Malpositioning of total knee replacements leads to an increased risk
of loosening, instability, and
pain1-5.
Computer-assisted navigation systems are designed to increase the precision of
implantation. Several studies have demonstrated this to be the case for the
mechanical
axis6-12,
which was measured on full-length standing radiographs of the lower extremity.
Limitations to this method are imaging errors caused by images that are not
strictly frontal or sagittal, extension deficits, and rotation between the
femur and the
tibia13-15.
Therefore, the spatial positioning of an arthroplasty component can only be
precisely described with use of a three-dimensional imaging procedure,
separately evaluating the position of the tibial component in reference to the
tibia (from the knee to the ankle center) and the position of the femoral
component in reference to the femur (from the hip to the knee center).
Apart from the mechanical axis, the rotational alignments of the femoral
and tibial components are important for the functional outcome. Even small
deviations have a considerable influence on patellar tracking, stability, and
the overall biomechanics of the
joint16-20.
It has been unclear whether rotational alignment can be improved with the use
of navigation systems. Although there is no consensus about the best landmark
for adjusting femoral
rotation21,
alignment according to the surgical epicondylar axis would appear to come
closest to allowing natural
biomechanics22-24.
The main problem is the considerable interobserver and intraobserver
variability in the determination of the femoral
epicondyles25-27.
Therefore, in addition to the epicondylar axis, modern computer-assisted
navigation systems provide the option of using the posterior condyles, the
Whiteside line28,
and the symmetry of the flexion gap for determining the rotation of the
femoral component.
Whether computer-assisted navigation leads to improved implant positioning
can only be determined through postoperative three-dimensional images of the
knee, on which the epicondylar axis can be identified with a high degree of
accuracy. Depending on the prominence of the epicondyles, there is
considerable inaccuracy in determining the epicondylar axis. The error of
tactile identification of the epicondyles is excluded by postoperative
analysis with computed tomography.
The present study was designed to establish whether imageless
computer-assisted navigation can be used to improve the precision of implant
positioning with regard to the mechanical axis, determined separately for the
tibia and femur, and with regard to component rotation compared with
conventional surgical implantation.
Study Subjects
Sixty patients (twenty men and forty women) were included in this
prospective study. All patients provided informed consent to participate in
the study, which was approved by our institutional review board. The inclusion
criterion was the presence of primary arthritis of the knee. Patients who had
undergone previous surgery on the joint or who could not be treated with an
unconstrained total knee arthroplasty with a short stem were excluded. A
random-number generator was used to assign patients to navigated knee
arthroplasty (thirty-two knees) or conventional arthroplasty (twenty-eight
knees). Preoperatively, the age and body mass index of the patients were
documented. At six months postoperatively, the range of motion was measured
and the Knee Society score was
calculated29.
Operative Procedure
In all patients, an unconstrained total knee replacement with a
Tricompartmental Plus Solution-type rotating platform (TC-PLUS; Endoplus,
Marl, Germany) was implanted with cement. The posterior cruciate ligament was
resected in all knees. No patella was resurfaced, and no lateral release was
performed. For the conventional implantation method, femoral intramedullary
and tibial extramedullary alignment guides were used. Computer-assisted
implantation was performed with the PiGalileo System (Plus Orthopedics AG,
Rotkreuz, Switzerland). This is an imageless navigation system with optical
pointer and tracker identification.
The operations were performed with use of a tourniquet after a single
injection of antibiotics (2 g of ampicillin and 1 g of sulbactam). After a
medial parapatellar approach, the femoral and tibial reference markers were
fixed and the anatomical landmarks were registered for the computer-assisted
implantation. For both the navigated and the conventional method, the
implantation technique included preparation of the femur first.
The sagittal alignment of the femoral component was referenced to the
distal femoral canal with the use of the intramedullary guide in the
conventional implantation. In the computer-assisted technique, the femoral
component was referenced parallel to the distal femoral anterior cortical bone
by locating the femoral reference marker there and the tibial reference marker
distal and medial to the tuberosity.
