Specimens
Nine hemipelves with a cementless porous-coated acetabular total hip
component and osteolytic bone defects that had been detected on posthumous
computed tomography scans were selected from the autopsy retrieval program at
our institution. All nine hemipelves were retrieved from individuals who had
been treated and followed at our institution and who had had a
well-functioning total hip prosthesis at the time of death. The index
operation was a primary total hip replacement performed for the treatment of
osteoarthritis in eight patients and for the treatment of rheumatoid arthritis
in one. The acetabular components, which had been implanted by the senior
author (C.A.E.), included five Duraloc 100 cups (DePuy, Warsaw, Indiana) and
four Arthropor cups (Joint Medical Products, Stamford, Connecticut), two
designs with different rationales. The specimens were retrieved from five men
and four women with an average age of sixty-seven years (range, forty-nine to
seventy-eight years) at the time of surgery and seventy-five years (range,
fifty-four to eighty-five years) at the time of death. The implants had been
in situ for an average of 106 months (range, forty-one to 178 months).
Specimen Preparation
All of the donors had provided appropriate consent, through the autopsy
retrieval program at our institution, for the harvest of the specimens after
their death. Only autopsy retrieval specimens with sufficient pelvic bone,
including the entire iliac wing, pubis, and ischium, were evaluated. The fresh
specimens underwent conventional radiographic examination and were then fixed
in 10% neutral buffered formalin, dehydrated in a series of graduated ethanol
concentrations, and then cleared in acetone. The duration of each of these
steps was at least one week. Following the completion of dehydration and
clearing, a plastic rod was placed into the ischium of each specimen
perpendicular to the coronal plane of the pelvis. Each specimen was examined
with computed tomography, after which the polyethylene liner was removed from
the hemipelvis and the specimen was embedded in clear polymethylmethacrylate.
Three aluminum rods were inserted into the polymethylmethacrylate block
parallel to the ischial plastic rod in order to fix the specimen parallel to
the coronal plane of the pelvis in a saw. All specimens were then cut at 3-mm
intervals in the coronal plane with a water-cooled diamond saw (LSM 250; Leco,
St. Joseph, Michigan) into 1.5-mm-thick slices in order to make slab
radiographs of each slice.
Radiographic Evaluation
Anteroposterior and iliac oblique radiographs of each patient had been made
prior to the index arthroplasty and at multiple postoperative intervals, and
these images were available for comparison with the same views of the
retrieved specimens. The senior author (C.A.E.), who was blinded to the
findings on the computed tomography scans, examined all of the plain
radiographs for evidence of periacetabular osteolysis. Osteolysis was defined
as a sharply demarcated lucent area adjacent to the acetabular component that
was not evident on the immediate postoperative
radiograph17,18.
The location of each lesion was classified with a scheme devised for this
study. Each acetabular component was divided into five zones—anterior,
posterior, superior, inferior, and dome
(Fig. 1)—and each lesion
was assigned to one of these five locations. If a lesion spanned more than one
region, both regions were designated as containing a lesion. The volumes of
the lesions were not measured on the plain radiographs.
Computed Tomography Evaluation
Prior to embedding, each retrieved specimen was placed into a container of
alcohol and was scanned in 1-mm axial intervals at 120 kV and 220 mA, with a
field of view of 22 cm. Either a single-detector, single-slice, helical
machine (GE High Speed Advantage; General Electric, Waukesha, Wisconsin) or a
multidetector, multi-slice machine (Somotom 4; Siemens, Munich, Germany) was
employed, depending on availability. Coronal and sagittal images were
generated from the axial images by a post-processing imaging program
(Muscular-Skeleton Analysis Software; VirtualScopics, Rochester, New York).
All of the specimens were initially scanned without a metal femoral head
articulating with the acetabular component, and then five specimens were
rescanned with a metal femoral head articulating with the acetabular
component.
