The complexity of revision total knee arthroplasty is related to the extent of damage to the underlying bone. Conventional augments that attach to modular revision components may not adequately address the larger osseous defects encountered in association with osteolysis, implant migration, or bone loss resulting from multiple revision procedures. The surgeon must choose a method for repairing large bone defects that provides optimal implant fixation and long-term support for the components.
The current methods for repairing severe bone loss during revision knee arthroplasty include the use of structural allografts, impaction grafting with morselized allograft, and metal augments that are not an integral part of the implant. We and several other authors have described promising results with the use of structural allografts to manage severe bone loss1-5. These studies included small numbers of patients with a relatively short follow-up. An exception was a report by Clatworthy et al.6 on the results of the use of structural allografts to repair large uncontained defects. However, similar to other studies on this subject, the study by Clatworthy et al. included cases with confounding variables that make it difficult to analyze the results of different options for managing bone loss during revision arthroplasty.
It was our hypothesis that structural allograft used to reconstruct a tibial bone deficiency during revision knee arthroplasty would enhance component fixation and provide long-term support for the revision components. The purpose of this study was to determine the results of one surgeon's treatment of forty-seven patients (forty-nine knees) with revision knee arthroplasty with use of structural allograft to repair an osseous defect of the proximal part of the tibia.
From January 1985 through September 1999, one of us (G.A.E.) performed revision arthroplasty in forty-nine knees in forty-seven patients with a severe tibial bone defect. At the time of follow-up, the status of the knee implants was known for forty-six of the forty-nine knees. It was unknown for one patient (one knee) who was considered to be lost to follow-up after all attempts to contact the patient had been unsuccessful and for two patients (two knees) who had died before the minimum five-year follow-up period had elapsed. Six other patients (six knees) had died before the minimum five-year follow-up period had elapsed, and the implant was in situ at the time of death in all six. Four of the revision procedures (in four patients) failed within five years. Thirty-three knees (in thirty-two patients) were evaluated at a minimum of five years after the revision with the structural allograft. The senior author (G.A.E.) obtained an update on the remaining three patients (three knees) by means of a telephone interview. One patient with a bilateral revision knee arthroplasty had a clinical evaluation of one knee during the required follow-up interval and a telephone interview to determine the status of the other knee.
One or more structural allografts were used for the repair of both contained and uncontained bone defects. All patients who had had a revision knee arthroplasty with use of a structural tibial allograft during this time interval were included in this study. These arthroplasties represent 17% (forty-nine) of the 295 tibial component revisions performed during the same time interval. Institutional review board approval was obtained for this retrospective analysis of this prospective clinical database.
Nineteen of the patients were women, and twenty-eight were men. The mean age of the patients at the time of the revision surgery was sixty-seven years (range, thirty-nine years to eighty-six years), and the mean body mass index was 29.4 (range, 19 to 40). The procedure was performed in twenty-seven right knees and twenty-two left knees. The mean preoperative arc of motion was 88° (range, 12° to 125°). Prior to the index revision with the allograft, eleven patients had had one (nine patients) or two (two patients) revision procedures in the knee. The primary reasons for the index revision included polyethylene wear and osteolysis in twenty-four knees, aseptic loosening in seventeen, infection in five, instability in one, a femoral fracture in one, and a failed patellar component in one. In the knees that had the index procedure because of instability, femoral fracture, or a failed patellar component, the tibial component was revised to be compatible with the revision knee system. Thirty-two knees required revision on both the femoral and the tibial side, and seventeen knees required revision on the tibial side only.
Preoperative Planning
Standing anteroposterior and lateral radiographs (with a 35.56 × 43.18-cm cassette) were made to assess the tibia for the possibility of a severe bone defect. Bone loss secondary to osteolysis often was far more severe than was apparent on standard radiographs. Bone damage that occurred with implant loosening and component migration included not only the segment of absent bone that was clearly evident on radiographs but usually also a segment of necrotic bone that appeared densely sclerotic beneath the subsided component.
