The proximal part of the tibia is a relatively common site for malignant bone tumors1,2. Surgical treatment for primary sarcomas and certain aggressive benign tumors in this area may involve wide excision of the proximal aspect of the tibia. Reconstruction of these surgically created defects is challenging, and no single technique has been found to be clearly superior to others. Current techniques include rebuilding these defects with endoprostheses, osteoarticular allografts, and allograft-prosthetic composites3-5. The early experience with osteoarticular allografts of the proximal part of the tibia demonstrated that this surgical location was associated with an especially high rate of complications, including infection and fracture6. Given the potentially high complication rate with tibial reconstructions, there are proponents of alternative surgical treatments, including amputation, arthrodesis, and rotationplasty, for young, active patients. Advances in modern prosthetics have improved the function of amputees, and some patients opt for an above-the-knee amputation or rotationplasty in order to have fewer restrictions with regard to sports and physical labor.
One of the main challenges in the proximal aspect of the tibia is the restoration of active knee extension. The integrity of the extensor mechanism is compromised by resection of the tibial tubercle, and reattachment of the patellar tendon has had variable results7-12. Abboud et al. reported a mean extensor lag of 7.5° in a series of twenty-two patients who had segmental endoprosthetic replacements of the proximal part of the tibia13. Ayerza et al. found that twenty-four of thirty-four patients who had osteoarticular allograft reconstructions had no extensor lag, while ten patients had a mean extensor lag of 6.5°7. Biau et al. reported failure of the extensor mechanism in six (24%) of twenty-six patients with allograft-prosthetic composites9. Inadequate reattachment is manifested by patella alta and weakness of the extensor mechanism14,15. This can result in substantial functional limitations for the patient, such as chronic limp, pain, and dependence on external support9,11.
Allograft-prosthetic composites combine the theoretical advantages of enhanced tendon attachment and increased durability of mobile joint surfaces16. In the proximal aspect of the tibia, the native patellar tendon can heal to the allograft tendon, which can gradually become revascularized and reconstituted with the fibroblasts of the host. During this process, the allograft tendon remains firmly attached to the allograft bone7. Metal prostheses have the potential advantages of greater modularity, ease of insertion, and lower cost. There may also be a lower infection and wound complication rate as a result of faster operative times with metal prostheses. However, they lack a strong attachment point for biological ingrowth of the patellar tendon.
A comparison of osteoarticular allografts with allograft-prosthetic composites reveals relative advantages for the allograft-prosthetic composite reconstruction. Unlike osteoarticular allografts, which can undergo subchondral fragmentation and degeneration of the articular cartilage, allograft-prosthetic composites are resurfaced with prosthetic materials that are more resistant to these processes. Allograft-prosthetic composites also do not require careful ligamentous repair and balancing, which is an important part of osteoarticular allograft reconstruction. Proponents of osteoarticular allografts, however, have noted that preservation of the native articular cartilage may be important, particularly for young patients with open distal femoral physes. Furthermore, the osteoarticular allograft does not incur the additional expense of a prosthesis in addition to the allograft.
Previous reports on proximal tibial allograft-prosthetic composites have shown a relatively high prevalence of complications, the most notable of which include fracture, loosening, infection, and extensor mechanism rupture9,17. The literature on allograft-prosthetic composites of the proximal part of the tibia is, however, sparse. Sporadic cases are sometimes combined with other types of reconstruction, making it difficult to determine the outcome of the allograft-prosthetic composite specifically8,10,12,18,19. Recognizing the potential complications associated with resection of the proximal part of the tibia, we adopted a number of modifications to our technique in recent years. These include the routine use of muscle flaps for soft-tissue coverage and rigid fixation of the implant by cementation of a long-stemmed tibial component. The goal of the present study was to evaluate the functional results and complications of a series of proximal tibial allograft-prosthetic composite reconstructions.
Patients
A consecutive series of twelve patients had allograft-prosthetic composite reconstruction of the proximal part of the tibia following wide resection of a primary tumor. The index procedure was performed between 1999 and 2005. The mean duration of follow-up was forty-nine months (median, forty-nine months; range, thirty to 110 months). The medical records and radiographic data were reviewed for all patients. Approval for the study was obtained from the institutional review board.
Ten male patients and two female patients were included in the study (see Appendix). The mean age was 34.5 years (range, twelve to seventy-three years), and the mean weight was 85.9 kg (range, 55.7 to 135.1 kg). The diagnosis was osteosarcoma in six patients, malignant fibrous histiocytoma of bone in two, giant-cell tumor in two, chondrosarcoma in one, and a small-cell sarcoma in one patient. This last patient had an unusual small-cell variant of synovial sarcoma of the soft tissue with intramedullary bone involvement. The cells carried the SYT-SSX chromosomal translocation of synovial sarcoma, but had the morphological features of Ewing sarcoma under light microscopy.