The raw data were processed with the VirtualScopics software package to
suppress any metal-induced artifact. The data were then segmented by the
VirtualScopics software package, with each image classified according to its
properties. An experienced orthopaedic surgeon (N.K.) then reclassified the
segmented data and labeled any demarcated area adjacent to the acetabular
component without trabecular bone as an osteolytic
lesion13,14.
The volume of osteolysis was calculated with use of the VirtualScopics program
during three-dimensional rendering of the pelvis. The three-dimensional model
of the pelvis included the acetabular component and the osteolytic lesion,
which allowed us to determine the location of the lesion according to the
previously described classification scheme. The medial wall of each specimen
was also evaluated for any loss of integrity.
Slab Radiographic Evaluation
After the embedded specimens were sectioned, slab radiographs were made
with use of mammography film (UM-MA HC for mammography; Fuji Photo Film,
Greenwood, South Carolina) at 25 kV and 100 mA for 4/5 of a second.
Mammography film was employed rather than conventional film because it reveals
greater bone structure detail. The area of each osteolytic lesion was measured
with use of the Martell two-dimensional computerized radiographic analysis
program (version 4.5.0.1; Hip Analysis Suite, University of Chicago Medical
Center, Chicago, Illinois) from scanned images (150 dpi) of the slab
radiographs. The thickness of each slab was measured at several points with a
digital caliper (Digimatic Caliper; Mitsutoyo, Kanagawa, Japan), and the mean
thickness at the three points closest to each lesion was used as the slab
thickness. The total volume of each lesion was calculated by multiplying the
area of the lesion by the sum of the thickness of the slab and 1.41 mm (the
thickness of the saw blade). This method of volume measurement was validated
by embedding a block of polymethylmethacrylate of known volume (see Appendix).
The characteristics of each osteolytic lesion and the presence of any erosion
of the medial wall were also recorded on the slab radiographs.
Statistical Methods
The presence of osteolytic lesions and their volumes as well as the
presence of medial wall perforations were assessed on the computed tomography
scans. Absolute and relative errors between data derived from the computed
tomography scans and those derived from the slab radiographs were then
determined. The absolute error was considered to be the difference between the
volume, in cubic centimeters, of the lesion measured on the computed
tomography scan with use of the VirtualScopics software package and the actual
volume, in cubic centimeters, of the lesion measured on the slab radiographs
with use of the Martell program. The relative error was considered to be the
percent error between the measurement made on the computed tomography scan and
the measurement made on the slab radiograph. The volumes measured on the slab
radiographs were assumed to represent the true volumes of the periacetabular
osteolytic lesions. We also assumed that all of the osteolytic lesions and
medial wall perforations were detected on the slab radiographs.
Linear regression was performed to calculate the agreement between the
measurements made on the computed tomography scans and those made on the slab
radiographs. A slope of 1.0 and an intercept of 0.0 cm3 would
indicate that the volumes measured on the computed tomography scans and those
measured on the slab radiographs were identical. To determine if one set of
measurements had a tendency to underestimate or overestimate the other, we
used a one-sample t test to determine if the mean of the volumetric
differences was significantly different than zero. Finally, a chi-square
analysis was used to determine if there was a significant difference between
the detection rates of the computed tomography without a femoral head in place
and those of the computed tomography with a femoral head articulating in
place. All data analysis was performed with use of SPSS statistical software
(version 8.0; SPSS, Chicago, Illinois). Probability values of <0.05 were
considered to indicate a significant difference.
Atotal of twenty-three lesions were identified on the 249 slab radiographs
of the nine retrieved and sectioned hemipelves. Of these twenty-three lesions,
twenty were identified on the computed tomography scans made without a femoral
head in place; thus, the detection rate was 87%. Two of the lesions that were
missed by the computed tomography were in the dome region and did not have
sclerotic borders; the other missed lesion was very small (0.27
cm3) and in the posterior acetabular region. In comparison, when
only plain anteroposterior radiographs were evaluated, only nine of the
twenty-three lesions were identified, a detection rate of 39%. When both an
anteroposterior radiograph and an iliac oblique radiograph were examined,
twelve of the twenty-three lesions were identified, increasing the detection
rate to 52%. The distribution of the lesions identified on the slab
radiographs, computed tomography scans, and plain anteroposterior and iliac
oblique radiographs is shown in Figure
1. Medial wall perforations occurred in three specimens, all of
which were identified by computed tomography when no femoral head was present.
