There is no standard procedure for reconstruction of the defect after resection of the acetabulum and pelvis for treatment of a primary bone sarcoma. When the tumor is limited to the periacetabular area and does not extend to the hip joint, it may be possible to attach the femoral head to the ilium, ischium, or pubis in an attempt to achieve either a fusion or a pseudarthrosis2,11. When the acetabular tumor can be resected through the ilium but resection of the femoral head is necessary, thus preventing arthrodesis, reconstruction can be achieved with a saddle prosthesis1. When the extent of the sarcoma necessitates resection of most of the hemipelvis, reconstruction of the skeletal defect must include an osteoarticular allograft, a prosthesis, or an allograft and an implant or the patient can be managed without reconstruction2,6-12. The alternative to limb-salvage resection of a periacetabular tumor is hindquarter amputation5,14,15.
There are few data available describing the outcome of procedures for reconstruction of the pelvis, and it is difficult to advise patients about the long-term results of such operations. However, despite the lack of outcome studies, we believe that hindquarter amputation is cosmetically and functionally unacceptable to most patients if a safe alternative is available.
At our unit, we have adopted the following approach for adults who have a periacetabular tumor. When a limb-salvage resection is indicated on the basis of preoperative oncological staging studies, it is preferred to amputation. If it is possible to save the femoral head and to avoid shortening the limb more than five to six centimeters, an iliofemoral arthrodesis is performed. If it is necessary to resect the femoral head or more than five centimeters of the ilium, or both, allograft reconstruction is performed, generally in combination with a total hip arthroplasty. We describe the outcome of allograft reconstruction after resection of a stage-IIB sarcoma of the periacetabular region in seventeen consecutive patients.
Materials and Methods
All patients who were managed in our unit, from 1982 to 1994, with resection of an acetabular sarcoma and reconstruction of the defect with allograft bone were included in the study. In order to estimate the proportion of patients with a pelvic bone sarcoma who had had an allograft reconstruction, we searched the complete prospective operative database to identify all patients who had been managed operatively for a pelvic bone sarcoma by surgeons in our unit. As the prospective database was initiated in 1989, we assumed that the rate was constant during the time of the study (1982 to 1994) and the five-year period (1989 to 1994) in the database.
All seventeen patients were evaluated with radiographic and histological staging studies before the acetabular resection and allograft reconstruction. The first patient (who was managed in 1982) had local computerized tomographic imaging of the site of the tumor, and the remaining patients had magnetic resonance imaging (Fig. 1). Systemic staging studies included computerized tomography of the chest and a total-body bone scan for all patients. For the ten patients who had not had a biopsy before they were referred to us, an open biopsy was done, generally with use of an approach over the iliac crest and dissection along the inner table of the ilium.
The anatomical extent of the tumor was described according to the three-zone model proposed by Enneking and Dunham2, in which zone I refers to the supra-acetabular portion of the ilium; zone II, the periacetabular region; and zone III, the ischium and the inferior and superior pubic rami. The stage of the tumor was described with use of the system of Enneking et al.3. As the pathologists at our institution routinely grade chondrosarcomas with use of a three-grade system, histological grades II and III were considered as stage II according to the system of Enneking et al. for the purpose of the present study. High-grade tumors that had not metastasized were classified as stage IIA when they did not extend into the surrounding soft tissues and as stage IIB when soft-tissue extension was evident on imaging studies3.
After the initial staging and diagnosis, the patients who had osteosarcoma or Ewing sarcoma were managed with chemotherapy preoperatively and postoperatively. The patients who had chondrosarcoma were managed operatively without adjuvant therapy. No patient had radiation treatment. After a patient had been managed with neoadjuvant chemotherapy, the tumor was restaged with use of computerized tomography of the chest and local magnetic resonance imaging of the pelvis before the operation.
The tumor was resected through either an ilioinguinal approach or an iliofemoral approach. The iliofemoral approach was used when the proximal part of the femur was to be removed en bloc with the pelvis, and the ilioinguinal approach was used when only the pelvis was to be resected.
