Technique Article. Jeffrey B. Riley;* Gregory J. Schears; Gregory A. Nuttall; William C. Oliver, Jr.; Mark H. Ereth; Joseph A.

Transcription

1 The Journal of ExtraCorporeal Technology Technique Article Coagulation Parameter Thresholds Associated with Non-Bleeding in the Eighth Hour of Adult Cardiac Surgical Post-Cardiotomy Extracorporeal Membrane Oxygenation Jeffrey B. Riley;* Gregory J. Schears; Gregory A. Nuttall; William C. Oliver, Jr.; Mark H. Ereth; Joseph A. Dearani* *Cardiovascular Surgery Division, Department of Surgery; and Department of Anesthesiology, Mayo Clinic, Rochester, Minnesota Abstract: Excessive bleeding and allogeneic transfusion during adult post-cardiotomy venoarterial extracorporeal membrane oxygenation (ECMO) are potentially harmful and expensive. Balancing the inhibition of clotting and distinguishing surgical from non-surgical bleeding in post-operative period is difficult. The sensitivity of coagulation tests including Thromboelastography (TEG) to predict chest tube drainage in the early hours of ECMO was examined with the use of receiver-operating characteristics (ROC). The results are useful to incorporate in clinical evidence-based algorithms to guide management decisions. In the eighth hour of ECMO, 26 of the 53 adult patients (49%) studied were identified as non-bleeders (less than 2.0 ml/kg/h). All had experienced various types of cardiac surgical procedures. Fifty-two percent were female and the group was 54 ± 19 (mean ± 1 SD) years old. The coagulation parameter threshold with the maximum sensitivity and specificity to predict non-bleeding at 8 hours on ECMO was the kaolin plus heparinase TEG maximum amplitude (KH-TEG MA) at a significant ROC threshold (t) > 50 mm. The activated partial thromboplastin time (aptt) t < 49 seconds, KH-TEG alphaangle t > 51, and the kaolin activated clotting time (ACT) t < 148 seconds were sensitive predictors of non-bleeders. The whole-blood KH-TEG MA was superior to the plasma-based aptt or International Normalization Ratio (INR) to predict bleeding in the eighth hour of ECMO. Using coagulation laboratory thresholds that predict non-bleeding can begin a process of identifying patients earlier that are likely to bleed. Awareness of these parameter thresholds may improve care through patient protection from unnecessary transfusion and prolonging the life of the ECMO circuit. An algorithm incorporating the ROC thresholds was created to help recognize surgical bleeding to minimize unnecessary transfusions. Keywords: extracorporeal membrane oxygenation, extracorporeal life support, coagulation, anticoagulation, hemorrhage, blood component transfusion, receiver operating characteristics, Thromboelastography. The principle causes of morbidity and mortality with adult post-cardiotomy venoarterial (V-A) extracorporeal membrane oxygenation (ECMO) remain surgical bleeding and circuit thrombosis (1). The goals for administration of anticoagulant drugs during adult ECMO are to prevent life-threatening surgical wound bleeding and avoid extracorporeal circuit (ECC) clot with potential thromboembolic events (1). Little information is available to recommend guidelines for minimal safe coagulation parameter values and allogeneic blood product transfusion during ECMO for Received for publication October 17, 2015; accepted June 06, Address correspondence to: Jeffrey B. Riley, Cardiovascular Surgery Division, Department of Surgery, Mayo Clinic, 200 First Street SW, Rochester, MN RileyJB@gmail.com post-cardiotomy cardiopulmonary failure. Unlike ECMO for respiratory assist, no controlled randomized trials for post-cardiotomy ECMO have been reported (2). In 1997, Nuttall et al. prospectively divided cardiopulmonary bypass (CPB) patients into bleeders and non-bleeders and used receiver operating characteristics (ROC) to discover the clotting test and Thromboelastography (TEG, Haemonetics Corp., Braintree, MA) thresholds and the parameters sensitivity and specificity to separate the two patient bleeding groups (3). Four years later in 2001, the same research team prospectively randomized bleeding CPB patients into two groups, one group where allogeneic blood product administration was managed by an algorithm based on the ROC thresholds and one group where transfusion was guided by the individual anesthesiologist s 71

