Prosthetic replacement of the elbow joint has been successful in a
variety of clinical situations, although it has not been put to the same test
as total knee replacement has in young, active patients with primary
osteoarthritis. The major concerns limiting its clinical application are
instability and
loosening1.
Many of the original elbow joint replacement designs were simple hinge
joints, whose intrinsically complete constraint of the articulation
predictably resulted in failure by
loosening2-8.
In retrospect, this was not surprising considering that the normal elbow does
not function as a simple
hinge9.
Subsequently, the development of elbow prostheses diverged into two general
types: loose-hinge (linked) and resurfacing (unlinked). The nomenclature and
descriptive terminology for total elbow replacements have become confusing
because of the use of the terms semiconstrained and unconstrained to imply
linked and unlinked designs,
respectively2-8,10-14.
It is important to distinguish between the physical attributes and the
biomechanical properties of the implants. The physical attributes of an
implant are described by the terms linked and unlinked as well as resurfacing
and nonresurfacing, whereas the biomechanical properties are described by the
terms constrained, semiconstrained, and unconstrained.
Unlinked joint replacements have the theoretical advantage of reduced
loosening rates due to lower bone-cement interface stresses relative to linked
joint
replacements11-13.
However, high loosening rates have been reported in association with some
unlinked implants, suggesting that the stresses across the articulation may
indeed be transferred to the prosthesis-cement-bone
interface14. Our
experience with the use of unlinked implants of different designs, under both
experimental and clinical conditions, has led us to conclude that the
"constraint" of a prosthesis is primarily related intrinsically to
its articular geometry. The articular geometries of commercially available
implants vary sufficiently to suggest that there is no consensus of opinion
regarding an optimal design. Furthermore, no biomechanical comparative data
are available to correlate the intrinsic constraint of different designs as
compared with the normal elbow.
We hypothesized that different designs of unlinked elbow implants have
different critical thresholds of displacement beyond which joint constraint is
no longer provided by the implant articulation. Furthermore, we postulated
that interface stresses, and hence implant loosening, are directly related to
intrinsic articular constraint in unlinked elbow implants. The primary purpose
of the present study was to compare the intrinsic constraints to displacement
of the ulnotrochlear components of unlinked elbow implants in a controlled
environment with reproducible mechanical tests and to ascribe design features
to constraint. A second purpose was to compare the intrinsic constraint of
prosthetic implants with that of the normal elbow.
Twelve fresh-frozen cadaveric elbows without macroscopic evidence of
osteoarthritis or instability were tested along with a single example of five
commercially available elbow replacements. The twelve elbows (including eight
right elbows and four left elbows) were obtained from nine female and three
male donors who had had a mean age of seventy-three years (range, sixty-three
to ninety-five years) at the time of death. The elbows were prepared by
transection of the humeral and ulnar shafts 8 cm from the articular margins.
All soft tissues were removed by sharp dissection, and the humeral and ulnar
shafts were embedded in metal fixtures with polymethyl-methacrylate
orthodontic resin (Coltène/Whaledent, Mahwah, New Jersey). The
alignment of the subcutaneous border of the ulna was maintained parallel to
the floor, as was the orientation of the prosthetic ulnar components. The
embedding of the implants in fixtures helped to facilitate rigid fixation and
alignment in the test apparatus. During cadaveric dissection and preparation,
saline solution was used to prevent specimen dehydration, but during testing,
the articular surfaces were lubricated with bovine calf serum (Sigma-Aldrich,
St. Louis, Missouri) and were maintained at room temperature.
The elbow implants that were studied included the Souter-Strathclyde
(Howmedica, Rutherford, New Jersey), Sorbie-Questor (Wright Medical
Technology, Arlington, Tennessee), Pritchard ERS (DePuy, Warsaw, Indiana),
Ewald (Johnson and Johnson Orthopedics, New Brunswick, New Jersey), and Kudo
type-5 (Howmedica) devices (Fig.
1). The prosthetic implants were the medium/standard size as
defined by each company, and the process that was used to embed the implants
into the fixtures was identical to that used for the cadaveric specimens. The
Ewald, Pritchard, and Sorbie-Questor prostheses were new implants, but the
Kudo and Souter-Strathclyde prostheses were explants that did not have any
visible damage to the metal or polyethylene surfaces.