At the time of the operation, the margins were assessed by evaluating the gross appearance of the resected specimen and, when indicated, by examining frozen sections. The final margins of resection were considered to be positive when tumor cells extended to the inked margins of resection. Positive margins were identified in only one patient at the time of the operation (as will be described). After the margins had been evaluated, the osseous defect was reconstructed with use of allograft bone consisting of a hemipelvis. Except for one osteoarticular graft in which the viability of the articular cartilage had been enhanced with dimethyl sulfoxide, all allografts had been irradiated with 25,000 gray before they were inserted. Specimens for culture were obtained from the allograft in the operating room, and it then was thawed in warm antiseptic solution.
When it was necessary to perform the resection through the sacral ala or the sacro-iliac joint, interfragmentary 6.5-millimeter cancellous-bone screws were inserted through the allograft into the bodies of the first and second sacral vertebrae (Fig. 2). When it was necessary to perform the resection through the ilium, interfragmentary cancellous-bone screws were inserted in the weight-bearing axis across the osteotomy line; this fixation was supplemented with neutralization plates, which are used to resist deforming forces across an osteotomy site, in the patients who were managed early in the study (Fig. 3). Only the interfragmentary screws were used in the patients who were managed later in the study. The interfragmentary compression screws were inserted through an acetabular roof ring when the size of the allograft acetabulum permitted the use of this device (Fig. 4-A). Autogenous bone obtained from the residual ilium was used, when available, for bone-grafting at the site of the osteotomy.
Before fixation of an allograft-implant composite, the articular cartilage was removed from the allograft with use of curets and careful reaming, with an attempt to leave the subchondral bone of the acetabulum intact. After fixation of the graft to the host bone, the acetabulum was resurfaced with a polyethylene cup that was cemented in place. In two patients who were managed early in the study, reconstruction of the pelvis with an osteochondral allograft was done without resection of the proximal part of the femur. In eleven of the remaining fifteen patients, the femur was reconstructed with use of a standard hip prosthesis, with or without cement (Fig. 2). Chemotherapy was not considered a contraindication to the use of a femoral component without cement, and an implant was inserted with cement only when the bone stock in the proximal part of the femur or the anatomical configuration was judged to be inappropriate for use of a prosthesis without cement. In the other four patients, substantial resection of the proximal part of the femur was necessary because the tumor extended from the hip joint into or along the proximal part of the femur; reconstruction was accomplished with use of a proximal femoral allograft and a long-stem femoral component that bridged the site of the femoral osteotomy. When a femoral allograft was used, the component was cemented into the graft before implantation and this composite then was press-fit into the host femoral canal (Fig. 4-B).
Capsular reconstruction after reduction of the joint was impossible in six patients because of the extent of the soft-tissue resection. In these patients, capsular stability was augmented with non-absorbable sutures placed through the allograft acetabular labrum before the acetabular cup was cemented. After reduction of the femoral component, the sutures were used to attach a synthetic vascular graft (Marlex; Davol, Cranston, Rhode Island, or Gore-Tex; W. L. Gore and Associates, Flagstaff, Arizona) to the acetabulum. The synthetic material then was sutured around the femoral neck and to adjacent femoral soft tissues to form a pseudocapsule that constrained the tendency of the joint to dislocate.
No patient had primary soft-tissue reconstruction with use of vascularized tissue transfer. Three patients had problems with wound-healing, and they were returned to the operating room early in the postoperative period for débridement and reclosure of the wound. All patients were confined to bed for at least four weeks postoperatively. During the last six years of the study, an air mattress was used on the bed in order to facilitate nursing care and to prevent breakdown of skin. A cephalosporin was administered prophylactically at the time of the operation and for three days postoperatively if the patient was not allergic to this drug. Suction drains were maintained for at least seventy-two hours, and the antibiotic was given for as long as the drains were in situ. Neither prophylactic anticoagulation therapy nor mechanical prophylaxis for thromboembolism was used. In the last five years of the study, venous Doppler ultrasound was performed to monitor for thromboembolic disease at least once a week while the patient was confined to bed. When a clot was identified with ultrasound, anticoagulation therapy was initiated and was continued for three months. All patients who had been managed with chemotherapy preoperatively were so managed postoperatively. No patient had radiation therapy postoperatively.
