In past decades, tumors of the long bones that required resection of a portion of the involved bone quite often necessitated limb amputation1. More recently, the use of allograft reconstruction allows a greater rate of limb salvage for these patients2. These allograft reconstructions are most frequently in the form of an intercalary allograft or an allograft arthrodesis2,3. The rate of limb salvage with use of allograft reconstruction has been good and has reached 80%4,5. However, allograft reconstruction has not been a panacea and can be associated with substantial complications. The most common complications include nonunion at the host-allograft junction or junctions; fracture of the allograft, most typically at the host-allograft junction; and infection. These problems, similar to the oncologic problems themselves, have proven to be quite difficult and challenging to ameliorate. In general, the rate of failure with allograft reconstruction due to the aforementioned complications has been reported to range between 15% and 25%2-4,6-10. Revision reconstruction of allograft failures is challenging and has had substantial complications5,8,11. Recently, the use of the free vascularized fibular graft as a salvage method for complications of allograft reconstruction of the long bones has been reported12.
The purpose of this study was to evaluate outcomes from the use of the onlay free vascularized fibular graft as a treatment for complications of long-bone allograft reconstruction following tumor extirpation. These complications include host-allograft nonunion, allograft fracture, and allograft infection. It was our hypothesis that this treatment modality enables limb salvage in most patients with these complications.
After institutional review board approval was obtained, a retrospective review of all oncologic long-bone allograft salvage procedures with a free vascularized fibular graft performed between January 1987 and July 2006 was conducted. Our institution began using this particular operation to treat allograft reconstruction complications in 1987. To facilitate this review, our institution's tumor registry was queried for the search terms "allograft" and "vascularized fibula." Following this query, all records isolated with these terms were searched for instances of allograft complications (host-graft nonunion, allograft fracture, or allograft infection) that were treated with vascularized fibular graft salvage. Once these patients were identified, the medical records were reviewed for the location of the complication, history of chemotherapy, specifics of the allograft reconstruction salvage, nature of the complication(s), number of subsequent surgeries, time to union, and outcome in terms of the function of the affected extremity after reconstruction. The characteristics of the patients in this series are listed in a table in the Appendix. Time to union, recorded in months, was determined by the presence of fibular graft incorporation in the proximal and distal aspects of the host-allograft junction, as well as union of the host-allograft junction that was spanned by the fibular graft. This was determined by the reconstructing surgeon on the basis of the presence or absence of bridging cortical and trabecular bone on both computed tomography scans and multiple plain radiographs10. In general, patients were seen ten to fourteen days following hospital discharge, every two to three months until the time of osseous union, and finally every year thereafter for surveillance. Follow-up intervals varied slightly according to surgeon and the distance from the patient's home to our facility; therefore, the number of months to osseous incorporation is an approximation. Postoperatively, an attempt was made in all patients to assess the viability of the free vascularized fibula with a technetium-99 methylene diphosphonate bone scintigraph. This diagnostic modality has been shown in both animal and human studies to be a sound method of graft assessment13,14.
Complications were divided into major and minor categories. A major complication was defined as one that required additional operation(s) or that caused substantial morbidity such as a deep venous thrombosis. A minor complication was one that did not fall into the aforementioned criteria.
The analysis of functionality after salvage was determined with use of the grading scale designed by Mankin et al., which is a general assessment of the ability of the patient to perform work and sports activities following oncologic resection and reconstruction15. This scale is commonly used in orthopaedic oncologic studies, but it has not been internally or externally validated. This analysis was completed by an author who was not involved in the care of these patients and was based on the postoperative functional narrative by the reconstructing surgeon in the medical record. Patients were not contacted specifically for this study. A result was considered to be excellent when there was no apparent disease and the patient had returned to full function without pain or disability and to some participation in sporting activities. It was considered to be good when the patient had no further disease, had only slight limitation with daily function, had no pain or major disability, and was able to perform a limited amount of sports activities. A fair result was achieved when there was no further disease, but there were major limitations requiring some form of assistance, including use of a crutch, walker, or brace, and the patient was unable to return to the preoperative level of work or activity. Finally, the reconstruction was considered a failure when the patient died as a result of tumor recurrence, the affected limb was amputated, or the allograft needed to be removed because of a complication.
