- Email: [email protected]
The HAT Score—A Simple Risk Stratification Score for Coagulopathic Bleeding During Adult Extracorporeal Membrane Oxygenation
The HAT Score—A Simple Risk Stratification Score for Coagulopathic Bleeding During Adult Extracorporeal Membrane Oxygenation Terence Lonergan, MD,* Daniel Herr, MD,* Zachary Kon, MD,† Jay Menaker, MD,* Raymond Rector, CCP,† Kenichi Tanaka, MD, MSc,‡ and Michael Mazzeffi, MD, MPH‡ Objective: The study objective was to create an adult extracorporeal membrane oxygenation (ECMO) coagulopathic bleeding risk score. Design: Secondary analysis was performed on an existing retrospective cohort. Pre-ECMO variables were tested for association with coagulopathic bleeding, and those with the strongest association were included in a multivariable model. Using this model, a risk stratification score was created. The score’s utility was validated by comparing bleeding and transfusion rates between score levels. Bleeding also was examined after stratifying by nadir platelet count and overanticoagulation. Predictive power of the score was compared against the risk score for major bleeding during anti-coagulation for atrial fibrillation (HAS-BLED). Setting: Tertiary care academic medical center. Participants: The study comprised patients who received venoarterial or venovenous ECMO over a 3-year period, excluding those with an identified source of surgical bleeding during exploration. Interventions: None.
Measurements and Main Results: Fifty-three (47.3%) of 112 patients experienced coagulopathic bleeding. A 3-variable score—hypertension, age greater than 65, and ECMO type (HAT)—had fair predictive value (area under the receiver operating characteristic curve [AUC] ¼ 0.66) and was superior to HAS-BLED (AUC ¼ 0.64). As the HAT score increased from 0 to 3, bleeding rates also increased as follows: 30.8%, 48.7%, 63.0%, and 71.4%, respectively. Platelet and fresh frozen plasma transfusion tended to increase with the HAT score, but red blood cell transfusion did not. Nadir platelet count less than 50 103/lL and overanticoagulation during ECMO increased the AUC for the model to 0.73, suggesting additive risk. Conclusions: The HAT score may allow for bleeding risk stratification in adult ECMO patients. Future studies in larger cohorts are necessary to confirm these findings. & 2016 Elsevier Inc. All rights reserved.
E
bleeding risk. The authors hypothesized that a risk stratification score using pre-ECMO variables could identify which patients were at high risk for bleeding during ECMO.
XTRACORPOREAL MEMBRANE OXYGENATION (ECMO) is a lifesaving therapy that has been used increasingly in adults with cardiopulmonary failure. One of the most frequent complications that occurs during ECMO is serious bleeding, which occurs in 33% to 56% of patients and can require large amounts of blood transfusion.1–3 In a previous cohort of ECMO patients, the incidence rate for serious bleeding events was 10 events per 100 ECLS days. Patients undergoing venoarterial (VA) ECMO and postcardiotomy shock patients appeared to have particularly high bleeding risk.1 The pathophysiology of bleeding during ECMO is complex and not fully understood but appears to be due in part to qualitative platelet dysfunction, fibrinolysis, and loss of large von Willebrand factor multimers.4–6 Other factors that may contribute to bleeding include overanticoagulation and loss of coagulation factors due to hemodilution and consumption. Unfortunately, it remains difficult to predict which patients are likely to experience bleeding during ECMO, and there are no large clinical trials on which to base anticoagulation practice or hemostasis management. Developing tools for risk stratification is important so that future interventions can target the highest-risk patients. To the authors’ knowledge, there are no adult ECMO bleeding risk scores, but there are bleeding risk scores for medical patients taking anticoagulation for atrial fibrillation, such as the risk score for major bleeding during anticoagulation for atrial fibrillation (HAS-BLED).7 The purpose of this study was to evaluate whether pre-ECMO variables could help determine the risk of coagulopathic bleeding events during ECMO. Furthermore, the authors sought to evaluate how thrombocytopenia and overanticoagulation during ECMO interacted with baseline variables in determining overall
KEY WORDS: bleeding, extracorporeal oxygenation, risk stratification
membrane
MATERIALS AND METHODS
Patients The institutional review board at the University of Maryland, Baltimore, approved the study. Secondary data analysis was performed on an existing database containing information recorded about bleeding events during ECMO for all adult patients who experienced ECMO between March 1, 2010, and August 31, 2013.1 Inclusion criteria for the initial cohort were age greater than 18 and receiving ECMO. There were no exclusion criteria. Patients in the initial cohort with an identified source of surgical bleeding were excluded from the analysis because surgical bleeding is distinct from coagulopathic bleeding and is unlikely to be predicted using the same variables. Surgical bleeding was defined as bleeding for which a surgeon identified a clear source during exploration. All other patients from the initial cohort were included in the analysis.
