Materials and Methods
The design of the cardiac simulator has been described previously and has demonstrated the ability to reproduce VAD patient hemodynamics in vitro.6 A MicroMed Debakey (MicroMed Cardiovascular; Houston, TX) continuous flow VAD is integrated into the simulator with an anatomically accurate geometry, which utilizes a flexible control interface to allow adjustment of VAD speed.
A bioprosthetic valve (Medtronic Mosaic; Medtronic, Minneapolis, MN) was used to simulate the aortic valve of a heart failure patient. These valves are prepared from porcine aortic valve leaflets that are chemically fixed with a process designed to preserve the native tissue properties and then attached to a plastic strut that is covered in fabric. Before attaching the construct to the device, approximately 40–60 small markers (0.25 mm) are attached to the ventricular surface of the leaflets using a needle and small amount of cyanoacrylate glue, as described for our previous study.7 Large markers were placed around the annulus to provide a reference for an anatomical coordinate system, with circumferential, radial, and axial coordinates. Thin rubber gaskets are used to mount the valve in the cardiac simulator and prevent leakage through the sewing ring. A mechanical valve (Medtronic, tilting disc) was used in the mitral position as the inflow valve to the left ventricle.
The aortic valve is mounted on the top of the simulator and the camera placed underneath, viewing the valve through the water in the tank. A fluid-filled chamber with an attached piston pump is cyclically pressurized to reproduce the pressure-volume relation of the LV. The Micromed Debakey continuous flow VAD is placed outside the chamber, which is connected to the aorta with a short length of Tygon tubing. The branched connection merges the fluid from both paths—the aortic valve and the VAD—in the distal aorta in which an in-line flowmeter is placed to record the flow (Transonic Systems, Inc., Ithaca, NY). A clamp-on flowmeter is placed on the VAD conduit to record flow. Left atrial pressure, left ventricular pressure (LVP), and aortic root pressure are measured with medical grade sensors (Abbott Transpak IV, Abbott Park, IL). Hemodynamic data are measured continuously at a rate of 200 Hz with Biobench, a Labview-based software (National Instruments, Austin, TX).
The experiments on the SDSU cardiac simulator were conducted for a matrix of test conditions with variables of VAD speed (off, 7.6, 10, and 12.3 krpm) and cardiac support (CS). The cardiac simulator has several different CS settings: off (no pumping, corresponds to the series flow condition), low (stroke volume of 23 ml at 70 bpm for a flow rate of 1.6 L/min), medium (stroke volume of 35 ml at 70 bpm for a flow rate of 2.5 L/min). These settings represent ventricles with compromised contractile function as is typical for patients with cardiac failure, achieving ejection fractions of 12% and 20% in the absence of a VAD, respectively. As described in our previous study, the resistance and compliance of the cardiac simulator circuit are set at the beginning of each experiment to reach the target range of pressures and flows and then left at those settings for the entire duration of the experiment.6
The camera was positioned directly underneath the aortic valve, parallel to the annulus of the valve and viewing the ventricular side of the leaflets. The image is calibrated using a grid of 2-mm spaced markers, which are used not only to convert the imaging data to scale but also used to calculate a dewarping correction. Sequences of 90 images were obtained at 9 Hz for the flow conditions listed earlier. Three different paths for flow were evaluated: PreVAD or normal flow in which the VAD is off and the VAD conduit clamped; series flow in which the transvalvular pressure (TVP) is much greater than zero and the aortic valve is closed during the entire cycle; and parallel flow in which the TVP is zero during part of the cardiac cycle, allowing the aortic valve to open. The area and timing of opening were measured from these images using ImageJ software (NIH, http://rsbweb.nih.gov/ij/).
To obtain the reference state for the strain analysis, the camera was focused on the markers when the valve was under a small TVP (∼10–20 mm Hg). The transvalvular pressure is the difference between the aortic pressure and LVP. The TVP was reduced until the valve opened and then slowly increased until the valve just closed at which point the reference state image for the strain calculations was obtained. In most cases, this reference state corresponded to a TVP of ∼2 mm Hg. The flow conditions listed in Table 1 were tested, and hemodynamic and imaging data were recorded simultaneously.
