20 July 2015 Abstract OBJECTIVES: Previous research suggests that platelet transfusion is associated with adverse events after coronary artery bypass grafting (CABG). The aim of the current analysis was to verify this hypothesis. METHODS: Data from 6745 consecutive patients undergoing CABG from 2006 through 2012 were collected. Patients receiving platelet transfusions intraoperatively or postoperatively in the intensive care unit were compared with control patients. To adjust for possible confounders, propensity score matching and conditional regression analyses were performed. Short-term outcomes were 30-day mortality, in-hospital myocardial infarction and stroke. Mid-term outcomes were 6-month mortality, and need for coronary angiography or repeat coronary revascularization within 6 months after surgery. Data were retrieved from the Western Denmark Heart Registry. RESULTS: Using propensity scores, 982 patients exposed to platelets were matched with 982 control patients. Platelet transfusion was associated with a higher rate of postoperative coronary angiography (adjusted odds ratio 2.34, 95% confidence interval ). There was no significant association between platelet transfusion and postoperative mortality, myocardial infarction, stroke and need for repeat coronary revascularization. CONCLUSIONS: Platelet transfusion at the time of CABG is not associated with increased postoperative mortality, in-hospital myocardial infarction, stroke or need for repeat coronary revascularization. Keywords: Coronary artery bypass Myocardial ischaemia Platelet transfusion Stroke Thrombosis INTRODUCTION Coronary artery bypass grafting (CABG) is the most common cardiac surgical procedure, with procedures performed annually worldwide. Due to advances in surgical techniques and myocardial protection, CABG is now considered a procedure with a relatively low perioperative risk. This is, however, counterbalanced by the increasing age and comorbidities of patients presenting for CABG. As a result, perioperative ischaemic events, such as myocardial infarction or stroke, are still a significant cause of disability and death. Strategies have been sought to reduce the risk of perioperative organ ischaemia. In 2002, Mangano and colleagues [1]publisheda large cohort analysis addressing the importance of platelets in CABG. The authors demonstrated that early initiation of antiplatelet therapy after CABG was associated with a markedly reduced risk of death and postoperative ischaemic complications. At the same time, Mangano and colleagues observed a strong relationship between perioperative platelet transfusion and increased postoperative mortality. The latter finding raised concerns about the safety of platelet transfusions, which are commonly used in cardiac surgery [2]. Subsequent analyses on the risk of platelet transfusion, however, yielded conflicting results [3 7]. The objective of the current analysis was to examine the association between platelet transfusion at the time of CABG and adverse postoperative events, including death, myocardial infarction and stroke. MATERIALS AND METHODS Study design This is a cohort analysis of 6745 consecutive patients undergoing CABG from 1 April 2006 to 31 December All three university The Author Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved.

2 2 M. Kremke et al. / European Journal of Cardio-Thoracic Surgery hospitals in the Western Denmark region (Aalborg, Aarhus and Odense University Hospital) participated in the study. Western Denmark is home to 3 million inhabitants, more than half of the total Danish population. Eligible procedures were both isolated CABG and CABG combined with aortic valve replacement. Both, on- and off-pump procedures were included, regardless of the urgency status of surgery. We excluded patients undergoing repeat CABG (within 6 months after primary CABG), patients preoperatively exposed to antiplatelet agents other than aspirin or clopidogrel, and patients preoperatively treated with low molecular weight heparin in therapeutic dosage. Patients who received at least one unit of platelets either intraoperatively or during the primary postoperative stay in the intensive care unit were assigned to the platelet cohort. Patients who did not receive platelets were assigned to the control cohort. Both cohorts were subsequently matched in a 1:1 ratio using propensity scores. The study design and patient allocation are shown in Fig. 1. Material The Western Denmark Heart Registry was used as the data source [8]. This regional registry gathers prospectively collected clinical data from all cardiothoracic procedures performed at the participating hospitals. Reporting to this web-based registry has been mandatory since its establishment in Detailed patient-, surgery-, anaesthesia- and intensive care-related data are reported, as are data on perioperative transfusion requirements and inhospital complications. Since April 2006, all data input related to this analysis has been obligatory. Patient characteristics and surgical procedures are described primarily by the European System for Cardiac Operative Risk Evaluation (EuroSCORE) variables [9]. The Western Denmark Heart Registry obtains data on mortality from the Danish Civil Registration System [10]. This system keeps daily updated records on residence, migration, vital status and date of death for all Danish citizens. The study was approved by the Danish Data Protection Agency ( ). Patient and outcome characteristics All outcome measures were in accordance with the prespecified classifications used in the Western Denmark Heart Registry. Short-term outcomes were 30-day mortality (all deaths occurring within 30 days of the primary surgery) and the incidence of myocardial infarction and stroke during the index hospitalization. Figure 1: Study design and cohort allocation depending on perioperative platelet transfusion. AVR: aortic valve replacement; CABG: coronary bypass grafting; LMWH: low molecular weight heparin.

