When total knee components with a posterior-cruciate-substituting
design were introduced, there was concern that the cam-and-post
posterior-cruciate-substituting mechanism would transmit anteroposterior
sheer stresses to the cement-bone interfaces and cause loosening.
Subsequent clinical experience with posterior-cruciate-substituting
total knee replacement designs has produced good or excellent clinical
and functional results1-6. Few
knees fail to achieve 90° of flexion7.
Aseptic loosening of cemented posterior-cruciate-substituting total
knee replacements has been rarely reported8,9,
and in several studies with follow-up intervals of five to ten years
there was no aseptic loosening1-6.
Because of low rates of wear, osteolysis, and loosening1-4,6, posterior-cruciate-substituting
total knee replacements are currently being used more frequently
in younger, more active patients. With patients making higher demands
on the knee replacements and with longer durations of follow-up,
aseptic loosening may become more common.
Because of the low prevalence of failure, there is a paucity
of information regarding the mechanism or mechanisms of loosening
of posterior-cruciate-substituting total knee replacements. In the
present study, we performed an analysis of sixteen posterior-cruciate-substituting
total knee replacements that were revised because of aseptic loosening
and osteolysis. The goal was to determine, through a review of the
clinical history, radiographic analysis, and implant retrieval analysis,
the factors that predispose or contribute to aseptic loosening and osteolysis.
From July 20, 1989, to October 14, 1994, 557 consecutive total
knee arthroplasties were performed in 467 patients by a single surgeon
(O.M.M.). The press-fit condylar modular total knee prosthesis with
a posterior-cruciate-substituting design (PFC; Johnson and Johnson
Orthopaedics, Raynham, Massachusetts) was used in all knees. All
components were implanted with vacuum-mixed bone cement (Simplex
P; Howmedica, Rutherford, New Jersey), with the exception of eighty-three
porous-coated femoral components that were implanted without cement.
All patellae were resurfaced with an all-polyethylene dome-shaped
component implanted with bone cement.
The patients were treated with a standardized protocol. Thirty-six-inch
(91-cm) radiographs of the lower extremity were used in preoperative
planning in order to establish a neutral mechanical axis
in the frontal plane10. A subvastus
approach was used in all patients. Intramedullary femoral and extramedullary
tibial alignment guides were utilized. Soft-tissue releases were
performed as necessary. The cement was in a doughy state when it
was pressurized into bone. No cement was applied to any prosthetic
surface except the posterior condyles of the femoral component. The
postoperative regimen consisted of continuous passive motion that
was started in the recovery room, with gait-training beginning on
the first postoperative day and weight-bearing increased as tolerated.
At a mean of fifty-six months (range, thirty-seven to eighty-nine
months), twenty-two (3.9%) of the 557 primary knee arthroplasties
had been revised. Four (0.7%) were revised because of infection;
two (0.4%), because of a supracondylar femoral fracture;
and sixteen (2.9%), because of aseptic loosening and osteolysis.
The sixteen knee replacements (fifteen patients) that were revised
because of aseptic loosening and osteolysis are the subject of this report.
The preoperative clinical and functional Knee Society scores11 for these sixteen knees were a mean
of 37 points (range, 15 to 53 points) and 47 points (range, 30 to 55
points), respectively. The postoperative activity level was quantified
with use of the method of Zahiri et al.12.
Standing anteroposterior and lateral radiographs and patellar skyline
radiographs were assessed for alignment and for radiolucencies,
according to the system described by Ewald13,
preoperatively, at one year after the index arthroplasty, and immediately
prior to the revision surgery.
Retrieved implants were inspected for surface damage14. Damage to the tibial cruciate-substituting
polyethylene post was specifically assessed. The top of the post
as well as the anterior, posterior, medial, and lateral
faces were analyzed separately. Four polyethylene implants
were inadvertently discarded by the hospital pathology department.
The change in the minimum thickness of the medial and lateral condylar
surfaces of the polyethylene inserts was defined as the difference
between the nominal minimum thickness of the as-manufactured component and
that of the retrieved implant as measured with vernier calipers
and calculated as the mean of three measurements of each condylar
surface15. The nominal minimum
thickness of the polyethylene inserts used for the primary total
knee arthroplasties ranged from 6 to 13 mm. The mean length of time
between sterilization of the polyethylene insert and implantation,
the so-called shelf life, was 7.2 months (range, 0.9 to 35.9 months).