The hemodynamic data file is marked to show the timing of the image sequence, and coincident sequences were used for analysis of time series and average results. Averages of 8–10 complete cycles of left atrial, left ventricular, aortic pressures and VAD and total aortic flows are computed.
The analysis of the strain data applied an automated marker tracking system designed for measuring displacement fields of moving surfaces (DaVis; LaVision GmbH, Goettingen, Germany). The current version of the software does not allow for polar coordinate-based geometries, such as the valve, thus regions of interest were defined on the leaflets in which the x-y orientation of the analysis region corresponded to the circumferential and radial directions of the leaflet. Twenty-five to thirty markers were included in each ROI and used to calculate two-dimensional strain. The reference image was appended into the imaging sequence for each flow condition and used as the original marker positions from which displacement occurred. A shift and intensity correction was applied to align all the images in the set and avoid false marker repositioning. The DaVis software used a 2D time series cross-correlation method to detect the new position of each marker for each subsequent image in the sequence, resulting in a 2D displacement map for all markers that remained in the view.
Vector postprocessing enabled the elimination of outliers, and the three strain components were calculated from the displacement field: two normal strains Ecirc and Eradial and the shear strain Eshear. The Eshear component was small and the measurement error significant, yielding a poor signal to noise ratio. For each flow condition in each valve, each normal strain component was averaged for five cycles.
The hemodynamic data are summarized in Table 1 for normal, series, and parallel flow path conditions from three different valve studies. Two levels of cardiac function (low and medium) as controlled by the simulator and three VAD speeds (7.5, 10, and 12.5 krpm) were tested as shown. The normal flow conditions confirm that the CS pressure and flow range are in the physiological range for a heart failure patient. The means and standard errors are shown for the three valves studied. The standard errors reflect the differences in mean values among the three studies not the variation in magnitude of the signal. Aortic root pressure (AoP), LVP, and aortic flow (AoQ) all increased with the level of cardiac function (off, low, medium) as expected. Aortic root pressure and AoQ both increased with VAD speed and LVP decreased with VAD speed, consistent with clinical observations. Ventricular assist device flow (VADQ) also increased with both VAD speed and cardiac function except under parallel flow conditions. In the parallel case, the aortic pressure increased, producing higher flow in the distal aorta (AoQ) than the VAD (VADQ). The ratio of VADQ to AoQ provides an estimate of the distribution of flow under parallel conditions, showing that at the lowest VAD speed and the greatest level of cardiac function, approximately 10% of the flow is through the aortic valve and 90% is through the VAD.
The transvalvular pressure is calculated by subtracting LVP from AoP and is shown for the normal, series, and parallel conditions in Figure 3. The results illustrate the cyclic behavior for the normal and parallel conditions, with a notable increase in the diastolic TVP and a decrease in the period where TVP ∼0 when the valve opens. The series flow condition produced a high TVP throughout the cycle and the valve remained closed. It is notable that the TVP was lowest in the normal state and increased in both series and parallel flow conditions. In the series case, this was primarily due to the decrease in LVP corresponding to the unloading of the left ventricle.
Imaging of the aortic valve during these flow conditions reflect this behavior and are shown in Figure 4. The images were selected from the entire sequence to show the relative area and duration of aortic valve opening. For the normal flow path, the valve opened to its fullest extent regardless of the level of cardiac support (low or medium). The series flow condition did not produce any valve opening. During the parallel flow condition, the valve opened only slightly and briefly, as best seen in the third image in that sequence. The area and timing of opening results are shown for a single valve in Figure 5. The measurements reflect that, while the valve opens during parallel flow conditions, the area and duration of valve opening were greatly reduced from normal conditions. The maximum open area (max A) and opening time (O time) were determined for each of the three valves studied and are reported in Table 2. The maximum opening area of the aortic valve is reduced by 50% from normal to parallel flow conditions, and the maximum opening time is reduced by 20%. Integration of the opening area-time curve yields an additional variable that combines both opening and duration and is listed in Table 2.