Surgical Technique
Patients were offered onlay vascularized fibular grafting for osseous discontinuity because of host-allograft nonunion, allograft fracture, or removal of the allograft because of infection. For thirty of the thirty-three patients in this series, the fibular onlay technique described by Chang and Weber12 and Duffy et al.16 was used (Fig. 1). The remaining three patients had allograft exchange and placement of the fibular graft into the medullary canal of the allograft and host bones. Generally, if the allograft was removed, either at the time of fibular salvage or prior to it because of infection, a new allograft and the fibular graft were inset simultaneously.
Statistical Methods
Baseline demographics, functional outcomes, and complications were reported as the mean and range (minimum to maximum) for continuous variables and as count (percent) for categorical variables. Observed time to healing was estimated by the Kaplan-Meier method, and a Kaplan-Meier plot including 95% confidence intervals was constructed. The probability of healing (95% confidence intervals) at various time-points was calculated. These descriptions of demographic data, functional outcomes, and time to healing estimates were provided both overall and within each of the three subgroups according to the anatomic location (femur, humerus, and tibia). The overall association of time to osseous union with dichotomized functional outcome (fair or failed compared with good or excellent) was estimated with use of univariate logistic regression. Because of the small sample size, especially within the reconstruction subgroups, significance was neither expected nor of interest, and p values were not calculated. Instead, the focus was on estimates and confidence intervals.
Study Population
Over the nineteen-year period of interest, a total of 235 free vascularized fibular reconstructions were performed. From the eighteen-year period of 1988 to 2006 (the current database structure precluded inclusion of 1987 in this aspect of the search), a total of 141 long-bone allograft reconstructions were performed. All patients who had complications secondary to allograft reconstruction were evaluated by the reconstructive surgery group; however, on the basis of the data from the current study, we were unable to determine which allograft complications healed with interventions other than a vascularized fibular graft. The surgical database query yielded thirty-six patients whose surgical descriptions matched the search terms utilized. Three patients did not have an adequate length of follow-up (one year) or had critical information missing from the record and were excluded from the study, leaving thirty-three patients for analysis. The long bones involved were the femur in eighteen patients (Figs. 2-A and 2-B), the tibia in eight, and the humerus in seven. The mean age of the patients was twenty-eight years (range, seven to sixty-three years). Twenty-one patients underwent allograft reconstruction at this institution, while twelve underwent reconstruction elsewhere. If the one patient who underwent fibular salvage in 1987 is removed to correspond to the time frame of the allograft reconstruction search, then twenty (14%) of the 141 allograft reconstructions performed at this institution eventually required fibular graft salvage for allograft complications.
The original malignant tumors affecting the designated long bones in this study were osteosarcoma (sixteen patients), chondrosarcoma (six), adamantinoma (four), Ewing sarcoma (two), giant-cell tumor (two), and hemangioendothelioma, lymphoma, and metastasis from lung carcinoma (one patient each). Seventeen patients received chemotherapy either prior to or following tumor resection. Two patients underwent radiation therapy; however, only one of them received radiation prior to the fibular salvage of the allograft reconstruction. The other patient received radiation for a recurrence of tumor in the soft tissue following successful allograft salvage. The locations of the host-allograft failures included nonunion at the proximal host-allograft junction in nineteen patients, proximal fracture nonunion in one patient, distal nonunion in eight patients, allograft shaft fracture in three patients, and allograft infection in six patients. Two patients had a nonunion at both the proximal and distal host-allograft junctions. One patient (tibial reconstruction) had a concurrent proximal nonunion and infection. One patient had a fracture of the allograft shaft complicated by subsequent infection.
Twelve patients underwent procedures following allograft reconstruction and prior to treatment with the fibular graft. Six had allograft removal to treat infection. Four underwent nonvascularized bone autografting and/or fixation revision to treat a nonunion. Two patients sustained an allograft fracture, one at the host-allograft junction and one in the allograft shaft. The junction fracture was plated, and the fractured allograft shaft was replaced with new allograft. The patients who underwent fixation or bone-grafting had persistent nonunion, thereby requiring later vascularized fibular grafting. A total of seven patients (six who had an allograft infection and one who had an allograft fracture) underwent allograft removal and replacement at the time of fibular grafting. Seventeen of thirty-three patients were administered adjuvant chemotherapy after the allograft reconstruction but prior to fibular graft salvage.
