Extracorporeal membrane oxygenation (ECMO) supports patients with pulmonary or cardiac failure (or both) by means of vascular cannulas, a pump, and an extracorporeal oxygenator. Since the initial use in adults,1,2 ECMO has become a common support technology for neonates and children with cardiorespiratory failure of diverse etiologies.3,4 As survival rates with ECMO have improved, investigative focus has turned toward reduction in the complications associated with the use of the technology.5–7 Pump-related hemolysis, initially recognized as a complication of cardiopulmonary bypass (CPB) nearly 20 years ago, continues to remain a major problem in extracorporeal circuits including ECMO circuits.8 Excessive pump-related hemolysis causes increased complement activation and is associated with decreased diuresis on CPB, prolonged postoperative ventilation time, and prolonged intensive care unit (ICU) and hospital length of stay.9,10 Recent pediatric studies further highlight that hemolysis during ECMO is associated with significant renal injury, increased need for renal replacement therapy, and increased mortality.11,12
Hemolysis is recognized to occur in response to certain components of the circuit priming solution, mechanical forces related to the pump design, and initial acceleration of the pump.4–8,12 Pump-associated hemolysis has been demonstrated using in vitro models, animal studies, and in clinical series and occurs with both roller-head and centrifugal blood pumps.11,12,14 In an early series of seven neonates supported on ECMO with a Biomedicus BP-50 centrifugal pump, Steinhorn et al.8 described pump-related hemolysis manifested as a rise in the plasma hemoglobin level after the initial 48 hours on support. Later, in a larger case series of ECMO patients supported with a centrifugal pump, the same investigators demonstrated that hemolysis could be minimized by replacing the centrifugal pump head with another identical pump head. The investigators speculated that the negative pressure in the pump head caused the hemolysis.12 Based in part on concerns about hemolysis with centrifugal pumps, use of roller pumps became the norm in most pediatric ECMO centers and currently remains the choice of most ECMO practitioners in more than 80% of cases.13 Interestingly, there are very few direct comparisons of the two pump types in clinical use in ECMO. Most comparisons have been performed in animal models14–17 and show minimal differences between pump types.
We sought to compare our clinical experience with a newer model centrifugal pump (Levitronix Centrimag) to our concurrent experience with traditional roller pump ECMO (Stockert-Shiley SIII), with particular focus on hemolysis and circuit durability. We hypothesized that less hemolysis, as measured by plasma-free hemoglobin (PFH) concentrations, occurs in pediatric cardiac patients supported with ECMO using the Levitonix Centrimag pump when compared with the Stockert-Shiley roller-head pumps.
After receiving approval from the University of Arkansas for Medical Science Institutional Review Board, the hospital ECMO database was queried to identify the study cohort. All patients aged 0–18 years with a cardiac indication for ECMO who were supported at Arkansas Children's Hospital with the Levitronix Centrimag ECMO pump between the time period July 1, 2007, to July 1, 2009 were identified for inclusion in the study. Patients who had ECMO support initiated at another institution then were transported to our facility and those supported with ECMO for <48 hours were excluded from the study. Based on the inclusion and exclusion criteria, seven patients were identified and their detailed data collated.
Seven cases supported using the centrifugal pump (Levitronix Centrimag Pump, Zurich, Switzerland) (Figure 1A) were matched to a control group of patients who had received ECMO support with a roller-head pump (Stockert-Shiley SIII, Munich, Germany) (Figure 1B). Matching was performed using propensity scoring methods to mediate any differences in age, weight, gender, total ECMO time, and CPB in previous 72 hours between Centrimag cases and Stockert-Shiley controls. Propensity scoring was used twice in a one-to-one, without replacement manner to create a 2-1 matched Stockert-Shiley group.
PFH Measurement and Utility
Plasma-free hemoglobin was measured by derivative spectrophotometry at a wavelength of 570 nm on a Thermo Spectronic Unicam UV500 (Waltham, MA). Plasma-free hemoglobin concentrations were measured daily as part of routine clinical care.
All information was obtained from the electronic and/or paper patient care information. Data regarding hemolysis, as measured by daily PFH concentrations, days on first ECMO circuit, and total number of ECMO circuit failures were abstracted from the medical record. All laboratory information collected every morning as part of the routine laboratory draw was classified according to the current day on ECMO with day 0 being the day of cannulation for the first circuit. Circuit life was measured in hours from the hour of cannulation.