3 M. Kremke et al. / European Journal of Cardio-Thoracic Surgery 3 Myocardial infarction was defined as the occurrence of a new Q-wave and/or a creatine kinase MB isoenzyme (CK-MB) plasma level greater than 100 µg/l. As CK-MB values are affected by a variety of factors, clinicians were asked to register myocardial infarction also when the diagnosis of myocardial infarction had been made by other means. Stroke was a combined outcome, defined as either a transitory ischaemic attack lasting less than 24 h or a new neurological deficit lasting more than 24 h. Only in-hospital strokes were registered. We did not distinguish between haemorrhagic or ischaemic stroke. Mid-term outcomes were 6-month mortality (all deaths occurring within 6 months of the primary surgery), and the need for coronary angiography, repeat CABG or percutaneous coronary intervention (PCI) within 6 months after the primary operation. Only coronary angiographies with the indication suspected myocardial infarction, unstable angina pectoris, complication after CABG or control after CABG were included. Coronary angiographies with other indications, e.g. stable angina pectoris or cardiac arrhythmia, were excluded. Likewise, we excluded PCIs that were part of a planned hybrid procedure ( patients preoperatively scheduled for combined CABG/PCI). All repeat CABGs within 6 months after the index CABG were registered. Perioperative management During the study period, most patients undergoing elective CABG discontinued aspirin and oral adenosine diphosphate receptor antagonists 5 days prior to surgery. Patients at increased risk of acute ischaemic events, e.g. patients with acute coronary syndromes or recent PCI, typically continued aspirin until the day of surgery. Vitamin K antagonists were usually withheld 2 3 days preoperatively. A platelet count and an international normalized ratio were routinely measured prior to surgery. Whether CABG was done on- or off-pump was left to the discretion of the attending surgeon. Intraoperative surgical techniques were those routinely used at the participating institutions. Patients received general anaesthesia with invasive haemodynamic monitoring and standardized cardiopulmonary bypass (CPB). Before initiation of CPB, heparin was administered to achieve an activated clotting time (ACT) greater than 400 s. Lower ACT values were often accepted during off-pump procedures. If necessary, additional heparin was given during CPB. A haematocrit of greater than 24% was typically accepted during CPB. The CPB circuit consisted of tubing, a hollow fibre membrane oxygenator and arterial filter with a surface-modifying additive coating, and a cardiotomy reservoir. Myocardial protection was achieved by either intermittent cold crystalloid or blood cardioplegia. During CPB, most patients were maintained either normothermic or mildly hypothermic. In general, blood flow rates were 2.4 l/min/m 2 with mean arterial pressures between 50 and 60 mmhg. After weaning from CPB, heparin was neutralized using protamine sulphate (Leo Pharma AS, Ballerup, Denmark). Blood conservation techniques varied between the participating hospitals. The majority of patients received either tranexamic acid or aprotinin as the agent of antifibrinolysis. Residual blood from the CPB circuit was routinely retransfused at the end of surgery. A platelet count was routinely determined upon arrival in the intensive care unit. Extubated patients typically stayed in the intensive care unit overnight and were discharged the morning following surgery, provided that their cardiorespiratory and neurological status was stable. Aspirin therapy was, in general, initiated or resumed the morning following surgery. At the same time, thrombosis prophylaxis with low molecular weight heparin was initiated and administered for at least 5 days, or until the patient was fully mobilized. In general, dual antiplatelet therapy was not resumed immediately after surgery, but typically withheld for at least 3 days after CABG. At this time, a chest X-ray was taken to rule out relevant amounts of pleural effusion. Blood products were given at the discretion of the attending anaesthesiologist or surgeon. Local transfusion guidelines existed at all hospitals. Additionally, national recommendations regarding blood product use were available. Platelet concentrates were leucocyte depleted and prepared from four pooled buffy coats. Indications for platelet transfusion varied between participating hospitals. In bleeding patients, a