Effect of anatomical variation on extracorporeal membrane oxygenation circulatory support: A computational study

Abstract

Background:

Extracorporeal membrane oxygenation (ECMO) via femoral cannulation is a vital intervention capable of rapidly restoring perfusion for patients in shock. Despite increasing use to provide circulatory support, its hemodynamic effects are poorly understood and the impact of patient-specific anatomical variation on perfusion is unknown. This study investigates the complex failing heart-mechanical circulatory support circulation and analyzes the effect of patient-specific vascular anatomical variations on hemodynamics and end-organ perfusion.

Methods:

Patient-specific vascular geometries were constructed from segmenting clinical computerized tomography angiography images and quantitatively compared using tortuosity, curvature, torsion, and lumen diameter. Computational fluid dynamic simulations were performed on a subset of geometries selected to represent a range of anatomical variation. Heart failure severity was modeled by varying the relative fraction of total flow provided by the heart and the extracorporeal circuit. A 3-element lumped parameter model was applied to accurately and dynamically model distal perfusion boundary conditions. Hemodynamic parameters and end-organ perfusion were analyzed and compared to assess the effect of anatomical variation.

Results:

Pulsatile antegrade cardiac perfusion and ECMO retrograde perfusion collide in the aorta to form a dynamic watershed region. The size, position, and variation of this region over the cardiac cycle is substantially altered by patient anatomical region. Increased vascular tortuosity reduces the proximal extent of flow from circulatory support and decreases the size of the watershed region.

Conclusions:

Patient vascular anatomy is a key determinant of the ECMO-failing heart circulation that alters the location and extent of the watershed region and affects the tissues at risk for differential hypoxia and circuit-derived thromboemboli for a given level of support.

1. Introduction

Extracorporeal membrane oxygenation (ECMO) is a vital means of providing circulatory support and restoring perfusion for patients in shock [1–3]. Deployed as a mechanical circulatory support (MCS) device, ECMO shunts venous blood to the arterial system to create a complex ECMO-assisted:failing-heart circulation [4]. In its less invasive and increasingly used form in the adult, cannulae are percutaneously placed via the femoral vein (drainage) and artery (return). Continuous retrograde flow from the ECMO circuit return collides with pulsatile antegrade flow ejected by the failing heart to create a dynamic watershed region [5]. The characteristics of the watershed region and its impact on hemodynamics and end-organ perfusion are poorly understood and the effect of patient anatomical variance on the ECMO-failing heart circulation is unknown.

Despite growing case literature of successful ECMO support for those in shock, overall mortality in these patients remains near 50% –comparable to mortality observed with medical therapy alone [6–9]. Bleeding, infection, and clotting are well-described sequelae of ECMO support that contribute to overall mortality [10,11]. Less well understood is how complications such as renal failure, intestinal ischemia, and cerebral thromboembolic events occur despite support well-titrated to hemodynamic parameters. The complex watershed region and resultant hybrid flow domains may be drivers of complications and poor outcomes. Improved understanding of the ECMO-failing heart circulation and how patient-specific factors determine perfusion is required to improve ECMO clinical use and spur development of new metrics to guide support.

Insight into these issues is challenging given the complex array of driving forces – precluding rigorous evaluation in clinical settings, preclinical animal systems, or benchtop models. Computational fluid dynamics (CFD) can provide the resources to span the range of combinations and quantifiable metrics of outcomes. Prior work examined flow distribution in the ECMO-failing heart circulation using an idealized aortic geometry and identified how watershed region location and size varied with relative support [12]. The current study seeks to expand on these insights through use of patient-derived aortic geometries to investigate the effects of vascular anatomical variation on (1) watershed region dynamics and (2) end-organ perfusion as a function of the severity of heart failure during steady-state support. This study hypothesizes that location and extent of the watershed region is determined by patient vascular anatomy for a given level of relative ECMO support. To evaluate this hypothesis, CFD simulations were performed using patient-specific geometries extracted from medical images and quantitatively compared to an idealized geometry at varying levels of relative ECMO support. Perfusion of the main aortic branches at each support level were compared to a normal cardiac function scenario to determine the effects of extracorporeal support. Further understanding of the ECMO-failing heart circulation may lead to optimized clinical use of this important support system.