In all forty-nine knees, the tibial bone deficiency was classified as deficient metaphyseal bone or an Anderson Orthopaedic Research Institute (AORI) Type-3 tibia7. Femoral head allograft(s) were used to reconstruct the bone defect in forty-five knees, proximal tibial allograft was used in three knees, and distal femoral allograft was used in one knee. The bone deficiency was reconstructed with the structural allograft alone in thirty-five knees and with structural allograft in conjunction with a tibial augment in fourteen knees. Because the bone deficiency often was larger than expected and there was a possibility that the allograft could be damaged or contaminated during the preparation process, an extra allograft was available at the time of surgery. The metal augment to the revision tibial component consisted of an integral half-wedge in five knees, a modular 10° step-wedge in one knee, an integral full wedge in five knees, and a modular 7° full wedge in one knee. In all but one knee, the metal augmentation was used to address bone deficiency that was most prominent on the medial side of the knee. An extra-thick (10-mm), nonmodular cobalt-chromium tibial component was used in two knees.
The revision implants included thirty-three Coordinate components (DePuy, Warsaw, Indiana), seven Modular IB (Insall-Burstein) components (Zimmer, Warsaw, Indiana), three Arizona components (DePuy), two Press Fit Condylar (PFC) components (Johnson and Johnson, Raynham, Massachusetts), two custom Total Condylar III components (Johnson and Johnson), one Kinematic Superstabilizer component (Howmedica, Rutherford, New Jersey), and one Total Condylar component (Johnson and Johnson). None of the revision implants were hinged. Of the thirty-three Coordinate components, six had a varus-valgus constrained polyethylene tibial insert and twenty-seven had a posterior-stabilized insert.
A straight stem was used in forty-one knees; it was inserted without cement in thirty-six knees and with cement in five. A tapered stem was used in four knees, and a curved stem was used in three. All of the tapered and curved stems were cemented. No stem extension was used for the Total Condylar component. Thirty-nine tibial stems (including all thirty-six straight stems inserted without cement) were 140 mm or longer. The stems used in the thirty-three Coordinate knee revisions were made of cobalt-chromium and had multiple 0.5-mm fins. None of the stems had a clothes-pin-type end.
Operative Technique
An additional technique was necessary to achieve adequate exposure for the revision procedure in 35% (seventeen) of the forty-nine knees. These procedures included a quadriceps release in ten knees, a tibial tubercle osteotomy in four knees, and an epicondylar osteotomy8 in three knees.
The surgical technique was similar for the repairs of both the contained and the uncontained tibial defects. First, the bone defect was cleared of all friable soft tissue and osteolytic membranes. Any residual bone cement was also removed. The medullary canal was accessed with a blunt-tipped drill, or a sharp-tipped drill when necessary, and then the diaphyseal segment was opened with progressively larger rigid reamers. The diameter of the revision tibial stem was estimated from the preoperative radiographs and confirmed by reaming with reamers of increasing size until resistance in the medullary canal was encountered. In most cases, a tibial stem of 140 mm or longer was used in order to engage the diaphyseal segment of the tibia.
When the need for an allograft was apparent, the femoral head graft was selected, placed in warm saline solution, and allowed to thaw for at least twenty minutes. To stabilize the graft during preparation, the femoral head was placed in an Allogrip bone-holding device (DePuy) (Fig. 1). Once the graft was stabilized, a female-type cheese-grater reamer with a diameter that was just slightly larger than that of the femoral head was used to remove the dense bone from its dome. Next, a female-type reamer with a diameter that was smaller than the graft diameter was used to remove bone from the hemispheric region of the graft. Reaming continued until cancellous bone was exposed over the full hemisphere of the allograft femoral head. To achieve an interference fit, the graft was measured to confirm that its diameter was not smaller than the diameter of the bone defect. The graft was thoroughly washed with saline solution with use of a pulsed irrigation device to remove marrow elements and then was placed in warm saline solution while the defect was prepared for the allograft.
A male-type cheese-grater reamer with a diameter that was the same as or one size smaller than the diameter of the available allograft was used to prepare the bone defect in the host bone. In some instances, sclerotic bone made it difficult to control the male reamer from wandering during preparation. Creating a central recess by removing sclerotic bone at the floor of the defect permitted more controlled use of the reamer. Contained defects were the easiest to prepare by reaming until viable bone was encountered at the floor of the defect (Fig. 2). As the tibial diameter was greater in the medial-to-lateral direction than in the anterior-to-posterior direction, we attempted to avoid reaming away the anterior or posterior wall of the tibia. The femoral head was only partially captured in the more severe uncontained defects. In this situation, the reaming progressed until a solid base for the allograft was reached.
Kirschner wires were used to temporarily stabilize the allograft to the host bone. The wires were directed parallel to the joint surface and in a position that did not interfere with the stem of the tibial component. The wires were inserted at a level below the anticipated joint line for the revision tibial component. Usually, two wires were sufficient. Once the allograft was stabilized, a rough cut was made to remove the portion of the graft that protruded above the anticipated joint line. In all cases, the wires were removed after final component placement and fixation of the tibial component with bone cement.