The average interval between allograft reconstruction and allograft salvage was thirty-six months (range, five to 276 months). The average interval between allograft reconstruction and fibular placement in patients who underwent chemotherapy was thirty-seven months. For the three anatomic subgroups, the average interval between allograft reconstruction and allograft salvage was forty-five months in the femur group, forty-one months in the humerus group, and nine months in the tibia group. The overall average duration of follow-up was seventy-three months (range, twelve to 170 months). Within this group of thirty-three patients, all but three received fibular reconstructions with use of a free tissue transfer method. Lesions in three of the patients who had had tibial allograft salvage were suitable for reconstruction with pedicled vascularized fibular grafts. One patient in the humerus group had received, at the time of tumor excision and reconstruction, an intercalary allograft along with a simultaneous free fibular onlay graft. This reconstruction failed, and the patient required another operation with use of the other free fibula to salvage this reconstruction.
On the basis of the postoperative physical examination and bone scanning, there were no acute failures of the fibula. Twenty patients underwent a technetium-99 methylene diphosphonate bone scan within one week of the operation. Seventeen scans indicated graft viability, while three were equivocal. All three patients with equivocal bone scans went on to achieve osseous union as judged by plain radiographs and computed tomography scans.
All thirty-three patients in this series went on to demonstrate healing radiographically. Following fibular graft placement, two patients had a persistent nonunion, which healed after nonvascularized iliac crest bone graft was placed into the nonunion site. While the fibular grafts did not achieve their intended outcome for these two patients, limb salvage was eventually achieved. Five patients had reconstruction failure that was directly attributed to allograft complications following the fibular graft placement. All of these failures were due to infection of the allograft. Two patients who had a femoral reconstruction experienced an allograft infection, and, when given the option for another attempt at salvage, they chose amputation. Another patient in the femoral reconstruction group underwent allograft removal to treat infection but went on to have successful reconstruction with an intercalary free fibular graft in the defect. Despite the fact that the reconstruction was ultimately successful, we classified that allograft reconstruction as a failure. Finally, two patients in the tibial reconstruction group had an infection of the allograft and went on to have an amputation. Somewhat surprisingly, all five patients who had an allograft infection after salvage were judged to have had osseous host-allograft union at the time of allograft removal, which occurred at a mean of nineteen months (range, six to forty-three months) following the fibular graft salvage procedure.
Six patients had infection of the allograft as the chief complication prior to fibular grafting, and all of them underwent resection of the allograft. Four had placement of an antibiotic-impregnated spacer at the time of allograft removal. Following resolution of the infection, the patients had a new allograft and new fixation hardware placed at the time of fibular grafting. Of these six patients, only one had a postoperative recurrent infection. This infection was drained, and the allograft was removed. When the infection resolved, repeat fibular bone-graft surgery was done, but this time with the fibula being placed in the intercalary defect, without the further use of allograft. As stated previously, five patients had failure of the reconstruction after the fibular graft because of infection of the allograft. Four of these infections occurred in patients who had had no infection prior to the fibular reconstruction.
The mean time to healing as judged on plain radiographs and computed tomograms was 7.7 months. The Kaplan-Meier curve, in which the probability of healing is plotted as a function of time to osseous union, is displayed in Figure 3. Not surprisingly, the probability of osseous union increased such that it was 0.61 (95% confidence interval, 0.40 to 0.74) at six months after surgery and 0.91 (95% confidence interval, 0.73 to 0.97) at nine months after surgery.
Seventeen patients received adjuvant chemotherapy following allograft reconstruction. The mean time to osseous healing following fibular graft salvage in this subgroup was 8.2 months. Two of these patients had healing times well outside the average, at thirty-two and twenty-one months. These patients both had persistent nonunions at the proximal and distal host-allograft junctions even following the vascularized fibular grafting, and both required nonvascularized iliac crest bone-grafting at later times to achieve healing.