At our institution, when a decision to institute ECMO support occurs, the in-house ECMO coordinator prepares a circuit with either a roller-head or centrifugal pump, based on availability. Each circuit includes a Josta QuadroxD oxygenator (Maquet, Hirrlinger, Germany), tubing diameter based on patient size with a length that is minimized and similar regardless of the pump selected. The cardiac surgical team, which is available at all times, is summoned for ECMO cannulation. Initial ECMO parameters are similar, but some variation may occur based on the clinical context: activated clotting times are maintained 180 to 200 with a heparin infusion, ECMO flow is maintained 100 to 150 ml/kg/min based on perfusion, hemodynamic variables, lactic acidosis, and oxygen saturation. The difference in the manner in which flow is generated with a roller-head versus a centrifugal pump dictates that the venous return pressures could not be comparable. The negative force generation by the impeller in the Levitronix pump requires a negative venous return pressure, which was maintained >negative 50 ml of water. To minimize hemolysis and ensure adequate venous return, Stockert-Shiley patients' venous return was maintained at 30–50 cm of water
Circuit failure was defined as mechanical dysfunction or hematologic abnormalities, which necessitated ECMO circuit change. Most commonly, this was recognized as a triad of disseminated clots through the ECMO circuit associated with increased blood product transfusion requirements and decreasing platelet, fibrinogen, and antithrombin III concentrations. The presence of unstable clots on the arterial limb of the circuit would lead to circuit change alone. The full triad was not necessarily present in all cases and relied on clinical conditions and practitioner preference to a large extent.
Summary statistics comparing variables used for matching was analyzed by ECMO group. Group differences with respect to continuous variables were analyzed using Wilcoxon rank-sum test, whereas categorical data were analyzed using Fisher's exact test; both tests maintain appropriate statistical properties when assessing small sample sizes.
Plasma-free hemoglobin differences between the two groups were analyzed using both the Wilcoxon rank-sum test and beta regression at each time point. Beta regression differs from traditional linear regression as it models a dependent variable that follows a beta distribution, not a normal distribution,18 allowing for skewed distributions such as PFH to be modeled. A caveat to beta regression usage is that the dependent variable must fall within the (0,1) open interval. This property leads to beta regression being naturally adept for modeling percentages and allows associations between explanatory variables and outcomes to be easily interpreted as odds ratios (ORs).18 Subsequently, PFH was scaled into a percentage by dividing by 100, the PFH concentration which was the upper limit of the assay. Beta regression was also used to conduct a repeated measures analysis, estimating PFH differences between groups for all days on ECMO while simultaneously accounting for multiple observations per patient.
The difference in the proportion of patients that survived to discharge between groups was estimated using Fisher's exact test. A survival analysis was not chosen for this outcome considering the specific scope of the data abstraction with respect to ECMO utilization. Survival analysis was conducted to estimate any differences that may exist between groups in time to first circuit change. Differences in the proportion of patients needing any circuit change and total number of circuit changes were estimated using Fisher's exact test. Statistical analysis was completed using Stata 11.1 (College Station, TX).
The study includes 21 ECMO deployments in 13 individual patients who were supported with roller-head ECMO pumps (Roller group) and seven patients who were supported with centrifugal ECMO (Centrifugal group). The 14 Roller group ECMO deployments were matched from a total population of 43 patients who received roller-head ECMO and met inclusion criteria during the study period. Propensity scoring methods were able to select a comparable patient group with respect to the matching variables (Table 1). Matching variable and outcome data are presented in detail by individual in Table 2.
Plasma-free hemoglobin concentrations for ECMO days 1–6 by two groups are presented in Figure 2. Patients in the Roller group were twice as likely to have significantly higher plasma PFH compared with those in the Centrifugal group when accounting for all observations (OR = 1.96; 95% confidence interval [CI]: [1.58–2.43]; p < 0.001). Comparing plasma PFH at each day, less hemolysis was observed on average in the Centrifugal group at each day (Table 3); the difference between groups on day 2 was determined to be significant by both statistical methods (Wilcoxon rank-sum p = 0.019), with the Roller group estimated to have five times the odds of having higher PFH than the Centrifugal group (OR = 4.82; 95% CI: [2.52–9.21]; p < 0.001).