After the allograft had been stabilized and a rough cut had been made to remove the portion of the graft that protruded above the anticipated joint line, the allograft often blocked access to the medullary canal of the tibia. The medullary canal first was accessed with a high-speed burr (Midas Rex, Fort Worth, Texas), with which the surgeon made a pilot hole through portions of the allograft. A small intramedullary reamer was then passed through the pilot hole and extended through the full depth of the prepared medullary canal to identify where the pilot hole needed to be enlarged. Next, the pilot hole was enlarged to pass progressively larger rigid intramedullary reamers into the canal until the reamer diameter corresponded to the diameter of the revision tibial stem.
Finally, the proximal part of the tibia was prepared with use of the standard intramedullary alignment device for the revision system that had been selected. In some instances, the allograft did not adequately reconstruct the entire segment of the proximal part of the tibia. In this situation, an augment, a portion of a second allograft, or separate allografts were used for the medial and lateral tibial plateaus. At times, the unused portion of a femoral head allograft was placed with an impaction grafting technique to provide additional support for smaller bone defects in the less involved tibial plateau.
Characteristics of the Allografts
Thirty-nine of the forty-nine tibiae were reconstructed with one femoral head allograft. The bone defect was mostly contained in twenty-nine knees and was uncontained in the remaining ten knees. For the reconstructions of the contained defects, the single femoral head allograft was centered in the medial tibial plateau (thirteen knees), more centrally in the tibial plateau (twelve knees), or in the lateral plateau (four knees). Six knees required two femoral heads to reconstruct the entire proximal part of the tibia. In four additional knees, the proximal tibial bone was restored with a full-segment allograft; a proximal tibial allograft was used in three knees, and a distal femoral allograft was used in the fourth. In the ten knees that required two femoral heads or a segmental allograft, the bone loss involved the entire proximal surface of the tibia.
Statistical Methods
The Kaplan-Meier technique was used to estimate implant survivorship and the 95% confidence interval, with removal or revision of the implant for any reason as the criterion for failure. An additional survivorship analysis was performed with removal or revision of the tibial component as the criterion for failure.
The status of the revision implant was known for forty-six of the forty-nine knees in this study. Of the eight patients who died before the minimum five-year follow-up interval had elapsed, six (six knees) were known to have had the revision implant in situ at the time of death. Two of the six knees were retrieved at autopsy, and histological analysis confirmed union between the host bone and the allograft (Fig. 3) as well as a stable tibial component.
Thirty-three knees (in thirty-two patients) were evaluated clinically at a minimum of five years after the arthroplasty with the allograft. At a mean of ninety-seven months (range, sixty-one to 191 months), the mean Knee Society clinical score was 84 points (range, 33 to 100 points) for these thirty-three knees, and the mean arc of motion had improved to 103°, compared with 87° preoperatively. Ninety-one percent (thirty) of the thirty-three knees were fully stable in the anterior-posterior direction (<5 mm of motion) and 82% (twenty-seven) were stable mediolaterally (<6° of angulation with the knee in 30° of flexion) at the time of follow-up. Seven knees had a flexion contracture, which was between 5° and 10° in six of them and was 20° in one. Three knees had an extension lag, which was <10° in two of them and between 10° and 20° in one. Nine knees had a mild effusion, and one knee had a moderate-to-severe effusion but was functioning. The cause of the effusion was not known.
Radiographs of the thirty-three unrevised knees were examined, and no complete radiolucencies were identified on the tibial side. The radiographs of two knees revealed an osteolytic lesion (>2 cm) on the tibial side, and a small amount of osteolysis (<2 cm) was seen on the radiographs of an additional four knees. The osteolysis did not involve the metaphyseal area repaired with the structural allograft. None of these patients had required additional surgery because of the osteolysis by the time of follow-up. No evidence of graft resorption or collapse was identified on the radiographs. Graft remodeling, seen as a change in graft density along the uncontained border of the allograft, was identified in several knees. Figures 4-A, 4-B, and 4-C demonstrate the most dramatic radiographic changes in both graft density and subperiosteal remodeling evident in this series.
