Thirty-four skeletally mature, heartworm-free, mongrel dogs weighing
28.9 to 40.9 kg were used. There were twenty-three female and eleven
male animals. The animal protocol was reviewed and approved by the
institutional animal care and use committee. A pelvic radiograph
was made of each dog to screen for signs of bone disease or hip
dysplasia and to confirm skeletal maturity.
Experimental Design (Fig. 1)
In phase I, sixteen dogs were used to create the defect model.
Seven defects were left untreated (Group I), and nine were treated
with the strut-grafting technique (Group II). In phase II, the surgical
procedure was modified, in order to provide a more reproducible
collapse of the articular surface, by allowing the articular cartilage
of the trapdoor to collapse into the untreated defect intraoperatively in
eighteen dogs. The trapdoor was left collapsed in six femoral heads
(Group III), the collapsed segment was elevated with bone-grafting
in six (Group IV), and the collapsed segment was elevated with a combination
of bone-grafting and osteogenic protein-1 in six (Group V). All
animals in all five groups were killed at four months postoperatively.
Surgical Procedures
Dogs were tranquilized with acetylpromazine maleate (1 mg/kg
intramuscularly) followed by anesthesia with Ethrane (enflurane),
nitrous oxide, and oxygen. With use of aseptic technique, the right
or left hip (chosen randomly) of each experimental animal was exposed
through an anterolateral approach46.
In this approach, the major vessels supplying the head (the lateral
circumflex branch of the profunda femoris) are preserved. The hip
capsule was exposed, and a capsulotomy was performed without dislocating
the hip to expose the femoral head and neck. A 1-cm-diameter trapdoor
composed of articular cartilage and its underlying subchondral bone was
created on the anterolateral aspect of the femoral head (Fig. 1). The trapdoor
was removed, and a mass of trabecular bone was excised from the
femoral head to create a spherical defect 2 cm in diameter. The
size of the trapdoor corresponds to the size (70% to 80% of the
femoral neck diameter) that has been utilized in trapdoor procedures
in human clinical studies9,12,14,19.
Bone was removed until the subchondral bone and cartilage adjacent
to the defect was 2 mm thick. After the grafting procedures, the
cartilage flap was replaced flush with the surrounding articular
surface and held in place by two 6-0 sutures and by closing the
capsule with two 6-0 sutures. The hip abductors were then repaired.
In phase II, the surgical procedure described above was modified
after the bone was removed. All animals had intraoperative collapse
of the trapdoor (consisting of articular cartilage and its underlying subchondral
bone) by application of light pressure to the trapdoor with use
of an 8-mm-diameter bone tamp until the trapdoor was depressed into
the defect 3 mm. In Group III, this segment was left collapsed.
In Groups IV and V, the trapdoor was elevated, the defect was filled
with bone graft without or with osteogenic protein-1, and the trapdoor was
replaced, flush with the surrounding cartilage.
After all procedures, the animals were permitted to recover from
the general anesthesia and were allowed free activity. The animals
were observed daily for wound-healing and gait assessment. All animals
received one preoperative dose and two postoperative doses of antibiotics
(gentamicin, 4 mg/kg). When grafting was to be performed,
autogenous corticocancellous bone was taken from the ilium through
the same incision and was utilized to pack the defect. For animals
to be treated with osteogenic protein-1 (Stryker Biotech, Hopkinton, Massachusetts),
an equal volume of this material was mixed with the autogenous cancellous
bone graft (250 mg osteogenic protein-1/g of bone) and packed
into the defect after the cortical struts were placed.
Load-Bearing
The ground-reaction forces were determined before and at eight
and sixteen weeks after the surgery to assess functional walking
in the phase-I studies. Dogs were led along a 7-m-long and 1-m-wide
platform. Velocity was maintained consistently and was measured
as the time between forelimb and hindlimb strikes. A force-plate
(model OR-6-6; Advanced Mechanical Technology, Newton, Massachusetts)
was mounted level with the top of the platform. As the dog struck
the force-plate, an analog signal was digitized and processed by
specialized software designed to interpret canine gait47,48. Each individual force-plate-analysis
session consisted of a minimum of six trials for both the left and
the right hindlimb. Vertical ground-reaction force was determined
for each trial, and the mean value was calculated. Data were normalized
to body weight. Differences between the mean values in Groups I
and II and time-sequential changes in mean values in each group
were analyzed.