A custom-made multiple-axis materials testing machine was used to conduct
all tests15. The
apparatus was supported by a rigid aluminum frame and incorporated axes for
three translations and one rotation. The ulnar component was rigidly fixed
onto a six-component load-cell, which in turn was mounted onto the x-y stage
(Fig. 2). The potted humeral
component was positioned at 90° to the ulnar component and was secured to
a vertically unconstrained sliding z axis. The x, y, and rotatory stages were
motorized with microstepping motors (OEM series; Parker Hannifin, Rohnert
Park, California). The x and y axes could be locked (allowing primary
motor-driven motion in the plane of the locked axis) or unlocked (allowing
passive secondary motion as dictated by the implant). The rotatory stage was
locked without the capacity of being unlocked. A six-component load-cell
(Advanced Mechanical Technology, Watertown, Massachusetts), with force and
torque sensitivities ranging from 6 to 24 µV/N, was mounted to the superior
surface of the x-y stage. Axial loads were applied in line with the humeral
component by the addition of dead weights through an unconstrained z-axis
sliding stage. Linear potentiometers were attached to the x, y, and z axes in
order to allow for translational displacement measurements in the designated
directions. Rotational displacement was calculated from the step count of the
microstepping motor attached to the rotational stage. LabVIEW software
(National Instruments, Austin, Texas) was utilized to control motion by means
of a motion-controller card (National Instruments). In addition to controlling
motion, the program collected and converted potentiometer voltage output data
into either centimeters or degrees and converted load-cell voltage output into
newtons or newton-meters, with all results being stored in a data file.
Each prosthesis underwent five repetitions of the same test protocol, as
described below, on different days, to ensure repeatability. Each day, the
implants were removed from the test apparatus and then were repositioned and
realigned for a subsequent test. Each cadaveric elbow underwent only a single
sequence of tests as there was pilot evidence, following a second sequence, of
chondral damage. Each test specimen was axially compressed sequentially with
10, 50, and 100-N static loads. The test protocol consisted of valgus
(external) and varus (internal) rotation of the ulnar component about a fixed
humeral component, oriented at 90° of flexion, that was unconstrained in
the z translational axis. The rate of rotation was kept constant at 5° per
second. Each motion was terminated either at 15° of rotation or when a
predetermined torque limit of 2 Nm was achieved.
This cold-flow behavior was manifested by gradual reductions in the maximum
achieved torque values with repetitive testing, which would have biased the
interpretation of these results. This necessitated the artificial upper torque
limit of 2 Nm to be instigated, which we recognize departs from the in vivo
situation, in which forces can reach as high as 3
kN16. However, our
study aimed to determine the geometric influence of the perfect implants, not
specifically their wear characteristics relative to their behavior, as would
occur in vivo. At the end of formal testing, no plastic deformation was
observed, indicating that the test results were related to the geometry and
were not influenced by deformation. The range of testing encompassed phases of
subluxation and, in some cases, frank dislocation of the components.
The measurements that were recorded were rotational displacement and torque
in the rotatory axis (in degrees and newton-meters, respectively) and axial
translation (in millimeters) along the z axis. Force-displacement curves were
plotted with each change in test parameter, e.g., axial compressive load.
Further analysis was conducted for maximum achieved torque, the displacement
at this torque, and the peak constraint ratio. The maximum achieved torque was
defined as the greatest torque that was registered by the torque cell within
the defined range of motion. The peak constraint ratio was defined as the
maximum achieved torque divided by the angular displacement at that
torque.
Statistical analysis was performed with use of a one-factor analysis of
variance for each individual measurement (torque, angle, and constraint ratio)
at each of the axial compressive loads (10, 50, and 100 N). A Student t test
was used as a post hoc test, with the level of significance set at p <
0.05.
Axial Compression
The effect of increasing axial load on the human elbow was
threefold; specifically, increased axial load was associated with increased
maximum achieved torque, smaller angular displacement at maximum achieved
torque, and decreased relative variability
(Fig. 3 and Tables
I and
II). Torque increased from 1.2
Nm (at 10 N of axial load) to 1.83 Nm (at 50 N of axial load) and reached the
2 Nm torque limit at 100 N of axial load. However, the angular displacement at
this peak torque decreased from 6.9° (at 10 N of axial load) to 6.0°
(at 50 N of axial load) to 4.8° (at 100 N of axial load).
The normal elbow demonstrated an increase in the maximum achieved torque
with increased axial applied load, with a concurrent decrease in rotational
displacement. Both the Souter-Strathclyde and Kudo implants approximated this
behavioral pattern in valgus displacement, whereas the Ewald implant
demonstrated an increased maximum achieved torque but maintained an equivocal
displacement pattern and the Sorbie and Pritchard implants demonstrated
increased maximum achieved torque and displacement. In varus rotation, the
normal elbow and the Kudo, Souter-Strathclyde, and Sorbie implants all
demonstrated increased maximum achieved torque and concurrently decreased
displacement with increasing axial applied load; however, the Ewald implant
demonstrated increased maximum achieved torque and overall increased
displacement and the Pritchard implant demonstrated increased maximum achieved
torque and increased displacement.