Overall, there were a total of six minor and twelve major complications associated with the allograft salvage operations. The minor complications included a seroma at the reconstruction site, a flexor hallucis longus contracture due to harvest of the free fibular flap, transient global brachial plexus palsy in two patients who had received humeral reconstructions, prominent reconstruction hardware, and one transient peroneal nerve palsy at the fibular donor site. Two of the minor complications (the flexor hallucis longus contracture and the peroneal nerve palsy) were attributable directly to the fibular harvest. Major complications consisted of allograft infection necessitating removal of the allograft (five patients); a persistent nonunion at both the proximal and distal junctions of the allograft that required nonvascularized bone-grafting to achieve eventual union (two patients); and deep venous thrombosis in the reconstructed leg, infection around hardware requiring replacement, superficial femoral artery thrombosis after femoral reconstruction, a soft-tissue tumor recurrence requiring amputation, and degenerative arthritis of a knee after tibial reconstruction salvage requiring above-the-knee amputation (one patient each). Five of the major complications (allograft infections) were attributable to the allograft reconstruction, while none of the major complications were due to fibular graft harvest. All but two complications (the soft-tissue tumor recurrence and the degenerative arthritis of the knee) can be attributed to the overall fibular graft salvage procedure.
Regarding the functional results, eight patients had an excellent result; fifteen patients, a good result; three patients, a fair result; and seven patients, a failure. Five of the failures were directly related to the reconstruction itself, as stated before. The other two failures were not directly related to the fibular salvage: one patient who underwent humeral reconstruction had a soft-tissue tumor recurrence requiring forequarter amputation of the affected extremity, and another patient had severe degenerative arthritis of the knee develop following tibial reconstruction. Given the options, the latter patient opted for above-the-knee amputation rather than further attempts at limb salvage. As stated above, six patients underwent fibular reconstruction because of allograft infection and removal. Four of them had good function, one had fair function, and one had a failure.
Of the seven failures, five resulted in limb loss. Two other patients had failure of the reconstruction but retained the limbs with use of a second free fibular graft as an intercalary graft, without additional allograft.
Overall, fourteen patients required a total of twenty-two operations following fibular salvage of the allograft reconstruction. These ranged from correction of the flexor hallucis longus contracture to amputation as well as fixation revision and nonunion treatment.
After reconstruction, allograft problems such as fracture, infection, and nonunion continue to be challenging in these patients. In one series of nearly 1000 allograft reconstructions, Hornicek et al. reported a nonunion rate of 17% at the allograft-host junction8. Despite multiple reoperations, only 60% of the patients in that subgroup went on to have a successful outcome. Similar findings have been reported in other studies2,9.
At least two studies have demonstrated increased complication rates when adjuvant chemotherapy was used following allograft reconstruction8,17. Hazan et al., in a large series of osteoarticular allografts, noted that chemotherapy did not increase the rate of infection or graft fracture, but the rate of nonunion was substantially increased (32% compared with 12% when no chemotherapy was administered)17. Hornicek et al. noted a similar increase in the nonunion rate in the setting of adjuvant chemotherapy8. In our series, seventeen of thirty-three patients received adjuvant chemotherapy. The time to osseous union in the patients who had chemotherapy (eight months) was slightly longer than that for the group as a whole (seven months), and the two patients in this series with markedly longer healing times (twenty-one and thirty-two months) had received chemotherapy. However, given the inaccuracies associated with the estimation of osseous healing, this one-month difference in time to union does not allow conclusions to be made regarding the effects of chemotherapy.
