Secondary Logo

Journal Logo

Pediatric Circulatory Support

Sources of Circuit Thrombosis in Pediatric Extracorporeal Membrane Oxygenation

Hastings, Susan M.*; Ku, David N.*; Wagoner, Scott; Maher, Kevin O.; Deshpande, Shriprasad

Author Information
doi: 10.1097/MAT.0000000000000444
  • Free


The use of extracorporeal membrane oxygenation (ECMO) as a form of extracorporeal life support (ECLS) is now a well-established therapy for cardiorespiratory failure.

Currently, the use of ECLS is dominated by neonatal and pediatric patients over adult patients, and the use of ECLS is increasing for both adults and pediatric patients. The ELSO registry currently documents the use of ECMO in more than 48,000 pediatric patients since 1990. Neonatal patients (<30 days old) account for 60.2% of all ECLS cases, infants (from 31 days old to 1 year old) and pediatric patients (from 1 year old to 16 years old) account for 25.6% of total cases, and adult patients account for the remaining 14.2%.1 However, in the face of ever increasing experience, there continues to be significant morbidity and mortality related to clotting and bleeding-related complications, especially in the neonatal and pediatric population.1 In addition to the events captured on the patient side, issues related to clotting contribute significantly to equipment malfunctioning necessitating interventions such as circuit/oxygenator changes. In the neonatal and pediatric population, management and appropriate balancing of the anticoagulation to counteract clotting is a tremendous challenge. A reduction in the inherent generation of clots in an ECMO circuit could reduce the need for anticoagulation with its concomitant bleeding complications.

There is very limited data on the clot composition, contributing factors, or mechanism of clot generation within the ECMO circuit. The goals of this study were to characterize clot formation and location within the circuit, to understand the basic histologic composition of the clot, and to ascertain the relation between clot location and the local hemodynamic conditions in the extracorporeal circuit. We also compared the thrombogenecity of circuits based on the type of pump used, viz. centrifugal or roller head pump.

Materials and Methods

This study was approved by the Emory University IRB; informed consent was waived as no interaction or intervention with the patient occurred.

Extracorporeal Membrane Oxygenation Circuits

Extracorporeal membrane oxygenation circuits were prospectively collected from Children’s Healthcare of Atlanta (CHOA), Emory University (Atlanta, GA) over a period of 2.5 years between 2012 and 2014. Every patient supported with ECMO during this period was considered eligible for inclusion in the study. Based on convenience sampling, an effort was made to collect as many circuits as possible, given practical issues. No circuits were excluded based on patient or ECMO characteristics, and the study population was representative of the overall ECMO supported group in terms of age, diagnosis, type of ECMO, and duration of ECMO. During the ECMO support, all of these patients were managed by using our institutional protocol for anticoagulation and blood product administration. The anticoagulation is achieved using unfractionated heparin, with target anti-Xa levels between 0.3 and 0.7 as well as bedside activated clotting time (ACT) measurements using an i-STAT device (Abbott Laboratories, Abbott Park, IL) with a Kaolin ACT cartridge. The ACT targets were adjusted based on patient anti-Xa levels as well as clinical scenario of bleeding or clotting problems. Blood component therapies were administered as needed to maintain hematocrit between 35% and 45%, platelet count of greater than 100,000/cm2, and fibrinogen count of greater than 200 mg/dl.

After removal of a patient from ECMO support, the circuits were immediately drained of blood and gently irrigated and filled with normal saline. The circuits were refrigerated (4°C) at Georgia Tech 24–48 hours until inspection.

Each circuit was inspected for gross clots easily visible through the saline. The location of adhered clots was recorded and regions of interest were photographed. To determine whether clots were adherent, saline was perfused lightly in the direction of flow. A clot was deemed adherent if it remained completely or partially attached during perfusion. Tubing sections of interest were cut from the circuit, then labeled, and photographed. Clots were then excised from the sections and were immediately fixed in 10% formalin (VWR International, Radnor, PA) until histologic analysis. Oxygenator clots were removed by either flushing clots out of the oxygenator with saline or with forceps if a sample was in reach of the entry. Patient parameters of interest while on ECMO were also collected.