Prepared tissue samples from osteolytic regions of the knee that
had been obtained intraoperatively and stained with hematoxylin
and eosin were examined under plain and polarized light microscopy
for evidence of inflammation, barium sulfate crystals from the bone
cement, and metal and polyethylene particles16.
Given the small sample size, statistical analysis for regression
was performed with the Spearman nonparametric correlation, rather
than the Pearson product-moment correlation, for putative related phenomena.
Clinical Findings
The study population consisted of eight women and seven men,
with a mean age of sixty-five years (range, forty-six to eighty
years). All sixteen knees had good or excellent function at one
year after the primary total knee arthroplasty. The mean Knee Society
clinical and functional scores at that time had increased to 95
and 86 points, respectively, from preoperative means of 37 and 47
points. The mean activity score was 5.2 points (range, 1 to 9 points).
All but one knee had synovitis between two months and three years
prior to the radiographic appearance of loosening or osteolysis.
Prior to revision, the mean clinical and functional scores had declined
to 48 and 45 points, respectively.
The percentage of patients who had had a total knee arthroplasty
on the contralateral side was significantly greater in the group
that had had a revision because of loosening and osteolysis (nine
of fifteen patients) than in the group that did not have a revision
(eighty-two [18%] of 452 patients) (Fisher exact
test for independence, p = 0.026). The percentage of patients
who had had bilateral primary total knee arthroplasty under the
same anesthetic was also significantly greater in the group that
had had a revision because of loosening and osteolysis (five of
fifteen) than in the group that did not have a revision (thirty [7%] of
452) (Fisher exact test for independence, p = 0.045).
Two knees had isolated loosening of the tibial component with
osteolysis, three had isolated loosening of the femoral component
with osteolysis, six had loosening of the tibial and femoral components with
osteolysis, three had loosening of the femoral component with osteolysis
around a well-fixed tibial component, and two had loosening of the
tibial component with osteolysis around a well-fixed femoral component.
No knee had loosening of, or osteolysis associated with, the patellar
component.
Radiographic Findings
The mean coronal alignment (anatomic axis) preoperatively, following
the primary total knee arthroplasty, and immediately prior to the
revision total knee arthroplasty was 0°, 3° of valgus, and 3° of varus,
respectively. There was a mean 6° progression into varus (range,
1° of valgus movement to 15° of varus movement) postoperatively
(Figs. 1-A and 1-B). The femoral component
demonstrated the greatest amount of movement into varus, from a
mean of 5° of valgus to a mean of 1° of valgus. The tibial component
averaged 1.5° of varus change postoperatively (from a mean of 1.5°
to a mean of 3° of varus). Although there were individual changes
in femoral component flexion and tibial component slope, there was
no trend and the mean for the group was unchanged.
The mean scores for radiolucent lines seen around the femoral
component on the lateral radiograph, around the tibial component
on the anteroposterior radiograph, and around the tibial component
on the lateral radiograph increased from 2.0, 1.6, and 0.4, respectively,
at one year postoperatively to 8.9, 10.6, and 7.7, respectively,
just prior to revision. The highest scores were found consistently
in the peripheral zones. According to linear regression analysis,
the anterior and posterior peripheral femoral zones demonstrated
significantly higher scores for radiolucent lines than did the combined
internal zones (p = 0.004), and the medial and lateral
tibial peripheral zones had significantly higher scores for radiolucent
lines than did the internal zones (p < 0.001).
Revision
The femoral component had been inserted without cement in three
of the sixteen knees that had had a revision because of loosening
and in eighty (14.8%) of the 541 knees that had no loosening
(p > 0.05). Loose components (nine femoral and ten tibial)
that had been inserted with cement were easily removed by hand from
the cement mantle, indicating loosening at the metal-cement interface.
All knees had as much as 2 mm of gross motion of the polyethylene insert
in the tibial tray.
Retrieval
In six of the twelve retrieved polyethylene inserts, the machine
marks on the posterior aspect of the polyethylene post had been
preserved, indicating a lack of substantial or repetitive contact
between the tibial post and the femoral cam of the posterior-cruciate-substituting
mechanism. There was no apparent relationship between wear on the
posterior aspect of the post and the range of motion of that knee
(r = 0.02), the tibial slope (r = 0.2), or femoral flexion
(r = 0.15).
