A Comparison of Outcome of Osteoarticular Allograft Reconstruction and Shoulder Arthrodesis Following Resection of Primary Tumours of the Proximal Humerus

Purpose. The purpose of this study was to compare the oncologic, reconstructive and functional outcomes of patients who underwent osteoarticular allograft reconstruction with those who underwent arthrodesis for a primary malignant bone tumour of the proximal humerus. Patients. Eleven patients were treated with osteoarticular allograft reconstruction of the proximal humerus. Five of these reconstructions failed, necessitating revision to a secondary arthrodesis. Five patients underwent arthrodesis as a primary reconstruction, for a total of ten patients in the arthrodesis group. Methods. End points included local and systemic disease recurrence, complications and functional outcome (measured using the 1987 and 1993 Musculoskeletal Tumour Society Rating Scales (MSTS) and the Toronto Extremity Salvage Score (TESS)). Results. One patient died of systemic disease 2 years post-operatively and one patient had an axillary node recurrence resected 10 months post-operatively and remains free of disease 53 months later. The other 14 patients were alive with no evidence of disease at the time of the last follow-up. Complications after the osteochondral allografts (n=11) included two infections, four fractures and three subluxations in eight patients. Six of these patients required removal of the allograft; one had a repeat osteochondral allograft and five were converted to an arthrodesis. Complications after arthrodesis in the ten patients (five primary and five secondary arthrodeses) included two non-unions, one infection and one fracture in three patients. Patients who underwent shoulder arthrodesis scored better in all outcome measures and this was statistically significant in the MSTS 1993 (p=0.001, Mann–Whitney U Test). Discussion. In this study, there was a trend towards improved function following arthrodesis compared to osteochondral allograft reconstruction following proximal humerus bone tumour resection.


Introduction
Reconstruction of the shoulder following resection of a prim ary tum our of the proxim al hum erus is a challenging clinical problem , particularly since the resection m ay result in de® cits of the deltoid, rotator cuff, joint capsule, glenoid and scapula. T he reconstructive options following proxim al hum erus tum our resection include: preservation of a m obile glenohum eral joint using a prosthesis, osteoarticular allograft or an allograft± prosthesis com posite; arthrodesis using an allograft or ® bular transplant; or, a¯ail joint (Tikhoff± Linberg procedure). 1± 4 W hen suf® cient deltoid, rotator cuff and joint capsule can be preserved to pow er the glenohumeral joint, our group has undertaken reconstruction with osteoarticular allograft. W hen insuf® cient soft tissues remain, we have utilized an allograft arthrodesis. W hen resection of the proxim al hum erus, glenoid and scapula leaves insuf® cient bone stock to permit shoulder fusion, the T ikhoff± Linberg procedure is undertaken to avoid forequarter am putation w henever possible. 5,6 The perceived advantage of an osteoarticular allograft reconstruction com pared to shoulder fusion is based on the potential for im proved function due to a m obile glenohum eral joint. How ever, there is often considerable m obility im parted to the upper extremity by the scapulothoracic joint follow ing arthrodesis. T here is no inform ation in the literature com paring the results of a m obile shoulder versus 1357-714 X/98/030163± 08 $9.00 Ó 1998 Carfax Publishing Ltd arthrodesis reconstruction following proxim al hum erus tum our resection. The purpose of this study w as to com pare the oncologic, reconstructive and functional outcom es of patients who underw ent osteoarticular allograft reconstruction with those who underwent arthrodesis for a prim ary bone tum our of the proximal hum erus.

M ethods
Patients were eligible for this study if they had resection of a prim ary bone tum our of the proximal hum erus and reconstruction w ith an osteoarticular allograft or shoulder arthrodesis. All patients had a m inimum of two years' follow-up from their index surgery.
Between 1986 and 1995, eleven patients were identi® ed w ho underw ent osteoarticular allograft reconstruction of the proxim al hum erus. D uring the sam e period, ten patients underwent shoulder arthrodesis. Five of these ten were initially treated w ith an osteoarticular allograft that subsequently failed. T hese ® ve patients underwent arthrodesis as a secondary procedure.
All patients underwent clinical, radiograph ical (com puterized tomography in the initial years and then both computerized tom ography and m agnetic resonance im aging in the latter years), and pathological staging prior to surgery.
All resections were classi® ed according to the system of the M usculoskeletal T um our Society. 7 If the abductor m echanism (deltoid and rotator cuff) w as intact following tumour resection, the case was denoted with an A, or a B if the abductor m echanism w as disrupted. All of the osteoarticular allografts w ere classi® ed as S345A. At least a portion of the deltoid was disrupted in all of these cases but the rotator cuff rem ained intact. All of the prim ary arthrodeses (that is, arthrodeses done im m ediately following tumour resection, rather than to salvage a failed osteochondral allograft reconstruction) were classi® ed as S345B and had resection of their rotator cuff and joint capsule as well as the deltoid.

