Secondary Logo

Journal Logo

Adult Circulatory Support

Heat Generation in Axial and Centrifugal Flow Left Ventricular Assist Devices

Yost, Gardner*†; Joseph, Christine Rachel*†; Royston, Thomas*; Tatooles, Antone; Bhat, Geetha

Author Information
doi: 10.1097/MAT.0000000000000438
  • Free
  • Conference Article


The continued evolution of left ventricular assist device (LVAD) design and management has harkened a new era for treatment of advanced heart failure. Increasingly, miniaturization, durability, and integration with cardiovascular physiology has yielded improved survival and quality of life.1,2 The number of implantations of current generation, continuous-flow LVADs has seen continued growth as heart transplantation has remained extremely limited because of low organ availability.3 As the prevalence of these pumps continues to increase, the importance of improved understanding of the effects of long-term support will continue to be critical in optimizing patient care. At this time there are no studies analyzing heat generation by either axial flow or centrifugal flow LVADs in vitro or in vivo. We present the results of laboratory testing of heat generation in the HeartMate II (HM II) and HeartWare (HW) LVADs.

The HeartMate II Axial Flow Left Ventricular Assist Device

The HM II is the most prevalent rotary LVAD currently used for treatment of advanced heart failure.4 The pump utilizes an axial flow design with left ventricular and ascending aortic cannulation. Blood enters the HM II through a titanium cannula which sits within the cavity of the left ventricle.5 A flexible woven Dacron graft protected by silicone sheathing carries blood from the intraventricular cannula to a titanium inflow elbow which is directly attached to the pump body. The pump body, displacing 124 ml and weighing 290 g, is primarily constructed of titanium and houses the motor assembly. Consecutive charging of magnets fixed radially around the pump bore generates a spinning magnetic field and imparts rotational torque on the impeller, causing it to spin. As blood enters the 12 mm diameter pump bore, it first encounters a three vaned inflow stator.4 The neutral, airfoil-shaped vanes on the stator remove prerotation of the blood and suspend the impeller via one of two ball-and-cup bearings. Distal to the inflow stator, rotation of the three bladed impeller imparts radial and axial velocity to the blood flow field. The impeller is suspended by a second mechanical bearing at the outflow stator. Unlike the inflow stator, the vanes of the outflow stator are curved to convert radial blood rotation to axial velocity at the pump outflow conduit. Blood exits the pump through a second woven Dacron graft which typically is sewn, end-to-side, to the ascending aorta.6

The pump housing and the impeller surfaces are smooth titanium, however, the stators and intraventricular cannula are sintered to improve biocompatibility.4–6 Blood flow in the HM II can be thought of as fluid moving through a pipe. The impeller behaves like an Archimedes screw, advancing fluid toward the outflow conduit. Pump speed is set for each patient in the clinic and typically ranges between 8,600 and 9,800 RPM with the pump achieving flows of 3 to 7 L/min.6

The HeartWare HVAD Centrifugal Flow Left Ventricular Assist Device

The HW is a small centrifugal pump with an integrated pump housing and inflow cannula that permits surgical placement within the pericardial space. Blood enters this 50 ml, 145 g pump through a sintered titanium inflow cannula and is accelerated by a wide-bladed, disk-shaped impeller before exiting through a 10 mm tangential outflow conduit.7 The impeller utilizes electromagnetic and hydrodynamic thrust bearing levitation systems, enabling operation free of mechanical bearings. Rare earth magnets within the impeller permit its magnetic alignment and levitation within the pump housing. Magnetic coupling to the two motors contained within the front and rear housing assemblies is used to generate impeller rotation.

This device does not incorporate blood diffusing or straightening stator vanes like the HM II does. Rather, blood moving through the HW contacts only the pump housing and the spinning impeller.8,9 The HW is designed to operate at speeds of 1,800 to 4,000 RPM with a maximum output of 10 L/min.


