Artificial Organs

25(5):386–391, Blackwell Science, Inc.

© 2001 International Society for Artificial Organs

The HeartMate III: Design and In Vivo Studies of a

Maglev Centrifugal Left Ventricular Assist Device

*Howard M. Loree II, *Kevin Bourque, *David B. Gernes, *J. Scott Richardson,
*Victor L. Poirier, †Natale Barletta, †Andreas Fleischli, †Giampiero Foiera,
†Thomas M. Gempp, †Reto Schoeb, ‡Kenneth N. Litwak, ‡Takehide Akimoto,
‡Marina Kameneva, ‡Mary J. Watach, and ‡Philip Litwak
*Research Dept., Thermo Cardiosystems Inc. (now Thoratec Corp.), Woburn, Massachusetts, U.S.A.; †Sulzer Markets and
Technology, (now Levitronix GmbH), Zürich, Switzerland; and ‡McGowan Center for Artificial Organ Development,
University of Pittsburgh, Pittsburgh, Pennsylvania, U.S.A.

Abstract: A compact implantable centrifugal left ventricu- (2.5 ⱕ INR ⱕ 4.0) throughout the studies. Six studies were
lar assist device (LVAD) (HeartMate III) featuring a mag- electively terminated (at 27–61 days), 1 study was termi-
netically levitated impeller is under development. The goal nated after the development of severe pneumonia and
of our ongoing work is to demonstrate feasibility, low he- lung atelectasis (at 27 days) another study was terminated
molysis, and low thrombogenicity of the titanium pump in after cardiac arrest (at 2 days) while a final study is on-
chronic bovine in vivo studies. The LVAD is based on going (at ≈100 days). Mean pump flows ranged from 2 to
so-called bearingless motor technology and combines 7 L/min, except for brief periods of exercise at 6 to 9
pump rotor, drive, and magnetic bearing functions in a L/min. Plasma free hemoglobin ranged from 4 to 10 mg/dl.
single unit. The impeller is rotated (␪z) and levitated with All measured biochemical indicators of end organ function
both active (X, Y) and passive (Z, ␪x, ␪y) suspension. Six remained within normal range. The pumps have met per-
prototype systems have been built featuring an implant- formance requirements in all 9 implants with acceptable
able titanium pump (69 mm diameter, 30 mm height) with hemolysis and no mechanical failures. Key Words: Left
textured blood contacting surfaces and extracorporeal ventricular assist device—Rotary blood pump—Magnetic
electronics. The pumps were implanted in 9 calves (ⱕ100 suspension—Centrifugal pump—In vivo studies.
kg at implant) that were anticoagulated with Coumadin

Thermo Cardiosystems Inc. (TCI), in a research and control) below 10 W should be achievable under
partnership with the Swiss firm Sulzer AG, is devel- nominal operating conditions, in league with both
oping a compact implantable centrifugal left ven- HeartMates I and II.
tricular assist device (LVAD) (HeartMate III) fea- The HeartMate III represents a new, third-
turing a bearingless motor incorporating a generation LVAD, after the reciprocating (pusher-
magnetically levitated impeller. This device com- plate) HeartMate I, now in full production and clini-
bines the functions of impeller rotation and levita- cal use and the continuous axial flow HeartMate II,
tion in a single magnetic structure. The result is a now in clinical trials (Fig. 1). The distinctions among
pump that fits in an acceptably small package and the 3 generations have to do with 2 important pa-
can operate completely sheathed in titanium. Total rameters: size and longevity (Table 1).
power consumption (including rotation, levitation, The HeartMate I, although a relatively large de-
vice, is ideally matched in function to the natural
Received January 2001. heart and is therefore comparatively easy to adapt to
Presented in part at the 8th Congress of the International So-
ciety for Rotary Blood Pumps, held September 6–9, 2000, in the circulatory needs of the body, using the simplest
Aachen, Germany. of control systems (1,2). Because it includes several
Address correspondence and reprint requests to Dr. Howard moving parts and conventional ball bearings, how-
M. Loree II, Senior Scientist, Research & Development Dept.,
470 Wildwood Street, Woburn, MA 01888-2697. E-mail: ever, its estimated life in vivo is 2 to 3 years.
[email protected] The HeartMate II, the smallest of all TCI blood

386
 HEARTMATE III: DESIGN AND IN VIVO STUDIES 387

the volume of the HeartMate II. Being a centrifugal

pump, it is disk-shaped rather than cylindrical (as in
the case of the HeartMate II) and has the appear-
ance of a miniature HeartMate I.