2 72 J.B. RILEY ET AL. practice (4). The ROC threshold algorithm-guided patients received less blood in the operating room, bled less in the intensive care unit (ICU), and experienced a lower re-operation rate. In 2006, Welsby et al. confirmed the ability of the same coagulation parameters that Nuttall reported to correlate to post-operative bleeding and potential coagulopathies (5). The purpose of this retrospective chart review was to use the same methods as Nuttall et al. to determine the thresholds that were associated with acceptable blood loss for several coagulation monitoring parameters measured in the first hours of post-cardiotomy adult ECMO procedures. The purpose of applying ROC analysis is to identify which coagulation parameter and its threshold has the highest sensitivity to detect clinically significant bleeding during the eighth hour of adult post-cardiotomy ECMO. An algorithm based on the reference work is presented to demonstrate how the ROC thresholds apply to post-cardiotomy ECMO patients. METHODS AND PATIENTS Patients and the ECMO Circuit After the Mayo Clinic Institutional Review Board granted exemption, records from 53 post-cardiotomy ECMO patients treated during January 2007 to February 2010 were reviewed to collect demographic information and to quantify bleeding and coagulation parameter values in the early hours of support. Adult ECMO patients were supported with circuits that included a Revolution centrifugal blood pump (Sorin Group, Arvada, CO) and Trillium coated Affinity NT (Medtronic, Minneapolis, MN) or Jostra QuadroxD Safeline (Maquet, Inc., Wayne, NJ) blood oxygenator. The circuit contained safety devices, in-line pressure, percent oxygen saturation of hemoglobin and air detection monitoring, and a shunt with a manifold system. The ECMO patient management techniques are described in greater detail in a 2009 publication (6). Evaluation of Bleeding The hourly total chest tube drainage (CTD) and estimated blood loss were divided by the patient s body weight and reported in ml/kg/h. Figure 1 shows the distribution of hourly blood loss for the patients. Between the seventh and eighth hour, the median blood loss became less than 2.0 ml/kg/h. Therefore, at eight ECMO elapsed hours, the 53 patients were divided into two groups: 27 patients with greater than 2.0 ml/kg/h (bleeders) and 26 patients with less than 2.0 ml/kg/h (non-bleeders) blood loss. The coagulation parameter values were sorted between the two bleeding groups for determination of ROC thresholds for the fourth and eighth hours. Figure 1. Box plots for the distribution of observed patient estimated blood loss by ECMO hour. Collection of Lab Parameters Laboratory tests were collected at 4-hour time intervals when the circuit management was stable. Although this is a retrospective chart review, a prospective written clinical procedure guided the collection and timing of lab tests. Lab test results were compared with measures of total estimated blood loss and the appearance of thrombus on the ECC surfaces. Statistical Analysis The statistical methods identified the coagulation parameters that had the greatest predictive value, highest sensitivity, and the highest specificity for the patients bleeding at fourth and eighth hours of elapsed ECMO time. Excessive bleeding was defined as greater than 2.0 ml/kg/h. Coagulation parameter values were expressed as the mean ± 1 SD for continuous quantities. Two-sample t test was used to compare means between bleeding groups. Medians were compared by rank sum tests depending on the distribution of the values. Dichotomous