Information regarding the estimated blood loss and the duration of the operation was obtained from the patient's record. All patients were followed prospectively, and any postoperative complications were recorded. After discharge, the patients were seen at least every three months for three years and every six months thereafter. A functional evaluation, with use of the most recent functional rating system of the Musculoskeletal Tumor Society4, was performed for all of the surviving patients who had an intact allograft reconstruction. The patients were classified as unable to walk, able to walk indoors, or able to walk in the community. The patient's occupation and whether the patient was collecting a disability pension were recorded.
From 1982 to 1994, seventeen consecutive patients (approximately 1.5 patients per year) were managed with an allograft reconstruction for a periacetabular sarcoma. From 1989 (the first year that complete prospective records on all patients managed with an operation became available) to 1994, an additional twenty-five patients (approximately five patients each year) had an operation without an allograft reconstruction to treat a bone sarcoma of the pelvis. Six of the twenty-five patients (all of whom had tumor extending into zone II) were managed with hindquarter amputation. Five patients had resection of the periacetabular region (zone II) and reconstruction with use of an autogenous graft and an implant, iliofemoral arthrodesis, or a saddle prosthesis. In the remaining fourteen patients, the pelvic bone sarcoma was treated with resection that did not necessitate reconstruction of the acetabulum. Thus, reconstruction with use of an irradiated bone allograft to replace the acetabulum was performed in an average of one or two of the approximately seven patients with a pelvic bone sarcoma who were managed each year in our unit.
Nine of the seventeen patients who were managed with an allograft reconstruction were male, and eight were female. The average age at the time of the diagnosis of the bone sarcoma was forty years (range, sixteen to sixty-four years). The initial diagnosis was chondrosarcoma for nine patients, osteosarcoma for six, Ewing sarcoma for one, and leiomyosarcoma for one. All of the sarcomas were classified as stage IIB on the basis of histological grading and radiographic staging studies.
The mean operative blood loss was 4700 milliliters (range, 1500 to 11,000 milliliters), and the mean operative time was 9.1 hours (range, 7.0 to 11.5 hours). Surveillance with Doppler ultrasound demonstrated deep-vein thrombosis in the thigh of one patient (Case 7), and anticoagulation therapy was carried out to achieve an international normalized ratio of 2.2 for three months.
All seventeen patients had a high-grade (stage-IIB) bone sarcoma that extended into the periacetabular soft tissues (Table I). No patient had evidence of metastatic disease on systemic staging studies. The tumor extended into zones I, II, and III (the ilium, acetabulum, ischium, and pubis) in thirteen patients and into zones I and II (the ilium and acetabulum) in four. The proximal margin of resection was through the proximal part of the ilium in nine patients, through the sacro-iliac joint in four, and through the sacral ala in four. Resection of the proximal part of the femur was also done in four patients (Table I).
The margin of resection was positive in two patients (Cases 3 and 17). In one (Case 3), the positive margin was not identified until the final histological analysis of the specimen and no additional therapy was administered. The patient eventually had an amputation for local recurrence. In the other patient (Case 17), a positive margin was recognized at the anterior aspect of the sacro-iliac joint after the gross evaluation of the specimen in the operating room; the margin was revised with additional resection of the sacral ala. The patient had a hindquarter amputation at one month because of infection (Table I).
Four patients died, five, eight, eighteen, and twenty-four months after the operation. Three of these patients (Cases 11, 12, and 14) had osteosarcoma of the pelvis, and one (Case 10) had Ewing sarcoma. The cause of death was profound pancytopenia related to chemotherapy in one patient (Case 11) and pulmonary metastases in the other three (Cases 10, 12, and 14). The reconstruction was intact in all four patients at the time of death, and three had been able to walk with a cane before they died. The functional outcome had been evaluated for two of the three patients who died of metastatic disease (Table I).