2. Methods

2.1. Patient imaging data

Patient-specific aorta geometries were derived from a cohort of 253 adult patients with severe aortic stenosis who underwent computed tomography angiography (CTA) imaging as part of standard pre-procedure assessment prior to planned transcatheter aortic valve replacement at the University Hospital Case Medical Center between 2007 and 2010. Images were obtained for retrospective analysis in accordance with The University Hospitals Institutional Review Board (IRB) which waived the requirement for patient informed consent (IRB protocol: 03–15-26; approval date: June 6, 2016). De-identified images were provided to the study investigators and were determined to be exempt for review by the Massachusetts Institute of Technology Committee on the Use of Humans as Experimental Subjects.

Expert review categorized each of 253 patient anatomies into one of three geometrical archetypes: (1) Archetype S – characterized by a straight descending aorta and non-tortuous iliac arteries; (2) Archetype T – characterized by a tortuous descending aorta and highly tortuous iliac arteries; or (3) Archetype M – intermediate anatomies with variable tortuosity (Table 1). Specific individual patient cases were drawn from each of the categories to serve as representative anatomies and are referred to as Case S, Case T, and Case M (Fig. 1).

Table 1.

Curvature, torsion, and tortuosity of the geometries. Curvature and torsion are measured at many points along the centerlines and thus the medians are displayed. Tortuosity provides one number per given centerline. Curvature, torsion, and tortuosity of a straight line are all zero.

	Median Curvature [mm−1]				Median Torsion [mm−1]
	Ideal	Case S	Case M	Case T	Ideal		Case S	Case M	Case T
Proximal	0.026	0.053	0.035	0.040	0.021		0.718	0.633	0.640
Thoracic	0.007	0.038	0.030	0.036	0.019		0.832	0.704	0.748
Abdominal	0	0.071	0.051	0.057	0		0.637	0.591	0.632
Left Iliac	0	0.045	0.054	0.055	0		0.988	0.529	0.663
	Tortuosity Ideal	Case S		Case M		Case T
Whole Aorta	0.66	0.62		0.46		1.02
Thoracic	0.07	0.12		0.28		0.40
Abdominal	0	0.03		0.06		0.07
Left Iliac	0	0.04		0.58		0.61

Fig. 1.

Graphical representation of the different aortic geometries used in flow simulations. Panel (A) shows the computer-generated idealized geometry with Panels (B)–(D) showing the representative patient-specific geometries. Panels shown (B) Case S, (C) Case M, and (D) Case T.

Images were segmented to include major aortic outlets using the open source ITK-snap package [13]. The aorta lumen extended from the sinotubular junction to the bifurcation of the iliac arteries and included the aortic branches of brachiocephalic artery (BCA), left common carotid artery (LCCA), left subclavian artery (LSA), celiac artery (CA), superior mesenteric artery (SMA), left and right renal arteries (LRA and RRA), inferior mesenteric artery (IMA), and left and right common iliac arteries (LCI and RCI). A standard arterial cannula (17 French) was modeled geometrically to represent peripheral cannulation placed 15 cm distal to aortoiliac bifurcation inside the LCI artery. A reference geometry, created from average values of patient-derived measurements, was adopted from published work to serve as an anatomy for baseline measurements [14].

2.2. Vascular anatomy analysis

Quantitative metrics of anatomical variance were developed to facilitate analytical comparison of the vascular geometry archetypes and are detailed in Supplementary Material. Lumen diameter was measured in the proximal, thoracic, and abdominal sections of the aorta and in the left iliac artery at the site of the proximal tip of the arterial cannula. Curvature and torsion were calculated at the same anatomical locations to quantify vessel bending and twisting, respectively. Tortuosity was calculated for the aorta as a whole, each region of the aorta, and the left iliac artery. All metrics were calculated using the Vascular Modeling Toolkit (VMTK) [15]. Further information on metric selection criteria is presented in the supplementary materials.

2.3. Computational simulations

CFD simulations were performed using ANSYS Fluent (ANSYS, Canonsburg, PA). Blood under all conditions was modeled with density of 1060 kg/m³ and viscosity of 3.5 cP (centipoise) in a multiphase turbulence model to distinguish the interface of ECMO and heart supplied blood flow [16]. While use of anticoagulation, inflammatory mediators triggered by blood contact with the ECMO circuit, and circuit-derived clot may alter blood viscosity in patients undergoing ECMO support, their cumulative effects are poorly described and support the use of established values pending availability of validated data for this patient population.

