From 1975 to 1998, 405 preserved allografts were used to reconstruct
skeletal defects created by limb-salvaging resection for the treatment
of musculoskeletal tumors. Sixty specimens retrieved from these patients
and thirteen other specimens provided by other institutions form
the basis of these observations.
Reasons for Retrieval
Forty-two specimens (58%) were retrieved because of
a complication associated with the allograft. Twenty-four specimens
(33%) were retrieved because of a complication associated
with a tumor: twelve were retrieved following amputation because
of a local recurrence and twelve were retrieved at autopsy after
death from pulmonary metastases. Seven other specimens were retrieved because
of miscellaneous reasons.
Types of Grafts
Sixty-six (90%) of the specimens were massive allografts:
thirty-seven were osteoarticular, twenty-one were intercalary, and
eight were composite. Seven specimens (10%) were intercalary
grafts that had been either inlaid or onlaid in a large defect: four
were fibular grafts and three were corticocancellous iliac-crest
grafts.
Sites of Grafts
Sixty-five massive grafts were used in reconstruction of the
femur (forty-one), tibia (thirteen), or humerus (eleven), and eight
were used in other sites.
Duration in Situ (Fig. 1)
The grafts were retrieved from two to 156 months after implantation.
Fifty-four (74%) were retrieved during the first three
years after implantation. The average duration in situ was
twenty-eight months for those retrieved because of complications
associated with the graft and thirty months for those retrieved
because of complications associated with a tumor.
Patient Demographics
The specimens were retrieved from thirty-three male and forty
female patients. The average age at the time of retrieval was twenty-eight
years (range, five to sixty-nine years). Forty-three patients had received
preoperative and/or postoperative chemotherapy, and three
patients had received postoperative radiation therapy.
Methods
Various combinations of radiographic, macroscopic, and histological
studies were done on the specimens, and these observations were
correlated with the clinical and radiographic data, as described in
our earlier study1.
After the initial dissection, radiographs and photographs of
the intact specimens were made and then the specimens were dissected.
The cut surfaces of the bisected slabs were photographed under incandescent
light and, after in vivo labeling with tetracycline
had been done, under ultraviolet light.
The slabs were fixed, decalcified, embedded in celloidin, cut
into 20-m sections, and stained with hematoxylin and eosin. These
macrosections were studied with low-power microscopy (with a magnification
of four to ten times), and spatial maps were prepared from tracings
of the macrosections to determine the extent and distribution of
revascularization and repair. On these spatial maps, areas of host
callus, appositional new-bone formation on the surface of the allograft,
osteoclastic bone resorption, internal cortical revascularization,
and osteonal remodeling were color-coded for area measurements by
planimetry. These features were then converted to proportions of
the specimen as quantitative estimates of the various reparative
processes. Selected regions of the specimens were decalcified, embedded
in paraffin, cut into 5-m sections, and stained with hematoxylin
and eosin. When articular cartilage was examined, toluidine blue
and safranin-O stains were also used. These specimens were studied
with high-power microscopy.
Union
Two types of junctions were available for study. Most were cortical-cortical
(forty-four specimens), whereas a few were cancellous-cancellous
(twelve specimens). At ten of the cortical-cortical junctions, additional
autogenous cancellous grafts had been onlaid.
Cortical-Cortical Junctions
At the cortical-cortical junctions, healing took place by bridging
external callus that originated from the periosteum of the host
bone and extended for up to 3 cm on the surface of the allograft,
where it became annealed (Fig. 2-A). The callus did not appear to
have extended from both sides of the junctions (host and graft)
and then to have joined at the osteotomy site in any specimen. The
gap at the osteotomy site initially filled with reparative fibrovascular
tissue from the surrounding host soft tissues. This, in turn, was replaced
by immature trabeculae that gradually matured into haversian bone.
The external callus about the osteotomy site consistently was more mature
than the bone filling the gap. Although immature trabeculae were
observed in the gap in specimens retrieved at four months, maturation
into haversian bone was not consistently seen until twelve months.
Radiographs demonstrated a persistent radiolucent gap at the osteotomy
site, even when it was filled with mature lamellar trabeculae joining
the two cortices, until the uniting bone within the gap had converted
to haversian bone. When the bone filling the gap had matured into
cortical bone, the haversian canals in the uniting seam were oriented
perpendicularly to the long axis of the bone, presumably because
of the orientation of the initial ingrowth of blood vessels from
the periphery of the gap. The site of union between the uniting
seam and the abutting allograft remained clearly marked by a distinct
cement line. Little or no fibrovascular repair tissue penetrated
into the haversian canals of the allograft, even in specimens that
had been in situ for up to five years. The nonstress
orientation of the bone in the gap, however, persisted for years and
was seen as late as eleven years without remodeling parallel to
the long axis of the bone (Fig. 2-B).
