Abstract
Background: To our knowledge, the medium to
long-term outcome after revision knee arthroplasty with structural
allograft augmentation for reconstruction of uncontained defects
has not been determined. The purpose of the present study was to
assess the outcome for patients managed with such a procedure.
Methods: We prospectively followed fifty patients
who had fifty-two revision knee replacements with sixty-six structural
grafts performed at three institutions. Twenty-nine knees (twenty-seven
patients) were independently evaluated at a mean of 96.9 months (range,
sixty to 189 months) by an investigator who had not been involved
in the index procedure. Twelve knees (23%) had a repeat
revision at a mean of 70.7 months (range, twenty-six to 157 months). The
allograft was retained in two of these patients. Eleven patients
died at a mean of ninety-three months (range, sixty-one to 128 months)
after the procedure; the structural allograft and implants were
intact, and the patients were not awaiting revision at the time
of death.
Results: Clinical evaluation revealed that the mean
modified Hospital for Special Surgery knee score had improved from
32.5 points preoperatively to 75.6 points at the time of the review
and the mean range of motion had increased from 60.5° preoperatively to
88.6°. Failure was defined as an increase of less than 20 points
in the modified Hospital for Special Surgery knee score at the time
of the review or the need for an additional operation related to
the allograft. Thirteen knee replacements failed, yielding a 75% success
rate. Five knees had graft resorption, resulting in implant loosening.
Four knee replacements failed because of infection, and two knees
had nonunion between the host bone and the allograft. Two knees
(one patient) did not have a 20-point improvement in the knee score.
The survival rate of the allografts was 72% (95% confidence
interval, 69% to 75%) at ten years. On radiographic
analysis, none of the surviving grafts had severe resorption, one
had moderate resorption, and two had mild resorption. One knee had
a loose tibial component, and three knees had nonprogressive tibial
radiolucent lines. All four knees were asymptomatic.
Conclusions: Our results demonstrate that allografts
used in revision knee replacement in patients with the difficult problem
of massive bone loss have an encouraging medium-term rate of survival.
Surgeons who manage patients after multiple revision knee
arthroplasties may have to contend with large osseous defects. Such
defects may not be amenable to the use of augments and wedges that are
an integral part of modern knee revision systems. The use of structural
allografts is a viable alternative for the treatment of massive
bone loss.
Parks and Engh1, in a study
of specimens retrieved at a mean of three and one-half years postoperatively,
demonstrated that structural allografts in total knee arthroplasty
do not revascularize, resorb, or collapse. The senior authors (A.E.G.,
G.W.B., and H.P.C.) of the present study previously reported encouraging short-term
results in patients managed with structural allograft for the treatment
of a large uncontained defect2-6.
Others also have described promising results after short to medium-term
follow-up. Engh et al.7, in a
review of the results of total knee arthroplasty with use of allograft
in thirty patients (thirty-five knees) who had been followed for
a mean of fifty months, reported that twenty-six patients (87%) had
a good or excellent result. Twenty-nine (83%) of the thirty-five
knees had received a femoral-head allograft. Harris et al.8 reported satisfactory clinical and
radiographic results in fourteen of fifteen patients at a mean of forty-three
months postoperatively. Mow and Wiedel9 reported
that twelve of fifteen patients had no allograft-related complications
at forty-seven months postoperatively, whereas Mnaymneh et al.10 reported that five of ten patients
with massive bone loss managed with a whole distal femoral allograft or
a whole proximal tibial allograft, or both, had a high rate of complications.
We are unaware of any medium-term follow-up studies on the use
of structural allografts for the treatment of large uncontained
defects in revision knee arthroplasty. It is difficult to collect
a large group of patients because of the relative rarity of this
condition and the advanced age of the patients undergoing the procedure.
For these reasons, we collated the results of three surgeons,
in tertiary-care centers, who used a similar operative technique
and postoperative regimen. We report the results after revision
knee arthroplasty with use of structural allograft in fifty patients
(fifty-two knees) who had been followed for a minimum of five years (mean,
eight years; range, five to fifteen years).
Since 1983, sixty-two revision knee arthroplasties requiring
structural allograft for the treatment of an uncontained defect
were performed in fifty-eight patients by the three senior authors
(A.E.G., G.W.B., and H.P.C.). All patients who had been followed
for a minimum of five years were included in the study. Four patients
(five knees) died less than five years postoperatively, and four
patients (five knees) were lost to follow-up. Thus, fifty
patients (fifty-two knees) were eligible for the review.