Of the ten patients with primary sarcomas, one patient had Musculoskeletal Tumor Society stage-IIA disease, eight patients had stage-IIB disease, and one patient had stage-III disease20. The mean tumor size (and standard deviation), in maximal dimension, was 118 ± 56 mm (range, 50 to 245 mm). The mean resection length was 141 ± 58 mm (range, 75 to 270 mm). Nine patients received preoperative chemotherapy, and eight received postoperative chemotherapy. The patient with a chondrosarcoma and the two patients with a giant-cell tumor did not receive chemotherapy.
Surgical Technique
Surgical resection was performed with the intent of obtaining wide margins. A medial skin incision was used in two patients, an anterolateral incision was used in four patients, and a straight midline approach was used in six patients. Open surgical biopsy tracts were excised with the tumor, but needle biopsy tracts were not routinely excised. The mean length of the resection of the tibia was 141.3 mm (range, 75 to 270 mm). The anterior tibial vessels were ligated in five patients.
The reconstruction was performed with fresh-frozen nonirradiated allografts that were closely matched to the size of the bone, with the use of radiographs. Care was taken not to employ allografts whose condylar width was greater than the native tibia. Two different endoprostheses were used in this study: Stryker MRS (Modular Replacement System; Stryker Orthopaedics, Mahwah, New Jersey) in seven patients and Biomet OSS (Orthopaedic Salvage System; Biomet Orthopedics, Warsaw, Indiana) in five patients. The allograft-prosthetic composite was prepared on the back table by reaming the canal and trimming the bone and soft tissue to the proper size. A long-stemmed tibial component was cemented into the allograft tibia with tobramycin-impregnated methylmethacrylate (1.2 g of tobramycin per two bags of powder mix, which contains a total of 60 g of methylmethacrylate). Cementation of the composite allograft-tibial component into the host bone was then performed, making sure that all cement was removed from the allograft-host bone junction prior to final seating of the allograft. The native remnant patellar tendon was sutured with multiple interrupted nonabsorbable sutures to the allograft tendon in all patients. Number-2 and 5 braided polyester sutures were generally used to secure the remnant of the native patellar tendon over the allograft tendon in a pants-over-vest fashion. The native tendon overlapped the allograft tendon, creating a larger area for the allograft tendon to heal to the native tendon. Care was taken to restore the level of the patella with reference to the joint line. In nine patients, the patellar tendon was sewn to the medial gastrocnemius muscle flap, again with use of multiple interrupted braided polyester sutures. The medial gastrocnemius covered part or all of the allograft tendon.
All but one patient had supplemental bone graft placed at the host-allograft junction. Eleven patients received a combination of two or three of the following grafts: autogenous iliac crest bone graft, corticocancellous allograft, and demineralized bone matrix. One patient received autogenous iliac crest bone graft only. No secondary bone-grafting procedures had been performed at the time of the latest follow-up.
Consultation with a plastic surgeon was obtained preoperatively for all patients, and the type of flap used was determined by the size and location of the soft-tissue deficit. A medial gastrocnemius flap was used primarily unless it could not provide adequate coverage of the allograft, and nine patients had a medial gastrocnemius flap alone. One patient had medial and lateral gastrocnemius and soleus flaps. One patient had a medial gastrocnemius flap supplemented with a free rectus abdominis flap. One patient had a lateral gastrocnemius flap. Split-thickness skin grafts were applied over the muscle flaps on all patients.
Postoperative Management
Patients were restricted to touch-down weight-bearing until there was radiographic evidence of healing at the allograft-host junction. All patients were managed with the knee in a brace locked in extension for six weeks, and they were encouraged to do isotonic quadriceps-setting exercises with the knee in extension. Range-of-motion exercise was started at six weeks, with use of a hinged brace to protect the knee.
Radiographic Analysis
Postoperative radiographs were evaluated for the presence of bridging callus on anteroposterior and lateral radiographs of the tibia. The time to the formation of bridging callus was evaluated on each radiograph. In order for callus to be considered bridging, it had to appear to be bonded directly to both the allograft and the native bone. A sleeve of callus over the allograft with a thin radiolucent separation between callus and allograft was not scored as bridging callus. In addition, the most recent radiographs were evaluated for the presence or absence of a lucent line in the junctional cortex, and the number of cortices united of a possible four (medial, lateral, anterior, and posterior). Complete union was defined as four of four cortices united on orthogonal radiographs (Figs. 1-A, 1-B, and 1-C). Partial union was defined as the presence of bridging callus on at least one cortex (Figs. 2-A and 2-B).