MATERIALS AND METHODS

The HeartMate III LVAD is based on magnetic

technology that combines drive, magnetic bearing,
and pump rotor functions in a single unit (15). Three
spatial degrees of freedom are stabilized passively
and the remaining 3 by active control. Active control
of the rotational (␪z) and the radial (X and Y) posi-
tion of the rotor is provided by feedback electronics,
implemented by appropriately placed sensors. Axial
position (Z) and tilting (␪x and ␪y) are stabilized by
passive magnetic forces.
Six prototype systems were built featuring an im-
plantable titanium pump (69 mm diameter, 30 mm
height) with textured blood contacting surfaces and
extracorporeal electronics. An exploded view of the
pump is provided in Fig. 2. The simple centrifugal
pump housing is made of an upper housing contain-
ing inflow and outflow channels and a lower housing
containing the motor, which are held together by a
FIG. 1. A comparison of the HeartMate I, II, and III LVAD geom- screw ring. The rotor includes a permanent magnet
etry (bottom to top) is shown. structure and a shrouded titanium impeller with
blades of a number, size, and shape found to be most
pumps, incorporates a set of precision jeweled bear- hydrodynamically efficient near the design point. A
ings and a high-speed impeller that maintain the flexible cannula is fitted upstream of the pump in-
same blood pumping capacity as the HeartMate I, flow to improve anatomical fit, to decouple the
but in a package approximately one-eighth its vol- pump from cardiac motion, and to provide a point
ume (3,4). It could prove invaluable in providing car- for clamping and ligation for pump explant. An out-
diac assistance to an entire spectrum of patients,
from very young children to large adults. Animal
tests suggest that our operational goal of 5 years can
be met (3). Control of the pump flow is by an auto-
responsive algorithm based on various motor param-
eters.
The ultimate service life goal for LVADs intended
for alternative-to-transplant use is 10 to 15 years.
This may not be feasible with mechanical, blood im-
mersed bearings. Rotating mechanical seals are not
feasible for long-term applications either. Therefore,
we (5) and several other groups (6–14) are develop-
ing rotary blood pumps with magnetically levitated
rotors.
The HeartMate III is the result of an intense
search for a miniature low-cost blood pump design
that could be used, in similar fashion to the Heart-
Mate II, to provide cardiac assistance from child-
hood through old age. The HeartMate III implant-
able pump occupies approximately one-third the
volume of the HeartMate I and just less than 3 times FIG. 2. The exploded view is of the HeartMate III LVAD.

