Mechanical circulatory devices like blood pumps are being studied to support patients with end-stage heart failure. Continuous-flow blood pumps, both axial and centrifugal, have significant advantages compared with traditional pulsatile ventricular assist devices (VADs). These pumps are smaller, have no valves, are better designed for fluid dynamics and have the potential for better durability, improved safety and lower cost. For instance, already two small axial-flow pumps, the Jarvik 2000 device1 (Jarvik Research, Inc, New York, NY) and the MicroMed DeBakey Ventricular Assist device,2 have been extensively studied in animals and successfully implanted in humans as bridge to transplantation or as long-term support.
In radial-flow (centrifugal) blood pumps, the rotation of the impeller produces a centrifugal force that drags blood from the inlet port to the outlet port. In axial pumps, flow occurs from momentum energy transfer between the inlet and the outlet of the impeller. At 5 l/min against 100 mm Hg, the best efficiency points of these pumps’ topology will result respectively in a centrifugal pump of a larger size and lower rotational speed than the axial pump, which in turn will present a smaller size and higher rotational speed. A smaller pump with high rotational speed will generally induce a high shear rate. Dimensional analysis suggests a compromise in the design topology of pumps that results in flow occurring from the energy transfer by both centrifugal forces and momentum energy transfer. This characterizes a mixed-flow pump, which yields a smaller device than the centrifugal pump with lower rotational speed than the axial topology.
We have designed a dual-inlet mixed-flow pump combining the potential advantage of both the axial and centrifugal flow pumps. From a Cordier analysis, using the specific speed (Ns) and diameter (Ds) at the prescribed operating point (5 l/min against 100 mm Hg), we identified the mixed-flow pump topology as the topology providing the highest efficiency. We aim for a fully implantable device in accordance with the geometrical constraints associated with the left ventricle dimensions. Using a computational fluid dynamics (CFD)-based design and optimization tool, we designed and manufactured a paracardiac prototype of this dual-inlet mixed-flow pump.
We have built one prototype to test our design hypothesis and to provide data for its future optimization. This prototype was built with good surface finish but not sufficient for long-term in vivo implantations. This implies that thrombosis and emboli were not extensively analyzed, because this is related to the pump dynamic and surface finish. In this work, we are primarily interested in the hemodynamic behavior of this prototype in comparison to in vitro and in vivo testing.
Materials and Methods
A new mixed-flow pump for left ventricular assistance was designed and tested in vitro at the Department of Mechanical Engineering of École Polytechnique de Montréal and in vivo at the Montreal Heart Institute.
The new blood pump consists of a dual-inlet mixed-flow pump with three main components: inlet guide vanes, a rotor, and a stator. Figure 1 presents a cross section drawing (Figure 1a) and a picture (Figure 1b) of the implantable prototype. A DC brushless electric motor (EPFL, Switzerland) is part of the device with an attached inflow cannula and an outflow Dacron graft. The blood-contacting surface is titanium (Ti 6Al 4V). Overall CFD-predicted hydraulic efficiency of the pump is 44% at 5 l/min against 100 mm Hg with low incidence of hemolysis as previously shown (average of 70 ± 28 mg/l plasma free hemoglobin for six acute experiments in pigs over 3 hours).3,4 The pump operates at low power (<3 Wd).
First, the present pump was tested in vitro in a specially designed static test rig where flow rate, pump inlet and outlet pressures were monitored and controlled (Omega Engineering Inc., Stamford CT). A magnetic flowmeter was used for the pump outflow monitoring (Bio-probe TX40 from Medtronic Inc. Mississauga, Ontario, Canada). The tests were performed at 20°C with a blood analog consisting of 40% (by weight) aqueous glycerol to construct pressure/flow (HQ) curves at different pump speeds expressed in rpm (Figure 2). For a given pump speed, we varied the resistance or pressure head and noted the resulting flow rate.