None of the medial wall perforations were detected by the anteroposterior
pelvic and iliac oblique radiographs.
The volumes of the twenty-three osteolytic lesions as calculated from the
computed tomography scans and the slab radiographs are listed in
Table I. The mean volume (and
standard deviation) of the osteolytic lesions was 7.7 ± 13.6
cm3 (range, 0.5 to 54.6 cm3) when measured on the
computed tomography scans in the absence of a femoral head and 6.6 ±
11.9 cm3 (range, 0.3 to 50.4 cm3) when measured on the
slab radiographs. The average size of the three lesions not detected with
computed tomography was 1.8 ± 2.1 cm3 (0.27, 0.98, and 4.13
cm3), and all three lesions were smaller than 5.0 cm3.
With regard to the twenty lesions detected by computed tomography, the
difference in the volumes between the computed tomography scans and the slab
radiographs was 0.3 ± 1.1 cm3 (range, -1.6 to 4.2
cm3), which corresponded to a relative error of 7.1% ± 24.1%
(range, -10.61% to 100%). The average difference in the mean volumetric
measurements was significantly different from zero (p = 0.02), indicating that
the volumes measured on the computed tomography scans consistently
overestimated the volumes measured on the slab radiographs. Linear regression
analysis showed a strong and significant relationship (r2 = 0.997,
p < 0.001) between the volumes measured on the computed tomography scans
without a femoral head and those measured on the slab radiographs, with a
slope of 1.1 and an intercept of —0.18 cm3.
Computed tomography scans were made of five specimens with and without a
prosthetic femoral head in place to determine whether the femoral head created
metal-induced artifact that interfered with the accuracy of our computed
tomography technique. Eleven periacetabular osteolytic lesions and two medial
wall perforations were identified on the slab radiographs of these five
specimens. Both the computed tomography scans made with the femoral head in
place and those made without the femoral head detected the same nine lesions
and both medial wall perforations; both sets of scans failed to demonstrate
the same two lesions.
The mean volume of the eleven osteolytic lesions measured with computed
tomography with the femoral head in place was 12.7 ± 17.2
cm3 (range, 0.6 to 50.4 cm3), whereas the mean volume of
these lesions was 10.8 ± 15.9 cm3 (range, 0.57 to 50.4
cm3) when measured on the slab radiographs. The average error of
the computed tomography, when compared with the slab radiographs, was 0.86
± 3.6 cm3 (range, -20.2 to 10.41 cm3), which
corresponded to an average relative error of 4.6% ± 11.5% (range,
-11.5% to 20.7%). The average size of the lesions missed on the computed
tomography scans when a femoral head was present was 2.6 ± 2.2
cm3 (0.98 and 4.13 cm3). Both of the lesions that were
missed were less than 5.0 cm3. The average difference in the mean
volumetric measurements was not significantly different from zero (p >
0.05), indicating that the volumes measured on the computed tomography scans
with a femoral head in place did not consistently overestimate or
underestimate the volumes measured on the slab radiographs. The linear
regression analysis revealed another strong and significant relationship
(r2 = 0.985, p < 0.001), with a slope of 1.2 and an intercept of
—1.1 cm3.
Plain radiographs underestimate the extent and size of pelvic osteolytic
lesions behind metal-backed acetabular
components9-12.
It has been shown that, even with multiple radiographic views, up to 25% of
lesions are missed in regions that are poorly visualized on plain
radiographs17.
Computed tomography in conjunction with newly developed algorithms to diminish
streak artifact from metal components has recently been proposed as a more
sensitive and accurate method for detecting and measuring
lesions7,9,13,14.
Two recent studies have demonstrated the accuracy of computed tomography for
detecting and measuring artificially created osteolytic lesions in human and
bovine pelvic
models15,16.