Artif Organs, Vol. 25, No. 5, 2001

388 H.M. LOREE II ET AL.

flow cannula includes a woven polyester graft for (Model SPR524, Millar Instruments, Houston, TX,
aortic anastomosis, protected from collapse by a U.S.A.). In 5 other animals, a thermistor tempera-
bend-relief. Self-locking screw rings and polytetra- ture probe (Model 08442-15 probe with Model
fluorethylene washers are used at the joints between P-08502-14 meter, Cole-Parmer Instrument Co.,
the cannulas and the pump. An electronic cable (not Vernon Hills, IL, U.S.A.) was glued to the outer
shown) is exteriorized percutaneously to connect to surface of the pump. In 3 of these 5 animals, an
the shelf-mounted power and control electronics. identical temperature probe also was implanted in
Hydrodynamic characterization of 1 pump was the mediastinum.
performed in vitro using a steady flow mock circu- The pump was preset to run at 3,500 rpm until a
latory loop with water as the working fluid. The flow signal was received and then gradually in-
head-flow characteristic was determined from mea- creased to provide a pump flow of either 5 ± 1 or 6
surements using an ultrasonic flowmeter (H9XL ± 1 L/min thereafter. At approximately 30 day inter-
probe, T110 meter, Transonic Systems, Ithaca, NY, vals and at the end of the study, most animals had 10
U.S.A.) and 2 catheter-type pressure transducers to 15 min treadmill workouts, inducing pump flows
(CDX-III, Cobe Cardiovascular, Arvada, CO, greater than 6 L/min and allowing study of the he-
U.S.A.). The hydraulic efficiency also was deter- modynamic interactions between the pump and the
mined from these measurements using additional native heart.
motor torque and speed measurements performed
by proprietary motor controller software. Analog RESULTS
outputs of the previously mentioned values were col-
lected and analyzed using a computer data acquisi- In vitro testing has demonstrated the desired
tion system (LabVIEW 5, National Instruments, steady-state flow and pressure characteristics (the
Austin, TX, U.S.A.). design point 7 L/min at 135 mm Hg achieved at 4,800
Separate in vitro studies characterized the power rpm) and hydraulic efficiency (30% at the design
consumption of the bearingless motor as well as the
magnetic stiffness along both active and passive axes
of levitation. Instrumentation included a single
phase spectral power converter (PC10, Infratek AG,
Uetikon am See, Switzerland) and custom-made test
fixtures.
The 6 pumps were implanted sequentially in 9
calves (ⱕ100 kg at implant). The pumps were at-
tached to the circulation in a left ventricle to de-
scending thoracic aorta bypass, with placement in
the sulcus between the diaphragm and the rib cage.
The implants were done without cardiopulmonary
bypass. Anticoagulation was by heparin periopera-
tively (activated clotting time 250 s) and then by oral
Coumadin (2.5 ⱕ INR ⱕ 4.0) for the duration of the
study. Blood chemistry, coagulation properties, and
cell count were measured once or twice each week.
All in vivo studies were performed in accordance
with the Guide for the Care and Use of Laboratory
Animals (NIH publication No. 86–23, Revised 1985)
and the guidelines determined by the Institutional
Animal Care and Use Committee of the University
of Pittsburgh.
Instrumentation included a flocked, fluid-filled
pressure monitoring line in the carotid artery and
ultrasonic flow probes around the pump outflow
graft and pulmonary artery (Models 16SB and 24A
probes with Model T206 meter, Transonic Systems).
FIG. 3. The graphs show head versus flow characteristics (A)
In 2 animals, solid-state pressure probes were placed and hydraulic efficiency versus flow curves (B) for the HeartMate
in the left ventricle and the pump outflow graft III LVAD.

Artif Organs, Vol. 25, No. 5, 2001

 HEARTMATE III: DESIGN AND IN VIVO STUDIES 389

point in water) as illustrated in Figs. 3A and 3B. The TABLE 1. Comparison of the HeartMate I, II, and III
nominal maximum flow (10 L/min at 135 mm Hg) is left ventricular assist device volumes and masses
delivered at just over 5,000 rpm when hydraulic ef- Priming volume Displaced volume Mass
ficiency remains in the 27% to 30% range. Pump (ml)a (ml)a (g)a
The power consumption of the motor and the HeartMate I (VE) 153 511 1,050
bearing at the design point (not including control HeartMate II 7 63 281
HeartMate III 25 168 474
electronics) was on average 8.1 W. In addition, the
magnetic bearings can handle forces of up to 10.8 N a
Excluding inflow and outflow subassemblies.
radially and 7.9 N axially. With a rotor weight of only
34.9 g, a very robust system results. flow had been deliberately induced in a brief study
The pumps were implanted in 9 calves as illus- just prior to sacrifice, very fresh red thrombus was
trated in Fig. 4 and described in Table 2. Six studies found under the rotor and in the volute. In the first
were electively terminated (at 27–61 days), 1 study 6 studies, rings of white thrombus partially occluded
was terminated after the development of untreatable the joint between the pump and the outflow graft.
pneumonia secondary to lung atelectasis (at 27 After design and procedural changes, the 2 subse-
days), another study was terminated after cardiac quent pump explants have shown no such occlusion
arrest (at 2 days), and a final study is ongoing (at at this joint.
≈100 days). Mean pump flows ranged from 2 to 7
L/min except for brief periods of exercise at 6 to 9 DISCUSSION
L/min. Plasma free hemoglobin ranged from 4 to 10 A fundamental goal of the HeartMate III me-
mg/dl. In appropriately instrumented animals, the chanical design was to achieve optimal pumping per-
pump housing exterior temperature ranged from formance and extended life in the simplest possible
39.5 to 41.5°C, about 1 to 1.5°C higher than the sur- way. For example, it was important to maintain an
rounding mediastinal tissue. All measured biochemi- ample and straightforward flow path to minimize he-
cal indicators of end organ function remained within molysis and thrombus formation. Based on earlier
normal range. The pumps have met performance re- work designing clinical LVADs, it was considered
quirements in all 9 implants with no mechanical fail- advantageous to provide textured blood contacting
ures. surfaces within the pump, a flexible inflow section,
The pump linings upon explant appeared similar and a bend relief for the outflow graft.
to HeartMate I bovine implants. Figure 5 shows the Pump flows were maintained in the target range
lower pump housing from a pump explanted from for all studies except Study 7. This animal experi-
Study 3 after 59 days. The structured titanium sur- enced a kink in the outflow graft due to malposition
face is completely covered with a thin well-adherent of the pump flow probe. The outflow stenosis seen in
biological lining. Several pumps showed small well- the first 6 studies has been addressed with minor
adherent islands of fibrous tissue within the volute modifications to the outflow joint and the implant
and outflow port. In Study 4, in which retrograde procedure, which are being tested in ongoing studies.
The performance of the textured surfaces appears
favorable, and evaluation is ongoing. Fresh throm-
bus found in 1 pump (Study 4) under the rotor and in
the volute was probably caused by stagnant flow in-
duced in a brief study just prior to sacrifice. Plasma
free hemoglobin values that did not depart from
baseline in all 9 in vivo studies demonstrated that the
HeartMate III hemolysis is of no clinical importance.
Thermal dissipation appears well managed. Tem-
perature of the pump exterior was maintained less
than 42°C, the accepted maximum for the device-to-
tissue interface (16). This represents a temperature
rise of less than 2°C.
The pumps used in the first set of animal studies
had percutaneous leads that exited the pump directly
beneath the outflow (visible in Fig. 4). Although this
FIG. 4. The photograph shows the HeartMate III LVAD ready for is an ideal orientation for the human configuration, it
implant. was not effective for the bovine model. In order to