Radiographic Analysis
Anteroposterior and lateral radiographs were made preoperatively,
immediately postoperatively, and at one, two, and three months postoperatively.
After the animals were killed (four months postoperatively), final
radiographs of the dissected femoral heads were made with a Faxitron
machine (Hewlett-Packard, Palo Alto, California).
The radiographs were evaluated by four clinicians who were blinded
to the animal groups and who used a scoring system in which filling
of the defect was graded as none or scant, minimal, moderate, or excellent.
None or scant healing of the defect denoted minimal bone formation
(filling <25% of the defect) composed mostly of
noncontiguous areas of minimal density. Minimal healing indicated
mostly contiguous areas of normal density that filled 25% to
75% of the defect. Moderate healing encompassed normal
density in 76% to 95% of the defect. Excellent
healing denoted normal density in >95% of the
defect.
Harvesting of Femoral Heads
Specimens from the wound were taken for aerobic and anaerobic
bacterial cultures after the animals were killed with an overdose
of sodium pentobarbital (200 mg/kg). Femoral samples were
then harvested and were examined macroscopically and photographed.
The bone was then dissected free and processed for mechanical testing,
followed by histological examination.
Biomechanical Testing (Figs. 2-A and 2-B)
After the animals were killed, both femora were excised from
each dog and were placed in a bath of ice and lactated Ringer solution
until testing. The femora were cut approximately 5 cm distal to
the neck and were embedded in a cube of low-melting-point metal
(Cerro-Bend Metals, Belfonte, Pennsylvania) at an angle of 30° to
the vertical, to position the weight-bearing surface at approximately
the highest point of the embedded femur. Indentation tests were
performed after removal of the articular cartilage from the specific
test sites with a scalpel to expose the subchondral bone. A line
was drawn along the femoral head between the osseous prominence
on the posterior edge of the greater trochanter and the fovea. A
second line originating at the fovea was drawn perpendicular to
the first line. A circular template with a 0.5-in (1.27-cm) diameter
was aligned with the lines. The center of the circle marked the
center test site. For the Group-I and Group-II specimens, four additional
test sites evenly spaced along the circumference of the circle were
identified on each femoral head; they were identified as the anterior-medial,
anterior-lateral, posterior-medial, and posterior-lateral test sites. This
procedure resulted in a testing region that approximately spanned
the weight-bearing region of the femoral head. For the specimens
in Groups III, IV, and V, only three sites were used, to reduce the
variability in the stiffness measurements49-51.
The same central test site was used along with one site approximately
0.25 in (0.64 cm) anterior to it and another approximately 0.25
in posterior to it. Each embedded femur was secured within a fixation
jig fixed to the loading frame of a mechanical testing machine (Bionix
858; MTS, Eden Prairie, Minnesota) on a base-plate that allowed
translation in two dimensions and rotational adjustment in three.
The base-plate is a rotational joint within the fixation jig, which
was used to position each test site approximately perpendicular
to the coding actuator. A compressive load was applied to each test
site at a rate of 2 mm/min to a maximum load of 125 N with
use of a 3-mm-radius indentor52.
Preliminary experiments showed a failure strength of nearly 400
N for hollowed-out femoral heads. Each site was tested four times
in succession. The stiffness consistently increased from the first through
the third test. The maximum stiffness from the third and fourth
tests was used for statistical analysis. At each test site, the
stiffness value of the surgically altered head was normalized by
the stiffness of the untreated, contralateral control. During testing,
each femoral head was kept moist with saline solution.
Histological Study
After biomechanical testing, the proximal part of the femur was
isolated and then sectioned longitudinally and transversely. The
undecalcified bone was sequentially dehydrated in increasing concentrations
of ethanol. The specimens were then embedded in methylmethacrylate
according to the technique of Emmanual et al.53.
