The most challenging problem for reconstructive surgeons
treating hips that have had multiple revision arthroplasties is
bone loss on both the pelvic and the femoral side1.
As the degree of bone loss is associated with the number of total
hip revisions and as younger patients are having hip replacement
surgery, this problem is likely to increase2,3.
Severe circumferential bone loss of the femur of >5 cm
in length often makes conventional revision techniques difficult,
if not impossible, to perform, especially if adequate distal stabilization
cannot be obtained1,4-7. One alternative
is to use a proximal femoral allograft cemented to a long-stem prosthesis
to restore the proximal integrity of the femur. This technique has
attracted interest, especially with regard to its use in young patients,
because of its preservation of existing bone stock and its potential for
improving bone stock to facilitate future reconstruction6,8,9. While the published results
of this technique have been encouraging, they have generally involved
relatively short-term follow-up5,7,10-12.
There has been concern that allograft resorption, which occasionally
has led to early failure5,11,
may be a greater problem with longer follow-up.
Since 1984, the senior one of us (A.E.G.) has performed more
than 250 revision hip arthroplasties using a large femoral allograft-prosthesis
construct in patients with proximal femoral deficiency of >5 cm.
This paper describes the operative technique for using such proximal
femoral allografts and presents the radiographic and clinical results
in our early experience.
Between April 1984 and December 1989, the senior author performed
122 revision total hip arthroplasties that involved use of allograft
bone on the femoral side. The allografts included morselized bone,
cortical struts, calcar allografts (<5 cm long), and large
proximal femoral allografts (>5 cm long). In sixty patients
(sixty-three hips) with a proximal femoral deficiency of >5
cm in length, a proximal femoral allograft-prosthesis construct
was used. The minimum length of the proximal femoral allograft was
10 cm. This paper describes the follow-up of these sixty consecutive
patients.
Study Group
There were twenty men and forty women, with an average age of
62.5 years (range, 30.2 to 81.6 years) at the time of the revision.
All patients had undergone at least one previous total hip arthroplasty
on the side evaluated in the study, and an average of 3.8 (range,
one to nine) previous total hip arthroplasties had been performed
in the series. Three patients had had a prior resection arthroplasty:
two because of infection, and one because of aseptic failure.
The average length of the proximal femoral allograft was 15 cm
(range, 10 to 22 cm). The proximal femoral defects were all circumferential
cortical defects of >5 cm in length13;
the lengths ranged from 8 to 22 cm as measured from the tip of the
greater trochanter. According to the classification system of the
American Academy of Orthopaedic Surgeons Committee on the Hip, the
hips had level-II and level-III defects with grade-III bone loss14 (Figs. 1-A, 1-B, 2-A, and 2-B). Level II is the area up to 10
cm distal to the lesser trochanter, and level III is distal to this.
Grade III is loss of host-prosthesis contact with the need for structural
bone graft.
Clinical and Radiographic Evaluation
Each patient was evaluated before the operation with use of a
modified Harris hip score. Living patients either returned for clinical
and radiographic evaluation or saw a local orthopaedic surgeon who
performed the hip-score assessment and sent radiographs for review.
The radiographs and clinical notes of the patients who had died
were reviewed, and relatives were contacted if necessary to assess
the patient’s functional status prior to death. Clinical
success was defined as a postoperative increase in the Harris hip
score of greater than 20 points, a stable implant, and no need for
additional surgery related to the allograft at the time of the review.
Radiographs were examined for trochanteric union, allograft-host
union, endosteal and periosteal resorption, component loosening,
and fracture. The allograft was divided into zones similar to those described
by Gruen et al.15. However, zones
1 and 4 were excluded because of the absence of an allograft trochanter
(zone 1) and because of the allograft-host junction (zone 4); thus there
were five zones6. Implant stability
was assessed on the basis of lucent lines and implant migration:
definite loosening was defined as migration of the implant of greater
than 3 mm or fracture of the cement. Mild resorption was classified
as partial-thickness resorption of <1 cm in length; moderate
resorption, as partial-thickness resorption of 1 cm in length; and
severe resorption, as full-thickness resorption of any length1.
Statistical Evaluation
The Kaplan-Meier method was used for survivorship analysis, and
the 95% confidence limits were calculated with the Greenwood
formula for variance16. Failure
was defined as planned or actual removal of the original allograft-prosthesis
construct or severe radiographic resorption of the allograft. Patients
who had died or had been lost to follow-up were censored as of their
latest date of evaluation.
Operative Technique
All of the operations were done in a laminar flow operating room
with body exhaust systems. The allograft is stored in the hospital
bone bank at -70°C after being irradiated with 2.5 Mrad (25,000 Gy).
The bank is accredited by the American Association of Tissue Banks17. Preoperatively the approximate
allograft size is templated, and a longer graft is ordered to allow
for any intraoperative variations. The proximal femoral allograft
is brought into the operating room only after possible infection
of the hip to be revised has been ruled out. Any suspicion of infection
results in a two-stage procedure. The graft is unwrapped, specimens
are obtained for culture, and the graft is placed in a warm solution
of povidone-iodine. To reduce operative time, the graft is prepared
on a separate back table by part of the team while the revision
is being performed.
The transtrochanteric approach was most commonly used because
of the need for extensive exposure. We now prefer a longitudinal
trochanteric splitting osteotomy or a sliding osteotomy because both
are more stable than a transverse osteotomy and are associated with
a lower risk of trochanteric migration1.
The proximal part of the femur is exposed by reflecting the vastus
lateralis off the intermuscular septum anteriorly, with care taken
not to strip any residual bone of its soft tissue completely, as
it later is used as a vascularized autogenous graft. Prior to dislocation,
a Steinmann pin is inserted into the iliac crest and the distance
to the vastus tubercle is measured as a reference point to determine
leg length. In preparation for the femoral splitting osteotomy,
a transverse cut is made at the junction of deficient and healthy
femoral bone. It is to this level that the femoral split will extend
distally. This transverse osteotomy, which extends just through the
lateral half of the femur, must be done carefully with an oscillating
saw, with the medial aspect of the femur left intact. Then a midline
lateral femoral split is carried out, also with the oscillating
saw. The split is gently spread open with use of multiple osteotomes,
with careful levering against the femoral implant or cement in the
canal. The deficient femur cracks anteriorly and posteriorly and
falls open down to the site of the transverse osteotomy. When this
happens, often a spike of bone, usually on the medial side, remains
attached to the healthy distal part of the femur. This should not
be resected; instead the spike can be shaped into a step cut or
an oblique osteotomy to help stabilize the graft-host junction later.
Care is taken to preserve the soft-tissue attachments to the split
femur.
The acetabulum is reconstructed first so that the length of the
femoral allograft can be determined. The host femoral canal is nearly
always of a larger diameter than the allograft. A guide-wire is
inserted into the distal part of the host femur, and the canal is
gently reamed. It is important to emphasize that press-fit of the
implant into the host femur is not essential. A press-fit generally
leads to the use of a larger implant proximally, which reduces the
thickness of the cement mantle. It also results in excessive reaming
of the allograft to fit the implant, which weakens the graft and
reduces bone stock. Finally, a tight distal fit can result in distraction
at the host-allograft junction, which increases the risk of nonunion.
The approximate length of the graft required is assessed by a
trial reduction of the implant in the host femur into the acetabulum.
Stability and any preoperative limb-length discrepancy are taken
into account in determining the allograft length. The allograft
is cleaned of soft tissue and is reamed and broached until a good
fit for the implant is achieved while allowing for at least a 2-mm-thick
cement mantle around the stem. The long-stem femoral prosthesis
is narrow proximally so the graft does not have to be excessively
reamed. It is long enough to reach the distal femoral diaphyseal-metaphyseal junction.
Trial reduction of the allograft-prosthesis composite is carried
out, a step cut or oblique cut of approximately 2 by 2 cm is marked,
and the cut is adjusted for correct anteversion and length. On the back
table, the graft canal is cleaned and dried, and then cement is
pressurized into the graft. The stem then is inserted in the correct
anteversion. Great care is taken to ensure that the cement is recessed around
the step cut to allow host bone contact. The composite is then inserted
into the host femur, and the hip is reduced. Wires are passed around
or through the lesser trochanter of the allograft for later attachment
of the host greater trochanter. The implant is always cemented to
the allograft but not to the host. We have found that cementing
distally is unnecessary. Bone from the reaming and other available
host bone are applied to the host-graft junction to encourage union.
The step cut is stabilized with cerclage wires, and increasingly
we are reinforcing this with cortical struts fashioned from any
remnants from the allograft or from bone-bank bone. If there is
a large discrepancy between the diameters of the host femur and
allograft, stabilization can be difficult. Telescoping the graft
inside the host femoral canal is occasionally an option, but this
may require trimming of host bone. We generally try to minimize
resection of healthy host femur. As we do not obtain rigid press-fit
distally, intraoperative rigid stabilization of the graft-host junction is
essential to prevent nonunion. The residual proximal part of the
host femur is wrapped around the allograft and held with cerclage
wires (Fig. 3).
An attempt is made to bring these vascularized pieces distally to
wrap around the osteotomy junction to encourage union. Finally,
the trochanter is reattached to the allograft with the previously placed
wires, and autograft bone is applied if it is available. The average
operative time, including administration of anesthesia, positioning,
and performing the acetabular revision, was 4.2 hours (range, three
to six hours). The average blood loss was 2249 mL (range, 900 to
5600 mL). Currently, we routinely use a cell saver, but a cell saver
was not employed for this series.
Prophylactic antibiotics (usually a cephalosporin) are started
intravenously at the time of the operation and used for five days;
the antibiotics are then given orally for ten days. If the patient
is catheterized, gentamicin is given, in addition, for twenty-four
hours and then cotrimoxazole is given orally until the catheter
is removed. The patient is mobilized, non-weight-bearing on the
side of the operation until there is evidence of allograft-host
union, which is usually at three to six months.
Clinical Results
At the time of follow-up, a minimum of nine years and four months
after the operation, forty-five patients (75%) were alive,
fourteen patients (23%) had died, and one patient (2%)
had been lost to follow-up.
The fifteen patients who had died or been lost to follow-up had
a total of fifteen allografts (24%). The average duration
from the operation to the time of death or loss to follow-up was
five years and seven months (range, two years and four months to eight
years). The patient who was lost to follow-up moved out of the province
four years after the surgery but had had no complications up to
that point. All patients who died or were lost to follow-up thus had
been followed for a minimum of two years and four months. The failures
and complications in these patients up to the time of the last evaluation are
described below.
The living patients had a total of forty-eight allografts (76%)
and were followed for an average of eleven years (range, nine years
and four months to fifteen years). There were twenty-eight women and
seventeen men, with an average age at the time of follow-up of seventy-three
years and eleven months (range, fifty-three to ninety-two years).
The average preoperative Harris hip score was 30 points (range,
6 to 65 points); at the time of the latest follow-up of the patients
who had the original graft in situ, the average
score was 71 points (range, 47 to 95 points), with a mean score
of 70 points.
Radiographic Results
Radiographic analysis revealed four cases of nonunion (6%)
at the host-allograft junction. Two were treated at six months and
one was treated at twelve months with cortical struts and use of
autograft because of thigh pain. All three eventually had union
and resolution of symptoms. The fourth patient, who was seventy
years of age at the time of the operation, was only mildly symptomatic
and declined additional surgery. At the most recent follow-up evaluation,
at eleven years and six months, the nonunion persisted. The patient
had mechanical thigh pain and a Harris hip score of 57 points, but she
still declined additional surgery.
Trochanteric escape of >1 cm was seen in fourteen hips
(22%). No fractures occurred.
Peripheral allograft resorption in some form was seen in thirteen
(27%) of the forty-eight hips in living patients who were
followed for longer than nine years. There was mild resorption in
ten hips (21%) and moderate resorption in two (4%).
Resorption was most common in zone 7 (ten hips) and zone 2 (six
hips), with three hips having resorption in both zones. In nine
of the thirteen hips, the resorption occurred around the cerclage
wires. Interestingly, the resorption took several years to appear;
however, comparison with an extensive radiographic review performed
on twelve patients in 1994 (unpublished study) revealed the resorption
to be nonprogressive. One hip showed progressive, full-thickness
resorption of >1 cm in length at the site of a host-allograft
nonunion, in the patient discussed above. This resorption was associated
with subsidence of the allograft-prosthesis construct into the host
femur.
One other hip showed subsidence, which was 14 mm at the prosthesis-cement
interface, at eleven years. The graft remained intact with no evidence of
resorption. The patient had a Harris hip score of 47 points and
was awaiting femoral revision at the time of the latest follow-up.
There were no cases of cement fracture or endosteal resorption.
No revisions were performed because of graft resorption.
Complications
There were a total of thirteen complications (21%) related
to the allograft and requiring an additional operation in patients
with more than two years of follow-up. The complications included
five infections, three nonunions, two dislocations, and three cases
of late aseptic loosening. Another patient with a nonunion declined
additional surgery, as described, and two dislocations were treated
with closed reduction. Therefore, the total number of allograft-related
complications (including aseptic loosening) was sixteen (25%)
over the entire follow-up period. There were also six complications related
to the acetabular side of the revision.
Dislocation
There were four dislocations (6%) in the sixty-three hips
over the follow-up period or until the time of death. Two hips dislocated
within the first six months. One was treated with closed reduction
and had not dislocated again by eleven years postoperatively. Two
underwent acetabular revision and had not dislocated again by eleven
years postoperatively and by six years postoperatively (in a patient
who had died by the time of the current review). The fourth dislocation
occurred at eight years and was possibly related to wear; the hip
remained unstable, but the patient had severe dementia and was not medically
fit for another operation. On the pelvic side, there were four acetabular
failures due to failure of ingrowth of the uncemented cup.
Infection
Five hips (8%) had a deep infection. Four were treated
with removal of the allograft within two years after the operation.
Three had a second-stage reconstruction with another massive femoral allograft
and remained free of infection at an average of seven years and
six months postoperatively. The fourth infection was in a patient
who had had septic arthritis as a child and two prior infections
at the sites of hip arthroplasties before the femoral allograft
reconstruction. After the first-stage resection arthroplasty, the
patient did not wish to proceed to the second-stage reconstruction
as of the time of the latest follow-up. The final patient had a late
infection, seven years after the surgery, and was treated with two-stage
revision in another country.
Aseptic Femoral Failure
There were no cases of early aseptic loosening (within two years).
Three allograft-prosthesis constructs (5%) failed as a
result of symptomatic loosening on the femoral side after an average
of ten years and three months. In all three hips the loosening occurred
at the interface between the cement and the implant. Two hips were
successfully revised with repeat allografting at nine years and six
months and at ten years and two months after the original allograft
procedure. The third patient, discussed above, was awaiting revision
of a loose prosthesis eleven years after the operation. There was
mild resorption in one of these hips but no severe resorption or
endosteal resorption. There were no aseptic failures in the patients
who died.
Overall Results
According to our definition of clinical success, fourteen hips
(22%) were deemed failures. There were five infections,
four nonunions, two dislocations requiring a reoperation, and three
cases of late aseptic loosening. The success rate in the entire series
(including the patients who died and those who were lost to follow-up)
was 78% (forty-nine of sixty-three hips) at an average
of nine years (range, two years and four months to fifteen years).
The success rate for those who were alive at the time of follow-up
was 77% (thirty-seven of forty-eight hips) at an average
of eleven years. Of the clinical failures, three nonunions and both
dislocations were treated successfully with retention of the allograft
and without additional complications. Four of the five infections
and two aseptic failures were treated successfully with a new proximal
femoral allograft. According to the Kaplan-Meier method16, the probability of the construct
surviving for five years was 90% (95% confidence
limits, 80% to 95%) and the probability of the
construct surviving for ten years was 86% (95% confidence
limits, 74% to 93%) (Fig. 4).
Several methods of dealing with segmental femoral bone loss in
revision hip surgery have been described. Excision arthroplasty,
a salvage procedure, has poor results when there is severe bone loss9,18,19. Arthrodesis is difficult to
achieve in the presence of major bone loss20.
Diaphyseal fixation21-23 and impaction
grafting4,24-28 have had good
results. However, stem subsidence, stress-shielding, and failure
have all been associated with the degree of preoperative bone loss,
the canal diameter, and the amount of diaphyseal fixation obtained22,29-31.
For large segmental defects, as described in this paper, the
options are generally limited to substitution of the missing bone
with metal (a custom prosthesis)32-34 or
an allograft-prosthesis composite. The use of a metal component
alone has some disadvantages: instability because of poor soft-tissue
attachment35,36, late fatigue
fracture37,38, early loosening
as the stem is fixed only distally39,
severe stress-shielding, and difficulty with fixation especially
in a femoral diaphysis with a large diameter or only a short portion
remaining4,34. Finally, these
so-called megaprostheses are very difficult to revise, resulting
in even more destruction of the host bone stock and soft-tissue
attachments37.
The use of a large allograft-prosthesis construct in revision
arthroplasty remains somewhat controversial6.
Most of the reported clinical series have involved a small number
of patients, a variety of techniques, and relatively short-term
follow-up5,7,10,11.
Chandler et al.5 reported on
twenty-nine patients (thirty hips) at an average of twenty-two months
(range, two to forty-six months) after revision with an allograft-prosthesis
construct. There were two nonunions (7%), five dislocations
(17%), and three trochanteric escapes (10%) of
more than 1 cm. There were three failures (10%): one was
due to infection; one, resorption; and one, gross nonunion. Four
hips (13%) had a reoperation.
Masri et al.11 reported on
thirty-nine patients who had been followed for a minimum of two
years and an average of four years after a revision total hip arthroplasty with
a proximal femoral allograft. Complications included late allograft
fracture in one patient (3%), deep infection in two (5%),
allograft-host nonunion in four (10%), and nonunion of
the greater trochanter in eleven (28%). The total reoperation
rate was 26% (ten patients).
Previously, the senior one of us (A.E.G.) and colleagues reported
on 168 proximal femoral allografts followed for an average of 4.8
years40. The average Harris hip
score increased from 30 points preoperatively to 66 points at the
time of follow-up. There were seventeen revisions (10%)
in sixteen patients: three (2%) were due to infection; eight
(5%), dislocation; five (3%), nonunion; and one
(0.6%), pain. Radiographic analysis showed nonunion in
seven patients (4%), minor resorption in six (4%),
and severe resorption in one.
Major concerns regarding allograft-prosthesis constructs have
included a lack of long-term results, early reports of resorption
with the potential for subsequent graft failure, and a high complication rate6. We observed nonprogressive mild-to-moderate resorption
in 25% (twelve) of forty-eight hips that were followed
for a minimum of nine years and an average of eleven years. Only
one hip had progressive full-thickness resorption, which occurred around
a mobile nonunion site. Nine of the thirteen hips with resorption
had it at the site of a cerclage wire. This may indicate that the
process is related to local abrasion and a subsequent vascular reaction. No
hip had resorption of the entire allograft. The absence of complete
resorption has also been noted in long-term follow-up studies of
allografts used in reconstruction of skeletal defects following
tumor resection8,41.
The use of massive allografts requires specialized knowledge
and technical skill6,41,42, and
this series represents our early experience with the technique,
from which lessons have been learned. In a larger review, which
included our later experience, the infection rate was reduced from
8% (five of sixty-three [in the present study])
to 3% (six of 200)1.
We believe that this reduction was the result of careful screening
preoperatively and intraoperatively (frozen sections) for preexisting
infection, use of prophylactic antibiotics systemically and in the
cement, reduction of the operative time and blood loss due to an
increase in experience, minimization of soft-tissue dissection,
and intraoperative irrigation. Four of the five infections in the current
series were successfully treated with a two-stage revision, including
reinsertion of another large proximal femoral allograft.
The four nonunions in the current series were thought to be due
to inadequate host-allograft contact and insufficient stabilization
of the junction. The nonunion rate has subsequently been reduced from
6% (four of sixty-three) to 3.5% (seven of 200)1. This was done by ensuring maximal
allograft-host contact with a step or oblique-cut junction and by avoiding
cement at the junction; rigid stabilization of the junction with
strut grafts and cerclage wires (a step that cannot be overemphasized);
autografting of the junction site with any available host bone,
including femoral and acetabular bone from the reaming; and, finally,
careful attention to maintaining soft-tissue attachments to remnant
host bone that is specifically wrapped around the allograft-host
junction as a vascularized sleeve1.
The dislocation rate in the current series was 6% (four
of sixty-three). Retpen et al.43 reported
a rate of recurrent dislocation of 17% (ten of sixty) after
second revisions and 22% (four of eighteen) after third
revisions. A more constrained revision cup could possibly reduce
this rate further, but we remain concerned about the increased acetabular
stresses associated with such devices, especially as many of our
patients also have acetabular bone graft. The trochanteric nonunion
and escape rate was high in the current series (22%), but
this rate was later reduced by using a trochanteric slide approach44.
The complications of allograft fracture45 and
early stem subsidence46 that have
been reported in other series were avoided in our study by bypassing
the graft-host junction with the prosthetic stem so that the stem reached
the distal femoral diaphyseal-metaphyseal junction; pressurized
cementing of the prosthesis into a clean, dry allograft; using a
narrow prosthesis so that there was an adequate cement mantle; and avoiding
the use of screws or plates, which may weaken the graft47. There were three cases of late
subsidence of the stem (at an average of ten years and three months) into
the cement mantle in this series. These cases were all in young
active patients (less than fifty-five years old at the time of the
revision procedure). We could not identify any factors that may
have caused this subsidence, although we speculate that it may have
been related to the thickness of the cement mantle and its fatiguing
over time. We advocate use of the widest femoral allograft available
and a narrow stem that allows an adequate cement mantle.
The use of an allograft-prosthesis composite for treatment of
segmental loss is technically demanding. It requires ready access
to a bone bank and a surgeon and team experienced in revision. There are
major advantages to the use of an allograft-prosthesis composite
for hips with severe femoral bone loss, especially when there are
few surgical options. The technique allows the use of conventional
revision components. The allograft unites to the host, becoming
an integral part of the patient’s femur and allowing soft-tissue
attachment and increased stability. Host bone is preserved, especially
in the medullary canal, which permits additional revisions to be
performed with relative ease5,48,49.
This is especially important in younger patients, who will likely
face additional revisions.
As with any allograft bone there is always the concern about
the possible transmission of the human immunodeficiency virus or
other such agents. Adherence to the American Association of Tissue Banks’ standards
reduces the rate of transmission of human immunodeficiency virus
to one in 1,667,60050,51. The
risk is reduced further with the addition of radiation at a dose
of 2.5 Mrad (25,000 Gy) to all of the allografts52.
Although the risk is not completely eliminated, it still compares
favorably with the one in 493,000 risk of transmission in blood
transfusions53.
The proximal allograft needs to be of sufficient size to allow
an adequate cement mantle without excessive reaming. The implant
must be long enough to reach the distal diaphyseal-metaphyseal junction, but
distal press-fit is not essential and distal cementing is not advisable.
As distal press-fit is not always obtained, rigid stability of the
host-graft junction must be achieved intraoperatively to stabilize
the construct. Failure to do so results in nonunion, progressive
subsidence, local graft resorption, and failure of the construct.
Thus, increasingly we are using cortical strut allografts to stabilize
the step cut further. Long-term stability is obtained by host-allograft
union, so we recommend wrapping of the junction with vascularized
host bone as well as autografting of the junction if bone from the
reaming is available. When a junction does not progress to union
by six months, additional grafting should be done to prevent progressive
failure.
Revision hip replacements are associated with a high morbidity
rate2,9,35,43,54,55. They represent
the most complex cases of reconstruction of the hip, as the patients
have severe bone loss, have undergone multiple hip operations, and have
few other options. Our experience indicates that the use of an allograft-prosthesis
construct is a viable solution to this difficult problem.