In two specimens (retrieved at twenty-eight and forty-two months)
that were subjected to torque to failure at the junctions, the failure
occurred exactly along the cement line that denoted the histological line
of union (Figs. 3-A and 3-B). None of the specimens
had primary union between abutting host and allograft cortices.
Only one specimen had areas of endochondral ossification. Correlation
between the radiographic and histological findings showed agreement
in all but three specimens. These specimens, each in situ for
less than a year, showed a persistent radiolucent gap at the osteotomy
site. However, on histological examination, they demonstrated immature
trabeculae uniting the gap.
Three specimens that had been fixed with a plate demonstrated
no meaningful external callus at the cortical-cortical junctions.
Rather, a bridge of cancellous bone extended from the medullary
canal of the host, filled the osteotomy site, and extended up the
canal of the allograft for 2 to 3 cm (Figs. 4-A and 4-B).
Seventeen cortical-cortical junctions in specimens from patients
treated with preoperative and postoperative chemotherapy were available
for histological evaluation. Of the nine specimens retrieved within
twelve months, only one showed immature union. Of the eight specimens
retrieved between fourteen and twenty-eight months, five showed immature
union, two had haversian union, and one contained only mature fibrovascular
reparative tissue in the gap.
Cortical-Cortical Nonunions
Six allografts were retrieved because of nonunion, and one was
found to have nonunion when it was retrieved because of tumor recurrence.
Two (mentioned above) were retrieved from patients who had received
chemotherapy, and one (the graft that was retrieved because of local
recurrence) was from a patient who had received radiation therapy.
The remaining four were retrieved more than twelve months after
implantation from patients who had not received chemotherapy or
radiation therapy. All of these specimens had similar histological
findings; a varying amount of external callus extended from the
host cortex but was prevented from annealing to the allograft by
an envelope of fibrovascular tissue filling the gap and encasing
the end of the allograft. The tissue contained few inflammatory
cells and occasional histiocytes and multinucleated giant cells.
Gram stains and cultures of this tissue were negative. The surfaces
of the abutting allograft were intermittently pockmarked with Howship
lacunae, with some containing osteoclasts, which resulted in only
modest amounts of resorption radiographically.
Of three specimens that were retrieved because of infection and
had been in situ for more than twelve months, two
in which the infection involved the cortical-cortical junction did not
have union. In these two specimens, the osteotomy gap was filled
with chronic inflammatory tissue and there was intense osteoclastic
resorption of the adjacent allograft. In the third specimen, a composite
allograft-total hip replacement, the infection was confined to the
periprosthetic tissues and the uninvolved distal junction was well
healed.
Of the ten specimens in which supplementary autogenous cancellous
grafts had been placed about the cortical-cortical junction, eight
had new-bone formation uniting the autograft to the surface of the allograft
as early as six months after implantation. When the autogenous graft
was in contact with the allograft, there was not only intense new-bone
formation on the surface of the graft but also, surprisingly, more
extensive internal repair in the underlying cortex of the allograft
than in the remainder of the graft. In the two specimens in which
there was no union between the autogenous graft and the allograft
by either an external callus or a seam of bone between the abutting
cortices, the autograft was separated by a zone of fibrovascular repair
tissue from the external surface of the allograft. Similar observations
were noted in the four specimens in which a vascularized autogenous fibular
graft had been placed in juxtaposition to the allograft. In each
specimen, the vascularized graft had united to the allograft and
the region of the allograft beneath the site of union had been much more
extensively repaired than the remainder of the allograft.
Cancellous-Cancellous Junctions
Union at the cancellous-cancellous junctions took place uniformly
and was present as early as four months. These specimens had no
external callus. Fibrovascular repair tissue from the host bone
had invaded the marrow spaces of the allograft and deposited seams
of reparative bone on the surfaces of the trabeculae of the allograft,
uniting them to the trabeculae of the host bone. The combination
of new bone encasing the preexisting allograft trabeculae produced
substantially thickened trabeculae, which was reflected radiographically
by a narrow zone of increased radiodensity just on the allograft side
of the junction. In comparisons of the macrosections and the radiographs
of the cancellous junctions, the 2 to 4-mm zone of radiodensity accurately
portrayed the depth of penetration of trabecular repair. Deep to
this zone of thickened, repaired trabeculae, the original architecture
of the allograft was well preserved, but the marrow spaces were
filled with dense hypovascular fibrous tissue that appeared to block
additional penetration of the allograft by host reparative fibrovascular
tissue. Beyond this zone, deeper in the allograft, the acellular
trabeculae preserved the original architecture of the allograft,
while the marrow spaces were filled with the acellular remnants
of fatty marrow. Even after many years, the cancellous area deep
to the fibrous barrier remained composed of undisturbed necrotic
remnants of allograft marrow and intact but acellular trabeculae.