A patient-derived physiological aortic waveform was applied at the sinotubular junction to model heart output [17] and a continuous jet of blood from arterial cannula was generated to model ECMO flow. The physiological waveform was linearly scaled to model each of the different levels of cardiac output representing varying severity of heart failure. Simulations were designed to represent a stably supported patient and used 5 L per minute (LPM) of total perfusion in all conditions. The volume of blood over a cardiac cycle at each outlet was calculated to determine total perfusion which consisted of the sum of blood flow provided by the ECMO circuit and the heart.

Differing severities of heart failure and the effects of variable titration of extracorporeal support were modeled by simulating three different relative proportions of heart-to-ECMO supplied perfusion: 1:1 (50% ECMO); 1:3 (75% ECMO); and, 1:9 (90% ECMO). Dynamic boundary conditions at the outlets were employed using a lumped parameter model to ensure the consideration of peripheral and distal resistance and compliance of the truncated vasculature for each aortic outlet [14]. Further detail of computational domain generation, computational algorithms, and boundary conditions are provided in Supplementary Materials.

3. Results

Computational simulations of idealized and patient-specific models of the aortic tree were performed to quantify vascular anatomical variation and analyze the ECMO-failing heart circulation at differing pro-portions of heart:ECMO supplied perfusion. Analyses were performed in three patient-specific cases representative of the vascular archetypes.

3.1. Analysis of vascular geometry variation

The three patient-specific cases were analyzed to quantify vascular differences. Aorta diameter varied between patient geometries in both absolute measure and relative progression of lumen diameter advancing distally (Fig. 2). Case S had the smallest lumen throughout the aorta whereas both Case M and Case T exhibited variability in aortic diameter over the course of the aorta relative to the ascending aorta. Curvature, torsion, and tortuosity calculated along the centerline confirmed differentiation of the three distinctive archetypes as initially determined by expert review (Table 1). The iliac artery is the most tortuous of the sections analyzed and is of particular importance as the site of the ECMO cannula.

Fig. 2.

Distribution of the aorta and iliac artery vessel diameter in each of the three patient-specific cases compared against the idealized geometry. Panel (A) shows the median lumen diameter and boxplot distribution for each case. Panels (B)–(D) compare the thoracic aorta, abdominal aorta, and iliac artery lumen diameter for each case referenced to the median proximal aorta diameter.

3.2. Watershed region dynamics

The watershed region, defined as the mixing zone created by the collision of continuous retrograde flow of highly oxygenated blood from the ECMO circuit and pulsatile antegrade flow generated by native heart blood, was qualitatively assessed using volume fractions and streamlines. Data were analyzed at various time points over multiple cardiac cycles to fully capture time-dependent evolution of the resulting flow. Spatial variability of the watershed region was quantified by measuring variability of ECMO volume fraction contours along the aorta centerline.

Following quantitative assessment of each geometry, the three cases were simulated with varying proportions of heart:ECMO flow to model different clinical situations (Fig. 3). In each patient case, ECMO flow pushes the watershed region proximally with the location of the advance determined by the degree of distal vessel tortuosity. Only in the straight configuration of Case S are the abdominal vessels predominantly perfused by the ECMO circuit at 50% ECMO (Fig. 3A). The more torturous Case T and Case M demonstrate slow proximal progression of ECMO-derived flow impeded by distal vascular tortuosity (Video 1 and 2). In Case S, by comparison, ECMO-derived flow is least impeded with heart-derived flow reaching only to the BCA at 90% ECMO (Fig. 3C). As the proportion of ECMO-derived flow decreases, the watershed region moves distally in the aorta.

Fig. 3.

Volume fraction of flow within the aorta provided by the ECMO circuit as a fraction of total flow is shown at peak systole for each of the three patient-specific cases as labeled. Simulations were performed at different levels of ECMO support corresponding to increasing severity of heart failure with Panel (A) shows the distribution of flow for ECMO providing 50% of total flow; Panel (B) at 75%, and Panel (C) at 90%. Flow derived from the heart is colored blue and ECMO-derived flow is colored red.