Functional Outcome
The Musculoskeletal Tumor Society score21 was used to evaluate the functional outcome. It measures pain, range of motion of the lower extremities, and walking ability. It is based on physical examination by a physician or another medical practitioner. For the lower extremity, the assessment consists of six categories: pain, function (restriction of recreational or work activities), emotional acceptance of the procedure, gait (severity of limp), supports (use of a cane, crutch, or wheelchair), and walking (distance). Each factor is rated by the examiner on a scale from 0 (worst) to 5 (best), with a maximum total score of 30.
The scores were collected on a prospective basis as the patients were evaluated during their clinic visits. In addition, the active and passive range of motion of the knee was recorded. Extensor lag, which is the difference between passive and active extension, was measured at each postoperative visit.
Statistical Analysis
Time-dependent end points, including overall survival, local recurrence-free survival, and radiographic bone union, were assessed by Kaplan-Meier survivorship analysis. Comparison of means was performed with the Student t test. The level of significance was defined as p < 0.05.
Source of Funding
There was no external funding source.
Functional Outcome
The mean amount of knee flexion was 103° (range, 60° to 120°). Only one patient had <90° of flexion (Table I). Nine patients had no extensor lag. Three patients did not have full active extension. One of these three deserves special mention. After initially achieving a full return of active knee extension, the patient sustained a traumatic rupture of the patellar tendon reconstruction in a farm accident. Electing to forego repair, the patient exhibited a persistent 15° extensor lag at the time of the last follow-up.
Four patients used a cane for walking—two used it on a daily basis, and two used it occasionally. The remaining eight patients walked without the use of an assistive device. All patients were able to bear full weight on the reconstructed extremity. Ten of the patients required no pain medications, while two patients used non-narcotic analgesics. The mean Musculoskeletal Tumor Society score for the patients was 24.3 (Table II).
Radiographic Outcome
The median time to the formation of bridging callus of at least one cortex was eight months (range, three to twelve months; 95% confidence interval, six to ten months) (Fig. 3). At the time of the most recent follow-up, nine patients had complete union with healing of four of four possible cortices. Three patients had partial union. Two of these patients had three cortices bridged, and one patient had two cortices bridged. The median time to complete union was nineteen months (range, twelve to forty-eight months; 95% confidence interval, sixteen to twenty-two months) (Fig. 3). At the time of the most recent follow-up, there had been no instance of stem fracture in the patients with partial union.
We could not detect a significant effect of chemotherapy on time to union. The mean time to formation of bridging callus on both orthogonal radiographs was 9.8 months for nine patients who received chemotherapy and 10.6 months for three patients who did not receive chemotherapy (p = 0.13). However, it should be noted that the number of patients may have been too small to detect an effect of chemotherapy on bone-healing.
Oncologic Outcome
Nine patients were alive at the time of the most recent follow-up. Eight patients were continually free of disease, and one was alive without evidence of disease after treatment of recurrent disease. Three patients had died of the disease. There were two local recurrences. The local recurrence-free survival at five years by Kaplan-Meier analysis was 79%. In both patients who had a local recurrence, the surgical margins were negative. Both of these patients had pulmonary metastases develop. One patient had a local recurrence of osteosarcoma thirty-seven months after surgery and underwent an above-the-knee amputation forty-five months after the initial operation. This patient died of pulmonary metastases, which were detected over a year prior to the development of local recurrence (twenty-two months after the primary resection). The tumor necrosis rate in this patient was 50% after preoperative chemotherapy. The other patient had a wide local excision of a recurrent osteosarcoma in the superficial soft tissues of the calf fourteen months after surgery. The case of this patient was notable for the presence of venous invasion by the original tumor. The patient had a pulmonary metastasis concurrently with the local recurrence, but at the most recent follow-up evaluation was free of disease for a year after resection of locally recurrent and metastatic disease. The extent of tumor necrosis after preoperative chemotherapy was 95% in this patient.
Complications
There were two wound complications: one minor and one major (Table III). One patient had prolonged serous drainage from the wound lasting thirty-four days. It was treated with oral antibiotics as well as local wound care. Deep infection had not developed in this patient at the time of the last follow-up at thirty-four months after surgery. One patient had early necrosis of the gastrocnemius flap and was returned to the operating room ten days after the index procedure for revision of the soft-tissue reconstruction with use of a free rectus abdominis flap. The patient was doing well seven years after surgery and had not had a symptomatic deep infection or any further complications develop.