Cortical Repair
The extent and distribution of cortical revascularization and
repair were studied in fifty-eight massive allografts. Repair of
the cortices was both external and internal.
First Year
Sixteen specimens were retrieved at less than twelve months.
The external surface of the allografts had scattered areas of osteoclastic
resorption, which were particularly evident where there was close
contact with host soft tissue. The osteoclastic resorption formed
small erosive cavities that were filled with fibrovascular reparative
tissue and occasionally were still occupied by an osteoclast. These
cavities occupied, on average, approximately 20% of the
external surface of the graft. In between the resorption cavities,
a 1 to 2-mm seam of appositional new bone was annealed by a cement
line to the unresorbed surface of the allograft. This newly deposited
surface bone was seen as early as four months and by one year occupied
an average of approximately 40% of the surface of the graft. Associated
with these surface changes, buds of fibrovascular tissue penetrated
into the Volkmann canals on the surface of the graft (Fig. 5). Occasionally,
the penetrating fibrovascular buds were accompanied by osteoclasts
enlarging the canals, and, where the canals had been enlarged, thin
seams of new bone lined the canal for a distance of 1 to 2 mm. The
penetration by the revascularizing buds from the external surface
rarely exceeded 2 mm during the first year.
At the osteotomy site, penetration of the transected haversian
canals in the graft cortices by buds of fibrovascular repair tissue
was rarely seen, and, when present, it extended for only 1 to 2
mm. Osteoclastic resorption of the edges of the revascularized canals
was also unusual. Direct osteoclastic resorption into the exposed
surface, a so-called cutting cone, was also rarely seen during the
first year.
In the medullary canal, ingrowth of fibrovascular reparative
tissue surrounding necrotic cancellous trabeculae extended no more
than a few millimeters with little or no new-bone formation. Internal
repair was almost completely absent in eight specimens, involved
about 5% of the graft in five, and involved about 10% of
the graft in three.
Second Year
Twenty specimens were retrieved in the second year. During this
period, the extent of osteoclastic surface resorption almost doubled,
to approximately 40%. At the same time, the extent of surface apposition
of new bone also doubled, to approximately 80%, filling
in many of the resorption cavities seen in specimens retrieved earlier.
However, in scattered areas of some specimens, an even pattern of
repair was not seen. In these specimens, the external surface was
pockmarked with unrepaired cavities ranging in size from several
micrometers to a few millimeters. These cavities were filled with loosely
arranged mesenchymal proliferations peppered with chronic inflammatory
cells. Focal concentrations of inflammatory cells, foreign-body-type
giant cells, or large histiocytes were only occasionally seen in
these cavities. In some areas, such cavities were interspersed between
areas of external repair by appositional new bone, while in others they
were closely clumped together. Immediately beneath these areas of
unrepaired surface erosion, no revascularization or internal repair
was seen.
At two years, somewhat deeper internal repair was directed by
the haversian and Volkmann canals of the graft. Vessels and osteoclasts
enlarged the diameter of the canals but rarely penetrated deeper
than 5 mm from the cortical surfaces. Within the enlarged canals,
osteoclastic resorption seldom extended beyond the peripheral cement
lines of the surrounding osteons, and the interstitial lamellae beyond
remained completely undisturbed by repair. In many of the enlarged
canals, resorptive activity had ceased, and the walls of the canals
were lined by seams of newly deposited living bone. Such repair
occupied no more than 20% of the graft and was mostly confined
to the peripheral aspects of the graft.
At the osteotomy site, deeper penetration by revascularizing
repair tissue was evident but lagged behind the repair from the
surface in almost all specimens (Figs. 6-A and 6-B).
In the medullary canal, the repair tissue extended up to 2 cm
but only intermittently differentiated into osteoblasts with new
bone laid on old necrotic trabeculae on the endosteal surface of
the cortex. At the leading edge of this repair tissue, it frequently matured
into dense fibrous tissue, impeding further reparative incursion.