A single connector was counted as having two tubing-connector junctions (TCJs), one upstream and one downstream, and the TCJ was cataloged according to its tubing size. A typical expansion connector, for example, would have one 1/4″ TCJ upstream and one 3/8″ TCJ downstream.

Histologic Analysis

The dissected clots were embedded in paraffin and 5 μm thick slices were cut using a microtome (Thermo Fisher Scientific, Waltham, MA). The slices were then mounted on glass slides and dried for 24 hours at 37°C.

Before staining, the slides were first deparaffinized and rehydrated. Carstairs’ stain for fibrin and platelets was used for staining.2 After staining, the slides were dehydrated via ethanol, cleared via xylene, and mounted. The Carstairs’ method stains platelets grey blue to navy, fibrin red, muscle bright red, collagen bright blue, and red blood cells (RBCs) yellow to clear. Images of the slides of the stained clots were randomly selected and analyzed using a pixel count by color in Adobe Photoshop to quantify the clot composition.

Computational Analysis

To quantify the regions of extreme shear rate, flow separation, and other flow profiles of interest, computational fluid mechanics was used to analyze the flow through a segment of tubing with a connector.

The tubing connector junction geometry was represented as a 2D, axisymmetric cylinder in COMSOL Multiphysics. Representative ECMO flow rates (300–5,000 ml/min, Re = 500–1,014) were used to generate streamlines, velocity profiles, and shear rate profiles. The mesh is smaller at the boundary layer and expands in size in the lumen. The mesh contains 49,923 elements and has an area of 33,940 mm2. The dynamic whole blood viscosity was assumed to be 0.0004 Pa·s. Convergence was calculated with relative tolerance of 0.001.

Statistical Analysis

An unpaired Student’s t-test was used for significant differences (p < 0.05).


Patients and Parameters

A total of 50 ECMO circuits were collected after separation from patients at decannulation. These circuits were then processed and analyzed as described earlier.

Basic patient characteristics and relevant ECMO parameters are described in Table 1. The study cohort is representative of the ECMO population at this hospital, with 64% of the patients having a cardiac etiology and 36% with a noncardiac etiology for the support. The majority of the cardiac patients (70%) were postcardiotomy. There was a wide range of age represented from newborn to 16 years of age, with corresponding weight range of 2.2–80 kg. Mean duration of ECMO was 166.5 hours (range of 12–878 hours [36.6 days]). Management of ECMO during the course was based on the basic strategy described in the Methods section. Goal-directed anticoagulation was maintained with target anti-Xa levels as well as point of care ACT testing. This resulted in maintenance of target ACTs in all of the patients with a range of 130–190 seconds. ACT range was lower for immediate postoperative patients with significant bleeding (130–160 seconds) whereas a range of 160–190 seconds was expected for the rest of the patients.

Table 1.
Table 1.:
Basic Patient Characteristics and ECMO Parameters for the Study Group (N = 50)

Circuit Analysis

The recovered circuits were composed of S-97-E Tygon tubing (Saint-Gobain Corporation, Courbevoie, France) for the roller pump raceway (in the case of circuits with roller head pumps) and Class 6 bypass tubing (Medtronic, Minneapolis, MN) for the rest of the circuit. Oxygenators used were either an adult or pediatric Quadrox D Oxygenator (Maquet, Rastatt, Germany). The circuit also included an arterial filter (Medtronic) as well as a bladder system—the majority of the circuits (39/50) with the Better-Bladder (Circulatory Technologies Inc., NY). The circuits used either a SIII roller pump (Stockert-Shiley SIII, LivaNova, Munich, Germany) (68% of circuits) or a centrifugal pump (Sorin Revolution, LivaNova, Munich, Germany) (32% of the circuits). A system-wide change over for the institute from a roller to centrifugal technology occurred 2/3 of the way through the collection time period. Extracorporeal membrane oxygenation circuits at CHOA were primed with packed RBCs immediately before use.