All twelve inserts had burnishing of both the medial and lateral
side walls; the burnishing was always present at the base, with
a variable degree of burnishing over the remainder of the side wall
(Figs. 2-A, 2-B, and 2-C). In ten knees,
the polyethylene post of the insert had grooves of variable depth,
associated with burnishing and delamination, at the anteromedial
and anterolateral edges. A horizontal burnished groove across the
anteroinferior aspect of the polyethylene post was seen in nine
of the twelve inserts.
The superior aspect of the polyethylene post had beveled corners
on the medial side in ten components and on the lateral side in
eight components. With as little as a 1-mm reduction in the nominal thickness
of the tibial condylar surfaces, the top of the polyethylene post
impinged on the roof of the femoral box. An asymmetric loss of material
from the top of the post reflects asymmetric loading and condylar
wear, rotation, or lift-off17,
or a combination of these factors.
The change in the minimum thickness of the twelve retrieved components
was greater on the medial side (mean, 1.4 mm; range, 0.7 to 2.4
mm) than on the lateral side (mean, 0.7 mm; range, 0.5 to 1.6 mm)
(p = 0.001, paired t test). The change in the minimum polyethylene
thickness on the medial side was correlated with that on the lateral
side (r = 0.69, p < 0.05). The change on the medial
side was correlated with the size of the component (r = 0.58, p = 0.049)
but not with the thickness. The mean rate of change in the minimum
polyethylene thickness of 0.30 mm/yr on the medial side
was significantly greater than that of 0.16 mm/yr on the
lateral side (p = 0.001, paired t test).
With the numbers available, the time in situ was
not correlated with a reduction in the minimum polyethylene thickness.
The body weight of the patient was mildly related to the change
in the minimum thickness (r = 0.52, p = 0.08).
A reduction in the minimum thickness on the medial side was correlated
with patient activity (r = 0.67, p = 0.017). A younger
age and greater height both were correlated with wear on the medial
side (r = 0.64 and 0.58, respectively, p < 0.05).
The shelf life of the tibial insert was moderately correlated with
the reduction in the minimum thickness on the medial side (r = 0.55,
p = 0.067).
A rotational pattern of scratching and burnishing on the inferior
polyethylene surfaces, the so-called back side, was seen on all
twelve inserts. Longer scratches were present laterally, with the
rotational axis offset into the medial compartment (Figs. 3-A and 3-B). A similar pattern
of scratching and burnishing was seen on the superior surface of
the titanium-alloy tibial base-plates. A rotational pattern of surface
damage also was seen on the inferior surfaces and keels of the tibial
base-plates.
The present analysis did not reveal evidence of transmission
of substantial anteroposterior stresses from the posterior-cruciate-substituting
mechanism. On the contrary, six of the twelve retrieved tibial components
demonstrated little, if any, evidence of cam-post engagement, despite
a good range of motion. Of the six specimens that demonstrated wear
of the posterior aspect of the tibial polyethylene post,
none had an anteroposterior pattern of any type of surface damage
on the inferior surface of the tibial insert or on the
superior or inferior aspect of the tibial tray. This finding is
consistent with the design principle of the posterior-cruciate-substituting
knee: the vector sum of the tibiofemoral forces and the cam-post
forces is compression into the tibia18.
All twelve retrieved components had evidence of damage to the
anterolateral and anteromedial aspects of the posterior-cruciate-substituting
post as well as to the bases of the side walls of the post. The
damage to the inferior surface of the polyethylene inserts consistently
had a rotational pattern, with the axis of rotation in the medial
compartment. Surface damage due to relative rotational motion was
also found on the superior and inferior surfaces of the titanium
tibial trays. These observations indicated that rotational stresses
can be transmitted to the modular interfaces and to the metal-cement interfaces.
These stresses result in relative motion, mode-4 wear (that is,
wear particles generated from two surfaces that are not the primary
or intended bearing surfaces)19,
progressive loosening, and osteolysis.
With this design, rotational forces were generated by impingement
of the side walls of the intercondylar box on the polyethylene post.
This box-post impingement can occur throughout the range of motion.
