Efficacy and safety of a 3D-printed arthrodesis prosthesis for reconstruction after resection of the proximal humerus: preliminary outcomes with a minimum 2-year follow-up

This was a retrospective observational study that was approved by the institutional review board (IRB) of our hospital. Informed consent was obtained from every patient or guardian. A total of 23 surgery-naïve patients with primary malignancies in the proximal humerus underwent intra-articular en bloc resection and prosthetic replacement in our center between January 2017 and January 2019. In ten of them, a 3D-printed arthrodesis prosthesis was used for reconstruction. The baseline characteristics are summarized in Table 1. There were 9 males and 1 female with a mean age of 32.1 ± 16.1 years. The mean duration from symptom onset to admission was 10.8 ± 14.1 months, and the pathological diagnoses included osteosarcoma, chondrosarcoma, undifferentiated pleomorphic sarcoma and malignant myoepithelioma. Eight cases were localized, while the remaining two cases had axillary lymph node metastases on admission. Three cases were complicated by pathological fracture. The average size of the tumor was (65.2 ± 23.4) mm.

The arthrodesis prosthesis (Chunli®, Beijing, China) aimed to reconstruct the humeral defect and create a fixed glenohumeral joint. It was composed of three parts: the glenoid component, the intermediate segment, and the humeral component (Fig. 1). The new implant design was only applied to the glenoid and intermediate component, as the humeral component was off-the-shelf (Modular shoulder prosthesis, Chunli®, Beijing, China). The glenoid component was made of Ti6Al4V via the electron beam melting (EBM) technique (Fig. 1a). It was prepared in a modular manner with three consecutive sizes. The contour of the outer interface was designed to match the shape of the articular surface of the scapular glenoid. It had a proper pore size (500 μm) and porosity rate (60%) that facilitated bone ingrowth [19, 20]. There were three screw holes for fixation of the prosthesis to the scapula. The inner side of the prosthesis was a Morse taper to fix the intermediate segment. Several holes around the rim of the prosthesis were used for suturing. The intermediate segment was a metal plug with a Morse taper that connected the glenoid and humeral components (Fig. 1b). The three parts were assembled by impaction of the Morse tapers (Fig. 1c).

Resection of the proximal humerus followed the same procedures as previously reported [3, 21]. After removal of the tumor, we first dissected the joint capsule to expose the edge of the glenoid, after which we removed the articular cartilage and burnished the subchondral bone until spotty hemorrhage was seen. Then, the glenoid component was implanted (Fig. 2a, b, c). As the contour of the glenoid component matched the shape of the articular surface of the scapular glenoid very well, it was easy to locate the glenoid prosthesis. The screw trajectories were pre-determined and could be well fixed to the scapula. A tip for fixation of the glenoid component was to insert the middle screw first and then the upper and lower screws. Second, we assembled the intermediate segment and the proximal humeral component. The medullary canal of the residual humerus was prepared, and the proximal humeral component was cemented. During the period of cement polymerization, we reduced the shoulder joint, impacted the prosthesis to make the Morse tapers tight, adjusted the pronation angle of the forearm to 30° and maintained it until the cement hardened (Fig. 2d). The holes at the rims of both the glenoid component and proximal humeral component were sutured together with nonabsorbable sutures to prevent dislocation. After wrapping the prosthesis with a LARS ligament, the residual capsule and rotator cuff, as well as the insertions of the deltoid, pectoralis major and latissimus dorsi, were reattached to the LARS ligament to provide additional strength for suspension.

Drainage tubes were removed when the volume was ≤20 mL/24 h. A third-generation cephalosporin was given during the hospital stay, which was usually 5 to 7 days, followed by oral antibiotics for 1 week. The patients wore an abduction splint for 4 weeks while exercises of the elbows and hands were allowed. Four weeks postoperatively, a sling was used, and active-assisted exercises of the shoulder were initiated. Adjuvant chemotherapy was administered 3 weeks after the operation.

Patients had regularly scheduled follow-up visits, usually every 3 months for the first 2 years, every 6 months from the third to fifth year, and yearly thereafter. Oncological status (recurrence or metastasis) and prosthetic status were assessed during each follow-up by senior orthopedic surgeons. Specifically, radiological evaluation for implant osteointegration was performed by routine CT follow-up. Implant failure was defined as removal of any part of the implant for any reason. At the latest follow-up, the patients’ survival status, ranges of motion (ROM) of the shoulder, Musculoskeletal Tumor Society (MSTS)-93 upper extremity score, and American shoulder and elbow surgeons (ASES) score were recorded (Fig. 2e) [22, 23].