In general, both Ecirc and Eradial increased with increasing cardiac function in the absence of the VAD. Ecirc and Eradial increased with VAD speed for all conditions, but the effect of increasing cardiac function during VAD operation was speed dependent or, perhaps more accurately stated, flow path dependent. At the low VAD speeds, a decrease in strain was observed, corresponding to the shift from series to parallel flow, which relieves stress on the leaflets during valve opening. At the higher VAD speeds, the valve remained closed, and increasing the level of cardiac function produced an increase in the TVP as well as an increase in Ecirc and Eradial (Figure 6).
The aortic pressure and strain data for each of the flow conditions were compiled into a single graph as shown in Figure 7. In this figure, radial strain is consistently greater than circumferential strain and increases in a nonlinear manner with aortic pressure. The flow conditions producing the lower points on the pressure-strain curve are those with the cardiac simulator alone. Pressure and strain increased with increasing VAD contribution.
The results of our study clearly demonstrate that the biomechanics of the aortic valve are altered with VAD use. It is well known clinically that under conditions of high VAD flow and low cardiac function, the transvalvular pressure is high and the valve remains closed throughout the cardiac cycle (series flow). Thus, especially with continuous flow pumps, physicians try to maintain pulsatile flow to assure that the ventricle is a full reservoir for the VAD to draw from, with the condition that the aortic valve opens periodically. This is considered to be important when setting the VAD pump speed for each patient. With pulsatile VADs, the asynchronous relationship of the VAD to the native heart naturally causes episodic aortic valve opening during periods of parallel flow. The valve opening results from our study showed that the diastolic portion of the cycle is prolonged during VAD use. In addition to the reduction in valve opening time, there is a reduction in valve opening area, effectively producing a VAD-related functional stenosis. The implications for clinical use is that valve opening during parallel VAD flow conditions, with either continuous or pulsatile VADs, is reduced compared with valve opening during normal flow conditions. Consequently, current clinical practice for VAD patients may not be achieving sufficient valve opening to prevent changes such as fusion and incompetence.
The findings support our hypothesis that the altered biomechanics introduced by the VAD result in increased strain in the valve leaflets, especially during series flow. The average strain in the leaflets was higher during VAD support, primarily due to an increase in the minimum strain, during systole, rather than the maximum strain during diastole. The average strain during continuous series flow did increase with VAD speed and transvalvular pressure in a nonlinear manner, with radial strain exceeding circumferential strain.
The aortic valve leaflets are soft tissue structures that exhibit nonlinear, anisotropic material properties.8,9 The structure consists of a highly aligned network of crimped collagen fibers oriented along the circumferential direction of each leaflet.10 The material properties of the tissue reflect this fiber reinforcement, with the nonlinearity of the stress-strain response illustrating the uncrimping of collagen fibers as they are stretched. The collagen fibers provide increased stiffness along their length, resulting in a steeper rise in the stress-strain behavior along the circumferential direction, compared with the transverse (radial) direction. This anisotropy is seen in the results of our study in the relative circumferential and radial strain values, best shown in Figure 6, in which the aortic pressure-strain diagram bears a notable similarity to the stress-strain behavior exhibited by the aortic valve tissue in an equibiaxial stretch study.9 Thus, when the valves are loaded with a transvalvular pressure, as in these studies, they respond with a stretch pattern similar to that seen in an equibiaxial stretch study.
Previous investigators have experimentally measured the strain in aortic valve leaflets, without the influence of a VAD, in a mock physiological loop similar to ours. An earlier study by Smith et al.11 analyzed the strain in the leaflets of a single bioprosthetic heart valve and found a large range for the principal strain values. A more recent study by Yap et al. measured strain in a normal porcine valve using marker-based video imaging. They observed a large anisotropic deformation when the valve was closed during diastole, reaching a maximum of 20% stretch along the circumferential direction and a stretch of 60% along the radial direction.12 The anisotropic response is consistent with our results, but the magnitude of stretch in the native tissue is higher than the bioprosthetic leaflets used in our study, reflecting the stiffening of the leaflets after glutaraldehyde fixation.