Visual documentation of gross clots was performed for all circuit components. Overall, 94% of circuits exhibited thrombus formation. Clots were not evenly distributed within the various components of the ECMO circuit. Although the tubing accounts for more than 90% of the surface area exposed to blood, no clots were present on the free tubing surface. Instead, thrombi were focused at two locations: the TCJs and the oxygenator. The clots found at the TCJs were adherent and typically axisymmetric (Figure 1). The clots that were found in the oxygenator were not adherent to the membrane and were found loose on the deoxygenated or pre membrane side.

Figure 1.
Figure 1.:
Example of a typical adherent clot at both the downstream TCJs from a ¼″ diameter connecter section in an ECMO circuit. ECMO, extracorporeal membrane oxygenation; TCJ, tubing-connector junctions.

The ECMO circuits were composed of approximately 5 m of tubing, with 6–12 connectors that were used to control flow and insert devices (oxygenator, filters, vascular access, etc.). Depending on the entry size of the circuit components, circuits were either a single diameter throughout or sized up and down between two diameters. Overall, the connectors accounted for about 10% of the exposed surface area, yet exhibited 99% of the clots. The majority of ECMO circuit tubing diameter was 1/4″ (0.635 cm), and a typical circuit sized up and down between 1/4″ and 3/8″ (0.953 cm) tubing. A few larger roller pump circuits sized up and down between 3/8″ and 1/2″ (1.27 cm) tubing, which accounted for only 8% of the circuits in this study.

The TCJ is formed by the thickness of the connector wall and the region of tubing that is expanded to fit over the outer diameter of the connector until it returns to the relaxed diameter (Figure 2). Clots were found in the step right at the junction in the lumen expansion zone on both the inlet and outlet ends of the connectors.

Figure 2.
Figure 2.:
Computational fluid dynamics analysis of the TCJ. The tubing and connector are ¼″ in diameter. A: The inlet of the connector with streamlines shown (Q = 1,000 ml/min). The black, thin arrow points to a region of low shear rate. The low shear rate region correlates with blood clot presence in the circuit. B: The outlet of the connector with streamlines shown (Q = 1,000 ml/min). The arrow points to a recirculation region, which corresponds to most of the blood clots in the circuit. C: The shear rate map of the inlet of the connector. The arrow points to the low-shear region. In the corner, shear rate drops to < 50/s. Axes are in meters. D: The shear rate map of the outlet of the connector. In the corner, the shear rate drops to < 50/s. The arrow points to the low-shear region. E: A 4× view of a clot at the TCJ. The blood clot is adherent to the connector edge. TCJ, tubing-connector junction.

The incidence of clotting was high at certain connectors and regions of the ECMO circuit, and is correlated with areas of low shear (p << 0.05) (Figure 2). The downstream end of expansion connectors was the most thrombogenic region in the ECMO circuits, with a 74% incidence rate, while its upstream counterpart had a thrombi incidence rate of only 13%. The 3/8″ diameter TCJs in general had a higher incidence of thrombosis of 45% overall versus the 1/4″ TCJs, which were at 22%. In general, a downstream TCJ was significantly more likely to be thrombogenic than its upstream counterpart (33% vs. 25.2%, p = 0.00297).

Centrifugal pump TCJs were compared with roller pump TCJs, and it was found that the centrifugal pumps had a higher overall incidence of thrombosis (41% vs. 25%, p << 0.05). In addition, every centrifugal pump (100%) had a small clot at an area of exposed metal in the pump shaft, and in some cases, these pumps exhibited large clots (Figure 3).

Figure 3.
Figure 3.:
Centrifugal pump thrombosis. A: Clot was seen at the top of the shaft in all cases. B: Extensive clot was found in 25% of the centrifugal pump heads.

In some cases, clots from TCJs grew downstream and formed large masses greater than 2 cm2 (Figure 4). The clots were only adherent to the TCJ attachment point, and under perfusion, the portion downstream of the TCJ was mobile. It is likely that these large clots would eventually break off and migrate to the oxygenator.