After the Second Year
Twenty-two specimens retrieved more than two years after implantation
were available for study: fifteen had been retrieved between two
and five years after implantation and seven, between five and thirteen
years.
In the typical specimen, osteoclastic resorption was markedly
decreased at the osteotomy site where union had occurred and on
the external cortical surface. The thin layer of appositional bone
covered a gradually greater proportion of the surface but had not
increased in thickness. In contrast, revascularization and proliferation
of osteoblasts repairing the excavated haversian canals and the
Howship lacunae had increased. Typically, the osteoclastic erosion
within an individual osteon had rarely reached the peripheral cement
line or invaded the adjacent interstitial lamellae before it subsided
and osteoblastic apposition superseded it (Figs. 7-A and 7-B). The cancellous
trabeculae of the medullary canal and the endosteal surfaces of
the graft remained unrepaired throughout the duration of the study.
Although the extent of repair slowly progressed, it rarely penetrated >10
mm into the cortex. The variability in the extent of repair, however,
was much greater in the group of specimens retrieved later than
in the specimens retrieved after less than two years. In the typical
specimen, the total extent of internal repair remained between 20% and
30%. However, a specimen retrieved at five years showed <5% repair,
a specimen retrieved at thirteen years showed only 20% repair,
and a specimen retrieved at eight years showed 70% repair.
Soft-Tissue Attachment
The interface between the allograft and the adjacent soft tissues
was studied macroscopically in forty-three specimens. The soft tissues
were attached to between 20% and 40% of the surface
area in the specimens retrieved before six months, to between 50% and
80% of the surface area in specimens retrieved between
six and twelve months, and to >80% of the surface
area in those retrieved after twelve months.
Labeling with tetracycline demonstrated the areas in which a
thin seam of new external repair bone had been deposited upon the
surface. Macroscopic and histological study of these areas showed
that heavy strands of collagen ran from adjacent muscle, fascia,
or ligament into the seam of new bone. Histological study under
polarized light showed continuity of the collagen fibers as they
extended from the soft tissues into the appositional new bone, but the
collagen fibers did not cross the cement line that demarcated the
living appositional new bone from the underlying acellular necrotic
cortex of the allograft. In areas where the adherent soft tissues had
been forcibly stripped from the graft, the tissues debonded along
the demarcating cement line. In areas of substantial surface erosion
in specimens that had been retrieved earlier, there was no bonding
of the soft tissue as these areas were separated from the underlying
bone by a film composed of fibrovascular repair tissue, scattered
chronic inflammatory tissue, and an occasional foreign-body giant
cell. When these erosion cavities had been filled in with appositional
new bone in the specimens that had been retrieved later, the soft
tissues became bonded as in the other areas.
In the infected specimens, there was no seam of appositional
new bone on the graft surface, and the surrounding edematous tissues
were separated from the graft by a zone of chronic inflammatory
granulation tissue with no soft-tissue attachment directly to the
graft.
Resorption of Allografts
Eight specimens were retrieved because of resorption of substantial
portions of the allograft. Consistent findings in these specimens
were encasement of the graft by tissue composed of chronic inflammatory
cells, histiocytes, and foreign-body giant cells in a background
of proliferating mesenchymal cells. The acellular matrix was undergoing
active osteoclastic resorption. There was no revascularization or
internal repair of the remaining bone or any host-soft-tissue attachment.
The specimens that had been retrieved later showed marked fragmentation. Bacterial
cultures were consistently negative.
A particularly interesting case was that of a fifteen-year-old
girl who had reconstruction of the distal one-third of the femur
with an osteoarticular allograft (Fig. 8-A). Plate fixation failed, and nonunion
and fragmentation led to removal of the graft at fourteen months (Fig. 8-B). The defect
was reconstructed with a second osteoarticular graft, which was
securely united at eight months (Fig. 8-C). Gradually, over the next six
months, there was marked resorption of the proximal portion of the graft
(Fig. 8-D).
The distal portion, including the joint, remained intact, and the
patient had a well-functioning knee. In an effort to prevent a pathological
fracture and to preserve the knee, a vascularized fibular graft
was used to bypass the areas of greatest resorption. Curettings
from the areas of resorption demonstrated findings similar to those
described above. The vascular graft united to the host femur proximally
and the allograft metaphysis distally, and, surprisingly, during
the next twelve months the areas of resorption reossified and remained
so for the following twelve years (Fig. 8-E).
Allograft Fracture
Thirteen specimens were retrieved because of fracture. In twelve
specimens, the fracture was diaphyseal and was at the end of a plate
(seven), at a screw-hole (four), or at the tip of a prosthetic stem (one).