Three patients (Cases 2, 3, and 7) had a hindquarter amputation because of local recurrence, forty-eight, fifty-four, and fifty-nine months postoperatively. One of these patients (Case 3) had a positive margin of resection. Another (Case 2) had had an inappropriate biopsy (before referral to our unit); that biopsy substantially complicated the planning of the definitive resection. Neither this patient nor the third patient (Case 7) was found to have a positive margin of resection. All three patients had a high-grade chondrosarcoma, and each had been walking with one cane before the local recurrence (Table II). Each patient remained free of local disease for at least two years after the amputation. One (Case 7) had slowly progressive pulmonary metastases despite remaining free of local disease. All three patients could walk in the community with use of two crutches or a walker at the time of the most recent follow-up.
In summary, four patients died within two years after the operation—three as a result of metastatic disease and one as a result of complications of chemotherapy—and three patients had an amputation because of local recurrence.
Outcome of the Reconstruction and the Functional Result
The graft was removed because of infection in two patients (Cases 9 and 17), one of whom had chondrosarcoma and one, leiomyosarcoma of bone. In both patients, problems with wound-healing had been identified in the early postoperative period. The resection had extended to the sacrum in both patients, and femoral resection with an allograft reconstruction had been done in one patient (Case 9). The patient chose to have the pelvic and femoral allografts removed; this resulted in a massive internal skeletal defect, and the patient could not bear weight on the involved side at the time of the most recent follow-up. The other patient (Case 17) chose an amputation. Neither patient had long-term complications as a result of the infection and removal of the graft, and both could walk with a walker in the house but otherwise used a wheelchair, at the time of the most recent follow-up.
None of the four patients who had an amputation (because of local recurrence or infection) used a prosthesis. Three of them had found that a prosthesis hindered their ability to walk. In general, the use of a wheelchair seemed dependent on the age and fitness of the patient. The three patients who had an amputation because of local recurrence had been thirty-nine, forty-nine, and fifty years old at the time of the diagnosis, and all could walk in the community using crutches after the amputation. The two patients who had an amputation or removal of the allograft because of infection had been fifty-six and sixty-four years old at the time of the diagnosis, and both of them used a wheelchair.
The eight surviving patients with an intact reconstruction (Cases 1, 4, 5, 6, 8, 13, 15, and 16) were able to walk at the time of the latest follow-up (more than four years postoperatively). The osteoarticular allograft failed in one patient (Case 8) as a result of avascular necrosis of the femoral head and non-union of the graft; the extremity was salvaged with use of a saddle prosthesis seated on the sacrum and was laterally constrained by a new allograft (Fig. 5). At the time of the latest follow-up (approximately four years after the revision procedure), the patient was able to walk with a cane. None of the remaining seven patients needed to have the graft removed or revised. These patients were followed for a mean of seven years (range, four to fourteen years) postoperatively. Of the eight patients, five had had chondrosarcoma and three, osteosarcoma. The proximal margin of resection was at the sacro-iliac joint or the sacral ala in three patients and through the ilium in five patients. Two patients (Cases 13 and 15) had also had an allograft reconstruction of the femur (Table I).
In addition to the major complications of metastases, local recurrence, delayed wound-healing, infection, osteonecrosis of the femoral head, non-union of the graft, and fatal pancytopenia, a variety of other complications occurred (Table I). The most common complication was dislocation: five patients (Cases 1, 7, 10, 13, and 15) had early dislocation (occurring during the initial hospital stay), and the dislocation recurred in two of them (Cases 1 and 15). Trochanteric advancement eliminated the problem in one patient (Case 1), but chronic dislocation developed in the other patient (Case 15), who had a poor functional result.