platelet count below /l after termination of CPB was frequently considered an indication for platelet transfusion. In patients exposed to platelet inhibitors prior to surgery or patients with prolonged CPB times, platelets were commonly given irrespective of platelet counts when there were signs of microvascular bleeding after termination of CPB. Point-of-care coagulation tests became available at two of the three participating hospitals during the study period. Thromboelastometry (ROTEM, Tem International GmbH, Munich, Germany) was used in severely bleeding patients or patients with signs of microvascular bleeding. Platelet function tests were not routinely used. Statistical analysis In a pilot analysis, we compared patients who received platelets with patients who did not receive platelets and observed a difference in the incidence of overall non-discriminated postoperative events (24 vs 12%, respectively). A power calculation based on the pilot analysis with the (two) sided α error set at 0.05 and β error set at 0.2 (power of 80%) demonstrated that a minimum of 230 patients were needed in each cohort. Categorical variables were described as count and percentages, and compared using the χ 2 test. Continuous variables were described using median values with interquartile ranges. Analyses to compare continuous data were done using the Wilcoxon Mann Whitney U-test. A P-value less than 0.05 was considered to be statistically significant. We used propensity score matching to reduce the risk of bias by confounding and non-random assignment of platelet transfusion. By modelling the exposure rather than the outcome, propensity scores efficiently allow for simultaneous control of a large number of potentially confounding factors [11]. The propensity score was computed using multivariable logistic regression. The included covariates were age, sex, chronic obstructive pulmonary disease, peripheral artery disease, preoperative central nervous disease, preoperative serum creatinine >200 mmol/l, previous cardiac surgery, critical preoperative state, unstable angina pectoris, reduced left ventricular function, recent myocardial infarction, pulmonary hypertension, acute surgery (all above EuroSCORE criteria) [9], surgery type (CABG ± aortic valve replacement), extracorporeal circulation time (off-pump, 120 min, >120 min), insulin-dependent diabetes, perioperative aprotinin use, perioperative fibrinogen use, no antiplatelet therapy within 5 days prior to surgery, aspirin therapy within 5 days prior to surgery, clopidogrel therapy within 5 days prior to surgery and dual antiplatelet therapy within 5 days prior to surgery. ADULT CARDIAC

4 4 M. Kremke et al. / European Journal of Cardio-Thoracic Surgery We aimed to match each patient exposed to platelet transfusion with a control patient with the nearest propensity score within a maximum caliper range of ±0.025 and without replacement. Covariates were adequately balanced after propensity score matching, as evidenced by a standardized difference of each covariate to values below 0.1 (Fig. 2). Mortality was described by cumulative incidence and odds ratios (ORs), stratified on the matched pairs. The risk of the nonfatal outcomes was estimated using ORs. Using conditional regression analysis, we adjusted the OR for perioperative red blood cell transfusion (yes/no), fresh frozen plasma transfusion (yes/no), perioperative inotropic therapy and postoperative acute renal failure. By stratifying and conditioning, we took the non-independence within each matched pair into account [12]. Inotropic therapy was defined as the infusion of milrinone, dobutamine, dopamine, epinephrine and/or levosimendan for more than 1 h intraoperatively or postoperatively in the intensive care unit. Postoperative acute renal failure was defined as a new need for dialysis during the index hospitalization, without differentiating between the precise indication for dialysis or the actual mode of renal replacement therapy. The statistical analyses were performed with MedCalc Statistical Software version (MedCalc Software, Ostend, Belgium). Propensity score matching and conditional regression analyses Figure 2: Standardized differences of all matching criteria before and after propensity score matching. AVR: aorta valve replacement; CABG: coronary artery bypass grafting; CNS: central nervous system; COPD: chronic obstructive pulmonary disease; ECC: extracorporeal circulation; ES: EuroSCORE; LV: left ventricular; MI: myocardial infarction.