Figure 4.
Figure 4.:
Examples of clot appearance at TCJs. A: A tapered, long fibrin strand growing from an axisymmetric clot. B: A U-shaped growth attached at two points to the TCJ. C: A large, nearly occlusive mass with tethered strands attached to several points at TCJs. D: A strand connected between two close TCJs. This strand occurred downstream of the centrifugal pump in tubing that was large for the circuit (3/8″, 0.953 cm). TCJ, tubing-connector junctions.

Clot Histology

Histologic analysis revealed that the TCJ clots to be fibrin-rich and full of RBCs (Figure 5). The oxygenator clots were coiled, and when expanded reached a length of more than 5 cm (Figure 6). Oxygenator clots were present on the upstream pre-membrane side, and these clots were nonadherent. Under light perfusion of water, the clots dislodged and became mobile. With changes in oxygenator orientation (tipping), clots would slide in the direction of gravity. The composition of oxygenator clots is similar to the axisymmetric TCJ clots (Figure 5). The clots were on average 54% fibrin, 45% RBCs, and ~1% platelets. These red clots had a paucity of platelets and are unlikely to be triggered by platelet aggregation.

Figure 5.
Figure 5.:
Histologic composition of ECMO clots. These samples were stained using Carstairs’ stain, which colors red blood cells yellow/clear, fibrin red, and platelets blue/navy. These clots are predominately fibrin and red blood cells. Little to no platelets are present. ECMO, extracorporeal membrane oxygenation.
Figure 6.
Figure 6.:
Oxygenator Clots. These samples were stained using Carstairs’ stain, which colors red blood cells yellow/clear, fibrin red, and platelets blue/navy. At left, loose clots in the entry side of the oxygenator. At right, histology of oxygenator clot reveals fibrin ribbons and red blood cells (4×). Little to no platelets are present.

Computational Fluid Dynamics Analysis

Based on the frequency distribution for the localization of adherent clots, we identified the TCJ as the highest priority site. We then characterized the hemodynamics of the TCJ zones. CFD analysis of the TCJ region revealed distinct regions of low shear and a recirculation region on the outflow side of the TCJ. A figure of the streamlines and shear rates is shown in Figure 2. For the inlet to the connector, a zone of very low shear rates less than 50/s is present in the corner. At the outlet, the zone of very low shear rates is even larger and directly located in the corner of the junction. Note that the shear rates outside of the corners return to a normal shear rate range of 450–1,000/s. Thus, the clots colocate directly at the site where blood is virtually stagnant with shear rates less than 50/s, or a shear stress of less than 0.2 Pa (p << 0.05).


This is one of the first studies to examine in detail the location and histologic composition of thrombosis within the ECMO circuit as well as statistical relationship of thrombosis with hemodynamics in the clinical circuits.

Analyses of clinical ECMO circuits revealed that thrombosis occurs in nearly all of ECMO circuits at specific sites, which has not been previously reported.

The ECLS Registry collects information on clot formation at certain ECMO circuit components; however, clots are typically recorded when large enough to be seen from outside the circuit while blood is still flowing, or when large enough to be detrimental. There is some variation in frequencies of these clot-related complications. For example, in neonatal cardiac ECMO patients, 11.6% reported oxygenator clots, 3.9% reported bridge clots, 5.9% reported bladder clots, 4.3% reported hemofilter clots, and 13.6% reported clots at other locations.1 For pediatric cardiac ECMO patients, the frequencies were as follows: oxygenator 8.4%, bridge 3.0%, bladder 4.0%, hemofilter 3.5%, and other locations 10.2% reported clots.1 However, the ECMO circuits vary from facility to facility and the methods for recording clots are inconsistent. The development of a consistent method for identifying and reporting clots at specific sites in ECMO circuits would be of benefit to the patient population. In general, our data suggest that there is significant under-recognition and therefore under-reporting of circuit thrombosis.