Each block was sectioned at 200 mm with use of the EXAKT processing
system. The specimens were then ground to 70 to 100-mm sections.
Next, each block was cut into 5-mm sections with use of an Exakt-System
microtome (Exakt Apparatebau, Munich, Germany). The slides were
then stained with hematoxylin and eosin), MIBS (Villanueva mineralized
bone stain), Goldner trichrome, toluidine blue, and safranin O.
Data Analysis
Parametric comparisons of each treated head with the contralateral,
intact head were made with use of the Student t test for matched
pairs. Comparisons between independent groups (untreated defects, defects
treated with grafting, and defects treated with grafting and osteogenic
protein-1) were carried out with use of one-way analysis of variance (randomized)
with the post hoc Newman-Keuls between-group comparisons
test. Nonparametric testing between two independent groups was done
with the Kruskal-Wallis test. Comparison between groups with regard
to success or failure of healing was performed with the Fisher exact
test. For the biomechanical studies, a nested analysis of variance
was used to compare the normalized stiffness values between the
group treated with grafting and the untreated group. The nested analysis
of variance treated each normalized stiffness value as an equivalent
data point, regardless of its magnitude. A separate analysis of
variance, combined with a Tukey-Kramer test for multiple comparisons,
was performed to examine the variations in the stiffness magnitudes
among the five test sites in the control femora.
All animals tolerated the operation well. All wounds healed with
no dehiscence, and no other complication, such as hip dislocation
or deep infection, was encountered.
Load-Bearing
All animals were able to stand unassisted within the first twenty-four-hour
period. Ground-reaction forces were determined with a force-plate
for the Group-I animals (untreated defects) and the Group-II animals
(defects treated with grafting) at eight and sixteen weeks postoperatively.
Load-bearing on the treated limb was significantly higher in Group II
than it was in Group I (p < 0.01). In Group I, dynamic
loading of the involved limb had significantly decreased to a mean
of 87% of the preoperative value at eight weeks after the
surgery (p < 0.05). The level increased slightly, to a
mean of 91%, by sixteen weeks, but the increase was not significant.
In Group II, no significant decrease in dynamic loading of the involved
limb, as compared with that of the contralateral limb, could be detected
at eight or sixteen weeks after the surgery.
Gross Appearance of the Femoral Heads
Three of the seven femoral heads in Group I and one of the nine
heads in Group II appeared to have irregularities in the articular
cartilage and evidence of collapse at the defect site. Macroscopically,
the articular cartilage of these four collapsed heads was depressed
between 1 and 4 mm. The articular cartilage of the trapdoor and
the cartilage adjacent to the trapdoor were depressed, soft, friable,
and fissured. All of the other femoral heads in Groups I and II had
minimal cartilage irregularities. The perimeter of the trapdoor
had a grayish-white appearance, was soft, and appeared to be composed
of fibrous tissue. This approximately 0.5-mm band of tissue surrounded
essentially normal-appearing, nondepressed cartilage of the trapdoor.
The anatomy was restored and was not grossly different between the two
groups.
All of the Group-III animals (collapse of the trapdoor without
grafting) had collapse of the femoral head ranging from 1 to 4 mm.
The articular cartilage was soft, friable, and yellow and had a
rough surface with clefts and fissures.
In Group IV (collapse of the trapdoor followed by grafting without
osteogenic protein-1) and Group V (collapse of the trapdoor followed
by grafting with osteogenic protein-1), the trapdoor cartilage as
well as the adjacent cartilage overlying the filled defect had an
almost normal appearance. There were only minor fibrillations, which
for the most part resembled normal hyaline cartilage, in some specimens. The
perimeter of the trapdoor in these specimens was similar to that
in Groups I and II.
Radiographic Analysis (Figs. 3-A, 3-B, 3-C, 3-D, and 3-E)
Postoperatively, all of the Group-I defects were clearly visible
and sharply demarcated on standard anteroposterior and lateral radiographs
(Fig. 3-A).