In both the navigated and the conventional knee arthroplasties, the
rotation of the femoral component was adjusted parallel to the epicondylar
axis. In the conventional knee arthroplasties, the epicondylar axis was marked
with a cautery and the epicondyles were palpated simultaneously. In the
computer-assisted technique, the epicondylar axis was determined by
consecutively palpating the epicondyles. The medial epicondyle was defined as
the sulcus between the attachments of the medial collateral ligament, and the
lateral epicondyle was defined as the attachment of the lateral collateral
ligament. In both the navigated and conventional implantation techniques, the
rotation of the tibial component was adjusted parallel to the axis between the
medial third of the tuberosity and the center of the tibial plateau.
In knees with a preoperative slope of >5°, the goal was to achieve a
postoperative slope of 5°. In knees with an anterior slope, the goal was a
postoperative slope of 0°, and, in all other knees, the goal was the
degree of preoperative slope.
No intraoperative radiographs were made in order to verify prosthesis
alignment. The operative duration from skin incision to the placement of the
last suture was documented. The perioperative blood loss was calculated by the
addition of the intraoperative suction and the postoperative drainage
volumes.
Evaluation of Alignment and Rotation
Three-dimensional implant alignment was determined by computed tomography.
In accordance with a standard protocol, we acquired three scans of the
affected limb: the femoral head (a 50-mm scan with a 4-mm slice thickness),
the center of the ankle (a 50-mm scan with a 4-mm slice thickness), and the
knee joint (a 200-mm scan with a 1-mm slice thickness).
The following anatomical landmarks were determined in absolute spatial
coordinates (x, y, and z) with ImageJ (a publicdomain software program of the
National Institutes of Health): the center of the femoral head in the computed
tomography slice with the greatest diameter
(Fig. 1, A); the
lowest point of the sulcus between the attachments of the superficial and deep
medial collateral ligaments (Fig. 1,
B); the highest point of the lateral epicondyle
(Fig. 1, B); the
center of the knee joint, which was considered the geometric center of the
polyethylene inlay (Fig. 1,
D); the medial third of the tibial tuberosity
(Fig. 1, F); the
center of the tibial stem, which was considered the geometric center of the
tibia at the level of the tibial tuberosity
(Fig. 1, F); and the
center of the ankle, which was considered the geometric center of the trochlea
of the talus (Fig. 1,
G).
The femoral mechanical axis was defined as the connecting line between the
center of the femoral head and the center of the knee. The tibial mechanical
axis was the line connecting the center of the ankle and the center of the
knee.
The spatial positions of the femoral and tibial implants were each defined
by one vector, which was perpendicular to the distal femoral and the proximal
tibial cut. These were produced as compensation lines (least-squares
algorithm) of the geometric centers of the femoral fixation pins
(Fig. 1, C) and the
tibial stem over the scan area (Fig. 1,
E).
The rotational deviation of the femoral component from the referenced axis
was determined by the angle between the line connecting the femoral fixation
pins (Fig. 1, C) and
the surgical epicondylar axis (Fig. 1,
B). The tibial rotational error was defined as the angle
between the angle bisecting the line of the tibial component fins
(Fig. 1, E) and the
line between the medial third of the tibial tuberosity and the geometric
center of gravity of the tibia (Fig. 1,
F).
From the spatial relationship between the femoral and tibial components and
the femoral and tibial mechanical axis, the following angles were determined:
the varus or valgus position of the femoral component relative to the femoral
mechanical axis, the varus or valgus position of the tibial component relative
to the tibial mechanical axis, the varus or valgus position of the entire limb
as a sum of the tibial and femoral mechanical axes, the extension-flexion of
the femoral component in relation to the femoral mechanical axis, the tibial
posterior slope, the rotational deviation of the femoral component from the
epicondylar axis, and the rotational deviation of the tibial component from
the referenced axis.
Statistical Analysis
The arithmetic mean, standard deviation, and the distribution of values
were determined for each measure and in both groups. Differences between the
study groups were determined with use of the nonparametric Mann-Whitney U test
for independent samples with a level of significance of 0.05.
No significant differences in the demographic data were detected, on
the basis of the numbers, between the patients who had conventional
arthroplasty and those who had computer-assisted arthroplasty
(Table I). The range of motion
and the Knee Society scores did not differ between the study groups at six
months postoperatively. There was a tendency toward a longer operating time
with the navigated technique (101 compared with ninety-four minutes), but the
difference was not significant.
For the frontal plane, the overall mechanical axis showed an irregular
frequency distribution in both the computer-assisted and the conventional
arthroplasty group (Fig. 2).