From the *Departments of Shock Trauma Critical Care, Baltimore, MD, †Cardiothoracic Surgery, Baltimore, MD; and ‡Anesthesiology, University of Maryland School of Medicine, Baltimore, MD. Address reprint requests to Michael Mazzeffi, MD, MPH, 22 South Greene St., S11C00, Baltimore, MD 21201. E-mail: mmazzeffi@anes. umm.edu © 2016 Elsevier Inc. All rights reserved. 1053-0770/2601-0001$36.00/0 http://dx.doi.org/10.1053/j.jvca.2016.08.037
Journal of Cardiothoracic and Vascular Anesthesia, Vol ], No ] (Month), 2016: pp ]]]–]]]
1
2
Patient Data For all patients, the following data were collected: age; sex; diabetes mellitus; chronic arterial hypertension defined by repeated systolic blood pressure greater than 140 mmHg or diastolic blood pressure greater than 90 mmHg (on at least 3 separate measurements); surgery within 24 hours of starting ECMO; pre-ECMO platelet count; pre-ECMO international normalized ratio (INR); pre-ECMO plasma creatinine level, previous stroke; use of antiplatelet drugs within 5 days of starting ECMO (aspirin or thienopyridines); history of illicit drug use or greater than 8 alcoholic drinks per week; liver dysfunction (defined by plasma transaminases 3 times greater than the upper limit of normal or total bilirubin 2 times greater than the upper limit of normal); history of bleeding diathesis; pre-ECMO left ventricular ejection fraction; pre-ECMO moderate or severe right ventricular dysfunction; chronic lung disease (defined as chronic obstructive pulmonary disease, asthma, or interstitial lung disease); ECMO type; heparin monitoring modality used during ECMO; indication for ECMO (defined as infectious respiratory failure, noninfectious respiratory failure, or cardiac failure); total ECMO days; nadir platelet count during ECMO; number of times above the target anticoagulation level; total red blood cell (RBC) transfusion; total platelet transfusion; and total fresh frozen plasma (FFP) transfusion. ECMO Details The standard ECMO circuit at the authors’ institution includes a Quadrox-i oxygenator (Maquet, Rastatt, Germany) and Rotaflow centrifugal pump (Maquet). Unfractionated heparin was used for anticoagulation, but monitoring varied over the study period between the activated clotting time (ACT) and activated partial thromboplastin time (aPTT). From 2010 until 2012, ACTs were used, with goal ACTs between 180 and 200 seconds for VA ECMO and 160 and 180 seconds for veno-venous (VV) ECMO. Beginning in 2013, aPTTs became the primary monitoring modality, with goals of 60-to80 seconds for VA ECMO and 45-to-55 seconds for VV ECMO. Transfusion decisions were made by attending physicians and were not based on a specific protocol. Hemoglobin levels generally were targeted to be greater than 10 g/dL, which is in accordance with the Extracorporeal Life Support Organization (ELSO) anticoagulation guideline.8 Bleeding Events The definition for serious coagulopathic bleeding was based on the definition for major bleeding in the ELSO anticoagulation guideline8: (1) bleeding that required surgical exploration or (2) bleeding that required immediate transfusion of at least 2 U of packed RBCs because of a sudden fall in hemoglobin of 2 g/dL with new hemodynamic instability or ongoing visible blood loss. Overanticoagulation was defined as an ACT or aPTT above the target range. Statistical Analysis Patient characteristics were examined using histograms and frequency tables. Continuous variables were summarized as the median value and interquartile range, and categorical variables
LONERGAN ET AL
were summarized as the number and percentage of patients. To build a risk score for coagulopathic bleeding events, the following pre-ECMO variables were tested for univariate associations with bleeding: age greater than 65, sex, hypertension, surgery within 24 hours of starting ECMO, baseline INR 42.0, baseline platelet count less than 50 103/uL, plasma creatinine level 42.2 mg/dL (based on HAS-BLED), history of stroke, use of antiplatelet drugs within 5 days of starting ECMO, history of illicit drug use or alcohol abuse (based on HAS-BLED), liver dysfunction (based on HASBLED), bleeding diathesis (based on HAS-BLED), indication for ECMO, heparin monitoring modality, and ECMO type. Patients who underwent sequential VA and VV ECMO were categorized as VA ECMO because VA ECMO was believed to have the higher bleeding risk. Variables that had a p value of less than 0.20 on univariate testing were entered into a multivariable logistic regression model, and the area under the receiver operating characteristic curve (AUC) value was calculated to determine the model’s predictive value. To optimize model parsimony with best predictive value, each excluded variable was sequentially added back into the model, and if any variable increased the AUC by 0.01 or more, it was retained. In addition, each variable was removed sequentially from the multivariable model. If removal of the variable resulted in a decrease of the AUC by 0.01 or more, it was placed back into the model. Using the variables in the final multivariable model, a risk stratification score was created by taking each variable from the model and entering it into the risk score with a weight equal to its odds ratio point estimate. Bleeding event rates and transfusion volumes were reported for patients in each level of the risk score. Bleeding rates were compared between different levels of the score using the chi-square test, and transfusion volumes were compared using the Kruskall-Wallis test. As a comparison of predictive power, an AUC was calculated for the HAS-BLED score to determine whether it better predicted coagulopathic bleeding during ECMO. Bleeding rates and transfusion volumes also were calculated for each level of the HAS-BLED score and were compared using the chi-square and Kruskall-Wallis tests. To evaluate how thrombocytopenia and overanticoagulation during ECMO additionally contributed to bleeding, nadir platelet count and overanticoagulation during ECMO were added into the previously described multivariable logistic regression model. The AUC was reported for the model. Bleeding event rates in different levels of the risk score were reported after stratifying by nadir platelet count and whether overanticoagulation occurred during ECMO. RESULTS
The study comprised 112 patients who met the inclusion criteria. Of those, 64 patients underwent VV ECMO, and 48 patients underwent VA ECMO. In addition, 53 of the 112 patients (47.3%) experienced coagulopathic bleeding, with most bleeding occurring in the chest. Table 1 shows characteristics of the patients in the cohort. The median age for patients was 49 (Q1-Q3 ¼ 32-60), and the median number of ECMO days was 7 (Q1-Q3 ¼ 4-13). Surgery for 25% of the patients
3
HAT SCORE FOR COAGULOPATHIC BLEEDING DURING ECMO
Table 1. Patient Characteristics Variable
Demographics Age Male sex Comorbidities Diabetes mellitus LVEF, % Moderate or severe RV dysfunction Chronic lung disease ECMO details Primary indication Respiratory failure, infectious Respiratory failure, noninfectious Cardiac failure VA ECMO VV ECMO Total ECMO days Surgery 24 h before ECMO Heparin monitoring aPTT ACT No data Potential pre-ECMO bleeding risk factors Hypertension Previous stroke Heavy alcohol or illicit drug use Liver dysfunction Bleeding diathesis Platelet count INR Creatinine Antiplatelet drugs Potential on-ECMO bleeding risk factors Nadir platelet count Overanticoagulation events Transfusion during ECMO Red blood cell units Fresh frozen plasma units Platelet units Outcomes Bleeding event Primary site of bleeding Chest GI tract Cannula site Other In-hospital mortality
Median [Q1-Q3] or n (%)
49 [32-60] 66 (58.9) 16 55 31 31
(14.3) [35-60] (31.0) (27.7)
26 46 40 48 64 7 28
(23.2) (41.1) (35.7) (42.8) (57.2) [4-13] (25.5)
33 (29.5) 74 (66.1) 5 (4.4) 51 5 16 32 3 144 1.4 1.2 28
(45.5) (4.6) (14.4) (29.1) (2.7) [87-210] [1.2-1.7] [0.8-2.0] (25.7%)
51 [28-83] 4 [0-17] 16 [9-28] 2 [0-8] 2 [0-7] 53 (47.3) 23 16 7 7 52
(43.4) (30.2) (13.2) (13.2) (46.4)
Abbreviations: ACT, activated clotting time; aPTT, activated partial thromboplastin time; ECMO, extracorporeal membrane oxygenation; GI, gastrointestinal; INR, international normalized ratio; LVEF, left ventricular ejection fraction; Q, quartile; RV, right ventricular; VA, veno-arterial; VV, veno-venous.