5 M. Kremke et al. / European Journal of Cardio-Thoracic Surgery 5 were done using the statistical software package Stata 12.0 (StataCorp LP, TX, USA). RESULTS A total of 7812 consecutive patients were included (Fig. 1). Of them, 1067 (14%) were excluded from the study according to the exclusion criteria mentioned above. Of the 6745 remaining patients in the study cohort, 1220 (18%) received allogeneic platelet transfusion intraoperatively or postoperatively in the intensive care unit. Baseline characteristics and operative data are summarized in Table 1. Prior to the matching process, patients exposed to platelets were significantly older and had greater comorbidity than patients in the control cohort. They presented more often in a critical preoperative condition, more frequently with unstable angina pectoris, a recent myocardial infarction, reduced ventricular function ADULT CARDIAC Table 1: matching Baseline characteristics and operative data of the 6745 patients in the study cohorts, prior to and after propensity score Primary cohorts Control cohort (n = 5525) n (%) or median (IQR) Platelet cohort (n = 1220) P Propensity score-matched cohorts Control cohort Platelet cohort n (%) or median (IQR) P Female sex 1153 (20.9) 286 (23.4) (20.9) 223 (22.7) a Age 68 (62 74) 71 (63 77) < (64 76) 71 (64 77) b Insulin-dependent diabetes 407 (7.4) 95 (7.8) (8.4) 84 (8.6) a COPD 563 (10.2) 160 (13.1) (12.1) 119 (12.1) a Peripheral artery disease 712 (12.9) 161 (13.2) (12.4) 129 (13.1) a Preoperative neurological 353 (6.4) 97 (8.0) (6.6) 72 (7.3) a disease Previous cardiac surgery 92 (1.7) 37 (3.0) (2.4) 19 (1.9) a Serum creatinine >200 mmol/l 132 (2.4) 53 (4.3) < (4.3) 39 (4.0) a Endocarditis 61 (1.1) 24 (2.0) (1.6) 15 (1.5) a Critical preoperative condition 169 (3.1) 129 (10.6) < (6.9) 72 (7.3) a Unstable angina pectoris 394 (7.1) 256 (21.0) < (16.6) 167 (17.0) a Left ventricular function Normal (EF >50%) 3690 (66.8) 690 (56.6) < (59.6) 576 (58.7) a Reduced (EF 30 50%) 1501 (27.2) 387 (31.7) < (29.8) 310 (31.6) a Severely reduced (EF <30%) 329 (6.0) 143 (11.7) < (10.6) 96 (9.8) a Myocardial infarction <90 days 1140 (20.6) 453 (37.1) < (35.5) 351 (35.7) a Pulmonary hypertension 165 (3.0) 60 (4.9) (4.7) 42 (4.3) a Acute surgery 242 (4.4) 202 (15.6) < (12.0) 121 (12.3) a CABG 4615 (83.5) 897 (73.5) < (72.8) 734 (74.8) a CABG and AVR 910 (16.5) 323 (26.5) < (27.2) 248 (25.3) a Extracorporeal circulation time Off-pump 1033 (18.7) 132 (10.8) < (10.6) 118 (12.0) a 120 min 3518 (63.7) 688 (56.4) < (57.7) 560 (57.0) a >120 min 959 (17.4) 396 (32.5) < (31.7) 304 (31.0) a Perioperative aprotinin therapy 84 (1.5) 38 (3.1) < (3.2) 30 (3.1) a Perioperative fibrinogen 23 (0.4) 145 (11.9) < (2.0) 19 (1.9) a therapy Preoperative antiplatelet therapy None 4326 (78.3) 732 (60.0) < (64.2) 620 (63.1) a Aspirin only c 1010 (18.3) 320 (26.2) < (25.4) 254 (25.9) a Clodipogrel only c 56 (1.0) 24 (2.0) < (2.04) 18 (1.8) a Dual antiplatelet therapy c 133 (2.4) 144 (11.8) < (8.5) 90 (9.2) a Department Centre A 1249 (22.6) 454 (37.2) < (36.4) 346 (35.2) a Centre B 1834 (33.2) 308 (25.3) < (27.9) 278 (28.3) a Centre C 2442 (44.2) 458 (37.5) < (35.7) 357 (36.5) a Year of procedure (10.6) 114 (9.3) < (10.6) 105 (10.7) a (14.7) 175 (14.3) < (17.5) 164 (16.7) a (13.5) 190 (15.6) < (17.8) 154 (15.7) a (13.8) 214 (17.5) < (17.0) 172 (17.5) a (12.9) 183 (15.0) < (10.2) 105 (10.7) a (17.2) 178 (14.6) < (13.1) 142 (14.5) a (17.3) 166 (13.6) < (13.8) 140 (14.3) a AVR: aortic valve replacement; CABG: coronary artery bypass grafting; COPD: chronic obstructive pulmonary disease; EF: ejection fraction. a χ 2 test. b Wilcoxon Mann Whitney U-test. c Within 5 days prior to surgery.