Further, more circuit failures were observed in Roller group than the Centrifugal group. In the Centrifugal group, one patient (14%) required one circuit change, whereas in the Roller group, seven patients (50%) required a total of 14 circuit changes (Table 4). However, this proportion of patients needing at least one circuit change was not statistically significant between groups (Fisher's exact p = 0.174). The proportion of total number of circuit changes was not statistically different (Fisher's exact p = 0.543) nor was the time to first circuit failure (log rank p = 0.230) (Figure 3).
The differences in survival to discharge and successful weaning of ECMO were not statistically significant in between the two groups. Of the centrifugal group, three (43%) patients survived to hospital discharge, whereas eight (57%) patients in the Roller group survived to hospital discharge (Fisher's exact p = 0.659).
We report the first clinical comparison of hemolysis between roller-head versus centrifugal pumps in pediatric cardiac ECMO patients. Significantly less hemolysis as measured by PFH was evident in the centrifugal pumps by day 2 of ECMO support. Although there was no difference found by day 5, the number of patients requiring that duration of support was relatively low in each group, limiting the statistical power of the study. Our results contrast from previous reports with earlier model centrifugal pumps. The Levitronix Centrimag pump contains an impeller that is the only moving part of the pump. The impeller is magnetically levitated, thus the pump does not contain bearings in contact with blood. Such bearings in earlier centrifugal pumps may have had a role in heat generation and blood path turbulence leading to increased hemolysis, as seen in the series of neonatal ECMO patients described by Steinhorn et al.8 The Stockert-Shiley roller-head pump operates using fewer revolutions per minute than a centrifugal pump, but it requires an occlusive mechanism for blood propulsion. We speculate that, regardless of flow required, the magnetically levitated impeller in the centrifugal pump tested caused decreased hemolysis compared with the occlusion of a roller head.
Although it is possible that differences in individual patient pathophysiology and anticoagulation could have contributed to differences in hemolysis, we attempted to make the populations as uniform as possible. Only patients with cardiac diagnoses were considered. Furthermore, there were a similar proportion of patients who had undergone recent CPB in both groups (Table 1). In all patients, heparin was the sole anticoagulant, and activated clotting times were maintained in a similar range (kaolin-based ACT 160–200), depending on clinical bleeding. Recently, we have begun to measure antithrombin III (AT III) concentrations in ECMO patients and to treat concentrations below 70% of normal AT III concentrations. Six patients in the Stockert-Shiley group were supported before this change in our protocol, but their clinical course did not differ from the later Centrimag patients. It seems unlikely that AT III monitoring and replacement would lead to less hemolysis, but it could theoretically be associated with improved “circuit health.” We did not control for differences in blood product usage as the replacement of plasma and cryoprecipitate could influence the regional hemostasis and fibrinolytic pathways. In the cardiac population, more than the general medical population requiring ECMO, blood loss through chest tubes and consumption of procoagulant and anticoagulant factors and proteins could influence hemolysis. The small proportion of patients in both groups who had undergone recent CPB and/or surgery should make this a minimal factor contributing to the differences in hemolysis. In the future, these variables should be collected and analyzed for their impact on hemolysis.
This study was not powered to test for significance in patient survival or freedom from circuit change. There was a greater proportion of circuit changes in the roller-head group, but this difference was not statistically significant. The study was limited in scope and did not explore the mechanistic role of the coagulation-inflammation nexus in this study.
The difficulty of answering such apparently simple questions lies in the complexity of these critically ill patients, who often have some degree of multisystem organ failure. Furthermore, when an endpoint such as circuit failure or time to ECMO circuit change is established, such an endpoint is perhaps more indicative of physician behavior than any underlying biologic process. The scope of the population was limited, and the ability to extrapolate to more general noncardiac populations in children is unknown and will require further investigations to explore validity of these results in noncardiac patients.
We plan future studies of consecutive cardiac patients with a stable anticoagulation strategy to describe any differences in blood product requirements, renal replacement therapy, hemolysis, and circuit life in roller-head versus centrifugal pumps. The link between increased hemolysis and decreased circuit life is not direct, but with more patients and with information collected in a uniform manner, a plausible link could be established.