was performed within 24 hours of ECMO, with the majority of patients undergoing cardiac surgery and a smaller number undergoing other procedures. Table 2 shows the results of the univariate tests for association with coagulopathic bleeding. The only variable that demonstrated a significant association was VA ECMO (p ¼ 0.02). Four variables (age 465, hypertension, use of antiplatelet drugs, and VA ECMO) had a p value less than 0.2 and
Table 2. Univariate Associations of Pre-ECMO Variables With Bleeding Variable
Odds Ratio (95% CI)
Age 465 Male sex Hypertension Surgery within 24 h INR 42.0 Platelet count o50 103 platelets/uL Creatinine 42.2 mg/dL Previous stroke Antiplatelet drugs Illicit drugs or ETOH Liver dysfunction Bleeding diathesis Indication for ECMO ARDS infectious versus cardiac failure ARDS noninfectious versus cardiac failure aPTT versus ACT VA versus VV ECMO
2.51 1.30 2.03 1.16 0.97 2.38 0.80 0.73 2.14 1.16 1.17 2.32
(0.80-7.90) (0.61-2.77) (0.96-4.33) (0.49-2.73) (0.34-2.90) (0.42-13.53) (0.31-2.09) (0.12-4.57) (0.88-5.14) (0.40-3.35) (0.51-2.66) (0.20-26.35)
p Value
0.12 0.50 0.07 0.74 0.96 0.33 0.65 0.74 0.09 0.78 0.71 0.50
0.54 (0.20-1.47) 0.52 (0.22-1.23)
0.53 0.38
0.62 (0.27-1.42) 2.54 (1.18-5.49)
0.25 0.02
Abbreviations: ACT, activated clotting time; aPTT, activated partial thromboplastin time; ARDS, acute respiratory distress syndrome; CI, confidence interval; ECMO, extracorporeal membrane oxygenation; ETOH, alcohol; INR, international normalized ratio; VA, veno-arterial; VV, veno-venous.
were included in the initial multivariable logistic regression model. No excluded variable changed the AUC by more than 0.01 when added back into the model. Antiplatelet drug use was removed from the final multivariable model because its removal had no impact on the AUC. The final 3-variable model demonstrated an AUC of 0.66, suggesting it had fair predictive value for bleeding. The odds ratio point estimate for each variable in the final model was close to 2, so each variable was weighted equally in the bleeding risk score. The HAT score, which includes 3 variables—hypertension, age 465, and ECMO type—is shown in Fig 1. Table 3 shows bleeding rates in each level of the HAT score. There was a consistent increase in bleeding as the HAT score increased from 0 to 3, with HAT 0 patients having a bleeding rate of 30.8% and HAT 3 patients having a bleeding rate of 71.4%. FFP and platelet transfusion generally increased as the HAT score increased and was significantly different between different levels of the score (p ¼ 0.008 and 0.02, respectively). RBC transfusion, however, did not increase consistently with the HAT score and was not significantly different between different levels of the score (p ¼ 0.40). A HAT score Z2 had a positive predictive value of 64.7% and a negative predictive value of 60.3% for predicting coagulopathic bleeding. Table 4 shows bleeding rates by HAS-BLED score level. The HAS-BLED score did not predict bleeding events as well as the HAT score did (AUC ¼ 0.64 compared with 0.66), and bleeding rates did not increase consistently with the HASBLED score. Most patients in the cohort had a HAS-BLED score of 0, 1, 2, or 3. Only one patient had a HAS-BLED score of 5. Transfusion did not differ between different levels of the HAS-BLED score, with HAS-BLED 0 to 4 patients receiving similar amounts of RBC transfusion.
4
LONERGAN ET AL
Fig 1.
The HAT score for bleeding risk stratification during adult ECMO. The score ranges from 0 to 3.
Adding nadir platelet count less than 50 103 and overanticoagulation during ECMO to the HAT logistic regression model increased its AUC to 0.73. Fig 2 shows bleeding rates by HAT score in patients with a platelet nadir less than 50 103 platelets/uL during ECMO and in patients with a platelet nadir greater than or equal to 50 103 platelets/uL. Patients with higher platelet nadirs experienced less bleeding, but bleeding risk appeared to increase with a greater HAT score in both groups. Fig 2 also shows bleeding rates by HAT score in patients who experienced overanticoagulation and in those who did not; patients with overanticoagulation had more bleeding than those without overanticoagulation. Again, the HAT score was associated with increased bleeding risk in both groups. DISCUSSION
ECMO can be a lifesaving therapy and has been used increasingly in adult patients. Serious bleeding is one of the most significant complications that patients experience and it affects their survival. Although ECMO is being used more frequently in adult patients, there are no large clinical trials to guide best practices for anticoagulation or hemostasis management in these patients. There was one small study suggesting that overanticoagulation contributed to bleeding and may increase mortality by as much as 20%.9 Similarly, there are few published data about which patients are most likely to experience coagulopathic bleeding. One of the most important necessary steps before planning future clinical trials is to determine which patients are most likely to experience bleeding. Once high-risk patients are identified, clinical trials can target interventions toward them.
0 1 2 3 p value
Bleeding Rate
RBC
12/39 ¼ 30.8% 19/39 ¼ 48.7% 17/27 ¼ 63.0% 5/7 ¼ 71.4% 0.04
13 (8-21) 18 (7-33) 19 (9-29) 16 (8-55) 0.40
Platelet
0 1 6 3
(0-3) (0-7) (1-10) (0-9) 0.02
Table 4. Bleeding Rates and Transfusion by HAS-BLED HAS-BLED
Table 3. Bleeding Rates and Transfusion by HAT Score HAT Score
Using a cohort of 112 adult ECMO patients, a model that used pre-ECMO variables and had fair predictive value for coagulopathic bleeding was created during this study. Interestingly, pre-ECMO coagulation parameters such as the platelet count and INR had no association with coagulopathic bleeding during ECMO and were not included in the final risk score. The variables that were included are similar to those that are associated with bleeding in patients receiving anticoagulation for atrial fibrillation. In fact, 2 of the variables in the HAT score —age 465 and hypertension—also are included in the HASBLED score. Bleeding during ECMO is complex and not fully understood. One mechanism that has been described is loss of large von Willebrand factor multimers due to increased cleavage by the ADAMTS13 enzyme under conditions of high shear stress.10,11 This loss of von Willebrand factor multimers leads to poor platelet adherence at sites of endothelial injury and ultimately to poor platelet aggregation. In the study presented here, the variables that were included in the final bleeding risk score were plausible risk factors for bleeding during ECMO. Hypertension increases bleeding risk in multiple scenarios, including hemorrhagic stroke, gastrointestinal hemorrhage in patients with portal hypertension, and spontaneous bleeding in patients receiving anticoagulation for atrial fibrillation.12,13 Age also is a well-established risk factor for bleeding, and VA ECMO has a higher bleeding rate than does VV ECMO, given the need for higher levels of anticoagulation.1,14,15
FFP
0 (0-3) 3 (0-8) 6 (0-13) 6 (0-23) 0.008
NOTE: AUC for HAT bleeding model ¼ 0.66. Abbreviations: FFP, fresh frozen plasma; HAT, 3-variable score of hypertension, age greater than 65, and ECMO type; RBC, red blood cell.