The blood clots are found at specific locations within the circuit, primarily at the TCJs. It is noteworthy that although the tubing of the circuit constitutes a large portion of the surface area, there were no clots detected on the tubing surface. This is remarkable given that our standard circuit tubing (Medtronic Class VI) is not coated. This TCJ clot location corresponds directly to zones of very low shear rates of less than 50/s. Conversely, almost no clotting occurs with tubing material where shear rates are greater than 450/s (p << 0.05) based on the CFD analysis. These conditions correspond to a Virchow’s Triad requirements for foreign surface, blood coagulation proteins, and zones of virtually stagnant blood. The fibrin clots are also consistent with previous studies that have looked at the behavior of blood components in relation to differences in shear rates. Using rabbit aorta with endothelial disruption, Weis et al.3 showed that the degree of fibrin deposition directly correlated with the shear rates; high fibrin content at low shear rates of 50/s versus low fibrin content at high shear rates of more than 650/s. Similarly, Guy et al.4 have showed that low shear rates correspond directly to a higher degree of fibrin gel deposition on an injured blood vessel surface, using mathematical modeling. The height of the fibrin gel achieved was highest at shear rates less than 100/s yet was minimal at shear rates above 1,000/s. In the current study, the areas of TCJs that had shear rates of 50/s corresponded with high fibrin deposition and clot formation. The contribution of the local hemodynamics to clot formation is critical.

Growing clots with small attachment points at TCJs are particularly worrisome, as they imply the possibility of large emboli, which could cause devastating patient complications. These large clots would be subject to large drag forces to generate the large, loose thrombi seen in the oxygenator.

The histologic appearance of clots in our study demonstrates that clots forming in ECMO circuits are fibrin-rich with few platelets. Thus, fibrin coagulation is more likely the problem for ECMO circuits than platelet thrombosis. The observation that ECMO clots are fibrin-rich is also consistent with reports of fibrinogen consumption in these patients.5,6 These red clots form in the setting of adequate anticoagulation with heparin and with therapeutic ACTs, illustrating the criticality of shear rate as a new factor. Currently, there are no universally established guidelines for ECMO anticoagulation, although unfractionated heparin is used at most ECMO centers, and very few centers (6 in a recent 119 center survey) use antiplatelet agents (acetylsalicylic acid and prostacyclin).7 Unfractionated heparin is effective at binding with antithrombin III (AT III) and causing a conformational change that leads to acceleration of AT III-mediated inactivation of various coagulation factors including thrombin, factors IX, X, and XI.8 However, one of the major limitations of heparin in the setting of biomaterials is its inability to inactivate thrombin bound to fibrin or to biomaterial surfaces.9 Hence, we can speculate that once fibrin deposition is initiated within the circuit, further propagation at low shear rate zones may not be sufficiently prevented by heparin use.

Our data also indicate that circuits with centrifugal pumps have greater incidence of thrombus than the roller pumps. The circuit thrombi may be due to the size of tubing to fit the centrifugal pump, which typically sizes up and down between diameters to go through the pump and the oxygenator. Our data show these connectors of diameter increase and subsequent larger diameter regions to be highly thrombotic. Redesign of the circuit to avoid such “step-ups” may reduce the thrombogenecity of the circuit by reducing the amount of regions with extreme low shear rates. Indeed, elimination of just four TCJs could reduce circuit thrombus by 80%. In addition, the centrifugal pumps themselves demonstrate adherent thrombus, especially at the pin of the cone. The centrifugal pump also potentially poses other complications such as heat generation caused by the high resistance of pediatric and neonatal cannulas. Variation in designs of various available centrifugal pumps and its implication to thrombus generation will be the focus of a future study.

This study is methodologically limited in some ways. The inspection of ECMO circuits was done macroscopically, and microscopic examination of the entire tubing was not performed. This may have excluded smaller, subclinical clots. Circuit components such as the oxygenator and arterial filter were not dismantled for investigation, and thus assessment of these components was limited to outer visual inspection. Although the morphologic and histologic similarity of clots found in the oxygenator entry and the TCJs suggest that these mechanisms are linked, it is also possible that low shear in the oxygenator could also potentiate thrombosis. Our results are also in agreement with other oxygenator findings, i.e., that the thrombi accumulate in the venous side of the oxygenator.10 The histologic analysis performed was used only to ascertain whether the clot was predominately composed of fibrin or platelets, neglecting further details of clot morphology. The circuits were gently irrigated with saline before refrigeration. So it is possible that some lightly adherent clots may have washed out. However, the overall clot burden noted in our study is far more extensive than the “visual inspection” methodology used for registry reporting. We anticipate the overall analysis including histology is not affected by the procedure, based on staining and microscopic examination of the various sections. We did not evaluate the circuits at successive time intervals and cannot comment on the growth rate of clots. This is also a single-center study, which may limit applicability.