The functional rating system of the Musculoskeletal Tumor Society4 was used to evaluate the outcome for fourteen patients (Table I); of the remaining three patients, two (Cases 10 and 11) had died and one (Case 17) had had an amputation (because of infection) before the time of the evaluation. The score for two patients (Cases 12 and 14) was determined after rehabilitation and before the time of death; the score for two patients (Cases 2 and 3), before the amputation because of local recurrence or infection; and the score for one patient (Case 9), after removal of the allograft because of infection. For the remaining nine patients, the score was determined at the latest follow-up examination. The thirteen patients who had an intact allograft reconstruction when the functional evaluation was performed were last seen within six months before the time of writing.
The mean functional score (and standard deviation) was 65 ± 21.16 per cent (median, 70 per cent; range, 13 to 87 per cent) for all of the thirteen patients who were evaluated, and the mean score was 65 per cent for the eight patients who survived with an intact allograft reconstruction for the entire duration of the study (Table II). The patient who was followed for the longest amount of time (Case 1) had one of the highest scores, and there was nothing to suggest that the scores deteriorated with time. Of the thirteen patients, nine had a score of more than 3 (of 5) points for pain, nine had a score of more than 3 (of 5) points for emotional acceptance of the procedure, and seven had a score of more than 3 (of 5) points for walking ability (Table II). The persistent deficit in motor power at the hip was reflected in the poor outcomes for function (only three of the thirteen patients had a score of more than 3 [of 5] points) and abnormalities or alterations in gait (only one of the thirteen patients had a score of more than 3 [of 5] points). The scores for the use of walking aids were generally poor (only two of the thirteen patients had a score of more than 3 [of 5] points), but the low scores were due in part to the fact that the patients had been strongly advised always to use a cane.
The major factor that appeared to be associated with the functional outcome was the status of the abductor muscles after resection and reconstruction. When the insertion of the abductor muscles to the trochanter had been maintained (or reattached) along with the nerve supply to the abductor muscles, the mean score was 73 per cent. In contrast, when the abductor muscles were resected or their insertion was disrupted, the mean score was 46 per cent.
No patient was evaluated with use of the functional rating system4 after a hindquarter amputation. The scores would have been uniformly low, as a score of 0 would have been given for three of the six items that were related directly to walking (use of walking aids, walking ability, and abnormalities in gait). One patient (Case 15) had a score that probably was worse than it would have been if a hindquarter amputation had been performed. The chondrosarcoma had been discovered after a revision arthroplasty of the hip. The tumor had been disseminated extensively through the soft tissues by the revision operation, and there was little muscle remaining after the tumor was resected. In addition, the patient had a perforation of the bowel seven months after the index operation. The patient had recurrent and subsequent chronic dislocation and a very poor result (a score of 13 per cent). Despite the fact that the outcome probably would have improved after an amputation, the patient refused this treatment.
Subjective, observable functional ability deteriorated substantially after amputation because of local recurrence. One patient (Case 2) continued to work in a family business, but the other two patients (Cases 3 and 7) did not return to work after the amputation. Of the eight surviving patients with an intact allograft reconstruction, only one (Case 15) collected a disability pension. In contrast, of the five patients from whom the allograft was removed because of infection or local recurrence, all but one (Case 2) collected a disability pension.
On the basis of our experience with patients who were managed operatively for a bone sarcoma of the pelvis from 1982 to 1994, we believe that a limb-salvage operation and reconstruction with a pelvic allograft is indicated for a minority of patients. Most of the time, the tumor can be resected without removing the acetabulum or, if a portion of the acetabulum is removed, there is sufficient periacetabular bone stock to allow for an iliofemoral arthrodesis, use of an autogenous graft and hip replacement, or use of a saddle prosthesis. Alternatively, the tumor may be so extensive that only amputation provides an adequate margin of resection. A review of our prospective database from 1989 to 1994 revealed that about one of every seven patients who had a pelvic sarcoma was managed with a limb-salvage resection and allograft reconstruction.
Despite the careful selection of patients for allograft reconstruction in the present series, there was a high prevalence of serious complications. Three patients died of metastatic disease, and one patient died of chemotherapy-related complications. Amputation was done because of local recurrence in three patients, and the allograft was removed because of infection in two patients. Thus, nine of the seventeen patients either died or had an amputation or removal of the graft within five years after the index operation, and only eight patients were available for an intermediate-term functional evaluation. Despite these limitations, to our knowledge this is the largest reported series of patients who were managed with an allograft after resection of a stage-IIB pelvic bone sarcoma.