0 1 2 3 4 5 p value
Bleeding Rate
8/25 ¼ 32% 17/32 ¼ 53.1% 12/29 ¼ 41.4% 13/21 ¼ 61.9% 3/4 ¼ 75.0% 0/1 ¼ 0.0% 0.21
14 17 14 18 13 28
RBC
Platelet
(8-21) (8-40) (9-25) (11-27) (5-23) (28-28) 0.67
0 (0-3) 1 (0-7) 1 (0-7) 7 (2-9) 2 (1-3) 1 (1-1) 0.06
FFP
0 2 2 8 2 10
(0-3) (0-8) (0-6) (2-12) (1-14) (10-10) 0.07
NOTE. AUC for risk score for major bleeding during anti-coagulation for atrial fibrillation (HAS-BLED) bleeding model ¼ 0.64. Abbreviations: AUC, area under the receiver operating characteristic curve; FFP, fresh frozen plasma; RBC, red blood cell.
HAT SCORE FOR COAGULOPATHIC BLEEDING DURING ECMO
5
Fig 2. Bleeding rates by HAT score after stratification by nadir platelet count and overanticoagulation on ECMO. (A) Bleeding rate by HAT score in those with nadir platelet o50,000/uL. (B) Bleeding rate by HAT score in those with nadir platelet count Z50,000/uL. (C) Bleeding rate by HAT score in those with overanticoagulation during ECMO. (D) Bleeding rate by HAT score in those without overanticoagulation during ECMO.
These data suggested that low platelet counts and overanticoagulation during ECMO also played a significant role in bleeding because they increased the AUC of the HAT model. The authors believe that the data presented in Fig 2 are particularly important because they imply that in patients with a HAT score of 0, lower platelet counts during ECMO might be acceptable and still be associated with a relatively low rate of bleeding, whereas in patients with a HAT score of 3, higher platelet counts might be necessary to decrease bleeding risk. Similarly, overanticoagulation appeared to be associated with a relatively low risk of bleeding in HAT 0 patients compared with HAT 1, 2, and 3 patients. These data may be important for planning future clinical trials involving platelet transfusion and anticoagulation protocols. This study had several important limitations. First, the study population was representative of ECMO practices at a single center where anticoagulation management and transfusion may be considerably different from other centers. Second, the bleeding model only had fair predictive value compared with predictive models for other disease states, which have good or excellent predictive value. Third, the HAT score was not associated with RBC transfusion, which would seem to be a logical surrogate for the amount of bleeding during ECMO. This partly may be because anemia during ECMO can be caused by hemolysis and other factors, which the HAT score does not predict. Also, patients with severe bleeding may not undergo transfusion as much because they may die earlier. Fourth, the HAT score was not validated prospectively in a
separate cohort of patients, which will be important in definitively establishing its usefulness. Fifth, excluding patients with surgical bleeding may have introduced some selection bias into the sample, and it was possible that excluding these patients limited the generalizability of the data. Sixth, the HAT score does not examine changes in coagulation parameters over time and hence may not reflect the variable risk for bleeding during ECMO. Instead, the HAT score is based on fixed risk factors and predicts a baseline bleeding risk at the time of ECMO initiation. Finally, data for some important laboratory parameters were not included, such as D-dimer levels, which may help to identify patients with ongoing low-grade disseminated intravascular coagulation, and the many different types of surgeries performed for patients in the cohort were not reported. This study also had important strengths. Its most important strength was that it was the first study, to the authors’ knowledge, to describe a bleeding risk stratification score for adult ECMO patients. Also, it represented patients from a single busy ECMO center where medical practice is relatively consistent. Finally, this study suggested that the HAT score may help to predict and allocate transfusion in adult ECMO patients, particularly platelets and FFP. The main clinical implication of this study was that the HAT score may be used to triage patients at the time of ECMO initiation into low, medium, and high bleeding risk groups. Hence, future interventional studies may use the HAT score to identify patients with high bleeding risk. Also, the HAT score may be used to determine baseline bleeding risk, and this information may be
6
LONERGAN ET AL
used in conjunction with daily changes in coagulation parameters to identify overall bleeding risk for patients undergoing ECMO. In summary, in a cohort of 112 adult ECMO patients, a novel 3-variable score, the HAT score, allowed for fair prediction of coagulopathic bleeding and was superior to the HAS-BLED score. The HAT score also was associated with the amount of platelet and FFP transfusion during adult ECMO but
not RBC transfusion. This suggested that the HAT score might predict the incidence of bleeding events, but not the total amount of bleeding during these events or total RBC transfusion during ECMO. Prospective validation of the HAT score is necessary to confirm its usefulness for predicting bleeding and transfusion in ECMO patients. The predictive value of the HAT score also might be confirmed in large ECMO databases, such as the ELSO database.
REFERENCES 1. Mazzeffi M, Greenwood J, Tanaka K, et al: Bleeding, transfusion, and mortality on extracorporeal life support: ECLS working group on thrombosis and hemostasis. Ann Thorac Surg 101:682-689, 2015 2. Cheng R, Hachamovitch R, Kittleson M, et al: Complications of extracorporeal membrane oxygenation for treatment of cardiogenic shock and cardiac arrest: A meta-analysis of 1,866 adult patients. Ann Thorac Surg 97:610-616, 2014 3. Zangrillo A, Landoni G, Biondi-Zoccai G, et al: A meta-analysis of complications and mortality of extracorporeal membrane oxygenation. Crit Care Resusc 15:172-178, 2013 4. Cheung PY, Sawicki G, Salas E, et al: The mechanisms of platelet dysfunction during extracorporeal membrane oxygenation. Crit Care Med 28:2584-2590, 2000 5. Tauber H, Ott H, Streif W, et al: Extracorporeal membrane oxygenation induces short term loss of high von Willebrand factor multimers. Anesth Analg 120:730-736, 2015 6. McVeen RV, Lorch V, Carroll RC, et al: Changes in fibrinolytic factors in newborns during extracorporeal membrane oxygenation. Am J Hematol 38:254-255, 1991 7. Pisters R, Lane DA, Nieuwlaat R, et al: A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: The Euro Heart Survey. Chest 138:1093-1100, 2010
8. Lequier L., Annich G., Al-Ibrahim O., et al: ELSO anticoagulation guideline. Available at www.elso.org. Accessed on August 1, 2016. 9. Yeo HJ, Kim DH, Jeon D, et al: Low dose heparin during extracorporeal membrane oxygenation treatment in adults. Intensive Care Med 42:2020-2021, 2015 10. Sing I, Themistou, Porcar L, et al: Fluid shear induces conformation change in human blood protein von Willebrand factor in solution. Biophys J 96:2313-2320, 2009 11. Chan CH, Pieper IL, Fleming S, et al: The effect of shear stress on the size, structure, and function of human von Willebrand factor. Artif Organs 38:741-750, 2014 12. Juvela S, Hillborn M, Palomaki H: Risk factors for spontaneous intracerebral hemorrhage. Stroke 26:1558-1564, 1995 13. Biecker E: Portal hypertension and gastrointestinal bleeding: Diagnosis, prevention, and management. World J Gastroenterol 19: 5035-5050, 2013 14. Lip GY, Clementy N, Pericart L, et al: Stroke and major bleeding risk in elderly patients age 475 years with atrial fibrillation: The Loire Valley atrial fibrillation project. Stroke 46: 143-150, 2015 15. Shoeb M, Fang MC: Assessing bleeding risk in patients taking anticoagulants. J Throm Thrombolysis 35:312-319, 2013
Measurements and Main Results: Fifty-three (47.3%) of 112 patients experienced coagulopathic bleeding. A 3-variable score—hypertension, age greater than 65, and ECMO type (HAT)—had fair predictive value (area under the receiver operating characteristic curve [AUC] ¼ 0.66) and was superior to HAS-BLED (AUC ¼ 0.64). As the HAT score increased from 0 to 3, bleeding rates also increased as follows: 30.8%, 48.7%, 63.0%, and 71.4%, respectively. Platelet and fresh frozen plasma transfusion tended to increase with the HAT score, but red blood cell transfusion did not. Nadir platelet count less than 50 103/lL and overanticoagulation during ECMO increased the AUC for the model to 0.73, suggesting additive risk. Conclusions: The HAT score may allow for bleeding risk stratification in adult ECMO patients. Future studies in larger cohorts are necessary to confirm these findings. & 2016 Elsevier Inc. All rights reserved.