Artif Organs, Vol. 25, No. 5, 2001

390 H.M. LOREE II ET AL.

TABLE 2. Summary of cumulative HeartMate III left ventricular assist device in vivo study data
Study Plasma free Pump Pump
Implant duration Pump flow hemoglobin temperature temperature
number (days) (L/min)a (mg/dl)a (°C)a rise (°C)a Notes
1 40 5.3 ± 0.6 7.8 ± 1.6 — — Electively terminated
2 27 4.3 ± 1.1 5.8 ± 1.3 40.9 ± 0.5 — Electively terminated
3 59 4.0 ± 0.5 5.6 ± 2.4 40.3 ± 0.5 — Electively terminated
4 42 4.0 ± 0.6 5.3 ± 2.4 40.1 ± 0.3 1.4 ± 0.2 Electively terminated
5 27 5.7 ± 1.1 5.3 ± 4.7 40.5 ± 0.4 1.2 ± 0.3 Terminated due to pneumonia
6 49 5.4 ± 0.5 5.1 ± 1.9 40.3 ± 0.8 1.2 ± 0.4 Electively terminated
7 61 3.2 ± 1.1 6.7 ± 2.4 — — Electively terminated, kinked outflow graft
8 ∼100 5.5 ± 1.5 6.6 ± 2.5 — — Ongoing
9 2 — — — — Terminated due to cardiac arrest
a
Mean ± SD.