The latter group showed a range of between 4.8° of valgus and 6.6° of
varus alignment, with a mean deviation of 2.6° ± 1.7°. Knee
components implanted with use of navigation were implanted more precisely,
with a mechanical axis ranging from 2.9° of valgus to 3.1° of varus
alignment and a mean deviation of 1.4° ± 0.8° (p = 0.004). If
one assumes a tolerance level of 3°, seven conventional arthroplasties but
only one computer-assisted arthroplasty were outside this tolerance.
The tibial mechanical axis in the frontal plane also revealed an irregular
frequency distribution, regardless of the surgical method. No tibial component
in the computer-assisted navigation group but five in the conventional
arthroplasty group were outside a tolerance range of 3°
(Fig. 3). The mean deviation
was 1.4° ± 0.9° for tibial components implanted with
computer-assisted navigation and 2.0° ± 1.7° for those that had
been conventionally implanted (p = 0.646). The femoral mechanical axis showed
a comparably irregular distribution in the frontal plane. No component
implanted with navigation but three conventionally implanted joints were
outside the 3° tolerance range (Fig.
4). On the average, the deviation from the mechanical axis was
significantly lower for the femoral components implanted with computer
assistance (1.0° ± 0.6°) than for the conventionally implanted
femoral components (2.2° ± 3.2°) (p = 0.008).
The distribution of the alignment of the femoral component in the sagittal
plane reflects the anatomical alignment of the navigation system used in the
present study with respect to the anterior femoral cortical bone. The mean
flexion of the femoral component was 3.4° ± 2.7°, which
corresponds to the sagittal profile of the femur. The conventionally implanted
femoral components showed a comparable flexion of 3.8° ± 3.1°
in relation to the mechanical axis. A tolerance limit cannot be stated, as it
is as yet unclear whether an alignment corresponding to the mechanical or
anatomical femoral axis should be sought.
The deviation from the preoperatively planned tibial slope of the knee
replacements implanted with navigation (2.1° ± 1.3°) was not
significantly different from that of the conventionally implanted components
(3.4° ± 3.0°). The difference in the preoperatively planned
slope angle was <3° in twenty-five knee replacements implanted with
navigation (72%) and in fourteen knee replacements implanted conventionally
(53%) (Fig. 5).
In the conventional implantation technique, the rotational deviation of the
femoral component from the epicondylar axis showed an irregular distribution
between 3.3° of internal rotation and 5.0° of external rotation, with
a mean deviation of 0.1° ± 2.2°
(Fig. 6). Three knees were
outside a tolerance range of 3° of deviation. In contrast to this, all but
one of the femoral components implanted with use of computer-assisted
navigation were within the tolerance range and showed a deviation of between
4.7° of internal rotation and 2.2° of external rotation, with a mean
of 0.3° ± 1.4°. The difference in the distributions was not
significant, on the basis of the numbers.
Compared with the femoral components, the tibial components showed a
greater range of rotational deviation, with between 27.1° of internal
rotation and 15° of external rotation and a mean deviation of 7.5°
± 6.0° for conventionally implanted knee replacements. There was no
significant difference in the rotational alignment of the tibial components
implanted with computer-assisted navigation, which showed rotational
deviations of between 21.2° of internal rotation and 11.0° of external
rotation and a mean deviation of 6.9° ± 4.7°.
On the basis of the numbers, no association was found between the
deviations in the frontal and rotational planes or between component alignment
and the early postoperative range of motion or the Knee Society scores.
Most investigators have demonstrated that total knee replacements
implanted with computer-assisted navigation have more accurate component
alignment, on the basis of plain radiographs, than those implanted
conventionally6-12.
The improvement of accuracy through computer assistance has been shown to be a
few degrees, which is within the range of inaccuracy produced by
projection-related errors in standing radiographs. To avoid this, a new method
for three-dimensional determination of implant position and rotation of knee
arthroplasty components is presented in this study. This method makes it
possible to determine separately the mechanical axes of the tibia and of the
femur, as well as the overall mechanical axis in a virtually derotated and
fully extended knee free of projection errors. The limitations of this method
result from the fact that implant alignment on radiographs as well as on
computed tomograms is always related to bone landmarks that are defined by an
observer. While landmark identification errors are of minor impact on the
mechanical axis, which is a long axis, they may influence the evaluation of
implant rotation, which is calculated relative to short axes and can have a
substantial impact on functional
outcome16-20.