E
bleeding risk. The authors hypothesized that a risk stratification score using pre-ECMO variables could identify which patients were at high risk for bleeding during ECMO.
XTRACORPOREAL MEMBRANE OXYGENATION (ECMO) is a lifesaving therapy that has been used increasingly in adults with cardiopulmonary failure. One of the most frequent complications that occurs during ECMO is serious bleeding, which occurs in 33% to 56% of patients and can require large amounts of blood transfusion.1–3 In a previous cohort of ECMO patients, the incidence rate for serious bleeding events was 10 events per 100 ECLS days. Patients undergoing venoarterial (VA) ECMO and postcardiotomy shock patients appeared to have particularly high bleeding risk.1 The pathophysiology of bleeding during ECMO is complex and not fully understood but appears to be due in part to qualitative platelet dysfunction, fibrinolysis, and loss of large von Willebrand factor multimers.4–6 Other factors that may contribute to bleeding include overanticoagulation and loss of coagulation factors due to hemodilution and consumption. Unfortunately, it remains difficult to predict which patients are likely to experience bleeding during ECMO, and there are no large clinical trials on which to base anticoagulation practice or hemostasis management. Developing tools for risk stratification is important so that future interventions can target the highest-risk patients. To the authors’ knowledge, there are no adult ECMO bleeding risk scores, but there are bleeding risk scores for medical patients taking anticoagulation for atrial fibrillation, such as the risk score for major bleeding during anticoagulation for atrial fibrillation (HAS-BLED).7 The purpose of this study was to evaluate whether pre-ECMO variables could help determine the risk of coagulopathic bleeding events during ECMO. Furthermore, the authors sought to evaluate how thrombocytopenia and overanticoagulation during ECMO interacted with baseline variables in determining overall
KEY WORDS: bleeding, extracorporeal oxygenation, risk stratification
membrane
MATERIALS AND METHODS
Patients The institutional review board at the University of Maryland, Baltimore, approved the study. Secondary data analysis was performed on an existing database containing information recorded about bleeding events during ECMO for all adult patients who experienced ECMO between March 1, 2010, and August 31, 2013.1 Inclusion criteria for the initial cohort were age greater than 18 and receiving ECMO. There were no exclusion criteria. Patients in the initial cohort with an identified source of surgical bleeding were excluded from the analysis because surgical bleeding is distinct from coagulopathic bleeding and is unlikely to be predicted using the same variables. Surgical bleeding was defined as bleeding for which a surgeon identified a clear source during exploration. All other patients from the initial cohort were included in the analysis.
From the *Departments of Shock Trauma Critical Care, Baltimore, MD, †Cardiothoracic Surgery, Baltimore, MD; and ‡Anesthesiology, University of Maryland School of Medicine, Baltimore, MD. Address reprint requests to Michael Mazzeffi, MD, MPH, 22 South Greene St., S11C00, Baltimore, MD 21201. E-mail: mmazzeffi@anes. umm.edu © 2016 Elsevier Inc. All rights reserved. 1053-0770/2601-0001$36.00/0 http://dx.doi.org/10.1053/j.jvca.2016.08.037
Journal of Cardiothoracic and Vascular Anesthesia, Vol ], No ] (Month), 2016: pp ]]]–]]]
1
2
Patient Data For all patients, the following data were collected: age; sex; diabetes mellitus; chronic arterial hypertension defined by repeated systolic blood pressure greater than 140 mmHg or diastolic blood pressure greater than 90 mmHg (on at least 3 separate measurements); surgery within 24 hours of starting ECMO; pre-ECMO platelet count; pre-ECMO international normalized ratio (INR); pre-ECMO plasma creatinine level, previous stroke; use of antiplatelet drugs within 5 days of starting ECMO (aspirin or thienopyridines); history of illicit drug use or greater than 8 alcoholic drinks per week; liver dysfunction (defined by plasma transaminases 3 times greater than the upper limit of normal or total bilirubin 2 times greater than the upper limit of normal); history of bleeding diathesis; pre-ECMO left ventricular ejection fraction; pre-ECMO moderate or severe right ventricular dysfunction; chronic lung disease (defined as chronic obstructive pulmonary disease, asthma, or interstitial lung disease); ECMO type; heparin monitoring modality used during ECMO; indication for ECMO (defined as infectious respiratory failure, noninfectious respiratory failure, or cardiac failure); total ECMO days; nadir platelet count during ECMO; number of times above the target anticoagulation level; total red blood cell (RBC) transfusion; total platelet transfusion; and total fresh frozen plasma (FFP) transfusion. ECMO Details The standard ECMO circuit at the authors’ institution includes a Quadrox-i oxygenator (Maquet, Rastatt, Germany) and Rotaflow centrifugal pump (Maquet). Unfractionated heparin was used for anticoagulation, but monitoring varied over the study period between the activated clotting time (ACT) and activated partial thromboplastin time (aPTT). From 2010 until 2012, ACTs were used, with goal ACTs between 180 and 200 seconds for VA ECMO and 160 and 180 seconds for veno-venous (VV) ECMO. Beginning in 2013, aPTTs became the primary monitoring modality, with goals of 60-to80 seconds for VA ECMO and 45-to-55 seconds for VV ECMO. Transfusion decisions were made by attending physicians and were not based on a specific protocol. Hemoglobin levels generally were targeted to be greater than 10 g/dL, which is in accordance with the Extracorporeal Life Support Organization (ELSO) anticoagulation guideline.8 Bleeding Events The definition for serious coagulopathic bleeding was based on the definition for major bleeding in the ELSO anticoagulation guideline8: (1) bleeding that required surgical exploration or (2) bleeding that required immediate transfusion of at least 2 U of packed RBCs because of a sudden fall in hemoglobin of 2 g/dL with new hemodynamic instability or ongoing visible blood loss. Overanticoagulation was defined as an ACT or aPTT above the target range. Statistical Analysis Patient characteristics were examined using histograms and frequency tables. Continuous variables were summarized as the median value and interquartile range, and categorical variables