Failures and Additional Procedures
Four knees in four patients failed and required a reoperation within five years after the index revision procedure with the structural allograft (Table I). There were no additional failures in the study group. Of the four patients with a failure, one underwent a knee arthrodesis with an allograft and three were treated with a rerevision. In one of the three patients with a rerevision, the allograft and the tibial component were stable and were not revised. In another, the structural allograft was intact and provided excellent support for the new revision tibial component. The third patient had a full revision, but the status of the graft was unknown. The four failures did not appear to be directly related to collapse or failure of the allograft; instead, they were secondary to femoral osteolysis, instability, infection, or arthrofibrosis.
Reoperations that did not involve revision as well as other complications in this series are listed in Table II.
Survivorship
Kaplan-Meier analysis showed the ten-year survival rate to be 91% (95% confidence interval, 82% to 100%) when removal or revision of any component of the implant for any reason was used as the end point and 93% (95% confidence interval, 85% to 100%) when removal or revision of the tibial component was the end point. All revisions or reoperations involving the implant occurred within five years after the index revision.
The relatively high complication rate following revision knee arthroplasties in patients with major bone loss is not surprising. The need for ancillary procedures such as release of the quadriceps tendon and osteotomy of the tibial tubercle reflects the difficulty in achieving exposure because of the extensive scarring often encountered in these complex cases. However, the complications and repeat revisions in this study were mostly related to problems encountered with revision total knee arthroplasty in general and not specifically to the allograft repair of the large bone defect.
The unique aspects of this study include its focus on revision knee arthroplasties that required structural allograft to reconstruct substantial tibial bone deficiency and were done primarily with a component with a long, press-fit diaphyseal stem. No hinged components were used for these revision procedures, in contrast to the studies by Clatworthy et al.6 and Backstein et al.9, who reported failure rates of 23% (twelve of fifty-two) and 21% (thirteen of sixty-one), respectively. The failures in those studies included cases of graft resorption that resulted in implant loosening. We found no evidence of graft resorption and no knee required a repeat revision because of loosening of the tibial implant in our study. The high rates of loosening in the other studies may be related to the type of allograft (more segmental grafts were used) and the constraint of the revision component. In our study, segmental grafts were used in four knees. In the remaining forty-five knees, one or more femoral head allografts were used to manage a Type-3 tibial deficiency.
Infection was reported as a cause of failure both by Clatworthy et al.6 and by Backstein et al.9. In the current study, two knees failed and required revision because of infection. In one of these knees, the original revision had been done because of infection. Therefore, it is likely that the knee failed because the infection had not been eradicated when the revision was performed.
Other methods of managing large tibial bone defects include impaction grafting and use of the newer metal augments that are not integral to the implant. To our knowledge, the longest duration of follow-up in a study on impaction grafting was a mean 5.3 years, as reported by Barden et al.10. In that study, impaction grafting was used in sixteen of twenty knees with a tibial defect and structural allograft was used in the remaining four knees; only one knee failed. Short intramedullary stems ranging from 8.5 to 10.5 cm were used. The authors stated that the use of cement was restricted to the cut bone surface, but for the latter revisions that involved large bone defects the stem was anchored with cement or a press-fit diaphyseal-filling stem was used with the modular revision prosthesis.
There is a great deal of controversy about the best method for fixation of the components used for revision surgery. Both Fehring et al.11 and Whaley et al.12 reported results that support the use of fully cemented stems. However, all of their patients were treated with shorter cemented and cementless metaphyseal-filling stems and not with diaphyseal-engaging stems. It is difficult to revise a fully cemented stem, particularly a longer diaphyseal-engaging stem, and such a revision may jeopardize the outcome of repeat revision surgery. The majority (thirty-six) of the forty-nine knees in our study were managed with a diaphyseal-filling stem that was not cemented. We believe that repairing a bone defect with the cancellous bone of a structural allograft enhances implant fixation in the metaphyseal region, thereby obviating the need to fully cement the stem. In our study, no revision tibial implant inserted with this technique failed as a result of aseptic loosening. In the current study, an additional benefit of using uncemented stems was that removal was easier and less bone was lost in the three knees that required repeat revision of the tibial component.
In summary, we believe that use of structural allograft provides a stable and durable reconstruction of a tibial bone deficiency encountered during a revision total knee arthroplasty. At a mean of 7.9 years, we found no instance of graft collapse or aseptic loosening of the tibial component associated with the structural allografts used in these knees. We recommend the use of structural allograft and a press-fit long-stemmed component for revision knee arthroplasty in knees with severe tibial bone deficiency. 