data were compared by chi-squared test. Correlations were evaluated by the Spearman s rank tests. JMP 8.0 statistical software (SAS Institute Inc., Cary, NC) was used and statistical significance was set at p <.05. ROC curves were constructed for coagulation parameters and the threshold values to predict non-bleeding at ECMO elapsed times of 4 and 8 hours were discovered. The JMP 8.0 statistical software identified the sensitivity, specificity, and negative and positive predictive values of the coagulation parameters, and the ability to predict bleeding less than 2.0 ml/kg/h. The statistical software

3 POST CARDIOTOMY ECMO COAGULATION PARAMETERS 73 yields the coagulation parameter value (threshold, t) in which sensitivity and specificity together are maximized. Sensitivity was defined as the percentage of patients with bleeding less than 2.0 ml/kg/h, who also had a normal test result. Specificity was defined as the percentage of patients with excessive bleeding, who also had an abnormal test result. A perfect coagulation test threshold would have 100% sensitivity and specificity (3). RESULTS Of 7,268 CPB patients, 53 were placed on V-A ECMO. Between January 2007 and February 2010, 4 9 patients per year (approximately 7 in 1,000 adult CPB patients) required post-operative ECMO support. The median survival was 16 days with a 1-year survival rate of 32% (15 of 53 patients at risk). The ECMO procedures were performed by 1 of 12 surgeons. The types of surgical procedures are listed in Table 1. Table 2 describes the post-cardiotomy adult ECMO patients after division into the two bleeding groups during the eighth hour of ECMO. There was no significant difference for age, gender, and hours on ECMO between the two bleeding groups. Table 3 lists the descriptive statistics for the coagulation parameters divided into bleeding groups. The coagulation parameter ROC thresholds for non-bleeding are listed in Table 4. The coagulation parameter threshold with the maximum sensitivity and specificity to predict non-bleeding was the kaolin plus heparinase TEG maximum amplitude (KH-TEG MA) where the ROC curves are presented in Figure 2. The aptt, KH-TEG alphaangle, the kaolin ACT, the fibrinogen concentration, and Table 1. Type and distribution of surgical procedures requiring post-cardiotomy ECMO. Procedure Count (%) Redo procedure 17 (32) IABP insertion 14 (26) Valve replacement(s) and CABG 11 (21) Adult congenital procedure 9 (17) AVR 4 (8) Triple valve replacement 4 (8) Cardiac transplant 3 (6) Ascending aorta replacement 3 (6) Pulmonary embolus 2 (4) Double valve replacement 2 (4) Hemi-arch replacement 2 (4) CABG 2 (4) Lung transplant 1 (2) OpCAB 1 (2) Distribution of the type and percent of surgical procedures requiring post-cardiotomy ECMO. IABP, intra-aortic balloon pump; CABG, coronary artery bypass graft; AVR, aortic valve replacement; OpCAB, offpump coronary artery bypass. Table 2. Patient bleeding groups at the eighth hour of ECMO. Parameter (units) the kaolin + heparinase Thromboelastograph R time (KH-TEG R) time significantly predicted non-bleeding with varying degrees of sensitivity. In the eighth elapsed ECMO hour, the fibrinogen concentration, the KH-TEG R time, the platelet count, INR, KH-TEG fibrinolysis, kaolin TEG R time (K-TEG R) time, and the difference between the K-TEG R and KH-TEG R times was not sensitive to bleeding less than 2.0 ml/kg/h. Platelet count, INR, KH-TEG fibrinolysis, K-TEG R time, and the K-TEG R minus KH-TEG R difference were not significantly associated with nonbleeding in the eighth hour of ECMO. DISCUSSION Bleeding Group <2.0 ml/kg/h >2.0 ml/kg/h p Value Number (%) 26 (49) 27 (51) Male (%) * Age (years) 55 (56) ± (53) ± ECMO (hours) 160 (124) ± (96) ± CTD (ml/kg/h).7 (.6) ± (3.6) ± 5.0 <.0001 Bleeders defined as greater than 2.0 ml/kg/h blood loss. Values are mean (median) ± 1 SD, values collected in the eighth hour of ECMO. ECMO hours, time on ECMO support. *Probability