LONERGAN ET AL
were summarized as the number and percentage of patients. To build a risk score for coagulopathic bleeding events, the following pre-ECMO variables were tested for univariate associations with bleeding: age greater than 65, sex, hypertension, surgery within 24 hours of starting ECMO, baseline INR 42.0, baseline platelet count less than 50 103/uL, plasma creatinine level 42.2 mg/dL (based on HAS-BLED), history of stroke, use of antiplatelet drugs within 5 days of starting ECMO, history of illicit drug use or alcohol abuse (based on HAS-BLED), liver dysfunction (based on HASBLED), bleeding diathesis (based on HAS-BLED), indication for ECMO, heparin monitoring modality, and ECMO type. Patients who underwent sequential VA and VV ECMO were categorized as VA ECMO because VA ECMO was believed to have the higher bleeding risk. Variables that had a p value of less than 0.20 on univariate testing were entered into a multivariable logistic regression model, and the area under the receiver operating characteristic curve (AUC) value was calculated to determine the model’s predictive value. To optimize model parsimony with best predictive value, each excluded variable was sequentially added back into the model, and if any variable increased the AUC by 0.01 or more, it was retained. In addition, each variable was removed sequentially from the multivariable model. If removal of the variable resulted in a decrease of the AUC by 0.01 or more, it was placed back into the model. Using the variables in the final multivariable model, a risk stratification score was created by taking each variable from the model and entering it into the risk score with a weight equal to its odds ratio point estimate. Bleeding event rates and transfusion volumes were reported for patients in each level of the risk score. Bleeding rates were compared between different levels of the score using the chi-square test, and transfusion volumes were compared using the Kruskall-Wallis test. As a comparison of predictive power, an AUC was calculated for the HAS-BLED score to determine whether it better predicted coagulopathic bleeding during ECMO. Bleeding rates and transfusion volumes also were calculated for each level of the HAS-BLED score and were compared using the chi-square and Kruskall-Wallis tests. To evaluate how thrombocytopenia and overanticoagulation during ECMO additionally contributed to bleeding, nadir platelet count and overanticoagulation during ECMO were added into the previously described multivariable logistic regression model. The AUC was reported for the model. Bleeding event rates in different levels of the risk score were reported after stratifying by nadir platelet count and whether overanticoagulation occurred during ECMO. RESULTS
The study comprised 112 patients who met the inclusion criteria. Of those, 64 patients underwent VV ECMO, and 48 patients underwent VA ECMO. In addition, 53 of the 112 patients (47.3%) experienced coagulopathic bleeding, with most bleeding occurring in the chest. Table 1 shows characteristics of the patients in the cohort. The median age for patients was 49 (Q1-Q3 ¼ 32-60), and the median number of ECMO days was 7 (Q1-Q3 ¼ 4-13). Surgery for 25% of the patients
3
HAT SCORE FOR COAGULOPATHIC BLEEDING DURING ECMO
Table 1. Patient Characteristics Variable
Demographics Age Male sex Comorbidities Diabetes mellitus LVEF, % Moderate or severe RV dysfunction Chronic lung disease ECMO details Primary indication Respiratory failure, infectious Respiratory failure, noninfectious Cardiac failure VA ECMO VV ECMO Total ECMO days Surgery 24 h before ECMO Heparin monitoring aPTT ACT No data Potential pre-ECMO bleeding risk factors Hypertension Previous stroke Heavy alcohol or illicit drug use Liver dysfunction Bleeding diathesis Platelet count INR Creatinine Antiplatelet drugs Potential on-ECMO bleeding risk factors Nadir platelet count Overanticoagulation events Transfusion during ECMO Red blood cell units Fresh frozen plasma units Platelet units Outcomes Bleeding event Primary site of bleeding Chest GI tract Cannula site Other In-hospital mortality
Median [Q1-Q3] or n (%)
49 [32-60] 66 (58.9) 16 55 31 31
(14.3) [35-60] (31.0) (27.7)
26 46 40 48 64 7 28
(23.2) (41.1) (35.7) (42.8) (57.2) [4-13] (25.5)
33 (29.5) 74 (66.1) 5 (4.4) 51 5 16 32 3 144 1.4 1.2 28
(45.5) (4.6) (14.4) (29.1) (2.7) [87-210] [1.2-1.7] [0.8-2.0] (25.7%)
51 [28-83] 4 [0-17] 16 [9-28] 2 [0-8] 2 [0-7] 53 (47.3) 23 16 7 7 52
(43.4) (30.2) (13.2) (13.2) (46.4)
Abbreviations: ACT, activated clotting time; aPTT, activated partial thromboplastin time; ECMO, extracorporeal membrane oxygenation; GI, gastrointestinal; INR, international normalized ratio; LVEF, left ventricular ejection fraction; Q, quartile; RV, right ventricular; VA, veno-arterial; VV, veno-venous.
was performed within 24 hours of ECMO, with the majority of patients undergoing cardiac surgery and a smaller number undergoing other procedures. Table 2 shows the results of the univariate tests for association with coagulopathic bleeding. The only variable that demonstrated a significant association was VA ECMO (p ¼ 0.02). Four variables (age 465, hypertension, use of antiplatelet drugs, and VA ECMO) had a p value less than 0.2 and
Table 2. Univariate Associations of Pre-ECMO Variables With Bleeding Variable
Odds Ratio (95% CI)
Age 465 Male sex Hypertension Surgery within 24 h INR 42.0 Platelet count o50 103 platelets/uL Creatinine 42.2 mg/dL Previous stroke Antiplatelet drugs Illicit drugs or ETOH Liver dysfunction Bleeding diathesis Indication for ECMO ARDS infectious versus cardiac failure ARDS noninfectious versus cardiac failure aPTT versus ACT VA versus VV ECMO
2.51 1.30 2.03 1.16 0.97 2.38 0.80 0.73 2.14 1.16 1.17 2.32
(0.80-7.90) (0.61-2.77) (0.96-4.33) (0.49-2.73) (0.34-2.90) (0.42-13.53) (0.31-2.09) (0.12-4.57) (0.88-5.14) (0.40-3.35) (0.51-2.66) (0.20-26.35)
p Value
0.12 0.50 0.07 0.74 0.96 0.33 0.65 0.74 0.09 0.78 0.71 0.50
0.54 (0.20-1.47) 0.52 (0.22-1.23)
0.53 0.38
0.62 (0.27-1.42) 2.54 (1.18-5.49)
0.25 0.02
Abbreviations: ACT, activated clotting time; aPTT, activated partial thromboplastin time; ARDS, acute respiratory distress syndrome; CI, confidence interval; ECMO, extracorporeal membrane oxygenation; ETOH, alcohol; INR, international normalized ratio; VA, veno-arterial; VV, veno-venous.