Study Group
The mean age of the patients at the time of the allograft surgery
was sixty-eight years (range, twenty-four to eighty-five years).
There were thirty women and twenty men. The procedure was performed
in twenty-seven right knees and twenty-five left knees. The patients
had had a mean (and standard deviation) of 2.55 1.32 previous procedures (range,
one to six procedures). The primary condition of the patients and
the reasons for the allograft procedures are given in Table I.
All patients had an uncontained defect, defined as segmental
bone loss with no remaining cortex5.
The defects were classified as either circumferential (requiring
a segmental distal femoral or proximal tibial graft) or noncircumferential
(requiring a partial distal femoral, partial proximal tibial, or femoral-head
graft)8. Sixty-six grafts were
inserted in fifty-two knees; fourteen patients had a graft inserted
in both the distal part of the femur and the proximal part of the tibia.
One patient had a proximal tibial and patellar tendon construct,
and one had a distal femoral and medial collateral ligament construct.
Forty-eight defects were circumferential, and eighteen were noncircumferential.
Forty-eight (73%) of the sixty-six grafts were segmental,
involving a whole distal femoral or a whole proximal tibial graft
(Table II).
The mean length of the graft was 54 36 mm (range, 20 to 160 mm).
The implants included thirty-one press-fit condylar or total
condylar-III prostheses (PFC or TC-III; Johnson and Johnson Orthopaedics,
Raynham, Massachusetts), seven Insall-Burstein modular prostheses
(Zimmer, Warsaw, Indiana), six Porous-Coated Anatomic prostheses
(PCA; Howmedica, Rutherford, New Jersey), six Guepar prostheses (Benoist
Girard, Bagneaux, France), one Genesis prosthesis (Smith and Nephew
Richards, Memphis, Tennessee), and one Insall-Burstein-II
constrained condylar knee prosthesis (CCK; Zimmer). A long stem
was used to protect the structural allograft in all but two patients.
A cemented stem was used in thirteen procedures performed
in the 1980s. Thereafter, a press-fit stem was used in thirty-nine
procedures.
The results for twenty-seven patients (twenty-nine knees) who
were alive at the time of the study were independently reviewed
by one of the investigators (M.G.C.) who had not been involved in
the index procedure. These patients had been followed for a mean
(and standard deviation) of 96.9 36.8 months (range, sixty to 189
months). Twelve other patients (twelve knees) had a revision at
a mean of 70.7 40.4 months (range, twenty-six to 157 months). These
knee replacements were considered failures and thus were not included
in the review. In two of the failed knee replacements, the allograft was
well integrated at the time of revision, allowing the graft to be
retained. Eleven patients (eleven knees) with more than five years
of follow-up died, with the implants intact and with no plans for
a revision procedure, at a mean of 93 35.6 months (range, sixty-one
to 128 months) after the procedure.
Clinical Evaluation
The clinical assessment was made on the basis of the modified
Hospital for Special Surgery knee score4.
Failure was defined as an increase of less than 20 points in the
knee score postoperatively or the need for an additional operation
related to the allograft4.
Radiographic Evaluation
Twenty-seven knees in twenty-five patients who were alive and
had not had a revision were assessed radiographically. The evaluation
consisted of inspection for union of the allograft and host bone (the
presence of trabeculae bridging the host-graft junction and obliteration
of the allograft-host junction) on both anteroposterior and lateral
radiographs11. Radiolucent lines
were measured according to the system of the Knee Society11. A component was considered loose
if the radiolucent lines were circumferential and more than 2 mm
in thickness, if the lines were progressive, or if the component
had migrated or broken. Allograft resorption was classified as mild
(partial-thickness loss of less than 1 cm in length in one cortex),
moderate (partial-thickness loss of at least 1 cm in one cortex),
or major (full-thickness loss of any length in one cortex).
Statistical Analysis
Survivorship analysis was performed with the Kaplan-Meier method12, and the 95% confidence
limits were calculated with the Greenwood formula for variance.