All of the Group-II femoral heads exhibited a greater healing response
(sclerosis across the defect increasing at each time-interval, with
no femoral head appearing collapsed; Fig. 3-B) than the Group-I femoral heads
(persistence of the lucent defect with minimal sclerotic bridging).
By four months, seven of the nine Group-II femoral heads showed
excellent radiographic healing of the defect and two showed moderate
healing. There was minimal healing of the defect in all seven Group-I
animals.
The radiographs made when the animals were killed revealed minimal
healing in all six Group-III femoral heads (Fig. 3-C). The defects
in Group V healed faster radiographically than did those in Group
IV. Of the six femoral heads in Group IV, four showed excellent healing
and two showed moderate healing (Fig. 3-D). In Group V, all six femoral heads
demonstrated excellent healing (Fig. 3-E).
Histological Analysis (Figa. 4-A and 4-B)
Histological analysis demonstrated fibrillation of the cartilage
in all of the femoral heads. In Group III, the trapdoor remained
collapsed in all specimens (Fig. 4-A), whereas there were no areas of
cartilage depression in Group IV (Fig. 4-B) or V. There was no significant
difference in cartilage thickness among the three groups, although Group
IV had a slightly greater cartilage width (mean, 212.5 mm) than
did Groups V and III (mean, 167 and 166 mm, respectively). These
values were not significantly different from the control widths
(mean, 185 mm).
Biomechanical Indentation Testing
The mean normalized stiffness value was significantly greater
(p < 0.05) in Group II than it was in Group I at three
of the five indentation test sites (anterior-lateral, anterior-medial,
and posterior-medial) (Table I). The nested analysis of variance
did not indicate that the test site significantly influenced the
normalized stiffness. A separate analysis of variance combined with
a Tukey-Kramer analysis for post hoc testing was
used to compare the stiffness magnitudes among the test sites of
the untreated, control femoral head from each dog. The stiffness
values ranged from 397 151 N/mm for the anterior-lateral
test site to 1021 228 N/mm for the posterior-medial test
site. The two anterior sites were significantly less stiff than
the central site and the posterior-medial site. The anterior-lateral
site was also significantly less stiff than the posterior-lateral
site.
In contrast to the comparisons between Groups I and II, comparisons
among Groups III, IV, and V with use of nested analysis of variance
and individual comparisons of the normalized stiffness values at
each of the three test sites did not indicate a significant difference
in the normalized stiffness values (p = 0.40) (Table I). At each site,
the standard deviations of the mean normalized stiffness values
were similar to the standard deviations in Groups I and II. The
mean stiffness values for the untreated controls in Groups III, IV,
and V were 488 108 N/mm at the anterior site, 856 200
N/mm at the center site, and 1076 270 N/mm at
the posterior site.
Femoral Head Bone-Grafting and Enhancement
To our knowledge, Ganz and Büchler8 were
the first to mention cancellous grafting through a window in the
femoral neck combined with osteotomies; however, they did not report
their results. Various investigators in Japan modified this procedure
by using strut grafts through a window in the femoral neck. Rosenwasser
et al.14 utilized a similar technique,
in which they combined complete evacuation of the femoral head with replacement
with cancellous iliac-crest bone; they reported thirteen excellent
results in fifteen patients at a mean of twelve years after this
procedure. Nonvascularized bone-grafting through a trapdoor in the
cartilage of the femoral head was apparently first described by
Meyers et al.11. We12 utilized a modification of this
procedure and reported twenty good and excellent clinical results at
a mean of 4.6 years in twenty-four hips with stage-III disease.
Growth and differentiation factors can be added to autogenous bone
graft without modifying the procedure substantially. Any enhancement
of healing that results may shorten the period of restricted weight-bearing,
ensure better compliance with rehabilitation protocols, and hopefully
lead to more successful outcomes. These goals led us to perform
the present study.