were included in the initial multivariable logistic regression model. No excluded variable changed the AUC by more than 0.01 when added back into the model. Antiplatelet drug use was removed from the final multivariable model because its removal had no impact on the AUC. The final 3-variable model demonstrated an AUC of 0.66, suggesting it had fair predictive value for bleeding. The odds ratio point estimate for each variable in the final model was close to 2, so each variable was weighted equally in the bleeding risk score. The HAT score, which includes 3 variables—hypertension, age 465, and ECMO type—is shown in Fig 1. Table 3 shows bleeding rates in each level of the HAT score. There was a consistent increase in bleeding as the HAT score increased from 0 to 3, with HAT 0 patients having a bleeding rate of 30.8% and HAT 3 patients having a bleeding rate of 71.4%. FFP and platelet transfusion generally increased as the HAT score increased and was significantly different between different levels of the score (p ¼ 0.008 and 0.02, respectively). RBC transfusion, however, did not increase consistently with the HAT score and was not significantly different between different levels of the score (p ¼ 0.40). A HAT score Z2 had a positive predictive value of 64.7% and a negative predictive value of 60.3% for predicting coagulopathic bleeding. Table 4 shows bleeding rates by HAS-BLED score level. The HAS-BLED score did not predict bleeding events as well as the HAT score did (AUC ¼ 0.64 compared with 0.66), and bleeding rates did not increase consistently with the HASBLED score. Most patients in the cohort had a HAS-BLED score of 0, 1, 2, or 3. Only one patient had a HAS-BLED score of 5. Transfusion did not differ between different levels of the HAS-BLED score, with HAS-BLED 0 to 4 patients receiving similar amounts of RBC transfusion.
4
LONERGAN ET AL
Fig 1.
The HAT score for bleeding risk stratification during adult ECMO. The score ranges from 0 to 3.
Adding nadir platelet count less than 50 103 and overanticoagulation during ECMO to the HAT logistic regression model increased its AUC to 0.73. Fig 2 shows bleeding rates by HAT score in patients with a platelet nadir less than 50 103 platelets/uL during ECMO and in patients with a platelet nadir greater than or equal to 50 103 platelets/uL. Patients with higher platelet nadirs experienced less bleeding, but bleeding risk appeared to increase with a greater HAT score in both groups. Fig 2 also shows bleeding rates by HAT score in patients who experienced overanticoagulation and in those who did not; patients with overanticoagulation had more bleeding than those without overanticoagulation. Again, the HAT score was associated with increased bleeding risk in both groups. DISCUSSION
ECMO can be a lifesaving therapy and has been used increasingly in adult patients. Serious bleeding is one of the most significant complications that patients experience and it affects their survival. Although ECMO is being used more frequently in adult patients, there are no large clinical trials to guide best practices for anticoagulation or hemostasis management in these patients. There was one small study suggesting that overanticoagulation contributed to bleeding and may increase mortality by as much as 20%.9 Similarly, there are few published data about which patients are most likely to experience coagulopathic bleeding. One of the most important necessary steps before planning future clinical trials is to determine which patients are most likely to experience bleeding. Once high-risk patients are identified, clinical trials can target interventions toward them.
0 1 2 3 p value
Bleeding Rate
RBC
12/39 ¼ 30.8% 19/39 ¼ 48.7% 17/27 ¼ 63.0% 5/7 ¼ 71.4% 0.04
13 (8-21) 18 (7-33) 19 (9-29) 16 (8-55) 0.40
Platelet
0 1 6 3
(0-3) (0-7) (1-10) (0-9) 0.02
Table 4. Bleeding Rates and Transfusion by HAS-BLED HAS-BLED
Table 3. Bleeding Rates and Transfusion by HAT Score HAT Score
Using a cohort of 112 adult ECMO patients, a model that used pre-ECMO variables and had fair predictive value for coagulopathic bleeding was created during this study. Interestingly, pre-ECMO coagulation parameters such as the platelet count and INR had no association with coagulopathic bleeding during ECMO and were not included in the final risk score. The variables that were included are similar to those that are associated with bleeding in patients receiving anticoagulation for atrial fibrillation. In fact, 2 of the variables in the HAT score —age 465 and hypertension—also are included in the HASBLED score. Bleeding during ECMO is complex and not fully understood. One mechanism that has been described is loss of large von Willebrand factor multimers due to increased cleavage by the ADAMTS13 enzyme under conditions of high shear stress.10,11 This loss of von Willebrand factor multimers leads to poor platelet adherence at sites of endothelial injury and ultimately to poor platelet aggregation. In the study presented here, the variables that were included in the final bleeding risk score were plausible risk factors for bleeding during ECMO. Hypertension increases bleeding risk in multiple scenarios, including hemorrhagic stroke, gastrointestinal hemorrhage in patients with portal hypertension, and spontaneous bleeding in patients receiving anticoagulation for atrial fibrillation.12,13 Age also is a well-established risk factor for bleeding, and VA ECMO has a higher bleeding rate than does VV ECMO, given the need for higher levels of anticoagulation.1,14,15
FFP
0 (0-3) 3 (0-8) 6 (0-13) 6 (0-23) 0.008
NOTE: AUC for HAT bleeding model ¼ 0.66. Abbreviations: FFP, fresh frozen plasma; HAT, 3-variable score of hypertension, age greater than 65, and ECMO type; RBC, red blood cell.
0 1 2 3 4 5 p value
Bleeding Rate
8/25 ¼ 32% 17/32 ¼ 53.1% 12/29 ¼ 41.4% 13/21 ¼ 61.9% 3/4 ¼ 75.0% 0/1 ¼ 0.0% 0.21
14 17 14 18 13 28
RBC
Platelet
(8-21) (8-40) (9-25) (11-27) (5-23) (28-28) 0.67
0 (0-3) 1 (0-7) 1 (0-7) 7 (2-9) 2 (1-3) 1 (1-1) 0.06
FFP
0 2 2 8 2 10
(0-3) (0-8) (0-6) (2-12) (1-14) (10-10) 0.07
NOTE. AUC for risk score for major bleeding during anti-coagulation for atrial fibrillation (HAS-BLED) bleeding model ¼ 0.64. Abbreviations: AUC, area under the receiver operating characteristic curve; FFP, fresh frozen plasma; RBC, red blood cell.