The limited experience with use of osteochondral allografts (two of the seventeen patients) in the present series suggests that pelvic reconstruction with an allograft and an implant is more likely to be successful than reconstruction with an osteochondral allograft. The two patients who were managed with an osteochondral allograft had an unsatisfactory outcome. One (Case 8) had revision to an allograft-implant composite, and the other (Case 2) had an amputation because of local recurrence. Although there are not enough data to directly compare these two methods of reconstruction of the acetabulum with an allograft, we now invariably use an allograft-implant composite. Notwithstanding other issues, it is extremely difficult to obtain a pelvic allograft that is the perfect size for a given femoral head.
To summarize the results with the allograft-implant composites, approximately one-half of the patients either died (with the reconstruction intact) or needed an amputation because of local recurrence or infection. One of the eight surviving patients with an intact allograft reconstruction had a poor result and chronic dislocation of the hip; the result probably would have been better if the patient had had an amputation. The results for the remaining seven patients (average functional score4, 72 per cent) were much better than they would have been after an amputation. The comparison of the work ability of patients who did and did not have removal of the allograft is telling: only one of the eight patients who had an intact allograft reconstruction collected a disability pension compared with four of the five from whom the allograft had been removed because of infection or recurrence.
In the present study, there were many serious complications, including systemic and local recurrence, fatal pancytopenia, infection, dislocation, nerve palsy, deep-vein thrombosis, and pneumonia. There were relatively few mechanical complications related to the reconstructive procedure. One patient (Case 5) sustained a stable fracture of the allograft five years postoperatively; the fracture healed rapidly (Fig. 3). There was only one non-union, in a patient (Case 8) in whom the initial reconstruction with an osteoarticular allograft had prevented the insertion of screws through the acetabulum. It would be useful to review the technique of allograft reconstruction that was used in the present series before evaluating the mechanical results.
Our group has consistently used allografts that have been irradiated (with 25,000 gray) because we believe that irradiation may reduce the risk of transmission of viral disease from donor to recipient. Because this form of processing may result in mechanical weakening of the graft, we use reconstructive techniques that involve fixation with interfragmentary compression screws, when possible, and that necessitate a limited number of screw-holes in the graft. When the resection is through the ilium, these principles involve placement of interfragmentary compression screws through the dome of the acetabulum in the direction of the weight-bearing axis. We no longer place neutralization plates across the site of the iliac osteotomy as this technique may weaken the allograft through stress concentration at the sites of the plate-fixation screws. The only fracture of the graft in the present series seemed to have occurred at the site of the insertion of a screw for the neutralization plate (Fig. 3). Union of the allograft at the site of the iliac osteotomy seems predictable when the interfragmentary screws are placed through the acetabulum in the direction of weight-bearing load.
Although the rate of union after allograft reconstruction is reasonably high, and fractures of the graft in the pelvis are not common and may heal spontaneously, the high risk of other complications might suggest that this technique should not be used. In the present study, the most severe and frequent complications were metastasis, local recurrence, and infection. Three patients died of metastatic disease, and a fourth had slowly progressive pulmonary metastases. This rate of metastatic disease (four of seventeen patients) at the time of the intermediate-term follow-up evaluation is troubling but probably reflects the biology of the disease more than the operative approach. The rate of metastasis in our study is similar to that reported by O'Connor and Sim (twenty-one of sixty patients) and reflects the fact that high-grade pelvic sarcomas are almost always large lesions that are diagnosed at a later stage of progression than are smaller appendicular lesions. Fahey et al. found that all but one of twenty-four patients who had been managed operatively for a high-grade osteosarcoma subsequently died of disease; the authors suggested that a high prevalence of vascular invasion by the tumor may have been responsible for the high rate of metastasis. In the current series, the patients who had rapid metastasis had a poor response to neoadjuvant chemotherapy; this suggests that poor response to drug treatment rather than inadequate operative management was responsible for the metastatic progression.