Unlike strain, stress cannot be measured directly but must be estimated using computational models. Previous computational studies have focused mostly on stress in the evaluation of model results. High stress has been associated with pathological remodeling of leaflets including calcification and annular dilatation.13,14 The prediction of stress is very sensitive to the choice of geometry and material properties for the valve. Strain and displacement are less sensitive to these influences but depend greatly on the choice of reference state. Previous models that have included strain predictions have estimated approximately 10%–15% strain under normal pressure loading but did not distinguish whether an anisotropic distribution was noted.15,16 Overall, while there was limited information on the strain distribution in porcine bioprosthetic aortic valve leaflets during normal physiological loading, previous reports generally agree with our results.
The development of aortic valve problems in patients with VADs has been reported previously.2 Aortic insufficiency in VAD patients can be acquired. Rarely, aortic dissection after VAD placement can cause prolapse of the aortic valve and produce AI. Insidious onset and progression of AI may also occur for undefined reasons such as (1) leaflet elongation, (2) sinotubular dilatation, (3) annular dilatation, or (4) any combination of the above. It is estimated that approximately 25% of patients with VADs develop AI or AS.5 One institution reports routinely using ultrasound imaging to evaluate patients on VAD support for more than 6 months, finding that 50% of patients developed inappropriately increased rates and flows and are found to have mild to moderate degrees of native AI.1 Furthermore, unusual structural changes have been observed in the aortic valves of VAD patients. In particular, the formation of loose collagenous tissue in the valve leaflets of VAD patients has been observed within a few months of implantation and is known as fusion because it binds the leaflets together.17 Fusion has been correlated with the development of VAD-acquired AS.18 Connelly et al.5 found that acquired commissural fusion is common in VAD patients, particularly those with pulsatile VADs operated in automatic mode, ultimately resulting in permanent series flow with the heart. In another study, four of five patients were affected.4 A recent study by Mudd et al.3 found evidence of commissural fusion in patients with continuous flow VADs, which was correlated with decreased valve opening and an increasing prevalence of aortic insufficiency. Clearly, this is a problem in a significant fraction of the VAD population. Obviously, there are deleterious consequences for patients who develop aortic fusion and do not receive a transplant. For instance, parallel flow will be limited by a stenotic valve, reducing the range of cardiac output and exercise capacity. For patients who will either “bridge-to-recovery” or have the VAD as a permanent implant, a method for timely diagnosis and treatment of aortic valve disease is imperative to the success of mechanical assist. Our findings show that the influence of VAD on aortic valve biomechanics is to shift the pressure burden onto the aortic valve, which is not designed for this loading. The response of the leaflets are to remodel, perhaps due to a combination of increased stress and decreased flow of nutrients to the tissue, both of which hasten the development of disease processes.
The cardiac simulator enables the investigation of VAD-related cardiac mechanics under well-controlled flow conditions. Although it can successfully reproduce VAD patient hemodynamics over a wide range, it cannot respond or adapt to changes the same way the human system does. The lack of autoregulation in the system prevents the control of blood pressure by adjusting systemic resistance. Another limitation is that a bioprosthetic valve is used to simulate the aortic valve of a heart failure patient. A bioprosthetic valve cannot undergo annular dilatation, which may play a role in modulating the alterations in valve biomechanics, and was not present in our study. Prosthetic valves exclude the coronary sinus tissue as an integral part of the whole valve mechanism. This boundary condition likely affects the distribution of stress and strain when compared with the normally integrated leaflets. Furthermore, the leaflet properties are usually stiffened by the glutaraldehyde fixation treatment, although the Medtronic Mosaic valve has a special fixation process that preserves the properties of the native tissue. Porcine tissue has been shown to have a similar structure and material behavior as young adult humans19 and has been used for decades as a prosthetic replacement. Not much is known about the aortic valve properties of heart failure patients, but there is no reason to expect valve disease, although some age-related stiffening is probably with this population.