Osteonecrosis Models
Over the last seventy years, a great deal of work has been devoted
to developing a relevant model for osteonecrosis. These efforts
have met with little success. No animal model has produced histological
or morphological changes representative of the pathological findings
in humans33,40-45. The defect
model in the present study is useful as it simulates the defect
created in the treatment of late-stage osteonecrosis with the trapdoor approach.
The trapdoor procedure, or modifications of it employing various
bone-grafting procedures, have been utilized by multiple authors7,9,11,12,14,19. A main feature of
the trapdoor procedure in human patients is the removal of all necrotic
bone. The evacuated femoral head is left with viable and vascular
bone, as it was in our animal model. Knowledge gained from assessment
of the healing of this defect after treatment with strut bone-grafting
with or without bone morphogenetic protein in dogs could be clinically
useful when these same defects are created in patients. In summary,
this is an important model for the study of osteonecrosis because
it (1) uses the femoral head, (2) involves a subchondral defect,
(3) has structural compromise, and (4) is similar to defects made
in treatment involving vascularized bone-grafting.
Present and Related Studies
In the present study, although all of the Group-I (untreated)
defects persisted as seen radiographically, they did not consistently
result in collapse of the femoral head; in fact, only three of the
seven heads collapsed. As a result of this finding, it was decided
that all subsequent femoral heads would be collapsed after the defect
had been made at the primary operation and then either left collapsed (Group
III) or treated with bone-grafting (Groups IV and V). Both Group
IV and Group V (grafting without and with osteogenic protein-1)
exhibited excellent healing by four months, as demonstrated by the
radiographic and biomechanical studies.
Mazières29 carried
out preliminary experiments with use of recombinant human bone morphogenetic
protein-2 in a pig model of osteonecrosis. Three animals underwent
bilateral core decompression of the femoral head followed by introduction
of a mixture of recombinant human bone morphogenetic protein-2 and
blood clot (Genetics Institute, Cambridge, Massachusetts) into the
core channel on one side. Three months after the surgery, standard
radiographs and magnetic resonance images as well as histological
examination revealed evidence of excellent healing of the treated
channels compared with that of the open channels. The one-month
histological findings in the hips treated with bone morphogenetic
protein-2 were comparable with the three-month findings in the hips
treated without bone morphogenetic protein-2. These findings are consistent
with those of the present study, in which the use of adjunctive
bone morphogenetic protein was found to be associated with an acceleration
of healing at early time-points.
Scully et al.33 studied the
effects of recombinant human bone morphogenetic protein-2 and vascularized
fibular grafting in a canine model of osteonecrosis that had been
induced by a combination of soft-tissue dissection and freezing.
There were no radiographic differences between the animals treated
with bone morphogenetic protein-2 and control animals treated with
vascularized fibular graft alone, but quantitative histomorphometry
did reveal increased amounts of viable bone at both eight and twelve weeks
in the former group when compared with the latter. Scully et al.
suggested that revascularization after fibular grafting is accelerated
by the addition of bone morphogenetic protein-2. Our study also indicates
that growth and differentiation factors enhance the early healing
response in an animal model of osteonecrosis.
In summary, we used an experimental model that is potentially
valuable for evaluation of the efficacy of procedures performed
to elicit the healing of osseous defects in the femoral head and
to prevent collapse of the overlying articular cartilage. Radiographic
and biomechanical studies indicated that bone-grafting either with
or without osteogenic protein-1 resulted in excellent healing of
2-cm-diameter spherical defects and prevented the displacement and
collapse of the trapdoor and the surrounding cartilage. Untreated
defects did not heal. Additional studies investigating additional
time-periods between implantation of osteogenic protein-1 and assessment
of results as well as doses of osteogenic protein-1 are warranted.
Note: The authors thank David Ruegger and Jaime Kemler for their
assistance with the histological preparation. They also acknowledge
the following individuals for their technical and support contributions
to this investigation: Dawn M. LaPorte, MD, Henri Pierre-Jacques,
MD, Ivan H. Pacheco, MD, Rad K. Payman, MD, A.H. Reddi, PhD, Melissa Schlenker,
BA, Anthony Valdevitt, PhD, and James F. Wenz, MD.