that male/female ratio is different across bleeding groups. In the eighth hour of ECMO in our post-cardiac V-A ECMO surgical patient group, 26 of the 53 post-cardiotomy patients were identified as non-bleeders. The coagulation parameter threshold with the maximum sensitivity and specificity to predict non-bleeding in the eighth hour on ECMO was the kaolin- and heparinase-activated TEG maximum amplitude with a significant ROC threshold >50 mm. An aptt equal to 49 seconds, a KH-TEG alphaangle equal to 51, and a kaolin ACT equal to 148 seconds were sensitive cut points associated with non-bleeders. The ROC statistical method evaluates a parameter s diagnostic accuracy when there is a definition for a defined disease state (non-bleeding or no ECC thrombus) or abnormal test results (7,8). Our ROC results demonstrated the whole-blood KH-TEG MA was superior to the plasmabased aptt and INR to predict non-bleeding. ROC analysis provides more clinical decision-making information than linear regression (7). Armed with the coagulation ROC parameter thresholds that predict non-bleeding, clinicians may write practice guidelines for transfusion therapy and advise for surgical re-exploration with confidence during the first 8 hours of ECMO. Bleeding is consistently reported as the major complication of post-surgical ECMO support. Citations of ECMO

4 74 J.B. RILEY ET AL. Table 3. Coagulation parameters for 53 post-cardiotomy adult patients at 4 and 8 hours of post-cardiotomy ECMO. ECMO ET Parameter Normal Range Bleeders Non-bleeders p Value 4 hours (n = 30/53 bleeders) aptt (seconds) ± 8 (113) 80 ± 9 (86).0202 KH-TEG R (minutes) ± 5 (50) 28 ± 6 (20).0800 ACT (seconds) ± 8 (154) 153 ± 8 (147).3533 KH-TEG R (minutes) ± 2 (12) 17 ± 2 (12).1865 KH-TEG AA* ± 2 (47) 48 ± 3 (50).1277 KH-TEG MA (mm) ± 2 (49) 51 ± 2 (51).1747 KH-TEG LY 30% <7.5.0 ±.0 (.0).1 ±.0 (.0).1017 INR ±.1 (1.6) 1.8 ± 02 (1.6).7179 Fibrinogen (mg/dl) ± 17 (168) 200 ± 17 (179).4646 Platelet count (K/mm 3 ) ± 7 (82) 91 ± 8 (97).3698 K-TEG R (minutes) ± 5 (57) 45 ± 6 (70) hours (n = 27/53 bleeders) KH-TEG MA (mm) ± 1 (49) 54 ± 1 (54).0004 aptt (seconds) ± 7 (72) 61 ± 7 (48).0047 KH-TEG AA* ± 2 (48) 54 ± 2 (55).0007 ACT (seconds) ± 5 (152) 137 ± 5 (132).0173 Fibrinogen (mg/dl) ± 11 (200) 222 ± 11 (211).0765 KH-TEG R (minutes) ± 1 (11) 13 ± 1 (11).6724 Platelet count (K/mm 3 ) ± 6 (72) 87 ± 7 (87).2159 KH-TEG LY 30% <7.5.2 ± 01 (0).2 ±.1 (0).8911 INR ±.1 (1.4) 1.6 ±.1 (1.5).8091 K-TEG R (minutes) ± 6 (34) 46 ± 6 (41).8005 K-H-TEG R (minutes) ± 6 (22) 34 ± 6 (19).7009 Mean ± SE (Median). At 4 hours: n = 23/53 bleeding less than 2.0 ml/kg/h and n = 30/53 bleeding more than 2.0. At 8 hours: n = 27/53 bleeding less than 2.0 ml/kg/h and n = 26/53 bleeding more per hour. ET, elapsed time in hours; KH-TEG, kaolin plus heparinase activated Thrombelastograph ; K-TEG, kaolin-activated TEG; KH-TEG R minutes, difference between the K-TEG R and the KH-TEG R times in minutes. use after cardiotomy surgery report that about.9 1.2% of surgical patients require extracorporeal support. The rate of ECMO use in our adult CPB patient group was slightly lower at.7%. Post-cardiac surgery case series for ECMO use have reported 24 53% survival since ECMO insertion is typical for moribund patients (9 11). Case series reports list coagulopathies and bleeding at surgical and cannulation sites as frequent complications commonly due to heparin, Table 4. ROC analysis for bleeding <2.0 ml/kg/h for parameters from 53 post-cardiotomy adult patients. ECMO ET Parameter ROC Threshold ROC Area (SZ) p Value True Positive (Sens) True Negative False Positive False Negative 4 hours aptt (seconds) (.01) (.78) KH-TEG R (minutes) (.01) (.78) ACT (seconds) (.01) (.87) KH-TEG R (minutes) (.03) (.61) KH-TEG AA* (.02) (.82) KH-TEG MA (mm) (.03) (.65) I KH-TEG LY 30%.7.53 (.30) (014) INR (.38) (.78) Fibrinogen (mg/d)l (.01) (.65) Platelet count (K/mm 3 ) (.0l) (.64) K-TEG R (minutes) (.01) (.61) hours KH-TEG