Experimental Setup

Mock circulatory flow loops were used to model physiological operation and heat generation for HM II and the HW LVADs. The LVADs were fitted to an abbreviated loop of 1.9 cm diameter vinyl tubing with a capacity of 110 ml, including the device. Fluid volume remained constant in all trials. The device housings were implanted in EcoFlex, a polyorganosiloxane viscoelastic gel with material properties similar to soft tissue, with volume 1,800 cm3, and mass 1.87 kg10 (Smooth-On Inc, Easton, PA). The gel served as a heat sink during pump operation. The devices were connected to their respective pump monitors, permitting adjustment of speed and display of pump flow generation and power consumption. Temperature was measured once per minute at the surface of the LVAD using an infrared thermometer. Ambient air temperature was 25.4°C

Physiological and pathological hemodynamic conditions were simulated by varying pump speed, fluid viscosity, and by introducing artificial thrombi into the device housing. Pump speeds for the HM II ranged from 8,200 to 12,200 RPM with separate trials at 8,200, 9,200, 10,200, 11,200, and 12,200 RPM. The pump speeds for the HW ranged from 2,600 to 3,400 RPM with separate trials at 2,600, 2,800, 3,000, 3,200, and 3,400 RPM. Temperature measurements at the device housing were made every 60 seconds for up to 80 minutes, at each speed. The fluid viscosity was adjusted by adding pure glycerol (viscosity = 950 cP) to distilled water. Fluid viscosity ranged from pure water (0.89 cP) to twice the viscosity of blood at 37°C (7.0 cP) in both pumps. Temperature measurements at the device housing were made every 60 seconds for up to 80 minutes, for each viscosity. Thrombus was simulated by fixing 1 cm3 pieces of Ecoflex gel to the inflow stator, impeller, and outflow stator of the HM II pump. This was only feasible for the HM II pump as the HW pump was required for other experiments and could not be altered.

Mathematical Modeling

In principle, heat generation in simple pump designs like those used by the HM II and HW may be modeled using known values for brake power (Ps), pump efficiency (μ), specific heat of the fluid (Cp), volume flow (q), and fluid density (p).11

Practically, these variables cannot be measured in the context of the highly dynamic cardiovascular system in which pressure, flow, and viscosity are dynamic. Accordingly, the in vitro model was designed to minimize the variability inherent to biological systems. Heat generation and the consequent temperature increase in the fluid flowing through the pump can be modeled using Equation 1 at the LVAD speed and viscosity conditions used in the bench top trials. And in this system, a viscoelastic model may be used to calculate the change in pressure head (change in pressure across the pump) via the Hagen-Poiseuille equation (Equation 2).

ΔP is the head pressure, μ is the dynamic viscosity (1.81 for blood at room temperature), L is the length of tube, Q is the volumetric flow rate, and r is the radius of tube.12

However, as the fluid flows through the remaining loop heat will dissipate from it at a rate that will increase as the temperature of the fluid increases (heated each time it passes through the pump). As the temperature of the circulating fluid increases, eventually the heat dissipation in the remainder of the fluid circuit will equal the heat input from the pump. Thus, an upper bound on temperature in the fluid will be approached asymptotically. A higher upper bound thus corresponds to a greater rate of heat input from the pump to the fluid. Quantifying exactly what that heat input rate is based on the temperature rise would require a detailed thermal model of the entire flow circuit, not just the pump, which is beyond the current study.

Statistical Analysis

All data are presented at mean ± standard deviation. One-way analysis of variance tests were used for comparison of mean values. A p value <0.05 was considered significant.


Sustained operation of both the HW and HM II resulted in gradual generation of heat. Time-dependent temperature curves are presented in Figures 1–5. In most cases, temperature reached a plateau by 60 minutes. Increased loading, via higher pump speed, greater fluid viscosity, or artificial thrombus resulted in greater maximum temperature in all trials.

Figure 1.
Figure 1.:
Temperature change and power consumption at variable speeds for the HeartMate II LVAD. LVAD, left ventricular assist device.
Figure 2.
Figure 2.:
Temperature change and power consumption at variable viscosity for the HeartMate II LVAD. LVAD, left ventricular assist device.
Figure 3.
Figure 3.:
Temperature change and power consumption with artificial thrombi placed at the pump inflow, the pump outflow, and directly on the impeller (large and small thrombi).
Figure 4.
Figure 4.:
Temperature change and power consumption at variable speeds for the HeartWare HVAD LVAD. LVAD, left ventricular assist device.
Figure 5.
Figure 5.:
Temperature change and power consumption at variable viscosity for the HeartWare HVAD LVAD. LVAD, left ventricular assist device.