In one specimen, involving an entire femur, a fracture occurred
midway between the tips of the stems of the total hip and total
knee prostheses. The average interval to fracture was twenty-four
months (range, thirteen to seventy-two months).
In the specimens retrieved shortly following the fracture, histological
examination of the site showed ingrowth of fibrovascular repair
tissue into the gap with moderate osteoclastic resorption of the adjacent
necrotic bone. There was no evidence of callus emanating from the
graft fragments or unusual internal repair of the bone adjacent
to the fracture line.
Four specimens were retrieved several months after the fracture
had occurred. One was retrieved because of malunion seven months
after the fracture; one, after amputation because of a local recurrence
eight months after the fracture; one, as a result of a biopsy of
the site of a suspected tumor recurrence twelve months after the
fracture; and one, at autopsy thirteen months after the fracture. The
specimen retrieved at eight months had solid union of a diaphyseal
fracture of the proximal part of the humerus by a large external
callus. The specimen retrieved at autopsy was from a patient who had
sustained a fracture of the tibia while receiving chemotherapy for
pulmonary metastases. The fracture had been treated with immobilization
in a cast for seven months, and it had healed with an external callus
at the time of autopsy thirteen months later.
The capacity of repairing allografts to heal was illustrated
by the case of an eleven-year-old girl who had reconstruction of
the distal part of the femur with a 27-cm osteoarticular graft after
resection of a stage-IIB osteosarcoma2 (Fig. 9-A). Four years
later, she sustained a fracture through the metaphysis of the allograft
(Fig. 9-B),
which was treated with a plaster cast and healed with a substantial
malunion. Because of concomitant degenerative changes in the articular
cartilage, the distal 15 cm of the allograft, including the site of
the malunion, was resected, preserving the more proximal 12 cm of
the original allograft. This defect was reconstructed with a composite allograft-total
knee prosthesis fixed by a long intramedullary stem extending well
above the junction of the original allograft and the second allograft.
Supplementary autogenous iliac-crest grafts were secured about the
junction with wires. The junction had united at twelve months (Fig. 9-C), and the
reconstruction functioned well for the ensuing seven years (Fig. 9-D). Histological
study of the resected malunion site in the original graft showed
internal repair of almost 60% of the cortices at the fracture
site compared with <20% in the remainder of the
graft.
The common finding in these specimens was an unusual amount of
internal repair in the region about the fracture. Spatial maps showed
extensive revascularization in both the cortices and the trabeculae,
extending several centimeters on either side of the fracture, paralleling
the extent of the external callus.
Allograft-Cement Interface
Eight allograft-prosthesis composites were retrieved. Macrosections
of five of these specimens, retrieved at intervals of seven months
to eight years after implantation, were available for study. There was
no radiographic evidence of resorption of the allograft about the
cement. The interface between the allograft and the cement was grossly
secure in all specimens. On histological examination, none of the
specimens demonstrated revascularization of the allograft adjacent
to the cement, ingrowth of tissue between the cement and adjacent
allograft, or disruption of the allograft architecture.
Frozen Articular Cartilage
The macroscopic and histological features of the articular cartilage
were studied in twenty-eight preserved osteoarticular allografts.
Twenty-four specimens were totally devoid of chondrocytes in the lacunae
of the persisting articular cartilage. Four specimens, all retrieved
within eight months after implantation, had occasional lacunae with
remnants of cells that appeared to represent mummification of chondrocytes
during the freezing and storage process.
The five specimens retrieved prior to one year (average, seven
months) showed no change radiographically either in the apparent
thickness of the articular cartilage or in the underlying subchondral bone.
All had surface irregularities of varying severity with fibrillation
of the superficial surface of the cartilage, tangential clefts with
flap separations, erosions, and, occasionally, perpendicular clefts
reaching as deep as the tidemark (Fig. 10).
Eight specimens were retrieved between one and two years (average,
seventeen months) after implantation. Again, radiographically, the
thickness of the articular cartilage showed little or no narrowing,
and the subchondral bone appeared to be intact with no evidence
of repair or fragmentation. The most striking change was the extent
of a pannus of fibrovascular repair tissue covering some or all
of the articular surfaces. The pannus emanated from the periphery
of the joint and extended to varying degrees until the entire surface
was covered (Fig. 11). As the pannus first developed,
it covered the free surface, actually increasing the thickness of
the cartilage as seen radiographically, but, with time, the articular
cartilage beneath the pannus was resorbed, resulting in a combination
pannus-cartilage remnant that appeared to be thinner radiographically. As
the pannus covered the surface, it frequently extended down into
the clefts in the cartilage in an apparent attempt to repair them.