Still images generated at different points over the cardiac cycle at steady state for 50% ECMO (Fig. 4) demonstrate dynamic variability of the watershed region with visible changes in its size and shape. At end diastole, when flow into the aorta is provided solely by the ECMO circuit, the watershed region is pushed proximally. The straight configuration of Case S demonstrates the largest proximal movement as retrograde flow from the ECMO circuit advances during diastole. By comparison, proximal movement of the watershed region for the highly tortuous Case T is reduced and resides within a more limited region of the aorta over the cardiac cycle. Case T and Case M also exhibit reduced mixing of heart and ECMO-derived flow as retrograde ECMO flow is impeded by distal vascular tortuosity.

Fig. 4.

Comparison of the watershed region movement over the cardiac cycle in the three patient-specific cases with ECMO flow constituting 50% of total perfusion. Panel (A) shows the distribution of perfusion for Case S while Panel (B) shows Case M and Panel (C) corresponds with Case T.

The clinical importance of maintaining renal perfusion during circulatory support makes the renal arteries ideal for comparing the watershed region across the patient geometries and at differing ECMO flows (Fig. 5). By defining the watershed region location quantified by a given aortic cross-section where equal amounts of ECMO and heart blood exist, the location of the watershed region relative to the renal arteries for each case was assessed. At both 50% and 75% ECMO, the watershed region of each case remains proximal to each geometry’s renal arteries throughout the entire cardiac cycle. The extent of the proximal movement of the watershed region is shown to be determined by both the distal vascular tortuosity and the proportion of ECMO-derived flow. As the pulsatile component of heart-derived flow is reduced relative to ECMO-derived flow, the movement of the watershed region over the cardiac cycle decreases with the amount of the decrement dependent on the vascular tortuosity.

Fig. 5.

Quantitative assessment of the watershed region extent and movement over the cardiac cycle in each patient-specific case. Panel (A) shows the relative distance of the watershed region at 50% ECMO flow from the renal artery over the cardiac cycle. Panel (C) shows the relative motion of the watershed region at 50% ECMO flow at each time point in the cardia cycle. Panel (B) and Panel (D) show the watershed region movement for each case at ECMO flow of 75% of total perfusion.

3.3. ECMO flow and end-organ perfusion

The outflow rate of ECMO blood was calculated in each outlet artery to determine the distribution and variation of ECMO-derived perfusion in the different patient cases (Fig. 6). To facilitate comparison, branching arteries were grouped into proximal, abdominal, renal, and distal outlets. ECMO-derived perfusion is predominantly distal within the aorta, leading to renal and distal arteries receiving an increasing flow at lower proportions of ECMO flow. Perfusion of the great vessels remained derived from the failing heart in virtually all cases except when ECMO flow exceeded 90% of total flow.

Fig. 6.

Comparison of the relative blood flow rates into each branching artery deriving from the aorta for each patient-specific case at Panel (A) 90%, Panel (B) 75%, and Panel (C) 50% ECMO flow.

Distal vascular tortuosity resulted in a lower peak wall shear stress (WSS) near the site of initial ECMO-derived retrograde flow that then rapidly dissipated in comparison with both higher peak and more sustained elevations in WSS observed in the straight configuration of Case S (Fig. 7). For both Case T and Case M, the areas of high WSS were concentrated within the tortuous iliac artery and then markedly reduced within the aorta. This contrasts with the straight geometry of Case S in which elevated WSS was present into the distal aorta, likely correlating with the preserved higher blood flow velocity.

Fig. 7.

Representation of the wall shear stress for each patient-specific case calculated at ECMO flow rate of 90% and 50%.

4. Discussion

Clinical use of ECMO to provide circulatory support is rapidly growing despite limited understanding of the ECMO-failing heart circulation and how it is affected by patient-specific anatomical variation [18,19]. Using a femoral vessel cannulation strategy, ECMO provides continuous retrograde perfusion of the aorta which collides with pulsatile antegrade perfusion from the heart to create a dynamic watershed region of mixing flows with uncertain effects on hemodynamics and end-organ perfusion [5,20]. ECMO-generated retrograde perfusion presents risks to the failing heart as increased afterload may increase cardiac work, lead to pulmonary edema, and, in the setting of inadequate left ventricular ejection, promote life-threatening thrombus formation in the cardiopulmonary circulation and left heart [4].