The top of the polyethylene post can also impinge on the roof of
the box. Even without rotation, with reduction in the tibial condylar
polyethylene surfaces due to wear, the top of the polyethylene post
can contact the roof of the intercondylar box.
Normal tibiofemoral rotation encompasses approximately 5° of
internal rotation at heel-strike and toe-off and approximately 9°
of external rotation during swing phase. The screw-home motion has
been described as relative external rotation of the tibia during
terminal extension20. It has been
shown that some posterior-cruciate-substituting total knee replacements
have altered kinematics, demonstrating a negative screw-home movement
in approximately half of the knees in one study17.
As much as 12° each of internal and external rotation have been
shown to occur during level walking and stair-climbing, most commonly
at flexion angles of 0° to 40°. Up to 15° of external rotation has
been demonstrated following total knee replacement21.
The present retrieval analysis indicated that a reduction in
the rotational constraint caused by the design of the box-post mechanism
would be desirable. However, considering the low prevalence of loosening,
the rotation allowed may be sufficient in the majority of knees.
This type of rotational constraint could be reduced by modifications
in the design of the polyethylene post or the femoral box, or both.
The design of the roof of the femoral box, whether metal or bone,
should allow several millimeters of clearance throughout the entire
range of motion in order to accommodate reductions in the thickness
of the tibial condylar surfaces that can occur as a result of normal
use and wear.
Patients with bilateral total knee replacement appear to be at
greater risk (p = 0.026) for rotational loosening and osteolysis
than are patients with a unilateral total knee replacement. This
finding warrants additional investigation, as there is no precedent
or obvious explanation as to why these patients would be at greater
risk. Fourteen of the sixteen knees had progression into varus.
In nine of them, the varus mechanical axis was incompletely corrected,
but seven knees were well aligned. Varus malalignment could facilitate
box-post impingement; however, the same types of surface damage due
to rotational stresses were present in the well-aligned knees.
Rotational malalignment may play a role in the loosening of these
and other total knee replacements. The rotational positions of the
components, relative to each other and to the patient’s
bone and soft-tissue anatomy, are an important consideration in
total knee arthroplasty. Unfortunately, these relationships are
difficult to assess on physical examination and with plain radiographs.
Even computerized tomography has limited value because the scan
is static while the rotational relationships are dynamic.
As has been described for the femoral component of total hip
arthroplasties, loosening at the metal-cement interface22, so-called debonding23,24, was seen in the sixteen total
knee replacements in our study. Although the benefit of maintaining
the metal-cement bond in total hip replacement has been debated,
maintaining this bond in total knee replacement is desirable. Attention
should be given to the preparation of the cement-prosthesis interface.
In the present series, the prosthesis was inserted into the cement
when it was in a doughy state. The metal-cement bond appears to
be strongest when the cement is applied to the implant in a low-viscosity
state, soon after mixing25.
Wear between the modular tibial polyethylene and the base-plate,
so-called back-side wear, was observed in all knees. Damage was
also seen on the inferior surface of the base-plates. Factors that
contribute to this type of mode-4 wear19 include
relative motion, contact area, and wear characteristics of the apposed
materials26. The retrieved components
had evidence of relative motion of approximately 2 mm, which is
an order of magnitude greater than that reported in studies involving in
vitro testing27; however,
such laboratory tests have not reproduced the complex in
vivo loading indicated by the present retrieval analysis. Compared
with cobalt-chromium alloys, titanium alloys have lower hardness;
thus, the surface is more easily damaged and less wear resistant19,28,29. In order to minimize mode-4
wear in modular, cemented total knee replacements, use of locking mechanisms
that minimize relative motion during off-axis loading and base-plates
made of harder materials, such as cobalt-chromium alloys, is desirable.
In summary, loosening and osteolysis are rarely associated
with cemented posterior-cruciate-substituting total knee replacements.
In the knees in the present study, we found no evidence
that anteroposterior shear forces from the posterior-cruciate-substituting
mechanism contributed to failure. Our findings indicate that loosening
was a result of the transmission of tibiofemoral rotational stresses
to the modular interfaces and to the metal-cement interfaces. With
this design, the primary source of the rotational stress was box-post
impingement. A reduction in rotational constraint would be desirable.
Note: The authors thank Frederick J. Dorey, PhD, for his assistance
in the statistical analysis of these data.