In this study of children with cardiac failure requiring ECMO support, hemolysis was found to be significantly less in those supported with a centrifugal pump than in those supported by a roller pump. There is a suggestion that ECMO circuit lifespan is longer during centrifugal pump support, although this study was underpowered to investigate that endpoint. Because of the important clinical consequences of hemolysis reduction and extended circuit life, it is hoped that this study will suggest the equipoise necessary to design and conduct a randomized trial to definitively answer whether centrifugal pumps are superior to roller pumps for ECMO support.
1.Hill JD, O'Brien TG, Murray JJ, et al
: Prolonged extracorporeal oxygenation for acute post-traumatic respiratory failure (shock-lung syndrome). Use of the Bramson membrane lung. N Engl J Med
286: 629–634, 1972.
2.Zapol WM, Snider MT, Hill JD, et al
: Extracorporeal membrane oxygenation in severe acute respiratory failure. A randomized prospective study. JAMA
242: 2193–2196, 1979.
3.Bartlett RH, Gazzaniga AB, Jefferies MR, et al
: Extracorporeal membrane oxygenation (ECMO) cardiopulmonary support in infancy. Trans Am Soc Artif Intern Organs
22: 80–93, 1976.
4.Bartlett RH, Gazzaniga AB, Huxtable RH, et al
: Extracorporeal membrane oxygenation (ECMO) in newborn respiratory failure: Technical consideration. Trans Am Soc Artif Intern Organs
25: 473–475, 1979.
5.Watson JW, Brown DM, Lally KP, et al
: Complications of extracorporeal membrane oxygenation in neonates. South Med J
83: 1262–1265, 1990.
6.Frenckner B, Ehren H, Palmer K: Patient complications during extracorporeal membrane oxygenation (ECMO). Eur J Pediatr Surg
1: 339–342, 1991.
7.Zwischenberger JB, Nguyen TT, Upp JR, et al
: Complications of neonatal extracorporeal membrane oxygenation. Collective experience from the extracorporeal life support organization. J Thorac Cardiovasc Surg
107: 838–849, 1994.
8.Steinhorn RH, Isham-Schopf B, Smith C, Green TP: Hemolysis during long-term extracorporeal membrane oxygenation. J Pediatr
115: 625–630, 1989.
9.Groom RC: Pediatric cardiopulmonary bypass devices: Trends in device use for cardiopulmonary bypass and postcardiotomy support. ASAIO J
51: 525–529, 2005.
10.Martin R, McKenty S, Thisdale Y, et al
: Hemolysis during cardiopulmonary bypass. J Cardiothorac Anesth
3: 737–740, 1989.
11.Gbadegesin R, Zhao S, Charpie J, et al
: Significance of hemolysis on extracorporeal life support after cardiac surgery in children. Pediatr Nephrol
24: 589–595, 2009.
12.McDonald JV, Green TP, Steinhorn RH: The role of the centrifugal pump in hemolysis during neonatal extracorporeal support. ASAIO J
43: 35–38, 1997.
13.Lawson DS, Lawson AF, Walczak R, et al
: North American neonatal extracorporeal membrane oxygenation (ECMO) devices and team roles: 2008 survey results of extracorporeal life support organization (ELSO) centers. J Extra Corpor Technol
40: 166–174, 2008.
14.Kress DC, Cohen DJ, Swanson DK, et al
: Pump-induced hemolysis in a rabbit model of neonatal ECMO. ASAIO Trans
33: 446–452, 1987.
15.Valeri CR, MacGregor H, Ragno G, et al
: Effects of centrifugal and roller pumps on survival of autologous red cells in cardiopulmonary bypass surgery. Perfusion
21: 291–296, 2006.
16.Lawson DS, Ing R, Cheifetz IM, et al
: Hemolytic characteristics of three commercially available centrifugal blood pumps. Pediatr Crit Care Med
6: 573–577, 2005.
17.Moon YS, Ohtsubo S, Gomez MR, et al
: Comparison of centrifugal and roller pump hemolysis rates at low flow. Artif Organs
20: 579–581, 1996.
Copyright © 2011 by the American Society for Artificial Internal Organs
18.Smithson M, Verkuilen J: A better lemon squeezer? Maximum-likelihood regression with beta-distributed dependent variables. Psych Methods
11: 54–71, 2006.