The identification of the sulcus between the attachments of the medial
collateral ligament is particularly difficult and depends on the individual
prominence of this landmark. Postoperative measurement of femoral component
rotation through computed tomography is additionally hampered by metal
artifacts. Hence, alternative anatomical landmarks (for example, the femoral
neck) were chosen for the evaluation of component
rotation30. With
this method, Oberst et al. could not show a difference in component rotation
between femoral components implanted with computer-assisted navigation and
those implanted
conventionally30.
In contrast, Chauhan et al. demonstrated a higher accuracy of implant
alignment through the use of a navigation system in both the frontal and the
rotational planes using an improved computed tomography
protocol31. The
images were reconstructed for the frontal and sagittal plane but without
virtually derotating and extending the joint. Stockl et al. showed a slight
improvement of rotational alignment in computed tomography scans of the knee
through navigation, too, but they evaluated the mechanical axes in
conventional plain
radiographs32. Both
Chauhan et al. and Stockl et al. defined the rotation of the femoral component
through a compromise between different landmarks (the Whiteside line,
posterior condyles, and epicondyles) only in knees that had the components
implanted with computer-assisted navigation, but not in the knees that had
conventional implantation. That may have resulted in a bias toward better
alignment in the computer-assisted placement of the implants. This bias was
excluded in the present study by defining the femoral component rotation
solely in relation to the surgical epicondylar axis. On the basis of this
approach, rotational alignment was not improved through navigation. Although
the optimum rotational alignment of the femoral component is still a matter of
debate, the consideration of additional landmarks (the Whiteside line,
posterior condylar axis, and the symmetry of the flexion gap) by the
navigation system may lead to improved
alignment21.
Regardless of which implantation system is used, the rotation of the tibial
component had a considerable range, from 27.1° of internal rotation to
15° of external rotation. The intraoperative determination of the selected
landmarks (the tibial tuberosity and the center of the tibia) had a high
variance. Since reproducible landmarks have been lacking, rotating platforms
may partially compensate for malrotations of the tibial component. In
contrast, the navigated implantation resulted in a more precise positioning
relative to the mechanical axis than did the conventional surgical method
without consuming significantly more operative time. This effect is seen in
the overall mechanical axis of the entire extremity in the frontal plane as
well as in the mechanical axis for the femoral component. There was no
significant difference, on the basis of the numbers, in the accuracy of
implantation of the tibial component between the two methods. We believe that
this is attributable to the fact that, in navigation and conventional
techniques, the center of the ankle is identified through palpation of the
mortise. Again, the main inaccuracy seems to result from the definition of
anatomical landmarks by the surgeon independently from the use of computer
assistance. In contrast to the tibial component, the femoral component
alignment was significantly more precise in the knees managed with
computer-assisted navigation than in those managed conventionally. This shows
that the functional determination of the hip center by the navigation system
is more accurate than the conventional intramedullary
method33.
The lack of an association between deviations in the frontal and rotational
planes shows that there was no propagation of errors through the operative
technique, which is an advantage of the so-called femur first approach, in
which alignment of the femoral component is solely based on the anatomy of the
femur.
In contrast to the conventional surgical method, the navigation system
shows the overall mechanical axis after insertion of the trial implants.
Although deviations of >1° lead to secondary resections, the study
revealed errors of up to 3.1° in the overall axis. This can be explained
by the summation of many different errors in the sense of error propagation:
inaccuracies of the navigation system (dirty reflectors and camera or rounding
errors), mobility of the modular plug connections (reference clips and control
tray) and the reference pin itself, micromovements in the cutting guide during
sawing, a deviation of 1° accepted by the operator, and inhomogeneous
cement distribution between the implant and bone.
The fact that we found no association between implant alignment and early
postoperative range of motion and Knee Society scores may be explained by the
considerably low deviation from the designated axes in the collective group of
patients. In fact, a slight component malalignment may only be a cofactor
beside instability and soft-tissue trauma, leading to restricted range of
motion and
function34, but it
still may promote early loosening through increased wear caused by suboptimal
implant loading35.
It remains for future investigations to determine the extent to which these
results raise hopes of improving the long-term outcome of total knee
arthroplasties. ?