Share this article on:

Pediatric ECMO Outcomes: Comparison of Centrifugal Versus Roller Blood Pumps Using Propensity Score Matching

Barrett, Cindy S.*; Jaggers, James J.; Cook, E. Francis; Graham, Dionne A.§; Yarlagadda, Vasmi V.§; Teele, Sarah A.§; Almond, Christopher S.§; Bratton, Susan L.; Seeger, John D.; Dalton, Heidi J.#; Rycus, Peter T.**; Laussen, Peter C.§; Thiagarajan, Ravi R.§

doi: 10.1097/MAT.0b013e31828387cd
Pediatric Circulatory Support And Perfusion

Centrifugal blood pumps are being increasingly utilized in children supported with extracorporeal membrane oxygenation (ECMO). Our aim was to determine if survival and ECMO-related morbidities in children supported with venoarterial (VA) ECMO differed by blood pump type.

Children aged less than 18 years who underwent VA ECMO support from 2007 to 2009 and reported to the Extracorporeal Life Support Organization registry were propensity score matched (Greedy 1:1 matching) using pre-ECMO characteristics.

A total of 2,656 (centrifugal = 2,231, roller = 425) patients were identified and 548 patients (274 per pump type) were included in the propensity score-matched cohort. Children supported with centrifugal pumps had increased odds of hemolysis (odds ratio [OR], 4.03 95% confidence interval [CI], 2.37–6.87), hyperbilirubinemia (OR, 5.48; 95% CI, 2.62–11.49), need for inotropic support during ECMO (OR, 1.54; 95% CI, 1.09–2.17), metabolic alkalosis (blood pH > 7.6) during ECMO (OR, 3.13; 95% CI, 1.49–6.54), and acute renal failure (OR, 1.61; 95% CI, 1.10–2.39). Survival to hospital discharge did not differ by pump type.

In a propensity score-matched cohort of pediatric ECMO patients, children supported with centrifugal pumps had increased odds of ECMO-related complications. There was no difference in survival between groups.

From the *Department of Cardiology, Children’s Hospital Colorado and Department of Pediatrics, University of Colorado School of Medicine, Aurora, Colorado; Department of Surgery, Children’s Hospital Colorado, University of Colorado School of Medicine, Aurora, Colorado; Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts; §Department of Cardiology, Children’s Hospital Boston and Department of Pediatrics, Harvard Medical School, Boston, Massachusetts; Pediatric Critical Care Medicine, Department of Pediatrics, Primary Children’s Hospital and the University of Utah, Salt Lake City, Utah; Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts; #Department of Critical Care Medicine, Phoenix Children’s Hospital, Phoenix, Arizona; and **Extracorporeal Life Support Organization, University of Michigan, Ann Arbor, Michigan.

Submitted for consideration September 2012; accepted for publication in revised form December 2012.

Disclosure: John Seeger has a financial relationship with OptumInsight and WHISCON and Christopher Almond has a NHLBI PumpKIN Contract and is the Clinical PI for DCCC awarded to NERI (CRO). The authors have no conflicts of interest to report.

Reprint Requests: Cindy S. Barrett, MD, MPH, Children’s Hospital Colorado, Department of Cardiology, 13123 East 16th Avenue, Box 100, Aurora, CO 80045. Email:

Extracorporeal membrane oxygenation (ECMO) is the most commonly used form of mechanical circulatory support for children with cardiorespiratory failure in the USA.1 Complications during ECMO are common and are associated with increased risk of end-organ injury as well as mortality. Complications due to mechanical components of the ECMO circuit, including the ECMO pump, are also common and may contribute to patient morbidity. Centrifugal pumps are being increasingly used in pediatric patients and may offer several potential benefits, including smaller circuit size requiring lower circuit prime volumes, smaller surface area for blood–prosthetic surface interaction resulting in decreased activation of inflammatory factors.2–6 Whether advantages of centrifugal pumps translate into improved patient outcomes and decreased incidence of complications remains unknown.

The primary aim of our study was to compare survival to hospital discharge in pediatric ECMO patients supported with centrifugal versus roller pumps. Our secondary aim was to compare ECMO-related complications between pump types.

Back to Top | Article Outline


Data for purposes of this study were obtained from the Extracorporeal Life Support Organization’s (ELSO) ECMO data registry. The ELSO registry collects national and international ECMO data from 140 member centers. Data are reported voluntarily and include patient demographics, pre-ECMO support, ECMO details including equipment and complications, and survival to hospital discharge. Data are reported from the member centers after approval from their local scientific review boards, and the data user agreements allow release of limited de-identified data sets from ELSO to member centers for purposes of research. For this analysis, all patients ≤ 18 years of age undergoing venoarterial (VA) ECMO support from January 1, 2007, to December 31, 2009, were analyzed. Patients were excluded if documentation of pump type was missing. Patient demographics, pre-ECMO and ECMO variables, and ECMO-related complications were extracted.

Back to Top | Article Outline

Demographic and Pre-ECMO Support Data

Demographic data included age, weight, race, indication for ECMO, and arterial blood gas values prior to ECMO cannulation. Pre-ECMO support variables included ventilator type and duration of mechanical ventilation prior to initiation of ECMO, pre-ECMO cardiovascular support including use of inotropic support, and adjunct therapies including the use of inhaled nitric oxide and high-frequency oscillatory ventilation (HFOV). Pediatric age groups were separated into four categories: neonates (< 30 days of life), infant (> 30 days to < 1 year of age), young children (> 1 year to 5 years of age), and older children (> 5 years to 18 years). The patient’s primary physiologic indication for ECMO was placed in one of three categories, pulmonary, cardiac or ECMO used to support cardiopulmonary resuscitation (ECPR). Blood gas pH and serum bicarbonate were trichotomized as previously described by Thiagarajan et al.7

Back to Top | Article Outline


The type of pump used to configure the ECMO circuit was categorized based on pump technology into centrifugal or roller pump types (table of pump types included online). ECMO variables included number of hours from hospital admission to initiation of ECMO support, arterial and venous cannulation site, use of percutaneous cannulation strategy, ECMO pump flow rate at 4 and 24 hours, and total number of hours of ECMO support.

Back to Top | Article Outline

ECMO Complication Data

ECMO complications were categorized using complication codes created by the ELSO registry and were grouped into the following categories: 1) Mechanical complications including ECMO circuit complications that include oxygenator failure, raceway or other tubing rupture, pump malfunction, air in circuit, cracks in connectors or cannula problems, circuit thrombosis, evidence of disseminated intravascular coagulation, and evidence of hemolysis (plasma-free hemoglobin > 50 mg/dl); 2) ECMO support complications included cannula or surgical site bleeding, need for inotropes, cardiopulmonary resuscitation (CPR), metabolic alkalosis (blood pH > 7.60), or metabolic acidosis (blood pH < 7.20) while on ECMO support; 3) cardiac complications included cardiac arrhythmias that required treatment, cardiac tamponade, diagnosis of myocardial stun, and systemic hypertension during ECMO support; 4) respiratory complications included pneumothorax and pulmonary hemorrhage; 5) gastrointestinal (GI) complications included GI hemorrhage and hyperbilirubinemia (defined as serum direct bilirubin level > 2 mg/dl and/or indirect bilirubin level > 13 mg/dl, and/or total bilirubin level > 15 mg/dl); 6) acute renal failure was defined as serum creatinine value > 1.5 mg/dl and/or use of hemodialysis or continuous arteriovenous hemodialysis; and 7) neurologic injury included seizures (clinical or electroencephalographic evidence) and/or ultrasonographic or computed tomographic scan evidence of hemorrhage and/or infarction or brain death.

Back to Top | Article Outline

Statistical Analysis

Demographic, pre-ECMO, and ECMO support details as well as ECMO complications were compared for children undergoing ECMO with centrifugal and roller pumps. Due to significant differences in sample size and observed covariates between centrifugal and roller pump users, we performed propensity score matching to select a similar cohort of patients from the centrifugal and roller pump users to study outcomes.

Greedy propensity score matching was performed using SAS (version 9, Cary, NC) to select the matched study cohorts (1:1 ratio).8 For children who underwent more than one ECMO run, only data from the first ECMO run were analyzed. Patients were excluded from propensity score matching if data on any of the matching variables were missing. Pre-ECMO variables used for propensity scoring included age, weight, race, indication for support (pulmonary, cardiac, or ECPR), hours of mechanical ventilation prior to ECMO support, history of pre-ECMO arrest, arterial cannulation site (aorta, carotid, or femoral), presence of internal jugular cannulation, percutaneous cannulation, inotropic support, bicarbonate administration, surfactant administration, use of inhaled nitric oxide, HFOV or additional mechanical circulatory support, and serum blood gas parameters prior to ECMO (serum pH, partial pressure of arterial carbon dioxide, partial pressure of arterial oxygen, and serum bicarbonate level). Pre-ECMO variables mean airway pressure (MAP), fraction of inspired oxygen, and arterial oxygen saturation had > 20% missing data and were not used in building the propensity score model.

Some of the older generation centrifugal pumps included in the above analysis were known to be associated with increased incidence of hemolysis. The dataset was then limited to newer generation centrifugal pumps and a separate propensity score analysis using the same pre-ECMO variables was performed.

Continuous variables are reported as median values with interquartile ranges and compared using the Mann–Whitney U test, and categorical variables were compared using χ2 test or Fisher’s exact test (SPSS 12 for Windows, Release 12.0.7, SPSS Inc. Chicago, IL). Odds ratios (OR) with 95% confidence intervals (CI) are reported. A p-value < 0.05 was considered statistically significant.

Back to Top | Article Outline


Unmatched Pediatric Cohort

A total of 2,977 patients less than 18 years of age underwent VA ECMO from 2007 to 2009. Of the 2,656 patients whose pump type was identified, 2,231 were supported with roller and 425 with centrifugal pumps. A flow diagram of patient inclusion (Figure 1) is provided online. A total of 1,388 patients survived to hospital discharge, and there was no difference in mortality between the two pump groups (centrifugal = 215 (51%), roller = 1,173 (53%), p = 0.45). Multiple demographic, pre-ECMO, and ECMO support variables differed between the two groups (Tables 1 and 2).

Back to Top | Article Outline

Propensity Score-Matched Cohort

A total of 2,611 patients were entered into the analysis, of which 2,036 (centrifugal = 326, roller = 1,710) had no missing variables used propensity scoring. The propensity score matching identified 548 patients (centrifugal = 274, roller = 274; c-statistic = 0.81, Hosmer–Lemeshow lack-of-fit statistic = 0.55). Figure 2 provided online shows the distribution of predicted probabilities between unmatched and matched data. After matching, there were no significant differences in demographic, pre-ECMO, and ECMO support (Tables 1 and 2).

Children supported with centrifugal pumps had increased odds of hemolysis (OR, 4.03; 95% CI, 2.37–6.87), hyperbilirubinemia (OR, 5.48; 95% CI, 2.62–11.49), and acute renal failure (OR, 1.61; 95% CI, 1.10–2.39; Table 3). A higher proportion of patients in the centrifugal pump group received inotropic support during ECMO (OR, 1.54; 95% CI, 1.09–2.17), were diagnosed with myocardial stun (OR, 2.24; 95% CI 1.21–4.06), and had a serum pH > 7.6 during ECMO support (OR, 3.13; 95% CI, 1.49–6.54). Reason for discontinuation of ECMO support differed between the propensity scored cohort (p = 0.017), with a higher number of patients within the roller group having hemorrhage listed as a reason for discontinuation and a higher number of patients within the centrifugal group having end-organ injury listed as a reason for discontinuation (Table 4).

Back to Top | Article Outline

Propensity Score Matching Limited to Magnetically Stabilized Centrifugal Pumps

A total of 114 patients (n = 57 centrifugal, n = 57 roller) were propensity score matched (c-statistic = 0.89 and Hosmer–Lemeshow lack-of-fit statistic = 0.38). There were no significant differences in patient demographics or in pre-ECMO variables, with the exception of pre-ECMO MAP. Patients supported with centrifugal pumps had an increased odds of circuit thrombosis 9.0 (95% CI, 2.49–32.58), hemolysis 14.93 (95% CI, 1.87–119.22), diagnosis of myocardial stun 12.69 (95% CI, 2.78–57.89), hypertension 4.8 (95% CI, 1.63–14.05), hyperbilirubinemia 1.58 (95% CI, 1.30–1.93), acute renal failure 8.06 (95% CI, 3.38–19.22), and cerebral infarction 1.16 (95% CI, 1.05–1.29) (Table 5).

Back to Top | Article Outline


Children supported with ECMO utilizing centrifugal blood pumps had increased odds of ECMO-related complications including hemolysis, hyperbilirubinemia, and acute renal failure. Notably, there was no difference in in-hospital survival between pump types. The complications of hemolysis, hyperbilirubinemia, and acute renal failure are likely interrelated as hemolysis can lead to subsequent hyperbilirubinemia,9,10 and hemolysis has been shown to be associated with increased incidence of acute renal failure.11 Although hyperbilirubinemia is a known morbidity associated with ECMO support, ELSO categorization of hyperbilirubinemia in the ELSO database does not differentiate between increased serum levels of direct or indirect serum bilirubin. Therefore, the etiology for hyperbilirubinemia cannot be directly discerned.

Renal injury during ECMO is a common ECMO-related complication. The cause of renal injury in this population is likely multifactorial and may include factors such pre-ECMO hemodynamic instability. ECMO factors including nonpulsatile blood flow, exposure to nephrotoxic medications, and low intravascular volume may also contribute to renal injury during ECMO. The use of continuous renal replacement therapy during ECMO support may increase turbulent flow, causing hemolysis and thus contribute to renal injury.11–13, 18 The presence of acute renal failure is associated with mortality.14–18 Although we found that patients supported with centrifugal pumps were at increased odds of renal injury, there was no difference in survival between patients supported using centrifugal pumps versus those supported with roller pumps. The reason listed as the cause for discontinuation of ECMO support did differ between groups, with end-organ failure listed more often in patients supported with centrifugal pumps.

Hemolysis is also common in patients supported with ECMO and may be generated from the blood pump (both centrifugal and roller), components of the ECMO circuit such as the oxygenator, blood to air contact, or from an area of turbulent flow such as those associated with circuit thrombus.9,19,20 Red blood cell lysis leads to the release of plasma-free hemoglobin. Heme proteins are known to be nephrotoxic and may lead to kidney injury via decreased renal perfusion, direct cytotoxicity, and creation of intratubular casts formed from the interaction of heme proteins.12,21 Free hemoglobin also contributes to other end-organ injury by scavenging nitric oxide, adenosine, and a disintegrin-like and metalloproteinase protein with thrombospondin Type 1 repeats-13, leading to microvascular vasoconstriction and platelet thrombi.22

In patients supported with roller pumps, hemolysis may be induced by shear forces exerted by compression of the tubing by the roller head.2,9,11,19,20,23 Hemolysis secondary to the use of centrifugal pumps has been attributed to both the negative pressure created in the venous cannula causing cavitation of red cells and subsequent hemolysis, and turbulence caused by thrombus formation within the centrifugal pump.19,24 Results of prior studies comparing the incidence of hemolysis for roller and centrifugal pumps have shown varying results, with many studies showing increased incidence in roller head pumps. However, these studies have not shown any differences in patient outcomes and are limited by the small number of patients.2,9,11,19,20,23

The blood flow provided by centrifugal pumps may vary with changes in the patient systemic vascular resistance such that increased vascular resistance may lead to decreased pump flows for a given rotational speed, resulting in inadequate organ perfusion. ECMO operators may attempt to overcome this issue by increasing ECMO flows. Using the servo regulator for monitoring of pump inlet pressures, which prevents negative inlet pressures, has been shown to decrease the potential of hemolysis when using centrifugal pumps.23 ECMO circuits using roller pumps are limited by servo regulation of gravity siphon on the venous return and therefore may not generate enough pressure to cause blood cavitation and subsequent hemolysis. Whereas with centrifugal style blood pumps, when flow is increased, suction pressure increases leading to blood cavitation and hemolysis.24

The incidence of hemolysis has been shown to vary by centrifugal pump type, and newer generation pumps cause less blood trauma and hemolysis compared with older generation pumps.5 When our dataset was limited to newer generation centrifugal pumps, children supported with centrifugal pumps had increased odds of circuit thrombosis, hemolysis, diagnoses of myocardial stun, hypertension, hyperbilirubinemia, acute renal failure, and cerebral infarction. There was no difference in mortality between cohorts.

Children supported with centrifugal pumps were more likely to be diagnosed with myocardial stun, require inotropic support, and have evidence of metabolic derangement with a serum pH greater than 7.6 while supported with ECMO. Whereas when the dataset was limited to newer centrifugal pumps, children supported with centrifugal pumps had increased odds of diagnosis of myocardial stun and hypertension. Whether these additional findings were incidental or possibly indicative of a “sicker” patient population represented in the centrifugal pump cohort remains unknown.

We acknowledge several limitations related to our dataset. The data were obtained from the ELSO registry, a large database, and thus, our analysis may be limited by missing data and misclassification. There are 140 centers contributing data to the registry, and practice variability in ECMO cannulation and management between these centers cannot be accounted for as the data user agreement between ELSO and member centers do not allow the use of center-related details for purposes of analysis. These data were not specifically collected for the purposes of comparing centrifugal and roller pumps, and thus important covariates that may influence ECMO outcomes and complications may not have been collected. Finally, the ELSO database does not provide data concerning long-term clinical outcomes and these important issues could not be evaluated in these analyses.

We analyzed the impact of centrifugal versus roller pump technology as a class effect. Many of the centrifugal pumps included in the analysis have been associated with hemolysis and newer centrifugal pump models are less well represented. Outcomes of comparisons using propensity scoring of newer centrifugal pump models, due to the limited number of patients, may not be generalizable to a larger population. Data on inflow cannula size, oxygenator type, pre-ECMO hemoglobin levels, or transfusion practices by center were not available for analysis. Physiologic data on cardiac loading conditions or echocardiogram findings are not available for determining the cause of myocardial stun.

The use of ECMO to support pediatric patients continues to have significant associated morbidities. In this study, centrifugal blood pumps were associated with increased odds of ECMO-related morbidities including hemolysis, hyperbilirubinemia, and acute renal failure. Given the increasing interest in the use of centrifugal pumps to configure ECMO circuits, further research into the use of centrifugal blood pumps as well as mechanisms to decrease ECMO-related morbidities needs to be carried out.

Back to Top | Article Outline


1. Duncan BW. Pediatric mechanical circulatory support in the United States: Past, present, and future. ASAIO J. 2006;52:525–529
2. Klein M, Dauben HP, Schulte HD, Gams E. Centrifugal pumping during routine open heart surgery improves clinical outcome. Artif Organs. 1998;22:326–336
3. Jakob HG, Hafner G, Thelemann C, Stürer A, Prellwitz W, Oelert H. Routine extracorporeal circul.0.ation with a centrifugal or roller pump. ASAIO Trans. 1991;37:M487–M489
4. Wheeldon DR, Bethune DW, Gill RD. Vortex pumping for routine cardiac surgery: A comparative study. Perfusion. 1990;5:135–143
5. Lawson DS, Ing R, Cheifetz IM, et al. Hemolytic characteristics of three commercially available centrifugal blood pumps. Pediatr Crit Care Med. 2005;6:573–577
6. Masalunga C, Cruz M, Porter B, Roseff S, Chui B, Mainali E. Increased hemolysis from saline pre-washing RBCs or centrifugal pumps in neonatal ECMO. J Perinatol. 2007;27:380–384
7. Thiagarajan RR, Laussen PC, Rycus PT, Bartlett RH, Bratton SL. Extracorporeal membrane oxygenation to aid cardiopulmonary resuscitation in infants and children. Circulation. 2007;116:1693–1700
8. Parsons L Reducing Bias in a Propensity Score Matched-Pair Sample Using Greedy Matching Techniques. 2001 Cary, NC SAS institute
9. Gbadegesin R, Zhao S, Charpie J, Brophy PD, Smoyer WE, Lin JJ. Significance of hemolysis on extracorporeal life support after cardiac surgery in children. Pediatr Nephrol. 2009;24:589–595
10. Walsh-Sukys MC, Cornell DJ, Stork EK. The natural history of direct hyperbilirubinemia associated with extracorporeal membrane oxygenation. Am J Dis Child. 1992;146:1176–1180
11. Steinhorn RH, Isham-Schopf B, Smith C, Green TP. Hemolysis during long-term extracorporeal membrane oxygenation. J Pediatr. 1989;115:625–630
12. Betrus C, Remenapp R, Charpie J, et al. Enhanced hemolysis in pediatric patients requiring extracorporeal membrane oxygenation and continuous renal replacement therapy. Ann Thorac Cardiovasc Surg. 2007;13:378–383
13. Sakota D, Sakamoto R, Sobajima H, et al. Mechanical damage of red blood cells by rotary blood pumps: Selective destruction of aged red blood cells and subhemolytic trauma. Artif Organs. 2008;32:785–791
14. Shuhaiber J, Thiagarajan RR, Laussen PC, et al. Survival of children requiring repeat extracorporeal membrane oxygenation after congenital heart surgery. Ann Thorac Surg. 2011;91:1949–1955
15. Rood KL, Teele SA, Barrett CS, et al. Extracorporeal membrane oxygenation support after the Fontan operation. J Thorac Cardiovasc Surg. 2011;142:504–510
16. Chen YC, Tsai FC, Chang CH, et al. Prognosis of patients on extracorporeal membrane oxygenation: the impact of acute kidney injury on mortality. Ann Thorac Surg. 2011;91:137–142
17. Kumar TK, Zurakowski D, Dalton H, et al. Extracorporeal membrane oxygenation in postcardiotomy patients: Factors influencing outcome. J Thorac Cardiovasc Surg. 2010;140:330–336.e2
18. Askenazi DJ, Ambalavanan N, Hamilton K, et al. Acute kidney injury and renal replacement therapy independently predict mortality in neonatal and pediatric noncardiac patients on extracorporeal membrane oxygenation. Pediatr Crit Care Med. 2011;12:e1–e6
19. Thiara AP, Hoel TN, Kristiansen F, Karlsen HM, Fiane AE, Svennevig JL. Evaluation of oxygenators and centrifugal pumps for long-term pediatric extracorporeal membrane oxygenation. Perfusion. 2007;22:323–326
20. Segers PA, Heida JF, de Vries I, Maas C, Boogaart AJ, Eilander S. Clinical evaluation of nine hollow-fibre membrane oxygenators. Perfusion. 2001;16:95–106
21. Qian Q, Nath KA, Wu Y, Daoud TM, Sethi S. Hemolysis and acute kidney failure. Am J Kidney Dis. 2010;56:780–784
22. Carcillo JA. Multiple organ system extracorporeal support in critically ill children. Pediatr Clin North Am. 2008;55:617–46, x
23. Pedersen TH, Videm V, Svennevig JL, et al. Extracorporeal membrane oxygenation using a centrifugal pump and a servo regulator to prevent negative inlet pressure. Ann Thorac Surg. 1997;63:1333–1339
24. Toomasian JM, Bartlett RH. Hemolysis and ECMO pumps in the 21st Century. Perfusion. 2011;26:5–6

pediatric; ECMO

Copyright © 2013 by the American Society for Artificial Internal Organs