Reconstructive procedures
T um our resection w as perform ed following accepted oncologic principles aim ing for negative operative margins. T he status of the abductor m echanism w as the m ost im portant factor in selecting the reconstruction m ethod. 4 An osteoarticular allograft was perform ed if the rotator cuff and capsule (and rarely the deltoid) w ere intact, and an arthrodesis was perform ed if the rotator cuff, capsule and/or glenoid surface were resected (Figs 1 and 2). If there was involvem ent of the scapula including the glenoid w ith or without the proximal hum erus, a T ikhoff± Linberg procedure was perform ed providing the neurovascular structures could be preserved. The patients undergoing Tikhoff± Linberg procedures were not included in this study.

Osteoarticular allogra ft
An osteoarticular allograft of appropriate size (size m atch perform ed using standardized radiographs) was chosen from the bone bank prior to surgery. All allografts w ere harvested and stored according to the standards of the American Association of Tissue Bank 8 and all grafts were treated with 2.5 m egarads of radiation following harvest. O steoarticular allografts were thaw ed in the operating room and cut to ® t the hum eral defect. Prior to ® xing the allograft to host hum erus, a rotator cuff repair was perform ed  with the glenoid and acrom ion at thirty degrees exion, thirty degrees of abduction and thirty degrees of internal rotation. 9,10 Soft tissues and cartilage were cleaned from the under-surface of the acromion. A very long (18± 26 hole) 4.5 m m broad plate was contoured to lie along the spine of the scapula, over the acrom ion and dow n along the allograft onto the patient' s rem aining distal hum erus. Interfragm entary com pression screws were used following provisional stabilization with K-w ires and passe d through the plate and allograft, obtaining purchase into the glenoid and neck of scapula. Further interfragm entary screws w ere then passe d through the acrom ion into the graft to achieve ® rm contact with the superior portion of the allograft hum eral head. A dynamic com pression plating technique w as then used to stabilize the distal osteotom y between the allograft and the host hum erus. N eutralization of the reconstruction was accom plished by insertion of screws through the plate into the allograft and the scapular spine. Iliac crest bone graft was packed around the proxim al and distal osteotomies.
Three arthrodeses were perform ed using a vascularized ® bular graft instead of an allograft as each of these had developed an infection; two following an osteoarticular allograft (Fig. 3) and one following an allograft arthrodesis. T he infections were treated in a tw o-stage process w ith rem oval of the allograft and replacement with a cement spacer until the infection cleared and then subsequent arthrodesis with the vascularized ® bular graft. A broad plate was contoured along the spine of the scapula to the distal hum erus. A microvascular ® bula was harvested and a hole w as ream ed in the glenoid where the ® bular graft was slotted into place. T he distal hum erus was also ream ed and the ® bula w as inserted about 2± 3 cm into the hum erus. The ® bula was held in place proximally by a single screw through the ® bula into the glenoid. The plate was secured to the ® bula with screws. The microvascular repair was then perform ed with an end-to-sid e repair between the brachial artery and vein and the peroneal artery and vein. Iliac crest bone graft was app lied at both anastom osis sites.
A latissim us dorsi muscle¯ap w as required in ® ve of the eleven arthrodesis reconstructions to provide com plete soft tissue coverage over the graft site.
All patients in the osteoarticular and arthrodesis groups received peri-operative antibiotics and were im m obilized in a Velpeau sling for 6 weeks. T he osteoarticular allograft patients were started on physiotherapy at 6 w eeks post-su rgery w ith shoulder shrugs and active-assisted forward¯exion exercises. T heir rehabilitation program s were progressed as tolerated and according to radiographic evidence of healing. The arthrodesis patients were started on range of m otion (RO M ) exercises of the scapulothoracic joint when there w as radiograph ic evidence of healing of the glenoid osteotom y. All patients were between the host and the allograft tendons using non-ab sorbable sutures. A stable repair of the rotator cuff was achieved in all cases. F ixation of the allograft to host hum erus was then perform ed using a dynam ic com pression plating technique. Autograft from the iliac crest was placed around the osteotom y site in eight of the eleven cases. T he insertions of the latissim us dorsi and pectoralis major m uscles were repaired to the corresponding areas on the allograft if possible. The biceps, brachialis and triceps were advanced for soft tissue coverage of the allograft.

A rthrodesis
Eight of the eleven arthrodeses perform ed in ten patients w ere achieved using allograft. T he allograft w as prepared by osteotomizing the articular surface so that the rem aining hum eral head was congruent to function (activities of daily living, AD Ls, leisure activities, work, and pain levels.