The fluid used in the cardiac simulator is transparent for visualization and has a density and viscosity that is significantly lower than blood, similar to that used in previous mock loop studies.6,20,21 Compared with water, the density of blood is 10% higher and the viscosity is approximately five times higher.22 Blood is a non-Newtonian fluid, and viscosity changes with shear rate. However, in large vessels, the behavior of blood can be considered Newtonian, and this assumption is commonly made.23 Therefore, the fluid we used is less viscous and requires less energy to move through the system than blood would at the same rate. In our study, we are measuring valve opening biomechanics and strain in the valve leaflets when the valve is closed. Because the strain is affected by changes in pressure, not flow, we would expect little change to the strain values we have measured in this study if we used blood instead of water. However, the valve opening biomechanics might be affected by the decreased flow through the system. This effect is in general counteracted by the use of the bioprosthetic tissue, which is slightly stiffer and more difficult to open than the natural valve. The net contribution of these two effects to the behavior of the cardiac simulator is admittedly unknown.
A single camera was used to visualize the markers and calculate two-dimensional surface strain. Although the surface of the closed valve is fairly flat, the valve leaflets undergo a large excursion during opening. Thus, some error may have occurred in our strain measurement during opening because of the curved leaflet surface moving away from the focal plane. This error can be minimized by using a biplane (two-camera) system that is planned for future studies.
Studies of the biomechanics of the valve can provide conclusive information to predict the progression of aortic dysfunction for prospective patients, providing recommendations for surgeons to decide whether to repair, replace, or close off the aortic valve at the initial VAD implantation surgery. This approach can potentially lower costs and risk associated with reoperation of VAD patients to repair a damaged valve or to treat thromboembolic problems caused by the altered blood flow. As the reliability and performance of VADs improves, their use will be extended to more heart failure patients and the need for reoperation because of mechanical failure will be reduced. Long-term problems will become more prevalent, and the need will increase for a better understanding of the host-VAD interaction.
Currently there is no definitive treatment strategy for patients with AI or AS at the time of VAD implant surgery. Even initially normal aortic valves have been observed to become fused and insufficient after long-term VAD support.3 The results of this study demonstrate that as the VAD lowers ventricular pressure, the time of diastole increases for the aortic valve, which increases the duration of valve closure. In addition, the maximum opening area of the valve decreases with VAD use, producing an effective AS. When the valve is closed, it is subjected to high strains, thus the overall effect of the VAD is to increase the strain within the aortic valve leaflets and likely decrease the nutrient replenishment, which may explain observations of remodeling and dysfunction such as fusion and AI. When the heart is receiving partial VAD support, it is likely that the aortic valve—if opening in parallel—is only opening partially. Thus, it is recommended that the aortic valve be closed off for destination therapy patients who exhibit any signs of aortic valve incompetence at the time of VAD implant. If recovery is desired and the valve must remain patent, more studies are needed to determine the most effective strategy for maintaining normal aortic valve and root dynamics in VAD patients.
Supported by the American Heart Association Western States Grant-in-Aid (#0655102Y, PI: May-Newman). The work was also supported by Medtronic and Micromed by providing essential components for the experimental system.
The authors thank David Sims, Tu Nguyen, Michael Lester, and Greg Morris for their assistance in conducting experimental studies and analyzing the data.
1. Samuels LE, Thomas MP, Holmes EC, et al
: Insufficiency of the native aortic valve and left ventricular assist system inflow valve after support with an implantable left ventricular assist system: Signs, symptoms, and concerns. J Thorac Cardiovasc Surg
122: 380–381, 2001.
2. Park SJ, Liao KK, Segurola R, et al
: Management of aortic insufficiency in patients with left ventricular assist devices: A simple coaptation stitch method (Park's stitch). J Thorac Cardiovasc Surg
127: 264–266, 2004.
3. Mudd JO, Cuda JD, Halushka M, et al
: Fusion of aortic valve commissures in patients supported by a continuous axial flow ventricular assist device. J Heart Lung Transplant
27: 1269–1274, 2008.