While the risks of ECMO-generated retrograde perfusion are increasingly recognized, both antegrade perfusion from the heart and the patient-specific vascular anatomy represent loads on the ECMO pump that alter perfusion and behavior of the newly generated hybrid circulation [21]. The impact of patient-specific factors on the distribution of ECMO provided perfusion and on the complex watershed region are unknown and may be important determinants of clinical outcomes.

4.1. Anatomical variation and the ECMO:failing heart circulation

The proximal extent of ECMO-generated blood flow is determined by (1) antegrade flow limitation due to severity of cardiac pump impairment and (2) progression of retrograde flow impeded by pressure drops across distal vasculature. As blood ejected from the ECMO cannula encounters tortuous vessels, fluid momentum is lost and flow velocity decreases. This anatomy-dependent shaping of ECMO-derived flow determines the source of end-organ perfusion and highlights areas at risk for complications from mechanical support. Proximal tissues perfused by the heart risk inadequate oxygen delivery in the setting of concomitant lung disease, a condition referred to as differential hypoxia phenomenon [4,22,23].

The risk of differential hypoxia results from the complex hybrid systemic circulation created by ECMO cannulation. The extent of retrograde blood flow into the aorta is determined by ECMO flow rate, forward flow from the heart, rate of blood flow movement through the outlet arteries from the aorta, and patient vascular anatomical variation. In the presence of severe lung disease or maltitration of lung support, antegrade blood flow from the heart may be insufficiently oxygenated to provide for the metabolic needs of tissues supplied by the output of the left ventricle giving rise to differential hypoxia. The location of the watershed region determines tissues at risk for insufficient oxygen delivery which motivates the need for improved understanding of ECMO-generated flow dynamics. The aortic arch is entirely perfused by the heart in simulations of 75% ECMO while even at 90% ECMO, in which cardiac output is only 0.5 LPM and models profound heart failure, ECMO-derived flow variably reaches the BCA as determined by patient-specific distal vessel tortuosity (Fig. 6). These findings support the clinical need to monitor oxygen tension on the right upper extremity or right side of the head when seeking to ensure adequate cerebral and myocardial oxygenation and demonstrate the pitfall of using systemic hypoxia as an indicator for preferential use of ECMO over other forms of MCS. While ECMO returns oxygenated blood as circulatory support, the distribution of perfusion provided by femoral cannulation is inadequate to reliably achieve adequate delivery of oxygenated blood to the aortic arch and risks cerebral and myocardial ischemia in the setting of concomitant lung disease. In cases in which severe lung failure is not amenable to optimization of mechanical ventilatory support, the addition of a return cannula inserted into the right jugular vein and splicing of the post-oxygenator flow to provide oxygenated blood to the cardiopulmonary circulation may be required.

The extent of ECMO perfusion also determines risk of injury from circuit-derived emboli which are a constant concern in ECMO-supported patients. Computational results demonstrate significant differences in distribution of ECMO-derived flow between patient-specific cases. ECMO-derived flow at 50% ECMO advanced proximately above abdominal and renal arteries in Case S, a straight and non-tortuous aorta, in comparison to the much more tortuous Case T in which the watershed region encompassed the major abdominal and renal arteries (Figs. 3 and 4). The high shear stress at the site of ECMO flow from the return cannula may constitute a risk of atherosclerotic plaque embolization in patients with peripheral artery disease. This may be exacerbated by placement of the cannula itself which may disrupt local plaques and induce embolization or even vessel rupture as has been reported during aortic cannula placement during cardiopulmonary bypass [24].

The cumulative effect of increased tortuosity is likely an increase in risk factors contributing to end-organ malperfusion and worse clinical outcome.