Oncologic results
T here were seven males and four fem ales w ho underwent an osteochondral allograft, and their average age at surgery was 34 years (range, 10± 78 years). Seven of the patients had osteosarcom a (two IB parosteal and ® ve IIB central osteosarcom as) and four of the patients had chondrosarcom a (all IIB lesions). 14 T hree m ales and seven fem ales underwent an arthrodesis (® ve of the ten after failed osteoarticular allograft reconstruction). The average age at surgery was 22 years (range, 12± 35 years). Of the ® ve patients undergoing prim ary arthrodesis imm ediately after tum our rem oval, three had a IIB osteosarcom a, one a IB m alignant giant cell tum our and one a IB chondrosarcoma. T able 1 sum m arizes the results. The average length of bone resection was similar for the osteoarticular allograft and arthrodesis groups, 15.2 cm (range, 7.3± 25 cm) and 17.3 cm (range, 7.5± 25 cm ) respectively. Adjuvant treatm ent varied depending on the grade and stage of the tum our. For the osteochondral allografts, six of eleven patients received chem otherapy. For the prim ary arthrodesis group, three of ® ve patients received chemotherapy. N one of the patients in the study was treated with radiotherapy. N ine of the patients undergoing osteoarticular allografts had negative m argins and two patients had positive m icroscopic m argins. Four of the ® ve patients undergoing prim ary arthrodesis had negative m argins and one patient had positive m icroscopic m argins.
At the time of the last follow-up, all of the patients who had an osteochondral allograft were alive with no evidence of disease. O ne patient w ith osteosarcom a treated w ith an osteoarticular allograft (negative margin surgery at initial operation) developed a high axillary nodal recurrence ten m onths after initial resection. F ifty-th ree m onths after wide resection of this nodal recurrence, the patient rem ains disease-free. O ne osteosarcoma patient treated by arthrodesis developed lung and pelvic m etastases eight m onths post-operatively and died of disease at two years. This patient had positive resection m argins at the time of surgery but did not develop a local recurrence. T he other patients reconstructed by arthrodesis were alive with no evidence of disease.

Reconstructive results
O f the eleven patients who had an osteoarticular allograft, ® ve patients required revision, all to a glenohum eral arthrodesis. T wo of these patients sustained late infections (greater than six m onths started on early gentle ROM exercises of the elbow, w rist and hand. D emographic, oncological outcom e, treatm ent com plication and functional data at m ost recent follow-up w ere collected for all patients. Function w as recorded using the M usculoskeletal Tum our Society Rating Scale, 1987 version (M ST S, 1987), 11 the M usculoskeletal Tum our Society Rating Scale, 1993 version (MST S, 1993), 12 and the T oronto Extremity Salvage Score (TESS). 13 The functional data of the six patients, w ho had an osteoarticular allograft (without revision to an arthrodesis), were com pared to the 10 patients who had an arthrodesis using non-param etric statistics.
The ® ve patients who had a prim ary osteoarticular allograft and went on to a secondary arthrodesis w ere interview ed and asked speci® c questions comparing the two different types of surgery w ith respect Of the ® ve patients who had a prim ary arthrodesis, four patients have had no com plications. O ne patient had a fracture of the allograft and developed chronic infection which was treated with a cement spacer and eventual m icrovascular ® bular graft arthrodesis. O f the ® ve patients w ho had a secondary arthrodesis, three patients have had no complications. T wo patients sustained non-unions which healed after repeat bone grafting.
Functional data w ere availab le for all but one patient w ith an osteoarticular allograft who did not return for follow -up assessm ent. favou ring im proved function in the arthrodesis group. T he active range of m otion achieved by the two groups was especially striking. For all but one patient w ho experienced chronic, debilitating pain from an unstable osteoarticular reconstruction and continued to experience som e pain following conversion to an arthrodesis, patients with an arthrodesis had better active forward¯exion (range 45± 85) than any patient with an osteoarticular allograft (range 30± 70). T his w as re¯ected in the M ST S 1987 range of motion ratings in which no patient with an osteoarticular allograft received m ore than one of ® ve points w hereas eight of the ten patients in the arthrodesis group (® ve prim ary and ® ve secondary arthrodeses) received three of ® ve points for range of m otion.
The subjective description of the com parison of the osteochondral allograft versus the arthrodesis from the ® ve patients undergoing both surgeries seemed to support the trend favou ring arthrodesis (see Table 2). Increased stability w as noted after the fusion. The pain was also reported as the sam e or better after arthrodesis.