exteriorize at the dorsum, the lead was forced into a no mechanical failures. Current design efforts are
half-loop in the thorax that occupied a volume nor- focused on a fully sealed, single-fault-tolerant in-
mally occupied by the left lung. As a result, there plantable subsystem. In this version, the microelec-
was some degree of atelectasis in the first 6 animals, tronic control module is entirely contained within
and 1 animal had to be sacrificed due to untreatable the titanium LVAD housing.
pneumonia secondary to atelectasis (Study 5). The
orientation of the percutaneous lead was rotated by
75 degrees (visible in Fig. 1), and the problem has REFERENCES
been averted in all subsequent animal implants.
1. Goldstein DJ, Oz MC, Rose EA. Implantable left ventricular
assist devices. N Engl J Med 1998;339:1522–33.
CONCLUSIONS 2. Morales DL, Catanese KA, Helman DN, Williams MR,
Weinberg A, Goldstein DJ, Rose EA, Oz MC. Six-year ex-
The HeartMate III development effort has estab- perience of caring for forty-four patients with a left ventricu-
lished the feasibility of the Sulzer bearingless motor lar assist device at home: Safe, economical, necessary. J Tho-
rac Cardiovasc Surg 2000;119:251–9.
for use in an LVAD. Ongoing in vitro studies suggest
3. Butler KC, Dow JJ, Litwak P, Kormos RL, Borovetz HS.
that the desired hydrodynamic design point (a mean Development of the Nimbus/University of Pittsburgh innova-
flow of 7 L/min at a mean head of 135 mm Hg) tive ventricular assist system. Ann Thorac Surg 1999;68:790–4.
should be achievable with a power consumption (in- 4. Siegel-Itzkovich J. Israeli surgeons implant first permanent
cluding rotation, levitation, and control) below 10 W. artificial ventricle. BMJ 2000;321(7258):399.
5. Bourque K, Gernes DB, Loree HM, Richardson JS, Poirier
In vivo studies are well under way, with 7 bovine VL, Barletta N, Fleischli A, Foiera G, Gempp TM, Schoeb R,
implants of 27 to 61 days completed and another Litwak KN, Akimoto T, Watach MJ, Litwak P. HeartMate
ongoing at ≈100 days. The pumps have met perfor- III: Pump design for a centrifugal LVAD with a magnetically-
levitated rotor. ASAIO J (in press).
mance requirements with acceptable hemolysis and
6. Bearnson GB, Olsen DB, Khanwilkar PS, Long JW, Sinnott
M, Kumar A, Allaire PE, Baloh M, Decker J. Implantable
centrifugal pump with hybrid magnetic bearings. ASAIO J
1998;44:M733–6.
7. Allaire P, Decker P, Baloh M, Lee J, Long J. Magnetic sus-
pension of an artificial heart pump using EM/PM bearings.
Proceedings of the 5th International Symposium on Magnetic
Suspension Technology, Santa Barbara, CA, December 1999.
8. Paden B, Antaki J, Groom N. Animal trials of a magnetically
levitated left ventricular assist device. Proceedings of the 5th
International Symposium on Magnetic Suspension Technol-
ogy, Santa Barbara, CA, December 1999.
9. Antaki JF, Paden B, Burgreen G, Groom N. Magnetically
suspended miniature fluid pump and method of designing the
same. U.S. Patent 6,015,272, Jan. 18, 2000.
10. Nojiri C, Kijima T, Maekawa J, Horiuchi K, Kido T, Sugiyama
T, Mori T, Sugiura N, Asada T, Shimane H, Ozaki T, Suzuki
M, Akamatsu T, Akutsu T. Development of Terumo implant-
able left ventricular assist system. Journal of Congestive Heart
Failure and Circulatory Support 1999;1:73–8.
FIG. 5. The interior of the HeartMate III LVAD in Study 3 at 59 11. Nojiri C, Kijima T, Maekawa J, Horiuchi K, Kido T, Sugiyama
days is shown. T, Mori T, Sugiura N, Asada T, Ozaki T, Suzuki M, Akamatsu

Artif Organs, Vol. 25, No. 5, 2001

 HEARTMATE III: DESIGN AND IN VIVO STUDIES 391

T, Akutsu T. Terumo implantable left ventricular assist sys- 14. Wampler R, Lancisi D, Indravudh V, Gauthier R, Fine R. A
tem: Results of long-term animal study. ASAIO J 2000;46: sealless centrifugal blood pump with passive magnetic and
117–22. hydrodynamic bearings. Artif Organs 1999;23:780–4.
12. Okada Y, Ueno S, Ohishi T, Yamane T, Tan CC. Magneti- 15. Schöb R, Barletta N. Principle and application of a bearing-
cally levitated motor for rotary blood pumps. Artif Organs less slice motor. JSME Int J Ser C 1997;40:593–8.
1997;21:739–45. 16. Norman JC, Huffman FN. Annual Technical Progress Report:
13. Wampler R. Sealless rotary blood pump with passive mag- Study of the Effects of Additional Endogenous Heat.
netic radial bearings and blood immersed axial bearings. U.S. Waltham, MA: Thermo Electron Corp., NIH Report TE4064-
Patent 5,695,471, Dec. 9, 1997. 51-71, July 1970.

Artif Organs, Vol. 25, No. 5, 2001