As the pannus thickened, it occasionally contained nodules of fibrocartilage
derived from the proliferating host mesenchymal cells within it.
Nine specimens were retrieved between twenty-four and sixty months
(average, forty-four months) after implantation. The radiographs
showed moderate thinning of the pannus-cartilage remnant and increasing
radiolucency of the cancellous bone beneath the subchondral plate.
In many of these specimens, there were varying degrees of subchondral
fragmentation and deformation. The macroscopic and histological
pattern was one of increasing resorption of the remaining articular
cartilage, which, in some specimens, reached to the tidemark but
rarely extended across it to the subchondral plate.
All of these changes progressed with time and were more advanced
in specimens from the lower extremity than in those from the upper
extremity.
Four specimens were retrieved between five and ten years (average,
seven years) after implantation, and two were retrieved after more
than ten years. These specimens had considerable variation in the
amount of cartilage destruction and the extent of subchondral fragmentation
or deformation.
Fresh Articular Cartilage
One fresh osteoarticular graft was a hemiarticular femoral condyle
and distal metaphysis harvested from a heart-beating donor and implanted
following resection of a presumed stage-3 aneurysmal bone cyst3. Postoperatively, the patient was
managed with immunosuppression with prednisone and Imuran (azathioprine)
on the schedule then used for organ recipients. The graft was fixed
with multiple screws and was united at four months. The patient
regained 90° of motion and began bearing weight at seven months.
The patient died of metastatic giant-cell tumor, and the specimen
was retrieved at autopsy thirteen months after implantation. The
articular cartilage was well preserved with no degenerative changes.
Histological examination showed a composition of viable chondrocytes
without fibrillation or clefts and a distinct intact tidemark (Figs. 12-A and 12-B).
The other freshly transplanted osteoarticular graft, a fetal
femoral head that was used to replace a metacarpal head destroyed
by psoriatic arthropathy, was retrieved at three months during a
ray resection. No immunosuppression had been used. The specimen
was populated with viable-appearing chondrocytes but was surrounded
by a zone of inflammatory tissue.
The findings in our initial report1 were
supported and confirmed by the observations in the present study,
which involved a much larger group of specimens obtained over a
longer period of time. However, the present group of specimens, particularly
those retrieved at longer intervals, had much greater heterogeneity,
which provided additional insights that suggest that the interaction between
the host and allograft is influenced by several factors.
Since more than half of the specimens had been retrieved because
of a complication associated with the allograft that may have altered
the usual course of repair, the findings in the twenty-four specimens (33%)
retrieved because of amputation or at autopsy may be more representative
of the reparative mechanisms than the findings in the group as a whole.
However, when the two groups were analyzed with respect to the timing
and extent of the repair, no important differences were noted except in
the specimens removed because of early infection (five) or resorption
(eight). In those specimens, both external and internal repair as
well as soft-tissue adherence were either markedly diminished compared
with the findings in the specimens without graft complications or
they were absent. Overall, however, the majority of the specimens retrieved
because of graft complications had findings similar to those in
the specimens without such complications.
Accurate and intimate contact between host and allograft cortices
appears to promote and accelerate union, although healing was also
observed when gaps of up to 4 mm were present at securely immobilized
junctions. The degree of contact and the security of fixation appeared
to influence the size and extent of the external callus and its
maturation into haversian bone. However, in no instance was there
evidence of primary union, even when there was the most intimate
contact; this was due, at least in part, to the inability of the
host fibrovascular repair tissue to produce cutting cones or to
effectively penetrate the allograft haversian systems. The persistent
failure of the host bone that filled the gaps to remodel along stress
lines and the fact that the junction of this bone with the allograft
was identified as the site of induced debonding indicate that an
allograft-host junction may remain a potential site for subsequent
failure for an extended period of time. Whether such remodeling
would be induced by removal of the internal fixation devices remains
speculative, and, in view of this, the additional support of fixation
devices may well be advantageous in this regard. It would also appear
to be desirable to quantitate the extent of this weakness with biomechanical
studies so that effective means of prevention might be developed.