A better understanding of clot formation in ECMO may allow for targeted preventive treatments and thus better patient outcomes. Current anticoagulation regimens alone are not sufficient to eliminate thrombosis, and modification of the circuit itself may be necessary to minimize thromboembolic events. Identification of the areas of the circuit that are thrombogenic allows for improved circuit design. Our results demonstrate that the local hemodynamics, which create small zones of low shear rate, are strongly related to thrombus formation in extracorporeal circuits. We recommend that circuits should be designed to reduce the zones of low shear rate (<100/s) such as occurs at expansions and connectors. However, it is also still important to circuit design to exclude pathologically high shear rates that may induce hemolysis (>40,000/s) or platelet thrombosis (>10,000/s).11,12 Our data also suggest that on the basis on thrombogenecity, roller pumps appear to produce less thrombosis than centrifugal pumps. We conclude that blood clotting in ECMO circuits may be reduced through an understanding of the induction of coagulation combined with fluid mechanic design to eliminate zones of stagnation.


We express our gratitude to the ECMO clinical staff at Children’s Healthcare of Atlanta.


1. Paden ML, Conrad SA, Rycus PT, Thiagarajan RR, ELSO Registry: Extracorporeal Life Support Organization Registry Report 2012. ASAIO J 2013.59: 202210.
2. Carstairs KC: The identification of platelets and platelet antigens in histological sections. J Pathol Bacteriol 1965.90: 225231.
3. Weiss HJ, Turitto VT, Baumgartner HR: Role of shear rate and platelets in promoting fibrin formation on rabbit subendothelium. Studies utilizing patients with quantitative and qualitative platelet defects. J Clin Invest 1986.78: 10721082.
4. Guy RD, Fogelson AL, Keener JP: Fibrin gel formation in a shear flow. Math Med Biol 2007.24: 111130.
5. Urlesberger B, Zobel G, Zenz W, et al.: Activation of the clotting system during extracorporeal membrane oxygenation in term newborn infants. J Pediatr 1996.129: 264268.
6. Arnold P, Jackson S, Wallis J, Smith J, Bolton D, Haynes S: Coagulation factor activity during neonatal extra-corporeal membrane oxygenation. Intensive Care Med 2001.27: 13951400.
7. Lehle K, Philipp A, Gleich O, et al.: Efficiency in extracorporeal membrane oxygenation-cellular deposits on polymethylpentene membranes increase resistance to blood flow and reduce gas exchange capacity. ASAIO J 2008.54: 612617.
8. Gorbet MB, Sefton MV: Biomaterial-associated thrombosis: Roles of coagulation factors, complement, platelets and leukocytes. Biomaterials 2004.25: 56815703.
9. Hirsh J, Anand SS, Halperin JL, Fuster V: Mechanism of action and pharmacology of unfractionated heparin. Arterioscler Thromb Vasc Biol 2001.21: 10941096.
10. Dornia C, Philipp A, Bauer S, et al.: Analysis of thrombotic deposits in extracorporeal membrane oxygenators by multidetector computed tomography. ASAIO J 2014.60: 652656.
11. Leverett LB, Hellums JD, Alfrey CP, Lynch EC: Red blood cell damage by shear stress. Biophys J 1972.12: 257273.
12. Para A, Bark D, Lin A, Ku D: Rapid platelet accumulation leading to thrombotic occlusion. Ann Biomed Eng 2011.39: 19611971.

extracorporeal membrane oxygenation; thrombosis; blood; anticoagulation; blood pump; pediatric

Copyright © 2016 by the ASAIO