Operative Technique
Preoperative Planning
Prior to surgery, an attempt should be made to assess bone loss;
however, the extent of bone loss often is greater than that anticipated
from the radiographs. Sizing is important, and we found it useful to
compare radiographs of both the involved and the contralateral knee
with radiographs of the proposed allograft. For defects in the distal
part of the femur, it is essential to have an allograft that is smaller
than the host bone so that it can be placed within the host cortical
shell with its attached collateral ligaments. If there is concern
about the extensor mechanism, a proximal tibial allograft with an
extensor mechanism attached should be available.
It is also important to examine the knee to determine the degree
of sagittal and coronal instability so that an implant with appropriate
constraint can be chosen. All of the senior authors currently use the
Total Condylar III prosthesis (TC-III; Johnson and Johnson Orthopaedics)
when possible. If the soft-tissue envelope is maintained, a less
constraining implant rather than a hinged prosthesis can be used,
resulting in decreased forces at the implant-allograft and allograft-host
interfaces.
Allograft Procurement
The grafts were procured under sterile conditions, according
to the protocol of the American Association of Blood Banks13. The bone was deep-frozen at —70°C
and irradiated with 25,000 Gy.
Surgical Technique
For noncircumferential defects, a femoral-head, partial distal
femoral, or partial proximal tibial allograft is used. The structural
allograft is fixed to the bone with cancellous screws, and additional
fixation is obtained with a long press-fit stem. Cement is used
to secure the allograft-implant and host-implant interface but not
to enhance stem fixation in the diaphyseal region.
For circumferential defects, the margins of the host bone are
dissected out to reveal the extent of the bone loss. As much residual
host bone with soft-tissue attachments as possible is retained.
The size and shape of the osseous defect are evaluated, and a replacement
construct is fashioned on the back table. Appropriate cuts are made
with use of revision total knee arthroplasty instrumentation. A step-cut
is made in structurally sound host bone, and the allograft is fashioned
to complement the step-cut. A trial reduction is then performed
to ensure that there is good approximation of bone at the allograft-host
interface. The flexion and extension gaps are then balanced, the
construct is satisfactorily externally rotated, and it is verified
that the overall alignment of the limb is acceptable. The level
of the joint line must be carefully assessed as there is a tendency
to translate the joint line with allograft constructs. With a femoral
allograft the tendency is to depress the joint line, whereas with
a tibial allograft the tendency is to elevate it. The most accurate
way to measure the joint line is to measure the distance from the
proximal tip of the fibula to the joint line on the radiograph of
the normal, contralateral knee. If the patient has undergone a joint
replacement on the contralateral side, the joint line should be
established 1.5 cm proximal to the tip of the fibula or 2.5 cm distal
to the medial epicondyle or at the site of the residual meniscal rim
scar.
The components are then cemented to the allograft on the back
table (Fig. 1Fig.
1). Cement is utilized at the implant-allograft interface
and the stem-allograft interface; however, care should be taken
to ensure that no cement is present at the proposed allograft-host
interface. In the femur, the cortical shell of bone with its attached collateral
ligaments is secured to the construct with cerclage wires or screws
to act as a vascularized graft at the host-allograft junction (Figs. 2 and 3). In the tibia, excellent
stability is usually obtained with use of the step-cut and insertion
of the press-fit stem; however, if there is any doubt, fixation
can be augmented with cortical screws. A mixture of morselized autograft
and allograft is placed around the allograft and host junction.
We do not recommend plate fixation to enhance fixation of the allograft8,10, as the multiple drill-holes produce
channels in the allograft that may facilitate revascularization
and fracture of the graft14. If
there is concern regarding the stability of the construct in the
femur, plates or cortical struts may be used as a last resort. Plates
and screws are not recommended in the proximal part of the tibia because
of concerns regarding soft-tissue coverage.
Postoperative Management
Patients are managed with a hinged knee brace and are encouraged
to initiate early range-of-motion exercises. Only toe-touch weight-bearing
is allowed for six weeks, after which the patient progresses to
partial weight-bearing, which is continued until there is evidence
of union at the allograft-host interface, usually by three months.
Clinical Results
Twenty-seven patients with twenty-nine revision knee arthroplasties
were evaluated. Preoperatively, the mean knee score (and standard
deviation), according to the modified system of The Hospital for
Special Surgery, was 32.5 ± 18.95 points
(range, 5 to 74 points). Postoperatively, the mean score had increased
to 75.6 ± 14.1 points (range, 51 to 98 points).
One patient who had undergone bilateral revision knee arthroplasty
did not have a 20-point improvement in the score. He had severe
psoriatic arthritis and was confined to a wheelchair with both knees
fixed at 90°. Radiographically, the allografts had an excellent appearance,
with no evidence of resorption or of migration or loosening of the
components.
The mean range of motion was 60.5° ± 32.4°
(range, 0° to 120°) preoperatively and 88.6° ± 30.6°
(range, 0° to 120°) postoperatively. Extensor lag was not routinely
recorded preoperatively. However, an extensor lag ranging from 5°
to 30° was noted in eight knees postoperatively. The 30° lag was
seen in the patient who had a patellar tendon allograft.
Preoperatively, fifteen knees had grade-3 instability (that is,
opening or translation of at least 1 cm compared with that in the
contralateral, normal limb). Stability was provided with a medial-collateral-ligament
allograft in one knee and with a constrained implant in the others.
Postoperatively, only three knees had grade-3 instability, and in
no case was it troublesome for the patient.
Radiographic Results
All sixty-six allograft-host interfaces were evaluated for union.
Two (3%) had failed to unite. One of the patients with
a nonunion had insulin-dependent diabetes and rheumatoid arthritis
and had had a periprosthetic fracture. At thirty months, open reduction
and internal fixation augmented with bone graft was unsuccessful.
A revision allograft procedure was performed at eighty months. The other
patient initially underwent a two-stage revision, with use of cortical
strut allografts, because of infection. Despite the persistent nonunion,
the knee replacement was stable, there was no infection, and the
patient had not needed a repeat revision at the time of the latest
follow-up.
The twenty-five patients, with twenty-seven intact knee replacements,
who had not had a revision had thirty-five grafts that could be
evaluated radiographically for resorption. One proximal tibial graft exhibited
moderate resorption, and two demonstrated mild resorption. No femoral
graft showed signs of resorption (Figs. 4-A and 4-B).
The radiographs were analyzed for the presence of progressive
radiolucent lines, indicating loosening. No radiolucent
lines were seen about any femoral component. Three knees had nonprogressive
radiolucent lines about the tibial component, and one knee had a
loose tibial component. All four knees were asymptomatic.
Repeat Revisions
Twelve (23%) of the fifty-two knees had a repeat revision.
In two repeat revisions, the allograft was retained; however, late
flexion instability developed, necessitating an exchange of the
polyethylene spacer in one knee and the use of a larger femoral component
in the other. In both instances, the allograft was well united to
the host bone and was structurally sound. Both patients did well
postoperatively. The cases of the patients who had a repeat revision
were not reviewed clinically or radiographically.
Five grafts (three femoral and two tibial) had progressive resorption,
resulting in implant migration and loosening. Repeat revision because
of resorption was performed at a mean (and standard deviation) of
92.8 42.2 months (range, fifty-seven to 157 months).
Four knees (8%) failed because of infection. One of them
had had a previous infection. Three knees had an additional revision
procedure, and one knee had an arthrodesis. The mean time to revision
because of infection was 34.25 12.28 months (range, twenty-two
to forty-eight months). The remaining patient who required repeat
revision had a nonunion and was described above.
Overall Results
The revision was considered to have failed in thirteen knees.
Ten knees required a repeat revision. Two knees (one patient) did
not have a 20-point improvement in The Hospital for Special Surgery knee
score, and one knee had a nonunion at the allograft-host junction
that had required onlay cortical struts to achieve union. Thus,
thirty-nine knees (75%) had a successful result.
Survivorship analysis with use of Kaplan-Meier methodology showed
that the rate of survival of the allografts was 92% (95% confidence
interval, 89% to 95%) at five years and 72% (95% confidence interval,
69% to 75%) at ten years. If the five knees lost
to follow-up were included as failures, the five-year rate of survival
would be unchanged, whereas the ten-year rate of survival would
be 67% (95% confidence interval, 51% to
83%) (Fig. 5).
There is a paucity of information in the literature on the medium
to long-term survival of allografts in revision hip and knee arthroplasty.
The longest follow-up in the literature on total knee revision,
as far as we know, was reported by Engh et al.7,
who reviewed the results at a mean of fifty months after use of
thirty-five allografts in thirty patients. They reported a satisfactory
result in twenty-six (87%) of the thirty patients; however,
only six patients had a circumferential defect requiring a whole
distal femoral or proximal tibial allograft. In the literature on
total hip revision, the longest follow-up, to our knowledge,
was reported by Gross et al.15,
who reviewed the results at a minimum of two years (mean, 4.8 years)
after revision with a proximal femoral allograft in 130 patients.
They reported a success rate of 85%.
In 1997, Ghazavi et al.4 reported
the results of the use of allograft in thirty revision knee arthroplasties
in twenty-eight patients who had been followed for a minimum of
twenty-four months (mean, fifty months). The success rate in their
study was 77%, with use of the same criteria as those used
in the present study. Our study demonstrated a success rate of 75% in
a larger number of patients who had been followed, on the average,
for an additional four years; this indicates that allograft reconstructions
continue to do well after medium-term follow-up. A comparison of
the rates of graft survival confirms this observation. In the earlier
study, the probability of graft survival was 67% at five
years. In our series, with a larger group of patients, the rate
of graft survival was 92% (95% confidence interval,
89% to 95%) at five years and 72% (95% confidence
interval, 69% to 75%) at ten years.
Our results are comparable with those reported in the literature
on revision knee arthroplasty without allograft. We know of only
five studies with a mean follow-up interval of more than five years.
Friedman et al.16 described the
results at a mean of 5.2 years after use of so-called first-generation
revision knee prostheses in 137 knees. Using criteria for success
similar to those used in the present study, they reported that 63% (eighty-one)
of the 129 knees that had a single revision and 50% (four)
of the eight knees that had subsequent revisions had a successful result.
The rate of revision at five years was 6% (eight of 137
knees) in their study and 8% (four of fifty-two knees)
in our study. Hohl et al.17 evaluated
thirty-five knees at a mean of 6.1 years after reconstruction with
the Total Condylar III prosthesis (Johnson and Johnson Orthopaedics). Twenty-nine
of these reconstructions were revisions. They reported that 71% had
a satisfactory result and 9% needed a revision. Mow and
Wiedel9 reported the results at
9.8 years after revision with a Porous-Coated Anatomic prosthesis
(PCA; Howmedica) in thirty-three knees. Nineteen knees (58%)
had a satisfactory knee score according to the system of The Hospital
for Special Surgery, and six (18%) had a revision. Gustilo
et al.18 reported that forty-one
(73%) of fifty-six knees had a satisfactory result after
a mean follow-up interval of 8.3 years, and Goldberg et al.19 noted a satisfactory result in twenty-seven
(46%) of fifty-nine knees after a mean follow-up interval of
five years. In our patients, the use of allograft in the revision
total knee procedure did not result in a poorer outcome even though
forty-eight defects were circumferential and thirty-one whole distal femoral
grafts and seventeen whole proximal tibial grafts had been implanted.
There is concern that allografts may resorb over time as a result
of revascularization by creeping substitution. Five grafts (8%)
in the present series failed because of resorption. Thirty-five
additional grafts were evaluated radiographically for resorption.
None of the femoral grafts showed signs of resorption, one tibial
graft demonstrated moderate resorption, and two tibial grafts had
evidence of mild resorption. Thus, in the medium term, graft resorption
does not seem to be a major problem.
There is also concern about allograft implantation in the setting
of a previous infection. We are not aware of any studies in which
allograft implantation has been specifically evaluated in patients
with a previous infection around a total knee replacement, although
one of us (G.W.B.) and colleagues reported no subsequent infection
in three patients who had had a previous infection6. Six of our patients had had an infection
around a total knee replacement that was treated with a two-stage
revision with an antibiotic-impregnated spacer and a minimum of
six weeks of intravenous antibiotic therapy. Only one patient had
ongoing infection.
Nonvascularized allografts serve as an excellent nidus for the
growth of organisms5; hence, late
infection of an allograft is a concern20.
The infection rate of 8% (four of fifty-two knees) in our
study is slightly higher than those in series of revision knee arthroplasties
without allograft, which have ranged from 0% to 4.5%9,16,19. However, the rate in our series
is comparable with those associated with other allograft procedures4,5,7,10,21.
In summary, our results demonstrate that allograft-knee replacement
constructs have a good rate of survival in the medium term. We are
cautiously optimistic that this technique will continue to show good
results; however, the present cohort of patients will need to be
followed to determine whether graft resorption increases with time.
We believe that this technique provides a durable option for patients
requiring revision total knee arthroplasty in the setting of massive
bone loss.
Parks NL, and Engh GA: Histology of nine structural bone grafts used in total
knee arthroplasty. Clin Orthop,1997.345: 17-23, 34517
1997
[PubMed]
Chandler HP: Revision total knee arthroplasty-structural bone grafting:
when and how. Orthopedics,1996.19: 797-9, 19797
1996
[PubMed]
Chandler HP: Structural bone grafting about the knee. Orthopedics,1998.21: 1044-5, 211044
1998
[PubMed]
Ghazavi MT; Stockley I; Yee G; Davis A; and Gross AE: Reconstruction of massive bone defects with allograft
in revision total knee arthroplasty. J Bone Joint Surg Am.,1997.79: 17-25, 7917
1997
[PubMed]
Stockley I; McAuley JP; and Gross AE: Allograft reconstruction in total knee arthroplasty. J Bone Joint Surg Br,1992.74: 393-7, 74393
1992
[PubMed]
Tsahakis PJ; Beaver WB; and Brick GW: Technique and results of allograft reconstruction
in revision total knee arthroplasty. Clin Orthop,1994.303: 86-94, 30386
1994
[PubMed]
Engh GA; Herzwurm PJ; and Parks NL: Treatment of major defects of bone with bulk allografts
and stemmed components during total knee arthroplasty. J Bone Joint Surg Am,1997.79: 1030-9, 791030
1997
[PubMed]
Harris AI; Poddar S; Gitelis S; Sheinkop MB; and Rosenberg AG: Arthroplasty with a composite of an allograft and a prosthesis
for knees with severe deficiency of bone. J Bone Joint Surg Am,1995.77: 373-86, 77373
1995
[PubMed]
Mow CS, and Wiedel JD: Revision total knee arthroplasty using the porous-coated
anatomic revision prosthesis: six- to twelve-year results. J Arthroplasty,1998.13: 681-6, 13681
1998
[PubMed]
Mnaymneh W; Emerson RH; Borja F; Head WC; and Malinin TI: Massive allografts in salvage revisions of failed total
knee arthroplasties. Clin Orthop,1990.260: 144-53, 260144
1990
[PubMed]
Ewald FC: The Knee Society total knee arthroplasty roentgenographic
evaluation and scoring system. Clin Orthop,1989.248: 9-12, 2489
1989
[PubMed]
Kaplan EL, and Meier P: Nonparametric estimation from incomplete observations. Am Statist Assoc.,1958.53: 457-81, 53457
1958
Fawcett K, Barr AR [editors]. Tissue
banking. Arlington, VA: American Association of Blood Banks;
1987. p 97-107
Burchardt H: The biology of bone graft repair. Clin Orthop,1983.174: 28-42, 17428
1983
[PubMed]
Gross AE; Hutchison CR; Alexeeff M; Mahomed N; Leitch K; and Morsi E: Proximal femoral allografts for reconstruction of bone
stock in revision arthroplasty of the hip. Clin Orthop,1995.319: 151-8, 319151
1995
[PubMed]
Friedman RJ; Hirst P; Poss R; Kelley K; and Sledge CB: Results of revision total knee arthroplasty performed
for aseptic loosening. Clin Orthop,1990.255: 235-41, 255235
1990
[PubMed]
Hohl WM; Crawfurd E; Zelicof SB; and Ewald FC.: The Total Condylar III prosthesis in complex knee reconstruction. Clin Orthop,1991.273: 91-7, 27391
1991
[PubMed]
Gustilo T; Comadoll JL; and Gustilo RB: Long-term results of 56 revision total knee replacements. Orthopedics,1996.19: 99-103, 1999
1996
[PubMed]
Goldberg VM; Figgie MP; Figgie HE 3d; and Sobel M.: The results of revision total knee arthroplasty. Clin Orthop ,1988.226: 86-92, 22686
1988
[PubMed]
Tomford WW; Thongphasuk J; Mankin HJ; and Ferraro MJ: Frozen musculoskeletal allografts. A study of the clinical
incidence and causes of infection associated with their use. J Bone Joint Surg Am,1990.72: 1137-43, 721137
1990
[PubMed]
Lord CF; Gebhardt MC; Tomford WW; and Mankin HJ: Infection in bone allografts. Incidence, nature, and treatment. J Bone Joint Surg Am,1988.70: 369-76, 70369
1988
[PubMed]