In vivo, the pump was implanted in 11 pigs (80 ± 7 kg) and three calves (76 ± 7 kg) to analyze its hemodynamic response and the effect on animals in acute experiments. We used pigs to test feasibility and calves to determine geometry modifications of the pump associated with the thorax of this animal in order to plan for future long-term implantations. All experiments were conducted under general anesthesia with isoflurane and propofol, endotracheal intubation and mechanical ventilation. Following a left thoracotomy and systemic heparinization (1 mg/kg), the inflow cannula of the pump was inserted through the apex of the left ventricle and secured with sutures placed on the felt flange of the cannula. The outflow cannula with the terminal Dacron graft was anastomosed end-to-side to the descending thoracic aorta. A catheter was inserted in the left common carotid artery to monitor systemic pressure. Pressures in the left ventricle and in the left auricle were also monitored with the use of two fluid-filled catheters. The pump has two pressure probes, one to monitor the outlet pressure and one for the pressure head (Omega Engineering Inc.). An ultrasound flowmeter was also placed at the outlet of the pump in order to measure the pump flow rate (Transonic Systems Inc, Ithaca, NY). Serial transepicardial echocardiographic studies were conducted at each pump speed (Hewlett-Packard, Palo Alto, CA). These indicated that the pump speed should not exceed about 10,000 rpm against 70 mm Hg in order to keep the aortic valve open for a flow rate of 5 l/min.5,6 Of the 11 pigs, six were used for this echocardiographic study while data from all 11 pigs were used for the hemodynamic study.
Anticoagulation was monitored by serial measurements of activated clotting times (Sulzer Medical Canada), which was maintained above 300 seconds throughout the experiments. Body temperature was maintained at 37°C by the use of heating pad. Blood gases were monitored to insure proper oxygenation throughout the experiments. As a first step to assess the immediate hemodynamic behavior of in vivo support, the experiments were conducted for 3 to 12 hours in pigs and 4 to 9 hours in calves. The animals were killed by means of injection of a bolus of potassium under general anesthesia. A detailed autopsy was not performed; however the implantation sites were examined for acute thrombus formation.
Pump speed was varied to study the physiologic response. Before data recording, 20 minutes of stabilization was required between changes in speed. The data are presented as the mean and standard deviation. Animals were maintained and tested in accordance with the recommendations of the Guidelines on the Care and Use of Laboratory Animals issued by the Canadian Council on Animals and approved by the Montreal Heart Institute Ethics Committee. The present study was only descriptive and no statistical analysis comparing groups was performed.
In Vitro Dynamics: Pump Speed and Pressure Head
The dual-inlet mixed-flow pump was tested on a static test rig to provide pressure and flow rate relationship for a range of speeds. As shown in Figure 2, rotation speed from 7,000 to 12,000 rpm provide a flow rate range of 0–7 l/min against a predicted pressure head of 0–200 mm Hg. These ranges were extrapolated from experimental in vitro data (also shown in Figure 2) with range of 0–7 l/min in flow rate against 40–140 mm Hg. For the prescribed design point, a flow rate of 5 l/min against a pressure head of 100 mm Hg, the pump runs at a rotation speed of 11,500 rpm. For the pressure head range of 80–120 mm Hg, the pump will run at a speed varying from 10,500 to 12,000 rpm to maintain the same flow rate of 5 l/min. At 10,000 rpm, the pump will work against 60 mm Hg still at 5 l/min.
In Vivo Pump Speed and Output
In 11 pigs, pump outflow averaged 3.8 ± 0.4, 4.5 ± 0.4, 5.2 ± 0.8, 5.9 ± 0.3, and 6.5 l/min at pump speeds of 8,000, 9,000, 10,000, 11,000, and 12,000 rpm (Figure 3). We observed a linear relationship between flow rate and pump speed, as the pump outflow will increase with the rotation speed (N). At higher pump speeds of 11,000 and 12,000 rpm, echocardiography readings highlighted that the aortic valve remained closed throughout the cardiac cycle. Over a subset of these acute experiments (six pigs), Table 1 presents the main control parameters in order to keep the aortic valve open during support.5,6 A simple ratio of the pump head over the pressure difference between the pump outlet and the left auricle can indicate whether the aortic valve is open or closed (Figure 4). On this subset, a smaller pump speed of 10,000 rpm indicated a closed aortic valve.
Differential pressure through the pump averaged 45 ± 6, 54 ± 8, 68 ± 16, 70 ± 12, and 85 ± 7 mm Hg at 8,000, 9,000, 10,000, 11,000, and 12,000 rpm. Figure 5 illustrates the averaged mean aortic pressure (MAP), left ventricle pressure (LVP), and left auricle pressure evolution against pump speed. Mean aortic and left auricle pressures remained stable throughout the experiment and averaged 64 ± 15 and 13 ± 3 mm Hg, respectively. The left ventricular pressure decreased as pump speed increased.