In the current study, we examined specimens retrieved at autopsy from
individuals with a well-functioning total hip replacement at the time of
death. We believe that our model more accurately represents the true location,
size, and radiographic characteristics of osteolytic lesions found in
patients, as these lesions occurred naturally and reflected the in vivo host
response.
By comparing lesions identified on plain radiographs and on computed
tomography scans with those observed on slab radiographs of the retrieved
hemipelves, we found that computed tomography detected 87% of the lesions with
an error of 7.1% ± 24.1% (0.3 ± 1.1 cm3). In
contrast, anteroposterior and iliac oblique radiographs identified only 52% of
the lesions. Our ability to identify and measure osteolytic lesions that had
occurred in vivo was similar to the ability to identify artificially created
lesions in previous
reports15,16.
Using a cadaveric model, Claus et al. found a detection rate of 81% with an
error of 5.6% ± 24.8% (0.5 ± 2.3
cm3)16.
Stamenkov et al. reported that computed tomography had an accuracy of 96% for
identifying small and large lesions of known volumes in the ilium of a bovine
hemipelvis and a cadaveric
pelvis15. We are
encouraged that the streak artifacts resulting from the addition of the
chromium-cobalt femoral component did not diminish our detection rate, which
was 82% with an error of 4.6% ± 11.5% (0.86 ± 3.6
cm3).
We found that lesions that had occurred in vivo had several important
differences compared with artificially created lesions. First, although the
lesions that had occurred in vivo varied greatly in size, on the average they
were smaller than those created
artificially15,16.
The lesions that had occurred in vivo also were typically either found around
the rim, with small openings to the joint space, or had occurred directly
adjacent to holes in the cup and had ballooned out behind them. In fact, by
examining several computed tomography images, we could almost always detect
the communication pathway between the lesion and the joint space. Finally, the
lesions that had occurred in vivo typically had thickened, sclerotic edges,
which we consider a remodeling response of the remaining trabecular bone. We
attribute the detection of even the smallest in vivo bone defects to the
presence of a sclerotic border (Figs. 2-A
and 2-B).
While we believe that our autopsy retrieval model is more accurate than
models employing artificially created lesions, we acknowledge that our study
had several limitations. First, we had a limited number of specimens to
examine because of the small number of autopsy specimens that we are able to
obtain each year. Second, the quality of the computed tomography images
obtained in this study was optimal because all of the soft tissue had been
removed from the specimens before the plain radiographs and the computed
tomography scans were made. Finally, when we measured the actual volume of
each osteolytic lesion on the slab radiographs, we assumed that the area of
the lesion on each slab radiograph remained constant throughout the entire
3-mm thickness. This may not be true if the dimensions of the lesion rapidly
increased or decreased within the slab thickness. Moreover, we were able to
examine only half of each slab radiograph specimen because, for each 3-mm
increment that was cut, 1.41 mm was lost to the saw blade.
Nevertheless, we are encouraged by our findings. Although we still consider
plain radiography to be the most practical and important method of
postoperative evaluation following cementless total hip arthroplasty, we think
that computed tomography is becoming an increasingly useful tool with which to
examine selected patients. Because plain radiographs cannot detect all
periacetabular osteolytic lesions and are a poor predictor of lesion size, we
now use computed tomography to evaluate patients in whom we suspect osteolysis
on the basis of our plain radiographic examination. When we do obtain a
computed tomography scan, we examine the images to find the three-dimensional
location of the osteolytic lesions, determine the sizes of any lesions, and
identify which areas of the acetabulum are in fact supporting the acetabular
component. We have used serial computed tomography scans to determine the
timing of surgical intervention, and we have employed this information to
optimize preoperative planning to determine the best surgical approach and the
array of prosthetic and bone graft requirements at revision surgery.
The method of validating the volumetric measurements on the slab
radiographs is available with the electronic versions of this article, on our
web site at
(go to
the article citation and click on "Supplementary Material") and on
our quarterly CD-ROM (call our subscription department, at 781-449-9780, to
order the CD-ROM). ?