Peak Constraint Ratio
With valgus displacement, the peak constraint ratio for the normal elbow
changed from 0.19 Nm/deg (at 10 N of axial load) to 0.44 Nm/deg (at 50 N of
axial load) to 0.57 Nm/deg (at 100 N of axial load)
(Table III and
Fig. 4). The Souter-Strathclyde
implant became stiffer with increasing axial load, with a peak constraint
ratio of 0.21 Nm/deg at 10 N of axial load and 0.74 Nm/deg at 100 N of axial
load. The Kudo implant demonstrated a peak constraint ratio of 0.12 Nm/deg at
10 N of axial load, 0.23 Nm/deg at 50 N of axial load, and 0.40 Nm/deg at 100
N of axial load. With varus displacement, the peak constraint ratio for the
normal elbow changed from 0.17 Nm/deg at 10 N of axial load to 0.6 Nm/deg at
50 N of axial load to 1.01 Nm/deg at 100 N of axial load
(Table III and
Fig. 4). The peak constraint
ratio for the Souter-Strathclyde implant increased from 0.19 Nm/deg at 10 N of
axial load to 0.82 Nm/deg at 100 N of axial load. The Kudo implant
demonstrated a peak constraint ratio of 0.29 Nm/deg at 10 N of axial load,
0.52 Nm/deg at 50 N of axial load, and 1.18 Nm/deg at 100 N of axial load. The
Ewald, Pritchard, and Sorbie implants had lower values for the peak constraint
ratio, but these values were not uniformly significantly less than those for
the human elbow or the Souter-Strathclyde and Kudo implants in both valgus and
varus rotation (Table III).
Angular Displacement, Torque, and Axial Distraction
No implant was similar to the human elbow with regard to varus-valgus
rotation and torque (Fig. 5).
The human ulnotrochlear joint generated higher torque values at lesser
rotational displacement values up to 3° of valgus and varus rotation than
did most but not all of the prosthetic implants tested. This behavior appeared
to be asymmetrical. Although there were variations between specimens, the
asymmetrical behavior was consistent within a specimen. The human elbow was
stiffer in varus rotation (with the torque limit being achieved at 2.8°
± 1.6°) than in valgus rotation (with the maximum torque being
achieved at 4.8° ± 2°). The Souter-Strathclyde implant was
stiffer in valgus rotation (with the torque limit being achieved at 2.8°
± 0.5°) than in varus rotation (with the maximum torque of 1.75 Nm
being achieved at 2.7° ± 0.8°). The Kudo implant was
asymmetrical in both directions, with the maximum torque being achieved at
5.1° ± 0.3° in valgus and 2° ± 0.5° in varus.
The Ewald, Pritchard, and Sorbie ulnotrochlear joints were dissimilar in
quality and magnitude to the human joint.
Axial Distraction
The components of the Kudo implant distracted from their neutral position
in valgus rotation more than those of any other implant did (average
distraction for the Kudo implant, 1.7 mm)
(Fig. 6). The maximum torque
for this implant was achieved at 0.9 mm of distraction. This maximum achieved
torque was approximately constant between 0.9 and 1.45 mm of axial
distraction. Only the Pritchard implant maintained its maximum achieved torque
throughout a greater range of axial distraction (0.2 to 1 mm). In contrast,
both the human elbow and the Souter-Strathclyde implant axially distracted
0.57 and 0.65 mm, respectively, before the torque limit was reached.
Axial Distraction and Torque
The human specimen remained relatively symmetrical in varus rotation
relative to its behavior in valgus rotation and reached the torque limit at
0.49 mm of axial distraction. The Kudo prosthesis also was symmetrical in
varus rotation relative to its behavior in valgus rotation and was the implant
that demonstrated the most axial distraction, with the maximum torque being
achieved at 0.55 mm of axial distraction. However, this maximum achieved
torque was maintained over a much narrower range of axial distraction (from
0.55 to 0.65 mm). The Souter-Strathclyde implant revealed a marked difference
in its behavior in varus rotation as compared with valgus rotation. In varus
rotation, it did not reach to the torque limit and the maximum torque was
achieved at 0.88 mm of axial distraction.