The contrast between histologically united cortical-cortical
junctions and the persistence of radiolucency in the intercortical
gaps reaffirms the prudence of postponing operative intervention
because of presumed nonunion for at least one year until maturation
of the bone is reflected radiographically. This is especially true
in patients who receive postoperative chemotherapy. Cortical-cortical
union appeared to be substantially enhanced when autogenous bone
grafts had been added, with active callus formation extending onto
the external surface of the graft. The osteoinductive properties
of autogenous grafts are likely responsible for the increase in internal
repair of the adjacent cortex of the allograft4,
which raises the hypothesis that the use of more recently developed
osteoinductive substances may be similarly useful.
The antiblastic toxicity of chemotherapeutic drugs is known to
have an inhibitory effect on allograft union and repair in animals5. This is reflected histologically
by diminished osteoclastic resorption and decreased new-bone formation
resulting from suppression of osteoblastic activity and abolition
of mesenchymal proliferation. In our group of human allograft specimens, preoperative
chemotherapy was associated with retarded union. Union was rare
before one year, and maturation of callus required more than one
year. These observations suggest that, in humans, the deleterious
effect of chemotherapy is a reversible process, with restoration
of a biological environment favorable to allograft healing occurring
at the end of the chemotherapy period. In contrast, a specimen from
a patient who received substantial amounts of postoperative radiation
therapy showed no signs of union three years after implantation, suggesting
that radiation damage permanently jeopardizes allograft incorporation.
Repair of the grafts occurred in two fashions—surface
(or external) repair and internal repair. Surface repair was accomplished
by the deposition of a thin seam of appositional host bone, beginning
within the first three to six months, on the unresorbed surface
of the graft that served as the anchor for the attachment of host
soft tissues. However, the demarcating cement line was the site
of debonding when the soft tissues were forcibly pulled away. After
maturation into lamellar bone, usually by one year, no additional
thickening of the seam occurred. On average, by one year such deposition
covered approximately one-half of the exposed cortical surfaces.
The remaining cortical surfaces underwent superficial resorption
in a random fashion. These surfaces became pockmarked with Howship
lacunae, only a few of which contained osteoclasts at any given
time. Such surface resorption removed only a superficial millimeter
or two and, with cessation, was filled in by appositional new bone
so that by two years almost the entire surface of the graft was
covered by a seam of viable lamellar bone. Although too thin to
be apparent on either conventional radiography or computed tomography,
it was evidenced by a correspondingly thin rim of increased activity
on isotope scanning.
Internal repair began with invasion of the surface stoma of Volkmann
and haversian canals by fibrovascular host tissue that enlarged
the haversian canals by osteoclastic resorption and then rebuilt them
with appositional new bone. The pattern of such revascularization
was random, and the pace was slow, seldom exceeding more than a
few millimeters per year. A repaired osteon was often seen side
by side with one that was totally ignored by the reparative tissues.
Such internal repair rarely occurred from the cut ends of the graft,
proceeding longitudinally into the old haversian canals, and almost
never by cutting cones at either the surface or the osteotomy sites.
The domination of resorptive activity during the first two years
suggests that the second year after implantation is the critical period
of cortical weakening of massive allografts. The extent and pace
of internal repair in cortical bone were so limited that they did
not alter the radiographic density of the graft with areas of radiolucency
or bind enough isotope to produce visible increases in uptake on
isotope scanning. This general picture was common to all repairing
grafts for two to three years. Specimens retrieved after three years
had a much wider variation in the extent of cortical repair. In
one specimen, cortical repair virtually ceased; in some, it sputtered
along; and in a few, it progressed steadily until virtually the
entire graft was repaired.
In the present study, there was a strong histological suggestion
of an immunological mechanism in the specimens that had undergone
resorption without repair. Each one had aggressive resorption of
cortical bone by osteoclasts crossing the peripheral cement line
about the osteons and resorbing the interstitial lamellae, a pattern
not seen in autogenous cortical graft repair6,7 and
one that is common in xenograft resorption8,9.
Areas of unusually extensive repair in the cortices were seen
under several diverse circumstances: at sites of union with supplementary
conventional autogenous cancellous grafts, at sites of union with supplementary
autogenous vascularized grafts, about the sites of healing allograft
fractures, and at sites of involvement of the allograft by recurrent neoplastic
tissue. There was no apparent reason for this phenomenon, and the
mechanism appeared to be a localized nonspecific acceleration of
the usual reparative response.
As suggested by the histological findings and as reported by
others10-12, in our study allograft
fractures occurred more frequently during the second year after
implantation, a reflection of the domination of the resorptive pattern
during this phase of the remodeling process. Diaphyseal fractures
occurred in areas of stress concentration about the ends of fixation
devices or where the cortex had been perforated by screws. In two
patients in whom autogenous cancellous grafts had been placed at
the osteotomy site, a fracture occurred just above the region of
accelerated internal repair adjacent to the autogenous grafts. Similarly,
subchondral fractures occurred through the metaphyseal cortex where
fibrovascular repair tissue and osteoclastic resorption had extended
into the subchondral bone.