One acute, deep infection that was polymicrobial in nature occurred approximately eight weeks after surgery. The patient was managed with early, aggressive surgical débridement and intravenous antibiotics followed by long-term oral antibiotic suppression. The prosthesis was retained and, at the most recent follow-up evaluation at thirty-six months, the patient had no further problems related to infection. No patient underwent amputation or removal of the original prosthesis for infection during the period of follow-up covered by this retrospective review.
Other complications included two transient peroneal nerve palsies and one traumatic rupture of the patellar tendon. No instance of symptomatic deep-vein thromboses or pulmonary emboli was observed. No fracture of the host bone or allograft occurred. No patient had aseptic loosening of the implant, breakage of the prosthesis, dislocation, or other causes of implant failure. No patient underwent revision of the implant.
Although previous authors have questioned whether allograft-prosthetic composite reconstruction is appropriate for the proximal aspect of the tibia9,17, our results indicate that good functional results and a reasonable rate of complications are achievable with this type of reconstruction. Success seems to be dependent on a number of critical factors: preservation of extensor mechanism function, absence of infection, and osseous union of the host-allograft junction. Without achieving these goals, it is doubtful whether the functional outcome would be satisfactory.
It is worth emphasizing that sound oncologic surgery to prevent local recurrence is critical to the success of the operation. In this series, two patients had a local recurrence of osteosarcoma. The number of patients in the series was too small to determine whether there was a significant increase in local recurrence in comparison with other sites. In theory, the resection of the proximal part of the tibia ought to be the same regardless of whether the subsequent reconstruction involves an endoprosthesis, an allograft-prosthetic composite, or an osteoarticular allograft. The selection of patients appropriate for limb-sparing surgery of the proximal aspect of the tibia is vitally important and is more likely to affect the risk of local recurrence than is the choice of implant. For the two patients who had a local recurrence, there were no clear contraindications to limb-sparing surgery. One patient had a poor response to chemotherapy, and the tumor necrosis rate was only 50%. It is possible that this might have affected the risk of local recurrence22. Unfortunately, current imaging techniques do not permit accurate prediction of the percentage of tumor necrosis after preoperative chemotherapy. Therefore, there are no reliable ways to identify the extent of tumor necrosis except at the time of surgery.
Previous authors have noted increased complication rates with allograft-prosthetic composites involving the proximal part of the tibia9,17. The reason for the high frequency of unfavorable results may be related to a number of factors. The tibia is a more vulnerable location in comparison with the femur. The tibia is notable for diminished muscle coverage as well as slower bone-healing. The calf muscles swell markedly after surgery, and this could affect wound-healing and infection rates. Wunder et al.17 reported on a small series of five proximal tibial allograft-prosthetic composites in their study of knee reconstructions for malignant tumors. They noted two infections, poor functional outcome with a mean Musculoskeletal Tumor Society score of 15.7, and a high rate of reconstructive failure. Their technique utilized irradiated allografts and cementless fixation into host bone. More recently, Biau et al.9 reported on a series of twenty-six patients with proximal tibial allograft-prosthetic composites. They reported a fracture in 50% of the patients, an infection in 23%, and failure of the extensor mechanism in 23%. It is of note that irradiated allografts were also used in their study, and the authors recommended that irradiated allografts not be used in this application.
In our series, soft-tissue restoration was a key component of the surgery, both in terms of wound-healing and reconstitution of tendon attachment. One of the main objectives of our approach was to protect the allograft with healthy, well-vascularized muscle tissue. Wound complications in this area frequently lead to infection of the allograft, which has devastating consequences, including the loss of the implant and even the limb9,17,23. In the present study, all patients had muscle flap coverage as part of the planned procedure, and only one of twelve patients had a deep infection develop. The deep infection rate with osteoarticular allografts of the proximal part of the tibia has been reported to range from 13% to 25%6,24,25. The infection rate for endoprostheses of the proximal part of the tibia has ranged widely, from 0% in one study13 to 36% in the early part of another series26, but most studies have found relatively lower rates of between 3% and 14%14,24,26,27. The infection rate for allograft-prosthetic composites of the proximal part of the tibia was 23% and 40% in two small series9,17. Finally, in one recent report of a composite reconstruction with pasteurized autograft and prosthesis, the infection rate was 23% in thirteen patients28.
The medial gastrocnemius flap provides excellent muscle coverage for the proximal part of the tibia26,29,30. In one large study of 151 patients with endoprosthetic replacement of the proximal aspect of the tibia, adoption of the medial gastrocnemius flap was associated with a substantial reduction in the infection rate from 36% to 12%26. We used this flap in most of our patients. For larger defects that could not be covered with the medial gastrocnemius alone, we used additional flaps, such as the soleus or the free rectus abdominis flap as a supplement31. Instead of attempting a tight closure over the transferred muscle flaps, and placing the adjacent native skin at risk of necrosis, we applied split-thickness skin graft over the flaps32,33, and they healed promptly.