6 6 M. Kremke et al. / European Journal of Cardio-Thoracic Surgery or pulmonary hypertension. Likewise, they were more frequently exposed to platelet inhibitors prior to surgery and, notably, the fraction of patients on preoperative dual antiplatelet therapy was significantly greater. Patients exposed to platelet transfusion more often underwent acute surgery, had longer ECC times and were more likely to undergo combined CABG and aortic valve surgery. Using propensity scores, we were able to match 982 (80%) patients in the platelet cohort with 982 (18%) control patients. The matching process resulted in two cohorts with equally distributed covariates (Table 1). In the matched cohort, we found in univariate analysis a significant association between perioperative platelet transfusion and increased postoperative mortality. Both 30-day mortality (OR 3.11, 95% CI ) and 6-month mortality (OR 2.16, 95% CI ) were higher in patients exposed to platelet transfusions compared with control patients (Table 2). Likewise, platelet administration was associated with greater rates of postoperative stroke, need for coronary angiography and PCI (Table 3). Although our study cohorts were matched according to baseline data, we suspected that confounding by indication was present. To verify this, we examined whether platelet administration was a surrogate for the severity of bleeding and whether platelet-treated CABG patients had prolonged stays in the intensive care unit. We defined major postoperative bleeding as either chest tube drainage volumes of more than 1200 ml or perioperative transfusion of more than 2400 ml red blood cells. Using these cut-off values, we could demonstrate that 64% of the propensity scorematched patients exposed to platelet transfusion experienced major postoperative bleeding, compared with only 14% of matched control patients. Additionally, 86% of all platelet-treated patients received red blood cells or fresh frozen plasma as well, compared with only 28% of all control patients. Furthermore, 14% of patients exposed to platelets had postoperative intensive care unit stays of 3 days or longer, compared with only 3% of control patients. These data indicated that patients in the platelet cohort were more likely to suffer serious bleeding and more frequently had a complicated postoperative course. To mitigate this confounding, we adjusted our data for red blood cell and fresh frozen plasma transfusion. Subsequently, there remained no significant association between platelet transfusion and greater mortality, myocardial infarction, stroke and need for repeat coronary revascularization (Tables 2 and 3). However, the use of allogeneic platelets was still an independent predictor for the need for postoperative coronary angiography. We further adjusted our data for the perioperative use of inotropic agents and postoperative acute renal failure, as it has previously been demonstrated that those are risk factors for worse outcomes after cardiac surgery [13, 14]. After adjustment, there was no significant association between platelet transfusion and postoperative mortality, risk of stroke, myocardial infarction and need for repeat coronary revascularization (Tables 2 and 3). Again, platelet transfusion remained associated with an increased rate of coronary angiography. Table 2: Postoperative mortality in the propensity score-matched cohorts Platelet cohort Control cohort Odds ratio Adjusted odds ratio a Adjusted odds ratio b Deaths within 30 days (n) ( ), P < ( ), P = ( ), P = Deaths within 6 months (n) ( ), P < ( ), P = ( ), P = Given as numbers and odds ratios (conditional regression). CI: confidence interval. a Adjusted for red blood cell and fresh frozen plasma transfusion. b Adjusted for postoperative acute renal failure, perioperative inotropic treatment, red blood cell and fresh frozen plasma transfusion. Table 3: In-hospital myocardial infarction and stroke, and number of coronary angiographies and repeat coronary revascularization