The frequency of local recurrence is related to the adequacy of the limb-salvage resection. Three of the seventeen patients had local recurrence, and the margin of resection had been recognized as histologically positive at the time of the operation in only one of them. This high rate of local recurrence is an indictment not of allograft reconstruction of the pelvis but rather of a limb-salvage operation for a high-grade bone sarcoma of the pelvis, and this issue deserves discussion. The rate of recurrence in the present study was virtually identical to the rate of 17 per cent (ten of sixty patients) reported by O'Connor and Sim in association with pelvic sarcomas arising at various locations in the pelvis. The rate in our study was lower than the rate of 25 per cent (nineteen of seventy-seven) reported by Sheth et al. in association with chondrosarcomas at all pelvic sites.
In our experience with preoperative planning for a pelvic bone sarcoma and the subsequent histological examination of the specimen after a limb-salvage operation or a hindquarter amputation, we have found that there are two anatomical areas where it is most likely that the surgeon will leave a positive margin of resection. The first area of concern is the ilium or sacrum, as the site of the proximal margin of resection. Although magnetic resonance imaging provides an excellent approximation of the proximal extent of the disease, it is difficult to translate this information into the complex three-dimensional operative anatomy of the pelvis. When a surgeon plans the transosseous resection of a long bone for a limb-salvage procedure, the level of resection of the longitudinal bone usually is related to a readily palpable anatomical landmark such as the acromion or the knee joint. When a resection is to be done in the pelvis, it is necessary to plan with use of multiple osseous landmarks, and this is particularly difficult if the surgeon intends to preserve a portion of the ilium for reconstruction with use of a device similar to the saddle prosthesis. The location of the proximal margin of resection probably was the cause of local recurrence in two of our patients. The margin of resection in the ilium of one patient (Case 3) was noted to contain chondrosarcoma cells on the final histological review. In the other patient (Case 2), the margin was very close to tumor (although there was no histological evidence of involvement by the disease). This experience has caused us to be more generous with the proximal margin of resection and to perform more resections at the level of the sacro-iliac joint or the sacrum rather than through the ilium. At this level of resection, it is difficult to use a saddle prosthesis for reconstruction, and the potential for fixing the allograft to the sacrum with interfragmentary screws offers an advantage for patients who need a proximal osteotomy (Fig. 2).
The sciatic notch, just anterior-lateral to the anterior-inferior aspect of the sacro-iliac joint, is the other anatomical location where the risk for a positive margin and local recurrence is increased. In this region, the internal iliac artery divides into multiple branches, some of which pass laterally through the sciatic notch to emerge as the superior and inferior gluteal vessels. These vessels passing through the sciatic notch tether the internal iliac artery to the region of the sciatic notch and make dissection difficult. If there is a large medial soft-tissue mass in the sciatic notch, it may be very difficult to ligate and divide the internal iliac vessels safely without leaving a positive margin. Indeed, from an oncological perspective, it may be safer to ligate the common iliac vessels more proximally. When we find that the common iliac vessels need to be ligated, we recommend that the patient have an amputation even if a limb-salvage resection is possible with reconstruction of the vasculature between the common iliac and external iliac vessels.
Despite the difficulty of obtaining a negative margin of resection with a limb-salvage procedure on the pelvis, hindquarter amputation does not guarantee eradication of local disease. Fahey et al. reported local recurrence of pelvic osteosarcoma in seven of eight patients who had survived after hindquarter amputation. Sheth et al. found virtually identical rates of local recurrence of pelvic chondrosarcoma in patients who had been managed with a limb-salvage procedure (nine of thirty-two; 28 per cent) or with hindquarter amputation (ten of thirty-five; 29 per cent). Shin et al. found that limb-salvage resection and hemipelvectomy were associated with an equal risk of contaminated margins. It is important to describe such results to patients who must decide between amputation and a limb-salvage operation for a pelvic sarcoma, in order to avoid creating the impression that hindquarter amputation will definitely prevent local recurrence. We usually tell such patients that hindquarter amputation may be somewhat safer than a limb-salvage operation from the perspective of local control of disease but that this is debatable.