HAT SCORE FOR COAGULOPATHIC BLEEDING DURING ECMO
5
Fig 2. Bleeding rates by HAT score after stratification by nadir platelet count and overanticoagulation on ECMO. (A) Bleeding rate by HAT score in those with nadir platelet o50,000/uL. (B) Bleeding rate by HAT score in those with nadir platelet count Z50,000/uL. (C) Bleeding rate by HAT score in those with overanticoagulation during ECMO. (D) Bleeding rate by HAT score in those without overanticoagulation during ECMO.
These data suggested that low platelet counts and overanticoagulation during ECMO also played a significant role in bleeding because they increased the AUC of the HAT model. The authors believe that the data presented in Fig 2 are particularly important because they imply that in patients with a HAT score of 0, lower platelet counts during ECMO might be acceptable and still be associated with a relatively low rate of bleeding, whereas in patients with a HAT score of 3, higher platelet counts might be necessary to decrease bleeding risk. Similarly, overanticoagulation appeared to be associated with a relatively low risk of bleeding in HAT 0 patients compared with HAT 1, 2, and 3 patients. These data may be important for planning future clinical trials involving platelet transfusion and anticoagulation protocols. This study had several important limitations. First, the study population was representative of ECMO practices at a single center where anticoagulation management and transfusion may be considerably different from other centers. Second, the bleeding model only had fair predictive value compared with predictive models for other disease states, which have good or excellent predictive value. Third, the HAT score was not associated with RBC transfusion, which would seem to be a logical surrogate for the amount of bleeding during ECMO. This partly may be because anemia during ECMO can be caused by hemolysis and other factors, which the HAT score does not predict. Also, patients with severe bleeding may not undergo transfusion as much because they may die earlier. Fourth, the HAT score was not validated prospectively in a
separate cohort of patients, which will be important in definitively establishing its usefulness. Fifth, excluding patients with surgical bleeding may have introduced some selection bias into the sample, and it was possible that excluding these patients limited the generalizability of the data. Sixth, the HAT score does not examine changes in coagulation parameters over time and hence may not reflect the variable risk for bleeding during ECMO. Instead, the HAT score is based on fixed risk factors and predicts a baseline bleeding risk at the time of ECMO initiation. Finally, data for some important laboratory parameters were not included, such as D-dimer levels, which may help to identify patients with ongoing low-grade disseminated intravascular coagulation, and the many different types of surgeries performed for patients in the cohort were not reported. This study also had important strengths. Its most important strength was that it was the first study, to the authors’ knowledge, to describe a bleeding risk stratification score for adult ECMO patients. Also, it represented patients from a single busy ECMO center where medical practice is relatively consistent. Finally, this study suggested that the HAT score may help to predict and allocate transfusion in adult ECMO patients, particularly platelets and FFP. The main clinical implication of this study was that the HAT score may be used to triage patients at the time of ECMO initiation into low, medium, and high bleeding risk groups. Hence, future interventional studies may use the HAT score to identify patients with high bleeding risk. Also, the HAT score may be used to determine baseline bleeding risk, and this information may be
6
LONERGAN ET AL
used in conjunction with daily changes in coagulation parameters to identify overall bleeding risk for patients undergoing ECMO. In summary, in a cohort of 112 adult ECMO patients, a novel 3-variable score, the HAT score, allowed for fair prediction of coagulopathic bleeding and was superior to the HAS-BLED score. The HAT score also was associated with the amount of platelet and FFP transfusion during adult ECMO but
not RBC transfusion. This suggested that the HAT score might predict the incidence of bleeding events, but not the total amount of bleeding during these events or total RBC transfusion during ECMO. Prospective validation of the HAT score is necessary to confirm its usefulness for predicting bleeding and transfusion in ECMO patients. The predictive value of the HAT score also might be confirmed in large ECMO databases, such as the ELSO database.
REFERENCES 1. Mazzeffi M, Greenwood J, Tanaka K, et al: Bleeding, transfusion, and mortality on extracorporeal life support: ECLS working group on thrombosis and hemostasis. Ann Thorac Surg 101:682-689, 2015 2. Cheng R, Hachamovitch R, Kittleson M, et al: Complications of extracorporeal membrane oxygenation for treatment of cardiogenic shock and cardiac arrest: A meta-analysis of 1,866 adult patients. Ann Thorac Surg 97:610-616, 2014 3. Zangrillo A, Landoni G, Biondi-Zoccai G, et al: A meta-analysis of complications and mortality of extracorporeal membrane oxygenation. Crit Care Resusc 15:172-178, 2013 4. Cheung PY, Sawicki G, Salas E, et al: The mechanisms of platelet dysfunction during extracorporeal membrane oxygenation. Crit Care Med 28:2584-2590, 2000 5. Tauber H, Ott H, Streif W, et al: Extracorporeal membrane oxygenation induces short term loss of high von Willebrand factor multimers. Anesth Analg 120:730-736, 2015 6. McVeen RV, Lorch V, Carroll RC, et al: Changes in fibrinolytic factors in newborns during extracorporeal membrane oxygenation. Am J Hematol 38:254-255, 1991 7. Pisters R, Lane DA, Nieuwlaat R, et al: A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: The Euro Heart Survey. Chest 138:1093-1100, 2010
8. Lequier L., Annich G., Al-Ibrahim O., et al: ELSO anticoagulation guideline. Available at www.elso.org. Accessed on August 1, 2016. 9. Yeo HJ, Kim DH, Jeon D, et al: Low dose heparin during extracorporeal membrane oxygenation treatment in adults. Intensive Care Med 42:2020-2021, 2015 10. Sing I, Themistou, Porcar L, et al: Fluid shear induces conformation change in human blood protein von Willebrand factor in solution. Biophys J 96:2313-2320, 2009 11. Chan CH, Pieper IL, Fleming S, et al: The effect of shear stress on the size, structure, and function of human von Willebrand factor. Artif Organs 38:741-750, 2014 12. Juvela S, Hillborn M, Palomaki H: Risk factors for spontaneous intracerebral hemorrhage. Stroke 26:1558-1564, 1995 13. Biecker E: Portal hypertension and gastrointestinal bleeding: Diagnosis, prevention, and management. World J Gastroenterol 19: 5035-5050, 2013 14. Lip GY, Clementy N, Pericart L, et al: Stroke and major bleeding risk in elderly patients age 475 years with atrial fibrillation: The Loire Valley atrial fibrillation project. Stroke 46: 143-150, 2015 15. Shoeb M, Fang MC: Assessing bleeding risk in patients taking anticoagulants. J Throm Thrombolysis 35:312-319, 2013