procedures within 6 months postoperatively Platelet cohort n (%) Control cohort Odds ratio Adjusted odds ratio a Adjusted odds ratio b Myocardial infarction 36 (3.7) 28 (2.9) 1.30 ( ), P = ( ), P = ( ), P = Stroke 35 (3.6) 18 (1.8) 1.94 ( ), P = ( ), P = ( ), P = CAG 80 (8.1) 44 (4.5) 1.84 ( ), P = ( ), P = ( ), P = PCI 37 (3.8) 17 (1.7) 2.25 ( ), P = ( ), P = ( ), P = CABG 2 (0.2) 2 (0.2) 1.0 ( ) 1.0 (-) 1.0 (-) Given as univariate and adjusted odds ratio (conditional regression). CABG: coronary artery bypass grafting; CAG: coronary angiography; CI: confidence interval; PCI: percutaneous coronary intervention. a Adjusted for red blood cell and fresh frozen plasma transfusion. b Adjusted for postoperative acute renal failure, perioperative inotropic treatment, red blood cell and fresh frozen plasma transfusion.

7 M. Kremke et al. / European Journal of Cardio-Thoracic Surgery 7 DISCUSSION In the present analysis, we primarily found an association between perioperative platelet transfusion and risk of death, stroke and need for PCI, following adjustment for baseline differences using propensity scores. However, even though baseline variables were well balanced between cohorts, we observed that platelet-treated patients were more likely to suffer from major bleeding, had greater transfusion requirements and prolonged intensive care unit stays. Consequently, we took into account that confounding by indication was present and adjusted for relevant confounding variables. Subsequently, the association between platelets and postoperative death, stroke and need for PCI lost statistical significance. There did remain, however, an association between platelet transfusion and greater postoperative coronary angiography rates. Mangano and colleagues [1] were among the first who investigated the relationship between platelet transfusion at the time of CABG and postoperative ischaemic events. They described an almost four times increased mortality among platelet-treated patients and suggested that platelets fostered an ischaemic response to reperfusion, thereby worsening overall survival. The authors pointed out that the risk of death was substantially reduced, still being considerable, if patients concurrently received aspirin. Mangano and colleagues, however, did not take into consideration that platelets were administered in a critical care setting and almost exclusively to actively bleeding patients. It seems not unlikely that confounding by indication had biased Mangano s analysis, resulting in an overestimation of the risks of platelet administration. The observed worse outcome that the authors exclusively ascribed to platelet transfusion may, in fact, have been mediated by other factors, like the circulatory instability accompanying major blood loss or the concomitant administration of other blood products. Spiess et al. [3] analysed data of 1720 patients who underwent either elective primary or secondary CABG and reported a greater prevalence of both postoperative stroke and death among patients receiving platelet transfusions. The authors did not, however, fully adjust for confounding variables, such as the transfusion of other blood products. More recently, Mikolla et al. analysed 2234 CABG patients undergoing surgery at three hospitals in Finland and also found a greater risk of stroke in patients receiving large amounts of blood products, particularly platelet concentrates [6]. The association between platelets and stroke was strongest, moreover, in patients receiving more than eight units of platelets, which arguably should be a rare event in routine CABG. We assume that this subgroup of patients suffered severe bleeding complications or experienced prolonged stays in the intensive care unit, which would explain the extraordinary transfusion requirements and the greater stroke rates. Karkouti