The treatment of these allograft complications is a vexing problem. Small series of patients with nonunions that had been treated with cortical allografts, struts, and cancellous bone-grafting18 have been reported in the orthopaedic trauma literature; however, the fracture-healing physiology of such patients may be different from that in patients receiving oncologic treatment19. To our knowledge, the largest series of nonvascularized bone-grafting procedures for complications of allografting is a subset of the 945 patients described by Hornicek et al.8. The composition of that series is somewhat different from ours in that more allograft arthrodeses were performed and more allograft prostheses were used. In the current series, only one patient had an allograft arthrodesis and no patient received an allograft-prosthesis composite. The nonunion rate in their study was highest in the allograft arthrodesis group, and the patients who were treated with adjuvant chemotherapy had a higher rate of nonunion than those who were not. At least in terms of nonunion, the difficulty of treatment with use of nonvascularized bone-grafting is readily apparent in the study by Hornicek et al.: 163 of the 945 patients had a nonunion, and a total of 269 reoperations were needed to treat the nonunions. In the end, eventual union was achieved in 108 (66%) of the 163 patients who had a nonunion. In the current series of vascularized fibular grafts, osseous union was achieved in all patients. Obviously, however, bone-healing does not ensure a good outcome, and this is apparent when one notes that our series includes seven patients who had failure of allograft reconstruction and five patients with limb loss despite the fibular salvage procedure.
The practice of vascularized bone-grafting in the oncologic setting appears to be increasing in both the primary and secondary settings16,20-22. Other researchers have used vascularized bone grafts in immediate reconstruction ostensibly to avoid allograft complications, and the practice at our institution has also followed this trend12,14,20,21,23-26. Recently, a small series of patients who underwent long-bone segmental reconstruction with an immediate vascularized fibular graft that was placed in the medullary cavity of the allograft as well as the native bone was described20. This reconstruction appears to yield good results14,23. However, despite the success of primary reconstruction with vascularized bone grafts, this method of reconstruction may not be feasible at all centers.
The use of vascularized bone-grafting as a method of long-bone salvage has been reported over the last several years. Duffy et al. described a series of seventeen patients with pathologic fractures for which an onlay vascularized fibular graft was used to span the fracture site16. While most of the patients in that series obtained limb salvage and osseous healing, the complications were similar to those in our study. Four of those patients had persistent nonunion, and four had postoperative infections. Chang and Weber were the first, as far as we know, to report the use of an onlay vascularized fibular graft to salvage a host-allograft nonunion, with six patients undergoing this type of reconstruction12. All but one of these patients achieved osseous union following fibular grafting, although the methods used for the adjudication of healing were not described. In the single patient who had a persistent nonunion, healing was achieved after a second onlay vascularized fibular graft.
As with any autologous tissue transplantation, the morbidity of the donor site is a concern. The complications of flexor hallucis longus contracture and peroneal nerve palsy are well known16, not to mention the potential for flap loss that is possible with any free tissue transfer. The donor site-related complications in the present series appear similar to those reported by other investigators27.
We recognize the limitations of retrospective chart reviews without a control group and standardization of surgical techniques. However, our method of fibular harvest and inset for these allograft-host nonunions has been consistent as it has been performed by only four surgeons over the last nineteen years. Retrospective chart review has inherent inaccuracies, especially when outcomes are based on clinician narratives. This aspect of our study is a definite limitation. Some patients in this study had a relatively short follow-up period, and this is especially true in the femoral reconstruction subgroup in which three patients were followed for only twelve months. These three patients are like many referred to our center in that they lived quite a long distance away. Frequently, these patients undergo reconstruction and are then followed up by orthopaedic surgeons in their local communities. This obviously makes it difficult to get accurate long-term follow-up on these patients.
The judgment of osseous healing is an inexact science and entails some degree of subjectivity. Even with the use of computed tomography, it can be difficult to state definitively that a host-allograft junction has healed. Other groups have used radiographs to help judge whether osseous healing has occurred, but in the end, it seems to be an approximation12,16,21. Therefore, the average time to osseous healing that was obtained in this study (seven months) can only be roughly compared with the same outcome from other studies. Another limitation of the study is the ability to accurately analyze postoperative functionality on the basis of the Mankin scale. This analysis was completed at the time of the chart review and was based on the narrative in the medical record. Because this grading is based, to a degree, on extrapolation, inaccuracies are sure to exist. Another limitation of the study is the disproportionate number of femoral reconstructions compared with humeral and tibial reconstructions. Optimally, we would prefer larger numbers of tibial and humeral reconstructions in order to better compare the three subgroups.
In conclusion, the use of an onlay vascularized fibular graft as a method of salvage of an allograft complication can result in limb salvage in most patients. The relatively high complication rate following this procedure can be discouraging but should be balanced against the potential for limb loss.