Ventricular cardiac-assist devices in infants and children: anesthetic considerations

Ventricular Cardiac-Assist Devices in Infants and Children: Anesthetic Considerations Ehrenfried Schindler, MD,* Matthias Mu¨ller, MD,† Myron Kwapisz, MD,† Hakan Akintu¨rk, MD,‡ Klaus Valeske, MD,‡ Josef Thul, MD,§ and Gunter Hempelmann, MD* Objective: The application of a mechanical cardiac-assist device is now a common procedure in modern cardiac surgery in patients with end-stage failure, whereas in pediatric patients it is still a great challenge. In the recent literature, a broad range of survival and weaning rates have been reported, depending on the variety of mechanical devices and the choice of patients with different conditions before implantation or if the device is used in emergency surgery. In this article, the authors report their experience with pediatric cardiac-assist devices and the perioperative anesthesia management in this group of patients. Patients and Methods: From 1997 to 2001, 11 infants and children were supported with a left and/or right ventricularassist device. Diagnosis included myocarditis and complex cardiac malformations (hypoplastic left heart syndrome, tetralogy of Fallot with cardiomyopathy, and combined heart defects). The data sets of all patients were recorded using the online anesthesia record-keeping system NarkoData (IMESO GmbH, Hu¨ttenberg, Germany). The program collects all perioperative data during surgery and during a stay in the PACU, including vital signs, administered drugs, as well as the data set of the German Society of Anesthesiology and Intensive Care Medicine.

Results: All patients were divided into 2 groups: group 1 ⴝ survivors and group 2 ⴝ nonsurvivors. A total of 5 patients were in group 1, and group 2 consisted of 6 patients. The duration of anesthesia in group 1 patients (173.2 ⴞ 95.1 minutes) was significantly (p < 0.05) shorter than in group 2 (631.1 ⴞ 258.8 minutes) as well as the amount of packed red cells (group 1ⴝ 540.5 ⴞ 150.3 mL, group 2 ⴝ 880.6 ⴞ 400.3 mL). Cardiopulmonary bypass before implantation of a VAD was necessary only in 2 patients from group 1, whereas 5 patients in group 2 were on pump during the procedure. The rate of aortic cross-clamping was also significantly lower in group 1 than in group 2 (p < 0.05). Conclusions: The surgical outcome depends on the patient’s condition at the time of surgery. Emergency surgery, preoperative multiorgan failure, and the need for an extracorporeal circulation with aortic cross-clamping seem to predict a negative outcome in this group of patients. © 2003 Elsevier Inc. All rights reserved.

T

complex cardiac malformations (hypoplastic left heart syndrome, tetralogy of Fallot with cardiomyopathy, and combined heart defects). All infants (⬍4 kg) were unpremedicated. Children older than 6 months of age weighing more than 4 kg received 0.1 mg/kg of flunitrazepam orally (maximum dose 1 mg) and 0.3 mg/kg of morphine (maximum dose 5 mg) for premedication. For facilitation of cannulating a peripheral vein, if not already done, sevoflurane was given by mask. In all patients, induction of anesthesia was performed with fentanyl (15 ␮g/kg) and pancuronium (0.2 ␮g/kg), and the lungs were ventilated using a mask and 100% oxygen except in children with manifest pulmonary hypertension or hypoplastic left heart syndrome. After nasal intubation, all children were ventilated with a zero endexpiratory pressure and a fraction of inspired oxygen of 1.0. In children with hypoplastic left heart syndrome, the fraction of inspired oxygen was adjusted according to monitoring of arterial oxygen saturation (70%-80%) and frequently performed blood gas analyses. After bypass, all patients were slightly hyperventilated with 100% oxygen. Anesthesia was maintained with fentanyl, pancuronium, and midazolam. Doses were adapted to the actual hemodynamic situation of the patients. In all patients, 2 double-lumen central venous catheters (5F) were inserted in the right internal jugular vein and the left subclavian vein because of

HE USE OF A mechanical cardiac-assist device is now a common procedure in modern cardiac surgery in patients with end-stage failure,1-3 whereas in pediatric patients it is still a great challenge. The experience with the implantation of a ventricular-assist device (VAD) in infants is now growing because of the availability of small-size systems for this special group of patients.4-7 In addition to VAD, a various number of extracorporeal membrane oxygenation systems are available for pediatric patients with postoperative left and/or right ventricular failure.8,9 In the recent literature, a broad range of survival and weaning rates has been reported, depending on the variety of mechanical devices and the choice of patients with different myocardial conditions before implantation; furthermore, it is of importance whether the device is applied routinely or in case of emergency surgery.7,10-14 In pediatric patients, the implantation of a left ventricular device (LVAD) often ignores the fact of right ventricular failure because of pulmonary hypertension. With an LVAD, there is a significant risk of thromboembolic complications because of a decrease in filling and emptying caused by oversized chambers.6,7,15 The problems with right ventricular assist devices (RVADs) include the risks of lung injury because of overflow in the pulmonary vascular system. In addition, they are difficult to implant. Overall, it is still a challenge in postoperative intensive care because of sometimes excessive bleeding and subsequently a higher rate of reoperations. In this article, the authors report their experience with cardiac assist devices and the perioperative anesthetic management in neonates and children. PATIENTS AND METHODS From 1997 to 2001, 11 infants and children were supported with a left- or biventricular-assist device. Diagnosis included myocarditis and

KEY WORDS: heart transplantation, anesthesia, neonates, cardiac anesthesia, assist device

From the *Department of Anesthesiology and Intensive Care Medicine, German Pediatric Heart Center, Sankt Augustin, Germany; and †Department of Anesthesiology and Intensive Care Medicine; ‡Cardiovascular Surgery; and §Pediatric Cardiology, University Hospital Giessen, Giessen, Germany. Address reprint requests to Ehrenfried Schindler, MD, Department of Anesthesiology and Intensive Care Medicine, German Pediatric Heart Center, Asklepios Klinik Sankt Augustin, D-53713 Sankt Augustin, Germany. E-mail: [email protected] © 2003 Elsevier Inc. All rights reserved. 1053-0770/03/1705-0010$30.00/0 doi:10.1053/S1053-0770(03)00206-4

Journal of Cardiothoracic and Vascular Anesthesia, Vol 17, No 5 (October), 2003: pp 617-621

617

618

SCHINDLER ET AL

Table 1. Patient Characteristics and Outcome Patient

Diagnosis

Age (month)

Weight (kg)

Height (cm)

Outcome

1 2 3 4 5 6 7 8 9 10 11

CMP Myocarditis UVH AST/AOI CMP HTX/Vasculopathy Myocarditis HTX/primary graft failure Absent AOV HLHS TOF/CMP

15 11 65 1 135.6 54 23 10 0.3 42 55

9.2 8.8 19.0 3.3 24.1 18.0 11.2 5.9 4.1 15.0 15.1

85 71 106 64 134 105 81 70 36 94 110

HTX Recovered HTX RDS/died MOF/died MOF Recovered Re-HTX RDS/died RDS/bleeding/died Brain death/died

Abbreviations: CMP, cardiomyopathy; UVH, univentricular heart; AST/AOI, combined aortic stenosis and aortic insufficiency; HTX, heart transplantation; AOV, aortic valve; HLHS, hypoplastic left heart syndrome; TOF, tetralogy of Fallot; RDS, respiratory distress syndrome; MOF, multiorgan failure.

expected excessive postoperative bleeding. If possible, the left radial artery was cannulated (24 or 22 gauges) for on-line arterial blood pressure measurements and blood sampling immediately after induction of anesthesia. Alternatively, the right radial artery or 1 of the femoral arteries was cannulated. Monitoring in the perioperative period consisted of measuring arterial and central venous pressures, temperature (esophageal and rectal), oxygen saturation by pulse oximetry (immediately after entering the operation room), end-expiratory carbon dioxide, and urine output by a catheter. Blood gas analyses, electrolytes, colloid osmotic pressure, glucose, and osmolarity were measured from arterial samples. Special care was taken regarding some logistic aspects. For all patients, at least 4 units of red blood cells and 2 packs of fresh frozen plasma were ordered from the blood bank. All units of blood and plasma were irradiated before delivery because of the possibility of a future heart transplantation. The red blood cells were all leukocyte reduced and cytomegalovirus negative. In patients requiring conventional cardiopulmonary bypass (CPB), 300 U/kg of bovine heparin was administered to achieve anticoagulation. The activated coagulation time (ACT) was kept at more than 400 seconds during the entire bypass period. A COBE VPCMLplus membrane oxygenator (Cobe Laboratories, Lakewood, CO) was used, and a flow of 2.4 L/min/m2 was maintained during CPB. Priming of the extracorporeal circuit consisted of Ringer’s solution, 5% human albumin, and electrolytes. One unit of packed red blood cells was added to the prime in all children weighing less than 10 kg. In the group of children weighing more than 10 kg, packed red blood cells were added when the preoperative hematocrit was less than 35%. The amount of the priming volume was identical in all children (800 mL). Additional packed red blood cells were given when the hematocrit was less than

20%. When necessary, Ringer’s solution was added to maintain the filling of the circuit. After the end of CPB, the blood remaining in the extracorporeal oxygenation equipment was salvaged by hemofiltration. All autologous blood was retransfused in the postbypass period. Heparin was neutralized by administration of protamine chloride in a ratio of 1:1 with regard to the initially given dose of heparin. Further protamine was given when the ACT exceeded 200 seconds. To avoid clotting in the VAD, the ACT was kept around 180 seconds. Additional heparin was given according to the ACT. In all cases, a MEDOS HIA-VAD (MEDOS Medizintechnik AG, Stolberg, Germany) system was used. It is a pneumatically driven device, available with 9- and 22.5-mL ventricles. With the 9-mL ventricles, the goal was to achieve a mean flow of 1 L/min, and with the 22.5-mL ventricle a mean flow of 2 L/min was desired. The polyurethane ventricles are shaped with a double-layer inner displacement membrane. The 3-leaflet valves are incorporated seamlessly, and the ventricles are totally transparent to allow visual control of filling and emptying of the device as well as the observation of possible air or clot formation during prolonged pumping. In patients with an LVAD, the outflow cannula was sewn to the aorta and exteriorized through the left epigastrium. The inflow cannula was placed through the superior pulmonary vein and exteriorized. In case of biventricular-assist device placement, the right inflow cannula was inserted in the right atrial appendage and the pulmonary trunk was cannulated for the outflow. During the filling of the device, care was taken to remove all air. This could easily be done because of the transparent design of the ventricles. Arterial cannulae are available with flexible graft material of expanded polytetrafluoroethylene (ePTFE) in the sizes 1/4”: 6/8 mm; 3/8”: 8/10 mm; and 1/2”: 13 mm. The venous cannulae used are designed with malleable tip and distal basket in the sizes 1/4”: 14F/18F/24F and

Table 2. Perioperative Data, Blood, and Blood Products During and After Implantation of the VAD Group 1

Duration of anesthesia (min) Aortic cross-clamp (min) CPB (min) PRC (mL) FFP (mL) PC (mL)

Group 2

Mean

SD (Min-Max)

n

Mean

SD (Min-Max)

n

173.2 94 49.6 540.5 150.0 50

95.1(121-342)

5 1 2 5 5 4

631.1 154.0 512.1 880.6 250.0 150.0

258.8(150-880) 110.3(0-256) 246.7(0-758) 400.3(600-1200) 175.1(200-600) 60.8(50-150)

6 4 5 6 6 6

91.1(0-210) 150.3(100-600) 80.6(50-200) 50.0(0-50)

NOTE. Group 1⫽survivors and group 2⫽nonsurvivors. Abbreviations: CPB, cardiopulmonary bypass; PRC, packed red cells; FFP, fresh frozen plasma; PC, platelet concentrates.

PEDIATRIC VENTRICULAR ASSIST DEVICE

619

Table 3. Perioperative Drug Regimen Before and After Implantation of a VAD Before VAD

Catecholamines Dobutamine 5-10 ␮g/kg/min Epinephrine 0.1-1.5 ␮g/kg/min Norepinephrine 0.1-1.0 ␮g/kg/min PDE-III-Inhibitors Milrinone 1.0 ␮g/kg/min Enoximone 10 ␮g/kg/min Vasodilators Nitroglycerin 5 ␮g/kg/min Nitroprusside 0.5-10 ␮g/kg/min Prostaglandin E1 0.05-0.25 ␮g/kg/min NO 15-20 ppm

After VAD

⫹ (n ⫽ 8) ⫹ (n ⫽ 11) ⫹ (n ⫽ 6)

— — —

⫹ (n ⫽ 4)* ⫹ (n ⫽ 4)

⫹ (n ⫽ 1) —

⫹ ⫹ ⫹ ⫹

⫹ (n ⫽ 4) — ⫹ (n ⫽ 2) ⫹ (n ⫽ 4)

(n (n (n (n

⫽ ⫽ ⫽ ⫽

11) 3) 6) 9)

Abbreviations: PDE, phosphodiesterase inhibitor; NO, nitric oxide. *Enoximone was replaced by milrinone after 1999.

3/8”: 24F. At the time of the study, no coated cannulas or PVC drainage systems were available. After the implantation, the target ACT was between 180 and 200 seconds. The data sets of all patients were recorded using the online anesthesia record-keeping system NarkoData (IMESO GmbH, Hu¨ ttenberg, Germany). The program collects all perioperative data during surgery and during a stay in the PACU, including vital signs, administered drugs, and the data set of the German Society of Anesthesiology and Intensive Care Medicine. The anesthesiologists and nursing teams responsible for perioperative patient care documented both the premedication visitation and the entire perioperative procedure in real time. All data were collected and sorted on MS Excel (Microsoft Corp, Redmond, WA). The statistical analysis was performed by SPSS (SPSS Inc, Chicago, IL). Quantitative variables that approximated a normal distribution were reported as mean ⫾ SD and were analyzed by Student’s unpaired t test. If the variables did not fit a normal distribution, they were calculated by the Mann-Whitney U test (Wilcoxon rank sum test). RESULTS

The demographic data as well as the outcome of each patient are presented in Table 1. The oldest patient was 11 years and the youngest 20 days. The mean age of all patients was 3 years. More than half of the patients were younger than 2 years. The main diagnoses were cardiomyopathy (n ⫽ 2), myocarditis (n ⫽ 2), complex congenital heart defects (n ⫽ 5), and former heart transplantation (n ⫽ 2). The perioperative data are shown in Table 2. All patients were divided into 2 groups: group 1 was made up of survivors and group 2 of nonsurvivors. A total of 5 patients were in group

1, and group 2 consisted of 6 patients. The duration of anesthesia in group 1 patients (173.2 ⫾ 95.1 minutes) was significantly (p ⬍ 0.05) shorter than in group 2 (631.1 ⫾ 258.8 minutes), as well as the amount of packed red cells (group 1 ⫽ 540.5 ⫾ 150.3 mL, group 2 ⫽ 880.6 ⫾ 400.3 mL). Standard CPB during the implantation of a VAD was necessary only in 2 patients from group 1, whereas 5 patients in group 2 were on pump during the procedure. The rate of aortic cross-clamping was also significantly lower in group 1 compared with group 2 (p ⬍ 0.05). All patients were on catecholamine therapy before the implantation of the VAD (Table 3). Five patients had a triple combination with dobutamine, epinephrine, and norepinephrine. To lower the pulmonary vascular resistance, phosphodiesterase-III inhibitors were used as well as nitroglycerin (n ⫽ 3), prostaglandin E1 (n ⫽ 6), and nitroprusside (n ⫽ 3). Nitric oxide was given in various concentrations in 9 patients until starting the VAD. In all cases, the catecholamine therapy could be stopped when switching on the VAD device. When using an LVAD, some vasodilators were continued to protect the right heart. There was no difference in the number of cardiopulmonary resuscitations in both groups; eventhough most of the patients from group 2 came with multiorgan failure (n ⫽ 5). Although it was not significant, there was a higher need for massive transfusion after the operation in group 2 (n ⫽ 5) patients compared with group 1 (n ⫽ 3). The number of sepsis patients and the frequency of hemofiltration or peritoneal dialysis were not different in the groups (Table 4). In group 1, all patients were elective surgery in contrast to group 2 in which 4 patients out of 6 were emergencies (Table 5). The duration of the VAD was not statistically different in the groups. One patient in group 1 was treated for 30 days with a MEDOS (MEDOS GmbH, Steinfurt, Germany) device. This patient recovered from a severe myocarditis and went home about 1 month after discharge from the intensive care unit. The rate of surgical intervention because of excessive bleeding was the same in both groups. In nearly all patients, the chest was closed in a second operation. DISCUSSION

The use of an LVAD and RVAD in pediatric patients for mechanical circulatory support has become more common.1,10,16 This retrospective investigation is focused on the anesthetic impact and data from the intensive care unit.

Table 4. Intensive Care Data and Outcome Group 1

Group 2

Patient

1

2

3

4

5

6

7

8

9

10

11

CPR before VAD MOF before VAD Massive Transfusion Sepsis HF/CAPD

2 N Y Y N

— N N Y Y

1 Y N N N

1 N Y N N

— N Y N Y

1 Y Y N Y

2 Y Y Y Y

1 Y Y N Y

— Y Y N N

— Y Y N N

— N N N Y

Abbreviations: Group 1, survivors; group 2, nonsurvivors; Y, yes; N, no; HF, hemofiltration; CAPD, continuous peritoneal dialysis; CPR, cardiopulmonary resuscitation; MOF, multiorgan failure.

620

SCHINDLER ET AL

Table 5. Surgical Data and Type of VAD According to the Outcome Group 1

Group 2

Patient

1

2

3

4

5

6

7

Emergency surgery Duration of VAD (d) VAD type Secondary chest closure Operative redo Change of VAD

N 30 LVAD Y 2 1

N 13 LVAD N 0 0

N 0.5 LVAD N 0 0

N 15 LVAD Y 4 0

N 9 BVAD Y 4 0

Y 3 LVAD Y 2 0

N 21 BVAD Y 1 0

8

Y 2 BVAD Y 2 2

9

10

11

Y 1 BVAD Y 1 0

Y 0.8 LVAD Y 0 0

N 14 LVAD N 1 0

Abbreviations: Group 1, survivors; group 2, nonsurvivors; Y, yes; N, no; BVAD, biventricular assist device; LVAD, left ventricular assist device.

Many factors influence the anesthetic management during the perioperative period. These data suggest that the surgical outcome depends largely on the patient’s condition at the time of surgery. Emergency surgery, preoperative multiorgan failure, and the need for an extracorporeal circulation with aortic cross-clamping probably are predictive of a negative outcome in this group of patients. The frequency of intraoperative as well as postoperative excessive bleeding force the anesthesiologist to ensure the availability of an adequate amount of blood components and sufficient vascular access. When the patients were scheduled for cardiac transplantation, the blood and the blood plasma were irradiated, which was time consuming. These results are comparable to the data presented by the concerted action registry on heart assist and replacement of the European communities medical and health program-BIOMED I.17 These data were collected from European cardiac surgery departments and are based on voluntary contributions. The registry was set up to assess the global activities in the field of heart assist and replacement. According to this report, the indications for implantation of a VAD was cardiomyopathy in 69%, postacute myocardial infection in 15%, postcardiotomy cardiogenic shock in 8%, graft failure in 5%, and cardiac rejection in 3%, which seems to reflect the distribution in the authors’ patients. In 68% of the patients, the VADs were applied with the aid of CPB, which is comparable with the present data. Walters et al18 reported on a larger number of pediatric cases. In their analysis, the pre-VAD risk factors for negative outcome were pulmonary artery hypertension, the need for CPB, and the extracorporeal bypass time, which is supported by these results. The main reason for continuing the vasodilating medication for all of the patients was pulmonary hypertension, whereas most of the catecholamines were discontinued when starting the VAD. The authors’ strategy is in accordance with the observation of Waldenberger et al6 who found in an animal model that pulmonary hypertension even in biventricular-assist devices could lead to a significant decrease in inflow to the left heart. This always leads to a low LVAD output syndrome.6 To avoid

a negative impact on the pulmonary vascular system, the authors used pressure-controlled ventilation and adjusted the positive end-expiratory airway pressure to the PA pressure and the filling of the artificial left heart. With the transparent design of the device used, the filling and emptying could easily be monitored so the preload and flow could be adjusted. To allow adequate flow rates, phosphodiesterase inhibitors were used before starting the VAD. In Germany, phentolamine is not available, and the authors now import it and are supplementing their therapy with phentolamine (starting with 2 ␮g/kg/min) to achieve complete arterial vasodilatation. All explantations of the VAD were done in the operating room. A major problem after implantation of the VAD was prolonged bleeding (8 of 11 children were transported to the intensive care unit with open chest and just 1 child was not redone for bleeding). The ACT was adjusted to 180 to 220 seconds. At the present time, anticoagulant-coated systems, like those used in extracorporeal membrane oxygenator systems, are available, and they must be considered if a reduction of heparin could reduce the postoperative bleeding. Aprotinin was used in 5 of 11 children, but the small number of cases and the case mix is not suitable for a statistical analysis of the possible blood-saving effect of aprotinin. This study has some limitations. A basic impact on statistical analysis is the small number of patients. In addition, the age of these patients and the fact that this is a single-center report must be considered. The retrospective design is another problem in all of these investigations. Of course, a prospective, randomized trial is the favorable study design, but such trials are hard to do because of ethical reasons in these critical patients. In summary, prospective studies with greater numbers of patients are necessary to find predictors of right and left ventricular function and organ failure reversibility. Despite the small numbers of patients, these data suggest that the surgical outcome in pediatric patients is dependent on the patient’s condition at the time of surgery, the need for emergency surgery, preoperative multiorgan failure, and the use of extracorporeal circulation with aortic cross-clamping.

REFERENCES 1. Scherr K, Jensen L, Koshal A: Mechanical circulatory support as a bridge to cardiac transplantation: Toward the 21st century. Am J Crit Care 8:324-337, 1999 2. Dembitsky WP: Bridging from acute to chronic devices. Ann Thorac Surg 68:724-728, 1999 3. Kocandrle V, Fabian J, Firt P: State of the art and future trends in heart transplantation and ventricular assist devices. Cor Vasa 29:325-332, 1987

4. Rakhorst G, Hensens AG, Verkerke GJ, et al: In-vivo evaluation of the “HIA-VAD”: A new German ventricular assist device. Thorac Cardiovasc Surg 42:136-140, 1994 5. Herwig V, Severin M, Waldenberger FR, et al: Medos/HIA-assist system: First experiences with mechanical circulatory assist in infants and children. Int J Artif Organs 20:692-694, 1997 6. Waldenberger FR, Meyns B, Reul H, et al: Pre-clinical eval-

PEDIATRIC VENTRICULAR ASSIST DEVICE

uation of a novel, pneumatic, ventricular assist device (Medos(R) HIA-VAD(R)) under pathophysiological conditions. Int J Artif Organs 20:389-396, 1997 7. Reul H: The MEDOS/HIA system: Development, results, perspectives. J Thorac Cardiovasc Surg 47:311-315, 1999 (suppl) 8. Pennington DG, Reedy JE, Swartz MT, et al: Univentricular versus biventricular assist device support. J Heart Lung Transplant 10:258-263, 1991 9. Waldenberger FR: Pathophysiological considerations concerning uni- and biventricular mechanical cardiac assist. Int J Artif Organs 20:684-691, 1997 10. Duncan BW, Hraska V, Jonas RA, et al: Mechanical circulatory support in children with cardiac disease. J Thorac Cardiovasc Surg 117:529-542, 1999 11. Pennington DG, Swartz MT: Circulatory support in infants and children. Ann Thorac Surg 55:233-237, 1993 12. Termuhlen DF, Swartz MT, Ruzevich SA, et al: Hemodynamic predictors for weaning patients from ventricular assist devices (VADs). J Biomater Appl 4:374-390, 1990

621

13. Ibrahim AE, Duncan BW, Blume ED, et al: Long-term follow-up of pediatric cardiac patients requiring mechanical circulatory support. Ann Thorac Surg 69:186-192, 2000 14. El Banayosy A, Deng M, Loisance DY, et al: The European experience of Novacor left ventricular assist (LVAS) therapy as a bridge to transplant: A retrospective multi-centre study. Eur J Cardiothorac Surg 15:835-841, 1999 15. Walley VM, Masters RG, Boone SA, et al: Analysis of deaths after heart transplantation: The University of Ottawa Heart Institute experience. J Heart Lung Transplant 12:790-801, 1993 16. Weyand M, Kececioglu D, Kehl HG, et al: Neonatal mechanical bridging to total orthotopic heart transplantation. Ann Thorac Surg 66:519-522, 1998 17. Quaini E, Pavie A, Chieco S, et al: The Concerted Action ‘Heart’ European registry on clinical application of mechanical circulatory support systems: bridge to transplant. The Registry Scientific Committee. Eur J Cardiothorac Surg 11:182-188, 1997 18. Walters HL III, Hakimi M, Rice MD, et al: Pediatric cardiac surgical ECMO: Multivariate analysis of risk factors for hospital death. Ann Thorac Surg 60:329-336, 1995