Increased speed generated statistically significant increases in device temperature at each iteration in both the HM II and HW. Net temperature gain for the lowest and highest speeds was similar for the two devices. Power consumption increased with speed in a linear fashion in both pumps (Figures 1 and 4).

In the HM II, attachment of a small thrombus to either the inflow or outflow stator did not result in appreciably increased heat generation (Figure 3). However, when thrombi were affixed to the impeller itself, heat generation was much more rapid and resulted in higher sustained temperatures. Pump operation was significantly impaired by the presence of the large impeller thrombus, which prevented measurement of flow generation and power consumption.

Pressure and heat generation were closely related in the HM II. Hydrodynamic pump pressure at the outflow increased as speed and viscosity increased (Table 1). The trend of increased temperature with time asymptotically approaching an upper temperature bound Tb (at which point heat input from the pump equals heat dissipation in the remainder of the fluid circuit) is evident in all plots in Figures 1–5, with the mean times needed to reach 5% of the upper bound equal to 54.4 ± 6.09 minutes (range: 45–62 minutes) for the HM II and 54.25 ± 7.09 minutes (range: 45–64 minutes) for the HW. By assuming this process is governed by first order dynamics, the temperature rise ΔT as a function of time t can be modeled as

Table 1.
Table 1.:
Modeled Pressure for the HM II at Various Speeds and Viscosities

where τ is the system time constant and T0 is the initial (ambient) temperature. ΔT reaches 63.2% of its final value after a period of time, τ, has passed and 95.0% after 3τ have passed.


Normal operation of both the HM II and HW LVADs in our abbreviated flow loop resulted in appreciably increased system temperature, and all trials showed statistically significant increases in temperature between the initial temperature and peak system temperature. Heat generation was largely dependent on pumping conditions. Thermal equilibrium, the point at which the system reached its maximum temperature, was achieved within 80 minutes of pump activation in all experiments. At steady-state loading conditions, increased pump speed generated increased flow and required increased power supply. Elevated power consumption led to an increased rate of heat generation and higher peak system temperatures in both the axial flow HM II and centrifugal flow HW devices. It was initially hypothesized that the HM II pump would generate more heat than the HW because of mechanical friction incurred by rotation of the impeller within the two ball-in-cup bearings which suspend it. Despite these major differences in pump design, the rates of heat production and peak system temperatures for the two pumps were similar, suggesting that heat generation is a primarily a result of normal electrical inefficiencies in the electromagnetic pump motors employed by both devices, rather than mechanical friction.

When braking power was elevated by increasing fluid viscosity, heat generation increased accordingly, as indicated by Equation 1. With pump speed constant, power consumption increased with viscosity in both pump designs. The increased rates of heat generation and elevated peak system temperature at higher viscosity may be attributed to increased work of pumping and the generation of increased pressure head. Interestingly, the presence of a small artificial thrombus attached to either the inflow or outflow stator of the HM II did not result in increased heat generation. It is possible that, despite impaired flow in the presence of these artificial thrombi, heat dissipation in the miniaturized circulatory model used in these experiments was not significantly altered. Further, without significant increases in braking power, power requirements remained nominal. When artificial thrombi were attached directly to the impeller, significant increases in both power consumption and heat generation were noted. Presumably, these thrombi increased the energy needed to turn the impeller, requiring the usage of more power, and heat generation was resultant to inherent inefficiencies in the electromagnetic motor. In addition, impairment of flow through the device reduced heat dissipation and increased heat accumulation.

The loss of electrical power as heat in normal operation of an LVAD is an easily appreciated inevitability inherent to any electric motor, yet the magnitude and effects of heat dissipation into the blood and tissues surrounding the LVAD have not been reported at this time. Our results indicate that heat generation during normal operation of the HM II and HW is appreciable in vitro, and that heat generation is increased during conditions of increased load and resistance. Heat dissipation occurs as thermal radiation from the pump housing and heat removal by blood flowing through the device. Heat dissipation in vivo is a dynamic process and is assumed to be much more rapid compared with our experiments because of increased fluid volume in the human body and physiological homeostatic thermal regulation mechanisms. Further research to develop an experimentally validated model that incorporates the effects of heat dissipation, flow conditions, and the temperature of surrounding tissues may provide insight into heat generation during both in vitro and in vivo pump operation. Even if systemic blood temperature is not increased during prolonged LVAD operation, short-term increases in blood temperature caused by focal heat generation within the pump may have important hematologic consequences. Exposure of human blood to temperatures greater than 50°C has been shown to cause dramatic increase in platelet and eosinophil count and platelet aggregation.13 Red blood cells undergo hemolysis at temperatures greater than 42°C in vivo.14,15 Heat stress has been shown to lead to increased blood lactate, cholesterol, thrombin–antithrombin, fibrin monomers, and plasmin-alpha-2-antiplasmin which may further enhance coagulation potential.16,17 It is possible that short-term exposure of blood to high temperatures as it passes through the LVAD conduit damages the blood components and increases risk for thrombosis. Despite a growing body of research regarding its cause and improved algorithms for its detection, LVAD thrombosis remains a vexing challenge in long-term mechanical circulatory support.18 Further understanding of the role of heat generation may inform improved design in future LVAD models.