It is not universally accepted that muscle flaps are a necessary component of proximal tibial reconstruction. One series had no deep infections after primary closure over twenty-two proximal tibial endoprostheses, and those authors did not recommend the routine use of muscle flaps13. Nevertheless, the experience of most other studies has been less favorable with regard to the rate of infection, and there is greater concern for the risk of infection with the use of allografts compared with endoprosthetic replacement6,9,11,17.
Weakness of knee extension is a major problem following proximal tibial reconstruction. Previous reports have shown the rate of extensor lag in proximal tibial endoprosthetic reconstruction to range from 16% to 56%8,10,14,17,26,34. It is likely that weakness of extension is related to the adequacy of soft-tissue attachment. One study found that the patellar length gradually increased over two years following attachment of a native tendon to an endoprosthesis by means of a synthetic tendon graft and gastrocnemius muscle flap15. Theoretically, the host tendon should heal more readily to allograft tendon than to a metallic surface. The patellar height was reported to be stable over time following allograft reconstruction in one series of forty-two patients7. However, despite excellent quadriceps strength in a majority of patients in that study, a substantial proportion (27%) had some degree of extensor lag7.
A number of factors could impact the success of extensor mechanism restoration. Incorporation of the vascularized muscle flap may increase the strength of the reconstructed extensor tendon30. For nine of the twelve patients in this series, the medial gastrocnemius muscle was repaired directly to the patellar tendon, and this might have contributed to the strength of the repair. We made every effort to reestablish the level of the joint line and the height of the patella during surgery in order to optimize quadriceps strength. Finally, the program of rehabilitation following surgery is likely to have had a substantial effect. In our experience, a six-week period of immobilization in extension allows sufficient healing of the tendon repair and yet is not so long that patients cannot regain knee flexion. Only one of twelve patients did not achieve >90° of flexion. We believe that a vigorous regimen of physiotherapy for both knee flexion and quadriceps strengthening is necessary to attain the functional goals of rehabilitation.
In this study, we obtained an excellent rate of bone-healing (100%), and the median time to the formation of bridging callus was eight months, which is good in comparison with that reported in other allograft studies5,35-37. It is notable that in one large review of allograft reconstructions, allograft-prosthetic composites had the lowest rate of nonunion compared with the other options38. While it is difficult to determine the exact reasons for our favorable results, we believe that several factors might have had a positive influence on the outcome. In addition to primary bone-grafting at the osteotomy site, the use of a cemented, long-stemmed tibial component may have also promoted bone-healing. The cemented stem produces a rigid construct, which is important for healing of the allograft to the host bone. Rigid fixation at the osteotomy site and good cortical contact have been shown to be factors that positively affect allograft union39. One major drawback to our approach of cementing a long stem into both the allograft and host bone is that a subsequent revision may be more difficult to perform. Another drawback is that the technique for constructing and cementing a long-stemmed prosthesis into the allograft is technically demanding. Special care must be taken to ensure excellent cortical contact of the host and allograft bone during the setting of the cement. All of the cement must be removed from the bone interface, since it could impede bone-healing. There are alternative methods of fixation, but a comparison of different techniques is beyond the scope of this analysis.
It was quite striking that we did not observe any fractures of the allograft. Fractures have been noted to occur in 13% to 46% of proximal tibial osteoarticular allografts6,24,25 and in 28% to 36% of proximal tibial allograft-prosthetic composites9,17. The reason for this difference is not altogether clear. Unlike previous studies on tibial allograft-prosthetic composites, we used only fresh-frozen, nonirradiated allografts. It has been shown previously that the rate of fracture is higher with use of irradiated allografts40. In addition, the absence of screw holes may have helped to prevent fractures by avoiding stress risers. Finally, placement of the stem of the prosthesis in a nearly neutral position allows it to act as a permanent load-sharing device. Once union of the host bone-allograft junction occurs, load sharing is enhanced even more.
In summary, the results of this study support the use of the allograft-prosthetic composite in the proximal part of the tibia for reconstruction of the defects created by wide, en bloc, resection of tumors. Our results are encouraging with regard to function and complications. In contrast to previous reports, we found that most patients achieve full active knee extension and good function. All of the patients in this series received muscle flap coverage of the allograft, rigid cement fixation of the endoprosthesis, and nonirradiated allografts. These factors may have contributed to the relatively good overall results.