4.2. Effects of anatomy on watershed dynamics

Flow dynamics of the watershed region in vivo are not well understood and may constitute a risk of thrombus formation and hypoperfusion of vessels originating within the affected region of the aorta. Prior work demonstrated that size and movement of the watershed region is dependent on the degree of ECMO support provided [12]. The current work extends this insight further and demonstrates watershed location and extent are driven both by proportion of total flow provided by the ECMO circuit and by patient vascular geometry which profoundly alters the character of ECMO-derived flow. The complex interplay between antegrade and retrograde flow over the cardiac cycle is impacted by distal vascular tortuosity which impedes ECMO-derived flow thereby pushing the watershed region distally and reducing the extent of its proximal movement during diastole. In the straight configuration of Case S, ECMO-derived flow progresses relatively unimpeded allowing ECMO flow to advance proximally. The movement of the watershed region becomes a dynamic interplay between antegrade progression during cardiac systole and retrograde recession in diastole.

In Case S, the relative lack of distal vascular impedance allows ECMO flow to push the watershed region more proximally during diastole in comparison to the relatively more tortuous anatomies of Case M and Case T in which the watershed region exhibits less movement over the cardiac cycle. At 50% ECMO support, the watershed region remains proximal to the renal arteries in Case S throughout the cardiac cycle (Figs. 4 and 5) in comparison to the more distally located watershed region in Case M and Case T which engages with the renal arteries. As ECMO support increases with worsening heart failure, the extent of the watershed region further decreases as total momentum of antegrade blood flow from the heart decreases.

These observations also have ramifications for patients experiencing recovery and lessening need for mechanical support. As cardiac function returns, antegrade perfusion from the heart increases and the fraction of total perfusion provided by ECMO decreases. This functionally acts to both move the watershed region distal in the aorta and increase its size. The result of this movement may explain clinical observations of delayed onset renal failure or gastrointestinal bleeding in patients maintained on ECMO. The watershed region likely resides proximally within the descending thoracic aorta in early support before moving distally with improving cardiac function. Observations of improved outcomes for patients on dual support with ECMO and mechanical support devices that propel blood from the ventricle may partially be explained by improved end-organ perfusion resulting from a stable watershed regime in addition to protection of the left ventricle provided by dual support [25,26]. Dual support therefore both offloads the ventricle and allow for stable positioning of the watershed region while simultaneously decreasing its size by reducing cardiac pulsatility. An alternative approach may be to introduce pulsatile flow into the retro-grade perfusion provided by the ECMO circuit as a means of augmenting systemic pulsatility. Such variability could be timed with native contraction to augment LV unloading while increasing pulsatile flow to end-organs.

4.3. Vascular tortuosity and wall shear stress

The non-tortuous anatomy of Case S demonstrated a higher peak WSS and a more anatomically diffuse distribution of shear stress experienced by the vasculature in comparison to Case T in which peak WSS is lower and more focally concentrated within the tortuous iliac artery at the site of ECMO cannulation (Fig. 7). WSS is proportional to the flow velocity parallel to the vessel wall with anatomies that present limited impedance to ECMO-derived flow experiencing higher peak WSS values from retrograde perfusion. Higher flow rates may risk generation of atherosclerotic emboli from plaque rupture and increase the propensity for bleeding from destruction of von Willebrand Factor (vWF) [27].

5. Conclusions

This study advances understanding of the ECMO-failing heart circulation and how patient vascular anatomy alters perfusion and characteristics of the dynamic watershed region. Using CFD methods and patient-derived vascular anatomies, simulations were performed at varying levels of ECMO support to model different clinical conditions encompassing a range of cardiac failure stably maintained on mechanical support. Simulations demonstrated variations in the watershed region size, location, and response to pulsatile flow over the cardiac cycle. Wall shear stress at the site of cannulation varied across the patient-specific anatomies and highlighted the importance of patient factors on the effects of ECMO support and potential risks of inducing vascular injury or generating thromboemboli.

Glossary of Abbreviations

Fr

French (unit of measurement)

kg

kilogram

m

meter

cm

centimeter

cP

centiPoise

L

liter

LPM

liter per minute

ECMO

extracorporeal membrane oxygenation

CFD

computational fluid dynamics

MCS

mechanical circulatory support

VOF

Volume of Fluid

Q

flow

C

capacitance

R

resistance

BCA

brachiocephalic artery

LCCA

left common carotid artery

LSA

left subclavian artery

CA

celiac artery

SMA

superior mesenteric artery

LRA

left renal artery

RRA

right renal artery

IMA

inferior mesenteric artery

LCI

left common iliac artery

RCI

right common iliac artery

IRB

institutional review board

CTA

computed tomography angiography