4. Rose AG, Park SJ, Bank AJ, Miller LW: Partial aortic valve fusion induced by left ventricular assist device. Ann Thorac Surg
70: 1270–1274, 2000.
5. Connelly JH, Abrams J, Klima T, et al
: Acquired commissural fusion of aortic valves in patients with left ventricular assist devices. J Heart Lung Transplant
22: 1291–1295, 2003.
6. Zamarripa Garcia MA, Enriquez LA, Dembitsky W, May-Newman K: The effect of aortic valve incompetence on the hemodynamics of a continuous flow ventricular assist device in a mock circulation. ASAIO J
54: 237–244, 2008.
7. Chen L, McCulloch AD, May-Newman K: Nonhomogeneous deformation in the anterior leaflet of the mitral valve. Ann Biomed Eng
32: 1599–1606, 2004.
8. Billiar KL, Sacks MS: Biaxial mechanical properties of the native and glutaraldehyde-treated aortic valve cusp: Part II—A structural constitutive model. J Biomech Eng
122: 327–335, 2000.
9. May-Newman K, Lam C, Yin FC: A hyperelastic constitutive law for aortic valve tissue. J Biomech Eng
131: 081009, 2009.
10. Billiar KL, Sacks MS: Biaxial mechanical properties of the natural and glutaraldehyde-treated aortic valve cusp—Part I: Experimental results. J Biomech Eng
122: 23–30, 2000.
11. Smith DB, Sacks MS, Vorp DA, Thornton M: Surface geometric analysis of anatomic structures using biquintic finite element interpolation. Ann Biomed Eng
28: 598–611, 2000.
12. Yap CH, Kim HS, Balachandran K, et al
: Dynamic deformation characteristics of porcine aortic valve leaflet under normal and hypertensive conditions. Am J Physiol Heart Circ Physiol
298: H395–H405, 2010.
13. Weinberg EJ, Schoen FJ, Mofrad MR: A computational model of aging and calcification in the aortic heart valve. PLoS One
4: e5960, 2009.
14. Thubrikar M: The Aortic Valve
. Boca Raton, FL, CRC Press, 1989.
15. Grande-Allen KJ, Cochran RP, Reinhall PG, Kunzelman KS: Mechanisms of aortic valve incompetence: Finite element modeling of Marfan syndrome. J Thorac Cardiovasc Surg
122: 946–954, 2001.
16. Ranga A, Bouchot O, Mongrain R, et al
: Computational simulations of the aortic valve validated by imaging data: Evaluation of valve-sparing techniques. Interact Cardiovasc Thorac Surg
5: 373–378, 2006.
17. MacKenna D, Sommerour SR, Villarreal FJ: Role of mechanical factors in modulating cardiac fibroblast function and extracellular matrix synthesis. Cardiovasc Res
46: 257–263, 2000.
18. Rose AG, Park SJ: Pathology in patients with ventricular assist devices: A study of 21 autopsies, 24 ventricular apical core biopsies and 24 explanted heats. Cardiovasc Pathol
14: 19–23, 2005.
19. Kunzelman KS, Grande KJ, David TE, et al
: Aortic root and valve relationships. J Thorac Cardiovasc Surg
107: 162–170, 1994.
20. Litwak KN, Koenig SC, Cheng RC, et al
: Ascending aorta outflow graft location and pulsatile ventricular assist provide optimal hemodynamic support in an adult mock circulation. Artif Organs
29: 629–635, 2005.
21. Jimenez JH, Liou SW, Padala M, et al
: A saddle-shaped annulus reduces systolic strain on the central region of the mitral valve anterior leaflet. J Thorac Cardiovasc Surg
134: 1562–1568, 2007.
22. Fung YC: Biomechanics: Mechanical Properties of Living Tissues
. New York, NY, Springer-Verlag, 1993.
Copyright © 2010 by the American Society for Artificial Internal Organs
23. Perktold K, Peter RO, Resch M, Langs G: Pulsatile non-Newtonian flow in three-dimensional carotic bifurcation models: A numerical study of flow phenomenon under different bifurcation angles. J Biomed Eng
13: 507–515, 1991.