The observations documented in this study have potentially important
clinical implications with regard to mechanical loosening of total elbow
arthroplasty components. The most obvious finding is that there is great
variation in the degree of constraint inherent in the design of unlinked elbow
prostheses. This variation is relatively apparent from a visual assessment of
the articular geometries, which display varying degrees of conformity between
the ulnar and humeral components. In fact, the Souter-Strathclyde implant has
a highly congruous articular geometry, which increased the constraint of the
articulation beyond that of the normal elbow. In a previous kinematic study,
this implant was shown to function essentially in a relatively highly
constrained
manner17.
Joint constraint is a function of the congruity of the articular geometry
and the surrounding ligaments and muscles. Of the implants tested, the Ewald,
Pritchard ERS, and Sorbie-Questor prostheses all behaved in a manner that was
less constrained than that of the normal elbow. King et al. documented a
similar effect for the Ewald prosthesis, which therefore has a greater
reliance on the surrounding soft-tissue structures for its
stability18,19.
It should also be noted that the Sorbie-Questor and Pritchard ERS implants
have been designed to be used with radial head components, which logically
would alter the constraint behavior in valgus but not in varus. However, the
current study did not investigate the effect of the radial components.
In accordance with King's earlier study of the Ewald
prosthesis18, all
of our specimens, normal and prosthetic, behaved in a more constrained manner
with increasing axial load. Greater constraint can occur by increasing surface
contact area between
components20 as
well as the phenomenon that Harryman et al. referred to as
concavity-compression21.
Care should be exercised before assuming that clinical problems of
instability will be predictable on the basis of the degree of constraint. In
the present study, the most constrained of the implants was the
Souter-Strathclyde device; however, the clinical results that have been
reported for this prosthesis have demonstrated problems not only with
loosening but also with
instability22,23.
Schneeberger et al. studied the kinematics of the Souter-Strathclyde
prosthesis in cadaveric elbows and found that, because of its highly
conforming surfaces, even minor errors in component positioning caused
subluxation if the elbow was moved through the full original range of
motion17.
Beyond the goals of prosthetic implantation, many implant-related
complications have been reported, including subluxation/dislocation, bearing
surface wear, and component
loosening1,24.
Subluxations and dislocations are in part due to implant design, with our
results suggesting that the Kudo and Souter-Strathclyde implants most closely
reproduce the normal elbow constraint behavior in rotation. Conversely, the
Ewald, Pritchard ERS, and Sorbie-Questor implants have a lower inherent
resistance to rotational displacement, suggesting a greater propensity for
subluxation and dislocation. The latter two implants demonstrated constraint
ratios of approximately one-quarter that of the normal elbow in internal
rotation and 47% and 56%, respectively, of that of the human elbow in valgus
rotation, with axial loads of 100 N. The inclusion of a radial head component
may improve their constraint function in valgus rotation but should not impact
their constraint in varus rotation. Therefore, our observations of these
implants in internal rotation seem to be valid.
The present study had several limitations. Axial translation of the humeral
component was left unconstrained during testing. Axial translation in vivo is
limited by soft
tissue25. It is
reasonable to assume that beyond the point of tissue laxity, the prosthesis
behaves in a more constrained manner. However, our attempts to simulate this
situation experimentally were complicated by plastic deformation of the
polyethylene. Another limitation was the fact that axial rotation of the ulna
was constrained, thus preventing coupled motion patterns that are typical of
clinical instability patterns. It is probable that axial rotation has an
influence on joint constraint and should be borne in mind when interpreting
these findings.
Clinically, our findings should be interpreted with caution. Implants that
are physically unlinked do not necessarily demonstrate unconstrained
behavior17, and
these two terms should not be used interchangeably. More extensive studies
relating articular geometry, kinematic performance, and the material
properties of the individual implants will be required to precisely predict
the in vitro constraint behavior as a function of the in vitro characteristics
demonstrated in this study. An even more important task will be to transfer
our understanding of in vitro behavior for the purposes of predicting in vivo
function and implant survival. Our study provides data with which to gauge the
relative intrinsic constraint of various designs of unlinked ulnohumeral
replacement implants.
In conclusion, there is wide variation in the geometric designs and the in
vitro biomechanical behavior of the five unlinked prosthetic ulnohumeral
replacements that we studied. Implants that are made to resemble the human
elbow in appearance, notably the Souter-Strathclyde and Sorbie-Questor
prostheses, do not replicate normal behavior consistently, whereas other
implants that do not resemble the human elbow closely, such as the Kudo
prosthesis, do not deviate from human behavior markedly. Hence, it appears
that a considerable amount of basic information about elbow form and function
is needed in order for us to understand and improve the performance of total
elbow replacements. ?