Fifteen retrieved polyethylene tibial inserts of the same design
(Anatomic Modular Knee; DePuy) (Fig.
1) were selected from a large collection of inserts that had been
retrieved thirty-six to 146 months (average, ninety-one months) after
implantation. The thicknesses, as labeled by the manufacturer, were 8 mm (one
insert), 10 mm (nine), 12 mm (four), and 14 mm (one). Fourteen inserts were
from a posterior cruciate ligament-retaining insert, and one was from a
posterior stabilized insert. The components were retrieved from eight women
and seven men, with an average age of sixty-six years (range, fifty-four to
seventy-two years) at the time of the revision arthroplasty. The average
weight and height of the patients was 925 N (range, 690 to 1308 N) and 1.5 m
(range, 1.4 to 1.9 m), respectively, corresponding to an average body mass
index of 32.4 kg/m2 (range, 23.8 to 42.5 kg/m2). The
reason for the revision, reported preoperatively, was loosening (six knees),
wear (six), or osteolysis (three).
The capture mechanism of the Anatomic Modular Knee design consists of two
dovetail slides running from anterior to posterior with a locking pin
(Fig. 1). The backside of each
insert was inspected visually with use of stereomicroscopy (thirty-two times
magnification) to evaluate surface damage. The backside was divided into four
quadrants: anteromedial, anterolateral, posterolateral, and posteromedial.
Within each quadrant, the severity of burnishing, pitting, embedded acrylic
debris, scratching, abrasion, delamination, and surface deformation
(protrusions into screw-holes) was assessed. All damage modes were given a
score of 0 to 10, in each quadrant, on the basis of either the area of the
surface affected by burnishing, pitting, embedded debris, and abrasion or the
severity of that particular wear pattern (delamination, scratching, and
surface deformation). The damage was categorized as mild or negligible (a
score of 0, 1, or 2), moderate (a score of 3, 4, or 5), or severe (a score of
6 to 10). The widths of the pegs in the anterior-posterior and medial-lateral
directions were also measured.
To quantify the material loss due to backside wear, each retrieved insert
was scanned with use of a laser surface profilometer (Cyberware, Monterey,
California). Multiple linear scans of each insert, made as the insert was
rotated 360° in 10° increments, were aligned and merged, with each
individual scan having a resolution of 0.1 mm in the plane of the scan and a
resolution of 0.008 mm in depth. This process resulted in approximately
300,000 three-dimensional points defining each insert. An unused insert of
each size was also scanned. With use of computer-aided design software
(UniGraphics; EDS, Plano, Texas), individual three-dimensional surface models
of each retrieved insert were developed. For each insert, the
three-dimensional surface was geometrically matched to its corresponding
unused implant with use of unworn, undeformed patches of the retrieved insert.
These unworn fiducial surfaces most commonly existed around the periphery of
the implant and between the two articulating surfaces on the superior aspect
of the implant. The volume of material lost because of backside wear during
the time in situ was then calculated from the volume bounded by the final
(worn) backside surface and the initial (unworn) backside surface. The sources
of backside wear debris were separated into three sections, naturally defined
by the capture mechanism grooves: under the medial plateau, under the lateral
plateau, and in the center anterior-to-posterior strip between the two
dovetail grooves. For comparison with previous studies, the volume of the
material lost was divided by the time in situ of each implant to estimate the
wear rate. The planarity of a surface patch was determined by measuring the
root-mean-square deviation of all of the points in the patch from a plane fit
to those points.
The backside surfaces of two retrieved inserts (which had been in situ for
thirty-six and eighty-six months) were also examined with scanning electron
microscopy (Amray 1830; Amray, Bedford, Massachusetts) operating under
secondary electron mode. These two inserts were chosen as one represents
typical or average wear, whereas the other exhibits severe wear. The images
were used to document the microscopic appearance of the worn backside surface,
particularly the transition between the worn surface and the raised
protrusion.