MA (mm) (.07) (.85) aptt (seconds) (.01) (.54) KH-TEG AA* (.03) (.70) ACT (seconds) (.01) (.81) Fibrinogen (mg/dl) (.01) (.42) KH-TEG R min 6.55 (.04) (.22) Platelet count (K/mm 3 ) (.01) (.56) KH-TEG LY 30% (.56) (1.00) INR (.64) (.62) K-TEG R (minutes) (.01) (.54) KH-TZG R (minutes) (.01) (.15) At 4 hours: n = 23/53 (true positive) bleeding less than 2.0 ml/kg/h and n = 30/53 (true negative) bleeding more than 2.0 ml/kg/h. At 8 hours: n =27/53 (true positive) bleeding less than 2.0 ml/kg/h and n = 26/53 (true negative) bleeding more per hour at optimal ROC threshold. KH-TEG, heparinase Thrombelastograph ; K-TEG, kaolin Thromboelastograph ; KH-TEG R minutes, difference between the K-TEG R and the KH-TEG R times in minutes; SE, standard error; Sens, sensitivity.

5 POST CARDIOTOMY ECMO COAGULATION PARAMETERS 75 Figure 2. ROC analysis for kaolin and heparinaseactivated TEG MA with bleeding less than or greater than 2 ml/kg/h. ROC area =.77 and KH-TEG MA threshold = 50 mm (n = 53, p <.0001). thrombocytopenia, fibrinolysis, uremia, and hepatic dysfunction (12). It is important to gain control of bleeding while minimizing allogeneic transfusion exposures. We should be able to use the significant ROC parameters to guide transfusion in the first 8 hours of adult ECMO. Several reports of the use of post-cardiotomy ECMO to save the lives of high-risk cardiac surgical patients have been published (9). The use of smaller prime, lower surface area, and miniature ECMO circuits is being reported (10). Like many high-volume cardiac surgical centers, we are now using portable centrifugal pumpoxygenator ECMO circuits to salvage moribund patients during weaning from CPB. Our practice has used the smaller V-A ECMO circuits in place of biventricular centrifugal circuits (with or without oxygenators on the right side) which require four cannulation sites. Use of ECMO allows the surgical team to bridge the patient to a decision for optimal care. The significant parameter ROC thresholds will help differentiate between normal and excessive bleeding during post-cardiotomy ECMO (7). Interestingly, we found similar predictive thresholds for TEG parameters as those reported in previous studies on micro-bleeding patients during cardiac surgery (4). Bolliger et al. have called for point of care management algorithm for use during ECMO support (13). Stammers et al. have identified significant differences in TEG parameters between groups of bleeding and non-bleeding pediatric ECMO patients and recommended the use of the TEG during neonatal ECMO (14,15). Whole blood viscoelastic parameters are reported to be more sensitive and have more specificity to bleeding than plasma-based tests. Huang et al. reported on the use of TEG changes after bridging to left ventricular assist devices from ECMO in adult patients (16). Venema et al. studied the interchangeability of the tissue factor-activated RoTEM (Tem Systems Inc., Durham, NC) and TEG parameters in cardiac surgical patients and concluded that only the MA parameters correlated closely (17). Based on their report, the intrinsic RoTEM MA may exhibit the similar sensitivity and specificity to predict adult ECMO patient non-bleeding as the unheparinized blood K-TEG MA. Welsby et al. used correlation and multivariate linear regression modeling to describe relationships among coagulation tests, TEG parameters, and early post-operative bleeding. In their study, the TEG maximum amplitude correlated well with post-operative bleeding more so than platelet count, fibrinogen level, or prothrombin time (5). Our ROC results also showed the whole-blood KH-TEG MA was superior to the plasma-based aptt and INR to predict non-bleeding. ROC analysis provides more clinical decision-making information than linear regression. The added benefit of obtaining coagulation ROC parameter thresholds that predict non-bleeding is that clinicians may write practice guidelines for transfusion therapy. Perhaps even more