D iscussion
It has been well docum ented that limb-salvag e surgery for tumours of the shoulder is an alternative to am putation. 4 Careful performance of the biopsy, the use of neoadjuvant chem otherapy when indicated and proper attention to com plete surgical resection of the tum our are essential in perm itting lim b salvage without com prom ising tum our control. Successful local resection, avoiding am putation while controlling sarcoma, is demonstrated by the post-surgery) and underwent staged salvage procedures with removal of the osteochondral graft, implantation of an antibiotic cement spacer and subsequent reconstruction using a vascularized ® bula. Two patients with painful subluxation of the shoulder and one patient with an osteoarticular allograft fracture were also revised to an allograft arthrodesis. T he length of time between the osteochondral allograft and the arthrodesis varied from 12 to 81 m onths (average 48.6 m onths).
O f the rem aining six patients with osteochondral allografts, one patient developed an inferior dislocation of the allograft four months post-operatively. T his w as stable and the patient did not want further treatment despite lim ited glenohum eral m otion. O ne patient sustained a fracture of the allograft after a fall and required replacement of the graft while two patients developed a fracture of the allograft w ithout traum a. In one of these two patients, atraum atic fracture resulted in formation of callus around the allograft fracture site with eventual stabilization of the fracture, while the second patient is presently aw aiting revision. Two patients have not had speci® c com plications although one of these patients has moderate to severe shoulder pain but does not w ish to undergo further surgery.  fact that there w as only one nodal recurrence and one system ic relapse in the patients in this study. H ow ever, the choice of the best lim b-reconstruction m ethod for tumours of the proxim al hum erus is a dif® cult clinical decision. In this study, patients w ere selected for osteoarticular allograft reconstruction when suf® cient abductor m usculotendinous tissue rem ained to provide a stable soft tissue repair to the rotator cuff and tendon of the allograft. Fusion was undertaken w hen the osteoarticular allograft failed due to infection, instability or fracture, or if the glenoid bone stock or soft tissues after resection w ere insuf® cient to reconstruct a stable, m obile shoulder. In both of these types of reconstructions, com plications were frequent and serious. In the osteoarticular group, ® ve have been revised to an arthrodesis: two for infection, two for instability and one for fracture. In the rem aining six patients, three further fractures w ere docum ented (one requiring allograft replacem ent and one for which revision surgery is still pending), one patient dislocated the shoulder and one patient has chronic pain of m oderate intensity. In comparison, three of the ten arthrodeses were com plicated by non-union requiring repeat bone grafting (n 5 2) and fracture/infection requiring a staged m icrovascular ® bula arthrodesis.
The rates and types of com plications experienced by both the osteoarticular allograft and arthrodesis patients in this study are sim ilar to those reported by other authors. G ebhardt et al. 2 reported four complications in three of seven patients undergoing shoulder arthrodesis. T hese included a w ound slough, m edian nerve palsy and two patients with prominent hardware. Layton et al. 15 reported on nine patients treated by shoulder arthrodesis. T wo patients died of m etastases prior to healing of the arthrodesis and there were com plications consisting of infection and fracture in four of the rem aining seven patients. 15 G ebhardt et al. 3 also reported on twenty patients w ho underw ent osteoarticular allograft of the proximal hum erus. In this group, one had a non-union, four fractured and three becam e infected. 3  F our of the six items of the M STS 1993 are m ore likely to vary am ongst the two groups. These are pain, function, emotional acceptance and lifting ability, which were generally higher in the arthrodesis patients than the osteochondral allograft patients. There was less of a difference in hand positioning and dexterity. W ith regards to the T ESS scores, m ost of the patients in both groups w ere clustered around the m id to high range. T his m eans that all of the patients were able to do basic activities of daily living (eating, bathing, groom ing) but experienced dif® culty perform ing high-level activities such as sports, w ork, endurance and leisure activities. M ost of these patients also classi® ed themselves as`somewhat disabled' . Although the number of patients in the study w as relatively sm all and ® ve of the patients w ere in both groups, there is a trend towards im proved function after an arthrodesis com pared to an osteochondral allograft. Subjective com pariso n also favoured the arthrodesis. T hese results are sim ilar to those of O ' C onnor in which patients treated with prim ary arthrodesis had higher M STS 1993 scores than those reconstructed with osteoarticular allografts. 4 The arthrodesis provides a stable shoulder girdle w ith m otion of the scapulothoracic joint to position the arm in space. 2 Subjective responses from patients in this study indicated a feeling of increased stability and decreased pain after an arthrodesis com pared to an osteochondral allograft. T hese qualitative responses were obtained from patients w ho had undergone both procedures due to complications from an osteoarticular allograft and were based on patients rem embering their`best' function after their osteoarticular allograft. These results m ust be interpreted cautiously and in recognition of all the biases imposed in the m ethod.
It is recognized that this study did not include patients with a prosthesis or allograft± prosthesis com posite as these reconstructive techniques have not been utilized around the shoulder by our group. Recent work by O' C onnor et al. 4 , however, suggests that the com plications and function in patients reconstructed with a prosthesis or allograft± prosthesis com posite provides inferior results when com pared to an osteochondral allograft or arthrodesis.
In conclusion, com plications seemed to be m ore app arent in patients undergoing osteoarticular allograft. Patients treated with shoulder arthrodesis tended to have better functional scores.