Different statistical tests were performed to elucidate the complex
relationship between all of the measured variables. A logistic regression
model was used to evaluate the relationship between all of the continuous
variables (e.g., height, weight, body mass index, time in situ, and
polyethylene thickness) and the damage measurements. Both simple logistic
regression models for each damage mechanism as well as a multiple logistic
regression model fit to the entire data set were analyzed. Because this
technique works best with dichotomous dependent variables, the results of the
logistic likelihood ratios were used to determine which independent variables
significantly affected the dependent variables. For robustness in the
predictive capabilities of those significant independent variables, Pearson
correlation coefficients and their corresponding z transformations and p
values were then calculated for the variables showing potential correlations.
This analysis was performed on all possible combinations of independent
variables and damage measurements.
Three modes of surface damage were observed on the non-articulating
backside surface of the retrieved inserts: burnishing, pitting, and surface
deformation in the form of protrusions into screw-holes. Some degree of
burnishing was noted on the backside surface of every component. Pitting was
also very common (thirteen of the fifteen tibial inserts). The average
burnishing score (and standard deviation) was 7.9 ± 2.5 (range, 1 to
10), and the average pitting score for the entire backside surface was 2.2
± 1.8 (range, 0 to 5.5). All components exhibited polyethylene material
protruding from the undersurface of the polyethylene insert into screw-holes
within the tibial backing plate. The average protrusion height was 0.6
± 0.6 mm (range, 0.03 to 2.7 mm) above the surrounding surface.
Protrusions over the anteromedial screwhole of the tray (average height, 1.1
± 0.9 mm) were significantly larger than those over the other three
locations (p = 0.002) (Table
I). Fourteen of the fifteen components had a flange of
polyethylene that overlapped the peripheral margin of the base-plate on the
medial side (Fig. 2).
With the numbers available, the original thickness of the insert was found
to have no effect on the height of the protrusions (p = 0.45). Patient weight
was positively correlated with the average medial protrusion height
(r2 = 0.79, p < 0.05). Ninety percent of the pegs were out of
round with the anterior-posterior dimension, measuring smaller than the
medial-lateral dimension. The average difference in the two dimensions was
0.26 ± 0.21 mm (range, 0 to 0.83 mm).
Three-dimensional computer reconstructions of the retrieved inserts
demonstrated that, for all retrievals, the unworn surface of the remaining
protrusions, the rim of material protruding over the medial edge, and the
unworn surfaces on the anterior-lateral edge all lay in a single plane. The
average root-mean-square deviation from a plane of the points defining the
unworn surfaces was 0.07 ± 0.05 mm.
An example of a computer reconstruction of the backside surface of the
posterior cruciate ligament-sacrificing insert showed that the surfaces of the
material protruded into the four screw-holes and that the surface of the
unworn posterior aspect that protruded into the posterior cruciate ligament
notch of the tibial base-plate all had well-defined machining marks and all
were in one plane (root-mean-square error of points was 0.06 mm)
(Fig. 3). This example
demonstrates that the protrusions and their inferior surface represent the
original backside surface and that the worn surface surrounding the
protrusions demonstrates loss of material. Electron microscopy revealed that
the typical polyethylene protrusion had eroded, shredded edges, providing
evidence of wear as opposed to a deformation phenomenon
(Fig. 4, A through
D).
On the basis of the three-dimensional computer reconstructions, the average
volume of material lost during the time in situ was 925 ± 637
mm3 (range, 197 to 2720 mm3); 521 ± 354
mm3 (range, 112 to 1183 mm3) from the medial
compartment, 245 ± 325 mm3 (range, 16 to 1338
mm3) from the lateral compartment, and 159 ± 96
mm3 (range, 34 to 366 mm3) from the center area between
the medial and lateral grooves for the locking mechanism. The corresponding
average volumetric wear rate was 138 ± 95 mm3/yr from the
backside surface. Patient weight was positively correlated with increased
volumetric wear from the medial backside surface (r2 = 0.52, p <
0.01).