important is the ability to better determine the need for surgical re-exploration within the critical period after admission to the ICU. Ronald and Dunning, in their review of 14 relevant articles selected from 170, reported that TEG may be useful to predict post-operative bleeding (18). Nuttall et al. used ROC thresholds to generate an algorithm to minimize microvascular bleeding during cardiac surgery (4). Using Nuttall s model, we generated the algorithms in Figures 3 and 4 which illustrate the use of ROC threshold values in ECMO clinical decision-making to achieve non-bleeding patient status. We have adopted these algorithms to help manage post-operative bleeding in our cardiac surgical ECMO patients. Future studies will report on the efficacy of these algorithms when used in a prospective manner. There are several limitations with this study. One of the biases of this chart review, in addition to being retrospective, is that we believe our coagulation management

6 76 J.B. RILEY ET AL. Figure 3. The Bleeding Algorithm is to be used in the first 8 hours of post-cardiotomy ECMO to prepare for anticoagulant administration. AUC is ROC area under the curve for parameter threshold at bleeding less than 2.0 ml/kg/h. The Transfusion Algorithm is Figure 4. routine favors our primary focus to reduce bleeding in favor of allowing thrombus to form on the ECMO circuit surfaces. We use the K-TEG R time and K-TEG alpha angle to guide heparin administration as described by Agati et al. (19). We generally observed less ECC clot formation when the kaolin TEG alpha angle (K-TEG AA) is maintained less than 30 and the K-TEG R time is longer than 20 minutes. We are employing ROC analysis to verify these K-TEG parameter thresholds and the parameter threshold associated with the absence of circuit thrombus in the miniature ECMO circuits. It is difficult to define a specific coagulation parameter that predicts ECC thrombus formation. Until there is a clinically useful measure to determine excessive ECC thrombin formation similar to excessive bleeding, it will continue to be difficult to manage coagulation therapy for ECMO patients. In our experience, the smaller heparin-modified circuit surfaces appear to allow the use of less heparin before thrombus forms. The balance between managing patient bleeding and anticoagulation to avoid ECC thrombus formation is facilitated by employing coagulation test ROC thresholds. We are evaluating the use of smaller circuits and the use of less heparin on allogeneic blood product use and circuit thrombus formation during the care of all ECMO patients. Further research into additional parameters to assess excessive bleeding and thrombus formation will perhaps lower the mortality associated with institution of ECMO. SUMMARY The accuracy of coagulation profile laboratory tests and TEG parameters to predict lower CTD was examined with the use of ROC. The most sensitive and specific tests associated with bleeding were more stable at eight elapsed hours than four elapsed ECMO hours. In the eighth hour of ECMO, the KH-TEG MA, aptt, KH- TEG AA, and the ACT respectively exhibited the greatest sensitivity to predict non-bleeding. It is logical to generalize the use of algorithms to post-cardiotomy ECMO like the ones proposed in this communication that are based on ROC analyses to guide transfusions for ECMO to other ECMO patient groups. A national database perhaps supported by the