In three calves, MAP and LVP remained stable during 4, 6, and 9 hours of support at 9,500 to 11,500 rpm with a pump outflow averaging 5.2 ± 0.7 l/min. Figure 6 illustrates the variations of the averaged mean aortic and left ventricular pressures against the pump rotation speed. Both pressures remained stable and presented an average value for the mean aortic pressure of 83.1 ± 2.4 mm Hg and for the left ventricular pressure of 19.1 ± 1.7 mm Hg. No data was available from left auricle pressure monitoring because we experienced a malfunction with the sensors. In the three calves, aortic valves remained functional with partial emptying of the left ventricle in all animals.
Thrombus formation was found to occur around the ball bearing shaft of the rotor. It takes the appearance of a protein ring. This event supports the necessity of an ameliorated support technology design. The heart and aorta were free from any thrombus formation.
Combining In Vivo and In Vitro Data
Figure 7 shows typical pressure and flow rate relationships measured directly across the device at constant speeds of 8,000, 10,000, and 12,000 rpm (the clouds of points) from an acute implantation in a pig. The dark curves are the theoretical pump curves obtained from the pump performance analysis on a static test rig operating with a blood analog (from Figure 2). It is observed that the measured data are clustered around their respective theoretical pump curve (diamonds for 8,000 rpm, crosses for 10,000 rpm, and triangles for 12,000 rpm). The hysteresis can be reliably associated with the stability of the pump controller (±5% in rpm).
The present study suggests that, for short-term experiments lasting 3 to 12 hours, the dual-inlet mixed-flow pump is able to sustain appropriate outflow through its outlet graft and to carry the left ventricular load at high speed (rpm). The main observation is that the prescribed operating point of 5 l/min against 100 mm Hg may not be appropriate for support purposes. Indeed, the hemodynamic data compiled so far indicated that the support of 5 l/min against 70 mm Hg is sufficient in order to keep the aortic valve open. It is our understanding that we need to maintain the aortic valve open during support if we do not want to have the valve leaflets merge. We recalled main results (Table 1 and Figure 4) from previous investigations5,6 based on a subset of the acute experiments in pigs. They demonstrated that for pump speed of 10,000 rpm and higher, the aortic valve was found to be closed. A simple relationship between pressures indicated this aortic valve state, as a closing ratio:
During support, keeping this ratio lower than 100% will ensure that the aortic valve remains open. If we trace for various pump speeds the state of the aortic valve against pressures, we get a graph similar to Figure 4 that supports the previous relationship findings. This may translate into an optimal pump prototype operation range at lower speed and smaller pressure head than expected, while identifying relevant parameters for the future pump controller. Note that to close the aortic valve in calf acute experiments, a higher pump speed than that used in pig experiments was required. Hence, we cannot decrease a fixed pump speed as threshold for the aortic valve closure; we need to assess the closing ratio for each animal.
We can also determine the degree of support of the prototype in acute experiments as defined by the following ratio:
In both pig and calf experiments, the support ranged from 49–83%. This seems an acceptable range for the admitted purpose of this device, because we do not aim for the pump to totally carry the heart output but to partially support it.
Comparatively, Macris et al.7 showed that the Jarvik 2000 axial blood pump achieved speeds as high as 13,000 rpm with more than 11 l/min of flow. Operating at 10,000 rpm gave an average of 5 l/min with a differential pressure approximating 70 mm Hg in vitro and in vivo with the latter pump.7 Other investigators have also reported average pump flow rate of 5 l/min at pump speed of 10,000 rpm in studying different axial flow pump designs.8 In the case of the HeartMate III (Thoratec Corporation, Pleasanton, CA), a centrifugal VAD with magnetically levitated rotor, it can generate 7 l/min of flow against 135 mm Hg of pressure.9 The present dual-inlet mixed-flow pump prototype designed to achieve 5 l/min at 11,500 rpm against 100 mm Hg of differential pressure showed similar characteristics to our in vitro study (Figure 2). At 10,000 rpm, the pump prototype provides 5 l/min against 68 mm Hg in vivo and 60 mm Hg in vitro. This fact suggests that a redesign of the prototype, operating at lower rotational speed and smaller pressure head, would provide better support while keeping the aortic valve open. Hence the prescribed operating point would have to be revised.