Our study indicates that heat generation in the HMII and HW LVADs may be measured in the laboratory—further study is required to investigate the hematologic effects of heat generation during prolonged hemodynamic support with LVADs.


1. Miller LW, Pagani FD, Russell SD, et al.HeartMate II Clinical InvestigatorsUse of a continuous-flow device in patients awaiting heart transplantation.N Engl J Med2007357885896
2. Slaughter MS, Rogers JG, Milano CA, et al.HeartMate II InvestigatorsAdvanced heart failure treated with continuous-flow left ventricular assist device.N Engl J Med200936122412251
3. Kirlin JK, Naftel DC, Pagani FD, et al.Seventh INTERMACS annual report: 15,000 patients and counting.J Heart Lung Transplant20153414951504
4. Griffth BP, Kormos RL, Borovetz HS, Litwak K, Antaki JFHeartMate II left ventricular assist system: From concept to clinical use.Ann Thorac Surg200171S116120
5. Burke DJ, Burke E, Parsaie F, et al.The Heartmate II: Design and development of a fully sealed axial flow left ventricular assist system.Artif Organs200125380385
6. Moazami N, Milano CA, John R, et al.HeartMate II InvestigatorsPump replacement for left ventricular assist device failure can be done safely and is associated with low mortality.Ann Thorac Surg201395500505
7. Larose JA, Tamez D, Ashenuga M, Reyes CDesign concepts and principle of operation of the HeartWare ventricular assist system.ASAIO J201056285289
8. Pagani FD, Miller LW, Russell SD, et al.HeartMate II InvestigatorsExtended mechanical circulatory support with a continuous-flow rotary left ventricular assist device.J Am Coll Cardiol200954312321
9. Slaughter MS, Sobieski MA 2nd, Tamez D, et al.HeartWare miniature axial-flow ventricular assist device: design and initial feasibility test.Tex Heart Inst J2009361216
10. Dai Z, Peng Y, Mansy HA, Sandler RH, Royston TJComparison of poroviscoelastic models for sound and vibration in the lungs.J Vib Acoust201413605101215101211
11. Karassik IJ, Messina JP, Cooper P, Heald CCCentrifugal pump theory, analysis, and performance, Chapter 12 in Karassik IJ,Pump Handbook4th edMcGraw-Hill Professional, 2008 2001 1986 1976AccessEngineering
12. Sutera SP, Skalak RThe history of Poiseuille’s Law.Annu Rev Fluid Mech199325120
13. Choi JW, Pai SHChanges in hematologic parameters induced by thermal treatment of human blood.Ann Clin Lab Sci200232393398
14. Gershfeld NL, Murayama MThermal instability of red blood cell membrane bilayers: Temperature dependence of hemolysis.J Membr Biol19881016772
15. Bennett VThe spectrin-actin junction of erythrocyte membrane skeletons.Biochim Biophys Acta1989988107121
16. Nagelkirk PR, Hogan KB, Hoare JMAmbient temperature affects thrombotic potential at rest and following exercise.Thromb Res2012130248252
17. Bouchama A, Bridey F, Hammami MM, et al.Activation of coagulation and fibrinolysis in heatstroke.Thromb Haemost199676909915
18. Goldstein DJ, John R, Salerno C, et al.Algorithm for the diagnosis and management of suspected pump thrombus.J Heart Lung Transplant201332667670

left ventricular assist device; heat; temperature; mock circulatory loop; in vitro; thrombus

Copyright © 2016 by the ASAIO