et al. published the results of a cohort analysis of platelet administration and adverse outcomes after CABG. Consistent with our data, they found neither an excess morbidity nor mortality in platelet-treated CABG patients [4]. We believe that the greater mortality and morbidity reported in previous analyses were not primarily related to platelet transfusion, but rather a result of the greater rate of bleeding complications or the inherent critical state of platelet-treated patients. We demonstrated, however, a higher rate of coronary angiography in platelet-treated patients. This finding is difficult to interpret, since PCI rates were not significantly different in the study cohorts. We can only speculate whether platelet-treated patients more frequently had postoperative signs of cardiac ischaemia, like unexplained low cardiac output syndrome, which necessitated diagnostic catheterization. Interestingly, neither we nor other authors have found an association between platelet administration and perioperative myocardial infarction. It would seem indisputable, however, that platelets play a central role in the pathogenesis of myocardial infarction. The initial thrombus in acute myocardial infarction is primarily composed of activated platelets [15]. The initial step in thrombus formation, the rupture of an atherosclerotic coronary plaque, might be comparable with the breach in coronary vascular integrity occurring during CABG. Both events create endothelial damage and inevitably lead to exposure of the thrombogenic subendothelial matrix, including the potent platelet activators collagen and tissue factor [16]. One of the strengths of our analysis is our large cohort of patients, which permits robust estimates for a number of serious adverse events. Use of prospectively recorded and detailed clinical data from an almost complete medical database limits the risk of selection, information and surveillance bias. There were, however, some important limitations. One weakness is the lack of random treatment assignment, which may have introduced confounding by indication. Although we used propensity score matching to adjust for a large number of pre- and perioperative covariates, and likewise adjusted for the use of other blood products, inotropic therapy and acute renal failure, residual confounding cannot entirely be ruled out. Excessive bleeding after CABG may not only result in greater transfusion requirements, but also necessitate surgical re-exploration and lead to delayed resumption of oral antiplatelet therapy. These events are closely related, and adjusting for these factors may lead to multicollinearity, hindering proper interpretation of the multivariate analysis. Uncontrolled confounding by frailty may have influenced our results, as frail high-risk patients are probably more likely to receive platelets and, at the same time, are at higher risk for adverse postoperative events. Additionally, frail patients are more likely to experience prolonged intensive care unit stays which, again, may result in both greater transfusion requirements and greater rates of adverse events, without a causal link between these outcomes. Frailty is notoriously difficult to measure. Although it is difficult to separate the impact of different treatment strategies in actively bleeding and critically ill patients, platelet transfusion at the time of CABG seems not to be an independent risk factor for postoperative myocardial infarction, stroke or death. It is, however, beyond the scope of our analysis to exclude a causal relationship between platelet administration and perioperative adverse events. There is no doubt, however, that the transfusion of allogeneic platelets carries the risk of immunological reactions and infections. Besides, platelets are a scarce and costly resource. We therefore believe that the administration of allogeneic platelets should not solely be guided by clinical judgement, but, whenever possible, be restricted to patients with critical bleeding or those with laboratory evidence of platelet dysfunction or low platelet count. It is desirable to conduct a large randomized trial on the risks of platelet transfusion. It may, however, be ethically infeasible to randomize patients to different platelet transfusion strategies. ADULT CARDIAC