The saddle prosthesis often could not be used in the present series because most patients had a proximal iliac or a sacral resection; therefore, the options for reconstruction were a metallic pelvic prosthesis, an allograft-implant composite, or an osteoarticular allograft. These three procedures are associated with similar risks of local recurrence. The risks of infection are probably also similar, although, to our knowledge, no comparative data are available to answer this question.
The literature offers little guidance to a surgeon attempting to evaluate the functional results of various methods of pelvic reconstruction. Rosenberg and Mankin reported high rates of infection and recurrence after pelvic reconstruction with an allograft. They reported that five of nine reconstructions with an osteoarticular graft failed (according to a different outcome analysis than the one that we used), as did all of four reconstructions with an intercalary graft or an allograft-implant composite. Ozaki et al. reported that the allograft was removed, usually because of infection, after seven of nine reconstructions with an allograft-implant composite.
Huth et al. stated that most of their patients who were managed with internal hemipelvectomy without reconstruction had a fair result according to an earlier version of the functional rating system of the Musculoskeletal Tumor Society. Aboulafia et al. reported on eight patients who were managed with a saddle prosthesis for a primary sarcoma; one of the patients had a subsequent amputation, and two used crutches when walking. The functional result, determined with a different rating system than the one that we used, was excellent for five patients and good for two. These results may be superior to the functional outcomes after the allograft reconstructions in the present study, and this difference probably is related to the fact that reconstruction with the saddle prosthesis requires a more distal iliac osteotomy that permits better preservation of the origin and innervation of the abductor muscles.
Gradinger et al. reconstructed the pelvic defect using a metallic implant with a bone-ingrowth surface fixed to the posterior aspect of the pelvis in nine patients; the result was good in five patients, fair in three, and poor in one. O'Connor and Sim reported on fifteen patients who had been managed without bone replacement for an acetabular tumor. Two patients who had had a partial acetabular resection had a good result. Of four patients who had had an iliofemoral arthrodesis, two had a good result, one had a fair result, and one had a poor result. One of five patients who had an iliofemoral pseudarthrosis had a good result, three had a fair result, and one had a poor result. The remaining four patients had a poor result after no reconstruction. Thirteen of our seventeen patients had tumor extending into all three zones2 (the ilium, ischium, acetabulum, and pubis), and there was little if any pelvic bone remaining to support either an arthrodesis or a pseudarthrosis after resection. The only real alternative to reconstruction with an allograft or a prosthesis is to leave the limb flail. The result for one patient (Case 9) in our study, as well as the results reported by O'Connor and Sim, demonstrate that the patient usually has very poor function when the acetabular defect is not reconstructed.
On the basis of a limited number of reports, it seems reasonable to perform arthrodesis for reconstruction after resection of a periacetabular sarcoma if the femoral head and a reasonable limb length can be maintained. When resection through the supra-acetabular portion of the ilium is needed, reconstruction can be done with a saddle prosthesis or with an allograft-implant composite. However, resection of the tumor through the proximal part of the ilium or through the sacrum necessitates the use of either an allograft or a pelvic prosthesis.
Fortunately, few patients who are seen in a musculoskeletal oncology unit are candidates for reconstruction of the acetabulum and pelvis with an allograft. Such patients must be informed of the high risk of complications before they consent to a limb-salvage operation. Of the seventeen patients in our study, four died and five had a subsequent amputation because of infection or local recurrence. We believe that the intermediate-term functional result (at a minimum of four years) for the patients who had an intact reconstruction with an allograft-implant composite was better than it would have been if those patients had had a hindquarter amputation.
*No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. No funds were received in support of this study.
Investigation performed at the University Musculoskeletal Oncology Unit, Mount Sinai Hospital and the University of Toronto, Toronto