Choosing the appropriate operating point for design of these flow pumps is essential because inappropriate points will lead to decreased durability and increased energy consumption.10 Overall, the present data obtained during acute experiments show that the pump operated in vivo similarly as predicted in vitro (Figure 7). Although Figure 7 shows the pump could create negative left ventricular pressure or suction effect in some physiologic conditions, most in vivo data points obtained were in the appropriate position in the pressure/flow curve.
Our current pump controller aims at maintaining constant speed without any physiologic control. This has consequences on patient management and care. Hence, as the pump speed increases, the aortic pressure tends to a nearly constant mean pressure with very little pulse pressure. However, the left ventricular pressure remains with a significant time-dependent pressure wave form. This implies that the pressure head through the pump is also time-dependent and exhibits time-dependent flow rate through the pump at constant speed (hysteresis effect in Figure 7). Tuzun et al.11 showed similar results in vivo with the Jarvik 2000 pumps: a decrease in mean left ventricular pressure and stable mean aortic pressure with a stepwise increase in pump speed as shown in Figure 5. This effect is related to the increasingly predominant continuous flow associated with progressive reduction in native cardiac pulsatility.
In this study, we found no thrombus formation in the heart and the great vessels, and we recall that a stable ring of thrombus was found at the pump bearings location. Design modifications will yield an improved device, and we are planning to test this new pump in chronic studies for thrombus formation and embolism within the pump and the animal.
The hemodynamic behavior of our dual-inlet mixed-flow pump prototype appears satisfactory during short-term support in animals. It supports similarly to axial-flow blood pumps marketing clinical trials. However, the collected hemodynamic data have underlined the necessity of an ameliorated design of this mixed-flow pump running at smaller rotational speed against pressure head, and a redesign of the rotor bearings technology to minimize thrombus formation. Still, results of the present study support further long-term animal implantations and studies to assess the benefit of this design.
1.Macris MP, Myers TJ, Jarvik R, et al: In vivo evaluation of an intraventricularelectric axial flow pump for left ventricular assistance. ASAIO J
40: M719–M722, 1994.
2.Noon GP, Morley DL, Irwin S, et al: Clinical experience with the MicroMed DeBakey ventricular assist device. Ann Thorac Surg
71: S133–S138, 2001.
3.Carrier M, Garon A, Pelletier LC: Developmental phase of a new axial flow pump for left ventricular assist. ASAIO J
40: 152., 2002.
4.Carrier M, Garon A, Farinas MI, Jullien S: Acute in vivo hemolysis investigation of a new axial flow pump for left ventricular assist. Can J Cardiol
(Suppl B): 127B, 2002.
5.Farinas MI, Garon A, Jullien S, et al: In vivo assessment of LVAD work output. Int J Artif Organs
25: 699., 2002.
6.Farinas MI, Jullien S, Garon A, Carrier M: Identification of the controller parameters for a new LVAD through in vivo acute implantations. P19, in Program & Abstracts of 10th Congress of the International Society for Rotary Blood Pumps (ISRP2002)
, Osaka, September 11–14, 2002.
7.Macris MP, Parnis SM, Frazier OH, et al: Development of an implantable ventricular assist system. Ann Thorac Surg
63: 367–370, 1997.
8.Macha M, Litwak P, Yamazaki K, et al: Survival for up to 6 months in calves supported with an implantable axial flow ventricular assist device. ASAIO J
43: 311–315, 1997.
9.Frazier OH, Tuzun E, Cohn W, et al: Total heart replacement with dual centrifugal ventricular assist devices. ASAIO J
51: 224–229, 2005.
10.Kosaka R, Yanagi K, Sato T, et al: Operating point control system for a continuously flow artificial heart: In vivo study. ASAIO J
49: 259–264, 2003.
11.Tuzun E, Eya K, Chee HK, et al: Myocardial hemodynamics, physiology, and perfusion with an axial flow left ventricular assist device in the calf. ASAIO J
50: 47–53, 2004.