8 8 M. Kremke et al. / European Journal of Cardio-Thoracic Surgery CONCLUSIONS Our data do not support the hypothesis that platelet transfusion is associated with an increased mortality or greater risk of ischaemic events after CABG. Conflict of interest: none declared. REFERENCES [1] Mangano DT. Multicenter study of perioperative ischemia research group: aspirin and mortality from coronary artery bypass surgery. N Engl J Med 2002;347: [2] Kremke M, Tang M, Bak M, Kristensen KL, Hindsholm K, Andreasen JJ et al. Antiplatelet therapy at the time of coronary artery bypass grafting: a multicentre cohort study. Eur J Cardiothorac Surg 2013;44:e133 e40. [3] Spiess BD, Royston D, Levy JH, Fitch J, Dietrich W, Body S et al. Platelet transfusions during coronary artery bypass graft surgery are associated with serious adverse outcomes. Transfusion 2004;44: [4] Karkouti K, Wijeysundera DN, Yau TM, Callum JL, Meineri M, Wasowicz M et al. Platelet transfusions are not associated with increased morbidity or mortality in cardiac surgery. Can J Anaesth 2006;53: [5] McGrath T, Koch CG, Xu M, Li L, Mihaljevic T, Figueroa P et al. Platelet transfusion in cardiac surgery does not confer increased risk for adverse morbid outcomes. Ann Thorac Surg 2008;86: [6] Mikkola R, Gunn J, Heikkinen J, Wistbacka JO, Teittinen K, Kuttila K et al. Use of blood products and risk of stroke after coronary artery bypass surgery. Blood Transfus 2012;10: [7] Bilgin YM, van de Watering LM, Versteegh MI, van Oers MH, Vamvakas ECBrand A. Postoperative complications associated with transfusion of platelets and plasma in cardiac surgery. Transfusion 2011;51: [8] Schmidt M, Maeng M, Jakobsen CJ, Madsen M, Thuesen L, Nielsen PH et al. Existing data sources for clinical epidemiology: the Western Denmark Heart Registry. Clin Epidemiol 2010;2: [9] Roques F, Nashef SA, Michel P, Gauducheau E, de Vincentiis C, Baudet E et al. Risk factors and outcome in European cardiac surgery: analysis of the EuroSCORE multinational database of patients. Eur J Cardiothorac Surg 1999;15: [10] Pedersen CB. The Danish Civil Registration System. Scand J Public Health 2011;39:22 5. [11] Sturmer T, Joshi M, Glynn RJ, Avorn J, Rothman KJ, Schneeweiss S. A review of the application of propensity score methods yielded increasing use, advantages in specific settings, but not substantially different estimates compared with conventional multivariable methods. J Clin Epidemiol 2006;59: [12] Austin PC. Propensity-score matching in the cardiovascular surgery literature from 2004 to 2006: a systematic review and suggestions for improvement. J Thorac Cardiovasc Surg 2007;134: [13] Nielsen DV, Hansen MK, Johnsen SP, Hansen M, Hindsholm K, Jakobsen CJ. Health outcomes with and without use of inotropic therapy in cardiac surgery: results of a propensity score-matched analysis. Anesthesiology 2014;120: [14] Cooper WA, O Brien SM, Thourani VH, Guyton RA, Bridges CR, Szczech LA et al. Impact of renal dysfunction on outcomes of coronary artery bypass surgery: results from the Society of Thoracic Surgeons National Adult Cardiac Database. Circulation 2006;113: [15] Silvain J, Collet JP, Nagaswami C, Beygui F, Edmondson KE, Bellemain-Appaix A et al. Composition of coronary thrombus in acute myocardial infarction. J Am Coll Cardiol 2011;57: [16] Furie B, Furie BC. Mechanisms of thrombus formation. N Engl J Med 2008; 359: