Disruption of the extensor mechanism is a devastating complication of total
knee arthroplasty, with a prevalence of
0.17%1 to
2.5%2,3
in series ranging in size from
2812 to 8288
knees1. This problem
has many causes4.
Multiple techniques for repair or reconstruction of a deficient extensor
mechanism associated with a total knee arthroplasty have been described.
However, despite encouraging results reported following direct repair in
native
knees5-8,
attempts at primary
repair1,2
following total knee arthroplasty have rarely restored extensor
function9,10.
The use of local autogenous tissue to augment a primary repair has been
recommended3,11,12,
but these autogenous tissues may be compromised in patients who have undergone
multiple previous knee procedures.
Emerson et
al.13,14
reported on the use of an extensor mechanism allograft after total knee
arthroplasty. Although the early clinical results were promising, further
follow-up revealed that an extensor lag of 20° to 40° had developed in
three of nine knees. Nazarian and
Booth15 modified
the technique described by Emerson et al. by recommending that the allograft
be tightly tensioned in full extension. Thirty-four of their thirty-six
patients treated with this modified protocol had a successful clinical
result.
The purpose of the present study was to compare the results of
reconstruction with an extensor mechanism allograft as performed with the
initial technique of Emerson et al. (with the graft loosely tensioned) with
the results of the reconstruction as performed with the technique of Nazarian
and Booth (with the graft tightly tensioned in extension).
Twenty consecutive reconstructions with an extensor mechanism allograft
were performed in nineteen patients by three surgeons at the same institution.
All knees had a failed extensor mechanism associated with a total knee
arthroplasty. All patients were followed prospectively, and the data were
reviewed retrospectively at a minimum of twenty-four months (or until failure
of the allograft reconstruction). Many of the patients in both groups had had
a failed primary repair and/or augmented repair (with use of local autogenous
hamstring tendon graft) of the extensor mechanism. Nonoperative treatment,
including bracing and prolonged immobilization, as well as other operative
options, including additional attempts at primary repair and a knee
arthrodesis, were discussed with and offered to all of the patients. All
patients who had undergone revision with an extensor mechanism allograft in
the period from 1989 to 2001 were included in this retrospective review.
Group I consisted of seven knees in six patients (four women and two men).
Group II consisted of thirteen knees in thirteen patients (nine women and four
men). On the average, the patients in Group I were older than those in Group
II at the time of the index surgery (seventy-four years [range, sixty-two to
eighty-two years] compared with sixty-four years [range, fifty-one to
seventy-seven years]; p = 0.042). The patients in Group I were followed for a
mean of thirty-eight months (range, twelve to 115 months), and those in Group
II were followed for a mean of thirty-seven months (range, twenty-seven to
forty-six months) (p = 0.384). For the purposes of the analysis, the time to
revision or clinical failure was used as the end point for the four Group-I
knees in which the allograft had failed less than twenty-four months
postoperatively. No patient was lost to follow-up. A mean of 3.8 (range, one
to four) previous open knee procedures had been performed in Group I, and a
mean of 3.5 (range, one to eight) had been done in Group II (p = 0.60).
In the seven Group-I
knees16, operated
on between 1989 and 1998, the allograft was tensioned to allow for
approximately 60° of passive flexion intraoperatively, as described by
Emerson et
al.13,14.
Five knees had undergone previous, unsuccessful primary repairs of the
patellar tendon, which had ruptured following a total knee arthroplasty; one
knee had a delayed traumatic rupture of the patellar tendon; and one had
patellar fragmentation with quadriceps disruption and a failed previous
attempt at primary repair. In the thirteen Group-II knees, operated on between
April 2000 and November 2001, the allograft was tensioned tightly in full
extension, as described by Nazarian and
Booth15. Three
Group-II knees, which had had a patellectomy before the total knee
arthroplasty, had chronic attenuation of the extensor mechanism and
instability subsequent to the arthroplasty, three knees with a previous
infection at the site of the total knee arthroplasty had avulsion of the
patellar tendon from the tibial tubercle, four knees had chronic patellar
tendon disruption after the total knee arthroplasty and failure of a previous
primary repair, one arthrofibrotic knee with severe patella infera had had the
total knee arthroplasty following a prior high tibial osteotomy complicated by
infection, and two knees had a quadriceps tendon attenuation and clinical
failure of an attempt at primary repair.
At the time of the extensor mechanism allograft procedure, six of the seven
knees in Group I and eight of the thirteen knees in Group II underwent a
simultaneous revision total knee arthroplasty because of aseptic loosening
and/or component malrotation. Three patients in Group II initially underwent a
resection arthroplasty and insertion of an antibiotic-loaded cement spacer to
treat a deep periprosthetic infection and then underwent the extensor
allograft reconstruction at the time of the second-stage reimplantation. The
polyethylene liner was exchanged at the time of the allograft procedure in all
of the patients who did not undergo component revision. A constrained condylar
articulation was used in five of the seven knees in Group I and in four of the
thirteen knees in Group II.
Clinical and radiographic examination was performed at six weeks, twelve
weeks, six months, nine months, one year, and yearly thereafter. At each
follow-up visit, a surgeon (R.S.J.B. or C.J.D.V.) other than the operating
surgeon examined the patient. The extensor lag was calculated by subtracting
the value for maximum active extension against gravity from that for maximum
passive extension of the knee. The knee was graded according to the 100-point
system of The Hospital for Special
Surgery17,18
preoperatively and postoperatively, beginning at three months. A score of
>84 points is considered an excellent result; 70 to 84 points, a good
result; 60 to 69 points, a fair result; and <60 points, a poor result.
Because quadriceps muscle weakness is associated with an increased risk of
falls19, all
patients were questioned regarding, and the charts were reviewed for, a
history of falls. Postoperative complications and the need for additional
surgery were recorded. Clinical failure was defined as an extensor lag of
>30°, recurrent falls secondary to incompetence of the extensor
mechanism, or regression to a lower category of ambulatory status following
the reconstruction. Radiographs, including standing anteroposterior, lateral,
and patellar skyline views, were reviewed at each visit. The alignment of the
allograft patella within the trochlear groove was graded as centered,
subluxated, or dislocated.
Operative Technique and Postoperative Management
A midline arthrotomy was performed with preservation of all quadriceps and
residual patellar tendon tissue. If the patella remained, it was removed (as
per the recommendation of Nazarian and
Booth15) with
preservation of the continuity of the retinaculum. The proximal portion of the
tibia was exposed with subperiosteal elevation over the tibial tubercle.
Component removal and revision total knee arthroplasty were performed as
necessary.
All allografts were fresh-frozen, nonirradiated specimens, and the initial
specimen consisted of an allograft tibia with its attached patellar
tendon-patella-quadriceps tendon segment
(Fig. 1). At least 3 cm of
quadriceps tendon was used to maximize proximal fixation of the allograft to
the host quadriceps tissue. Before the allograft was prepared, full passive
extension of the total knee prosthesis was obtained intraoperatively. The
allograft tibial tubercle was then cut with a microsagittal saw into a
3-cm-long, 1.5-cm-wide, 1-cm-deep block with a dovetail oblique cut in the
proximal aspect to provide an interference fit into the host tibia. The host
tibial tubercle was removed with the microsagittal saw to accept the allograft
tibial tubercle bone block to create a lock-in-key fit, which provided
inherent stability prior to any internal fixation of the tibial bone block.
The allograft tubercle was then press-fit into place and was fixed with three
stainless-steel 18-gauge wires (Fig.
2) with the addition of a cancellous screw and washer if more
stability was required.
The quadriceps junction was then repaired with use of a locking Krackow
stitch20 with
heavy, nonabsorbable number-5 suture and with additional suture placed along
the medial and lateral sides of the allograft quadriceps tendon. The host
quadriceps was oversewn with the sealed vest-over-pants method over the top of
the allograft, with as much of the allograft covered as possible. In Group I,
the allograft patella was appropriately positioned on the femoral component
throughout the range of motion of the
knee13 and at the
anterior flange of the femoral component with the knee in extension.
In the first seven knees (Group I), the repair was then completed without
excessive tightening of the allograft, allowing up to 60° of passive
flexion against gravity intraoperatively on the table. In the next thirteen
knees (Group II), the allograft was tightly tensioned, in full extension; this
resulted in only 20° to 30° of flexion against gravity
intraoperatively. To achieve this tensioning, the native quadriceps with the
attached retinaculum was tensioned and pulled distally (with the use of
atraumatic retention sutures) by an assistant to maximally advance the host
soft tissues while the allograft was pulled proximally, also with the use of
retention sutures. The allograft was covered anteriorly by the host extensor
mechanism as much as possible, with a vest-over-pants repair. The surgeon then
repaired the graft with nonabsorbable sutures while this amount of tension was
maintained. No specific measurement tool or tensioning device was used for the
tensioning. The retinaculum was always repaired while tension was maintained
on the repair. The knee was protected from further flexion once the repair was
completed, and closure was completed routinely.
The patella was not resurfaced in either group. Skin closure was performed
in all patients without the need for local or free flap coverage. The knee was
immobilized in full extension postoperatively.
Postoperative management differed between the two groups. In Group I, the
knees were braced in full extension for six to eight weeks. Isometric
quadriceps exercises, under the supervision of a physical therapist, were
started. Only toe-touch weight-bearing was allowed until radiographs showed
incorporation of the tibial tubercle allograft, at which time full
weight-bearing was permitted. At six to eight weeks, a progressive range of
active flexion to 90° was allowed in a hinged knee brace. In Group II, the
protocol was changed. The knee was braced or immobilized in a cast in full
extension for eight weeks, and touch-down weight-bearing was allowed.
Isometric quadriceps exercises were started during this period. After eight
weeks, 30° of active flexion was permitted in a hinged knee brace, and the
patient advanced to weight-bearing as tolerated. At twelve weeks, further
active flexion up to 90° was allowed, and gentle quadriceps-strengthening
exercises were initiated. To minimize the chance of graft failure, passive
flexion was not permitted in either group.
Statistical Analysis
Nonparametric analysis was performed with use of the Mann-Whitney U test
for continuous variables, and nominal data were compared with use of the
Pearson chi-square cross-tabs method (SPSS, Chicago, Illinois). A paired
Wilcoxon test was used in the analysis of paired samples in Group II.
Significance was defined as a p value of <0.05.
With the numbers available, there was no difference between Groups I and II
with regard to the duration of follow-up or the preoperative knee scores, knee
extensor lag, flexion contracture, or arc of knee flexion
(Table I). At the most recent
evaluation, the mean extensor lag was 59° (range, 40° to 80°) in
Group I compared with 4.3° (range, 0° to 15°) in Group II (p <
0.0001 for the difference between the groups, and p = 0.003 for the
improvement in the extensor lag in Group II). All of the reconstructions in
Group I were considered to be clinical failures. The mean knee score at the
time of the latest follow-up was 52 points (range, 15 to 81 points) in Group I
compared with 88 points (range, 62 to 100 points) in Group II. Group II showed
a significant improvement in the postoperative knee score compared with the
preoperative score (p = 0.002), and the difference in the postoperative knee
scores between the two methods of graft fixation was also significant (p <
0.0001). One patient in Group II had a postoperative Hospital for Special
Surgery score of 62 points, which we attributed to a failed extensor mechanism
in the contralateral knee and debilitating lumbar spondylosis.
At the time of the latest follow-up, three of the six patients in Group I
were able to walk about the community, three were able to walk about the
house, and all but one required the use of an assistive device at all times.
Four of the knees in Group I required a hinged knee brace, and five of the six
patients had documented episodes of falling postoperatively, which they
attributed to knee instability. All thirteen patients in Group II were able to
walk about the community; two required the use of a cane, and two (one with
associated cervical myelopathy and one with severe lumbar spondylosis) used a
walker. No patient in Group II reported a fall or knee instability or required
the use of a brace on the operatively treated knee.
One patient in each in group required revision of the tibial tubercle
fixation. Radiographic analysis showed all other allografts to be incorporated
at the tibial tubercle by six months. All of the patellar allografts were
graded as centered within the trochlear groove of the femoral component on the
patellar skyline radiograph.
Complications
Six of the seven knees in Group I required repeat surgery. In three of
these knees, a revision was performed within the first year because of
attenuation and stretching of the allograft at the quadriceps or patellar
tendon anastomosis. At the time of the revision, severe attenuation of the
allograft tissue was noted at the proximal anastomosis. The fourth knee had a
rupture of the allograft patellar tendon. In each of these knees, the
attenuated regions were repaired with the existing allograft tensioned tightly
in full extension. The fifth knee had a nonunion of the tibial tubercle, which
was repaired with repeat internal fixation. The sixth knee failed at the
patellar tendon insertion and was revised with the use of another extensor
mechanism allograft tensioned to allow 60° of flexion. That reconstruction
later failed and required revision with use of an extended medial
gastrocnemius transposition
flap11. Two other
failed reconstructions in Group I were revised with the use of a local medial
gastrocnemius muscle rotation flap for soft-tissue coverage.
Three of the thirteen knees in Group II required a repeat procedure. One of
them required repeat internal fixation of the tibial tubercle six weeks after
the original procedure. The tibial tubercle united without complication. The
second patient had a reoperation because of the development of a non-traumatic
partial tear of the quadriceps tendon anastomosis, with a symptomatic extensor
lag and pain, twelve weeks postoperatively. At the time of the reoperation
(sixteen weeks postoperatively), a tear on the medial aspect of the
anastomosis was identified and repaired. The third patient required a
manipulation under anesthesia for a stiff knee and poor flexion at twelve
weeks postoperatively; that patient subsequently gained 90° of flexion. In
addition to those three cases, a complex regional pain syndrome developed
postoperatively in one patient and recurrent painless knee effusions developed
in another. An evaluation for infection in the knee with the effusions
revealed negative findings, and the knee remained painless and continued to
function well.
The results of this study suggest that tensioning of an extensor mechanism
allograft in full extension yields a higher rate of clinical success,
including the potential for restoration of full active extension and better
walking status, while not adversely affecting final flexion. However, in Group
II, two patients had a reoperation, one had manipulation under anesthesia, one
had a painful knee, and one had effusions.
The initial
experience16 with
extensor mechanism allograft reconstructions (Group I) was disappointing,
leading to a change in technique. Emerson et
al.13 reported
promising early results after use of an extensor mechanism allograft to
reconstruct a failed extensor mechanism in patients with a previous total knee
arthroplasty. However, the authors concluded that the long-term results needed
further evaluation. The original technique of Emerson et al. was modified by
Nazarian and
Booth15, who
tensioned the allograft tightly after ensuring full knee extension before
placement of the graft. Crossett et
al.21 recently
reported the results after using an Achilles tendon allograft for similar
indications.
Histological analysis of incorporation of an extensor mechanism allograft
at both bone and soft-tissue
interfaces22
indicated that mechanical injury to the allograft secondary to prosthetic
impingement may be one mode of failure of this reconstruction.
We chose a fresh-frozen nonirradiated allograft over a freeze-dried
allograft because of the results previously described by Emerson et
al.13,14
and because of concerns that freeze-drying may weaken the
allograft13,14,23,24
and that freeze-drying has a greater risk of generating a host immune
response25-28.
None of the patellae were resurfaced in our study. We agree with Emerson et
al.13,14
that, as the allograft tissue is insensate, patellar resurfacing both is
unnecessary and potentially weakens the graft.
Malalignment of the components in total knee arthroplasty predisposes a
patient to extensor
dysfunction29 and
may contribute to further failure of or stress in a weakened or devascularized
host extensor mechanism. Fourteen of the twenty knees in our series required
simultaneous component revision. A surgeon faced with an incompetent extensor
mechanism must search for factors, such as malalignment, that may have
contributed to the extensor mechanism failure and must correct these problems
at the time of the repeat surgery. The possibility of an infection being
associated with the extensor mechanism failure should also be considered, as
three of the twenty patients in our series required a two-stage exchange
protocol prior to the reconstruction of the extensor mechanism.
The current study confirms the belief of Nazarian and
Booth15 that the
technique of tensioning of the extensor mechanism allograft is a crucial
determinant of success. In Group II, we strictly followed the guidelines
outlined by Nazarian and Booth by tightly tensioning the graft in full
extension. We also followed a similar postoperative protocol, although we used
a longer period of immobilization (eight weeks compared with the six weeks
described by Nazarian and Booth).
The patients in Group I were older, on the average, than those in Group II
(mean, seventy-four compared with sixty-four years; p = 0.042). It is possible
that this factor contributed to the inferior clinical results seen in Group
I.
Revisions were required in both groups. In Group I, six of the seven knees
were revised, and substantial attenuation of the allograft, with the
graft-host junctions remaining intact, was noted at the time of the
reoperations. At the time of writing, there had been no revisions due to graft
attenuation in Group II. The site of failure of the grafts in Group I was
neither consistent nor predictable, and all of the Group-I knees that
presented with attenuation of the graft had had an initial period of good
extensor function with later deterioration into a progressive extensor
lag.
In summary, the technique of allograft tensioning plays a significant role
in the outcome of a reconstruction of a failed extensor mechanism with an
extensor mechanism allograft following a total knee arthroplasty. Careful
attention to graft preparation and handling, tensioning of the allograft, and
postoperative rehabilitation led to encouraging results in this study of the
management of a failed extensor mechanism following total knee arthroplasty.
?