The longer periods of observation of the interface between allograft
bone and bone cement in these specimens confirmed the earlier observation1 that there is little or no change
in the microarchitecture at the bone-cement interface and that the usual
lack of extensive cortical revascularization is advantageous in
this regard. Given these observations, it seems prudent to use intramedullary
fixation rather than plates and screws whenever practical, to use
bone cement to firmly anchor fixation devices and prosthetic stems
in composite reconstructions, and to leave fixation devices in
situ indefinitely unless device-related disability mandates
removal.
The survival of chondrocytes in the articular cartilage of frozen
allografts treated with cryoprotectants has been a controversial
subject1,4,13-16. Delloye et al.
cultivated frozen cartilage treated with dimethyl sulfoxide (DMSO)
and observed cell growth with the morphological appearance of living chondrocytes
but did not observe any differences with the control group (allografts
frozen without cryoprotectant)13.
Malinin et al. reported cell growth after cultivating cartilage
from an articular allograft stored in liquid nitrogen, although
they did not establish the cartilaginous nature of these cells14. Schachar and McGann studied the
viability of isolated chondrocytes both in vitro and
in animal models16. They asserted
that the persistent viability of cartilaginous cells was optimized
by the use of DMSO as a cryoprotectant and by utilizing slow freezing and
rapid thawing of the graft. However, we observed only scattered
remnants of nucleated chondrocytes with cytomorphological features
of viability in four of twenty-eight osteoarticular specimens. Quite
clearly, the majority of human chondrocytes stored by conventional
banking techniques do not survive despite the use of cryoprotectants.
In contrast, two fresh specimens had clear histological evidence
of chondrocyte survival, confirming the observation of Oakeshott
et al. that fresh, unfrozen allografts had substantial chondrocyte
survival17. Their small allografts
had little or no inflammatory response. In our study of larger specimens,
a marked inflammatory response had occurred in the fresh specimen
from the patient who had not had immunosuppression, whereas evidence
of widespread chondrocyte survival with a minimal inflammatory response
was seen in the specimen retrieved at thirteen months from the patient
who had had immunosuppression. These observations suggest that fresh,
unfrozen chondrocytes survive, but patients receiving a large osteochondral
graft may require immunosuppression for the chondrocytes to remain
viable.
In the osteoarticular specimens retrieved later (after two years
or more), formation of the fibrovascular pannus took two forms.
In some specimens, it formed a thin covering over the necrotic cartilage protruding
into and filling clefts and it appeared to be primarily reparative
in nature. In other specimens, the pannus was much thicker, was
more cellular, and consisted primarily of chronic inflammatory cells,
and it appeared to be primarily inflammatory in nature, again suggesting
that these differences are based on histocompatibility. It may be
that when early technical disruption exposes previously immunologically
sequestrated chondrocytes to the host immune system, those with
a major immunological mismatch incite an inflammatory response whereas
those with a minor mismatch evoke a lesser immunological stimulus
to form a reparative response. This hypothesis is also suggested
by the observation that, without exception, every specimen had mechanically
induced changes in the articular cartilage, whereas the formation
of pannus was an inconstant and later event.
Abnormal anatomical and dimensional matching of the articular
surfaces and capsuloligamentous instability appeared to play a role
in the rapidity and degree of articular cartilage degeneration.
In the specimens retrieved earlier (after less than two years),
anatomical mismatching and/or instability appeared to be
responsible for irregular distribution of mechanical forces on the
articular surface, which created stress-raising areas in which degeneration with
fibrillation and tangential flap separation, representing the initial
phase of osteoarthritis, took place. The hypothesis that there is
a mechanical role in the initiation of degenerative changes is reinforced
by the higher prevalence of such changes in the weight-bearing knee
than in the non-weight-bearing shoulder.
It was of considerable interest that for the first one to two
years, despite substantial clinical activity, the necrotic articular
cartilage functioned well and appeared to be normal radiographically.
In fact, on the radiographs of many specimens, it was not possible
to distinguish between persisting articular cartilage with a thin
pannus and major cartilage destruction and thick inflammatory pannus.
Only when substantial subchondral bone resorption and fragmentation
occurred were clinical and radiographic signs of substantial narrowing
of articular cartilage apparent. In some instances this occurred at
one year, and in others it did not occur for more than five years.