7 POST CARDIOTOMY ECMO COAGULATION PARAMETERS 77 Figure 4. Algorithm to be used in the first 8 hours of post-cardiotomy ECMO to adjust coagulation monitoring parameters to minimize patient bleeding. AUC, ROC area under the curve for parameter threshold at bleeding less than 2.0 ml/kg/h; KTEG, kaolin-activated TEG; KHTEG, kaolin- and heparinase-activated TEG. Adapted from Nuttall GA, Oliver WC, Santrach PJ, et al. Anesthesiology. 2001;94; Extracorporeal Life Support Organization (Ann Arbor, MI) to collect coagulation parameter values, patient bleeding data and ECMO circuit information would be useful. ECMO teams could use statistical methods like ROC and multiple logistic regression to provide parameter thresholds for ECMO clinical procedure guidelines for different ECMO circuits and for different patient indications. ACKNOWLEDGMENTS The authors wish to thank Tammy P. Friedrich, RN, Jodie L. Holmen,RN,JamieL.House,RN,andJeanS.Shue,RNfor their help to organize this quality improvement project, and James E. Baker, RRT for his accurate review of the patient records. REFERENCES 1. Oliver W. Anticoagulation and coagulation management for ECMO. Semin Cardiothorac Vasc Anesth. 2009;13: Sidebotham D, McGeorge A, McGuinness S, Edwards M, Willcox T, Beca J. Extracorporeal membrane oxygenation for treating severe cardiac and respiratory disease in adults: Part 1 Overview of extracorporeal membrane oxygenation. J Cardiothorac Vasc Anesth. 2009;23: Nuttall G, Oliver W, Ereth M, Santrach P. Coagulation tests predict bleeding after cardiopulmonary bypass. J Cardiothorac Vasc Anesth. 1997;11: Nuttall G, Oliver W, Santrach P, et al. Algorithm for nonerythrocyte component utilization after cardiopulmonary bypass. Anesthesiology. 2001;94: Welsby IJ, Jiao K, Ortel TL, et al. The kaolin-activated Thrombelastograph predicts bleeding after cardiac surgery. J Cardiothorac Vasc Anesth. 2006;20: Riley J, Scott P, Schears G. Update on safety equipment for extracorporeal life support (ECLS) circuits. Semin Cardiothorac Vasc Anesth. 2009;13: Fawcett T. An introduction to ROC analysis. Pattern Recognit Lett. 2006;27: vanerkel A, Pattynama P. Receiver operating characteristic (ROC) analysis: Basic principles and applications in radiology. Eur J Radiol. 1998;27: Doll N, Kiaii B, Borger M, et al. Five-year results of 219 consecutive patients treated with extracorporeal membrane oxygenation for refractory postoperative cardiogenic shock. Ann Thorac Surg. 2004;77:151 7.

8 78 J.B. RILEY ET AL. 10. Meyer A, Strueber M, Tomaszek S, et al. Temporary cardiac support with a mini-circuit system consisting of a centrifugal pump and a membrane ventilator. Interact Cardiovasc Thorac Surg. 2009;9: Bakhtiary F, Keller H, Dogan S, et al. Venoarterial extracorporeal membrane oxygenation for treatment of cardiogenic shock: Clinical experiences in 45 adult patients. J Thorac Cardiovasc Surg. 2008;135: Sidebotham D, McGeorge A, McGuinness S, Edwards M, Willcox T, Beca J. Extracorporeal membrane oxygenation for treating severe cardiac and respiratory failure in adults: Part 2-Technical considerations. J Cardiothorac Vasc Anesth. 2010;24: Bolliger D, Zenklusen U, Tanaka KA. Point-of-care coagulation management algorithms during ECMO support: Are we there yet? Minerva Anestesiol Mar 30 [Epub ahead of print]. 14. Zavadil D, Stammers A, Willett L, Deptula J, Christensen K, Sydzyik R. Hematological abnormalities in neonatal patients treated with extracorporeal membrane oxygenation (ECMO). J Extra Corpor Technol. 1998;30: Stammers A, Willett L, Fristoe L, et al. Coagulation monitoring during extracorporeal membrane oxygenation: The role of thrombelastography. J Extra Corpor Technol. 1995;27: Huang CY, Chen IM, Hsieh YC, et al. Thrombelastography change after bridging to left ventricular assist device from extracorporeal membrane oxygenation patients. J Chin Med Assoc. 2012;75: Venema L, Post W, Hendriks H, Huet R, Jd JT, devries A. An assessment of clinical interchangeability of TEG and RoTEM thromboelastographic variables in cardiac surgical patients. Anesth Analg. 2010;111: Ronald A, Dunning J. Can the use of thromboelastography predict and decrease bleeding and blood and blood product requirements in adult patients undergoing cardiac surgery? Interact Cardiovasc Thorac Surg. 2005;4: Agati S, Ciccarello G, Salvo D, Turla G, Undar A, Mignosa C. Use of a novel anticoagulation strategy during ECMO in a pediatric population: Single-center experience. ASAIO J. 2006;52:513 6.