All 10 patients underwent en bloc resection of the tumor with a wide margin. Four patients preserved the axillary nerve (Table 1). The mean proportion of resection was 40.3%. The mean duration of the operation was 151.5 ± 61.0 min, and the volume of intraoperative bleeding was 410.0 ± 353.4 ml. The mean postoperative length of hospitalization was 5.3 ± 1.9 d.

All of the patients were followed, and the mean duration was 29.3 ± 6.4 months (Table 2). Except for one patient who died at 16 months postoperatively, this cohort attained a minimum follow-up of 24 months. Two patients experienced local recurrence at 15 and 26 months postoperatively, with the former treated by forequater amputation and the latter by local resection. The former patient developed multiple skeletal metastases 10 months after amputation and received targeted therapy. Three other patients developed pulmonary metastases at 2, 19 and 22 months after definitive surgeries and were treated with chemotherapy. Two metastatic patients died at 16 and 31 months postoperatively, and the remaining two patients were alive with disease by the last follow-up.

Two cases (20.0%) experienced detachment at the taper (Fig. 3). One patient was disease-free and refused further operations for reduction of the prosthesis, while the other patient also had tumor recurrence and was treated by forequater amputation. There were no cases of aseptic loosening, breakage, fracture, or infection in this cohort (Table 3). The rate of implant survival was 90%.

In 4 of the 8 cases without complications, new bone formation around the porous interface of the glenoid component was observed via CT scan during the follow-up (Fig. 2f). Histological study of the glenoid component of the arthrodesis prosthesis was performed in patients who underwent forequarter amputation. This confirmed tight implant osseointegration at the bone-prosthesis interface (Fig. 4a, b). Photomicrographs showed new bone growing into the porous structure of the 3D-printed metallic trabeculae (Fig. 4c).

Outcomes of functional status were obtained from 8 patients at more than 24 months postoperatively. The mean MSTS-93 score and ASES score were 24.9 ± 3.1 and 79.4 ± 8.3, respectively. The mean forward flexion and abduction angles were 71.3 ± 19.4° and 61.3 ± 16.4°, respectively. Loss of the axillary nerve did not significantly decrease the MSTS-93 score (25.0 vs. 24.8, p = 0.919) or the ASES score (77.1 vs. 81.7, p = 0.469) or the ranges of flexion (66.3° vs. 76.3°, p = 0.509) and abduction (55.0° vs. 67.5°, p = 0.317) (Table 3).

Prosthetic reconstruction is convenient for intraoperative manipulation, but the function of the shoulder greatly depends on the status of the deltoid muscle [24]. In contrast, arthrodesis of the joint could achieve a permanent stable shoulder with a fixed glenohumeral position, which allowed the linkage motion of the scapula-humerus composite despite the loss of axillary nerve [25]. However, the procedures of arthrodesis are complex, and mechanical failures are common afterwards (Table 4) [1, 11, 26].

The starting point of this arthrodesis prosthesis was to create a fused shoulder via a convenient operation. The immediate arthrodesis status could be achieved by assembly of the three components, while permanent arthrodesis could be fulfilled by osseointegration at the interface of the glenoid component. Radiological signs of osseointegration at the bone-implant interface were observed in half of the available cases (Fig. 2f). Histological study of the glenoid component from a recurrent chondrosarcoma case also showed new bone growing into the 3D-printed metallic structure (Fig. 4), which supported the purpose of long-term arthrodesis via osseointegration at the porous interface. The postoperative functional status of this arthrodesis prosthesis was satisfactory, with a mean MSTS-93 score of 24.9 ± 3.1 and a mean ASES score of 79.4 ± 8.3. Patients achieved an average of 71.3° forward flexion and 61.3° abduction, which justified the purpose of linkage motion by glenohumeral arthrodesis. Moreover, the functional outcomes remained stable regardless of whether the axillary nerve was preserved. Based on the results of this study, we assumed that this new arthrodesis prosthesis was rational and practical for reconstruction of the proximal humerus, especially for those without preserving the axillary nerve.

However, arthrodesis prostheses did cause new forms of complications compared with PHR, i.e., detachment of the taper (Fig. 3). Detachment of the taper might be related to the weakness in resistance to torsion at the taper. A series of measures have been taken to reduce the risk of detachment. During implantation, we repeatedly examined the stability after engagement of the taper, after which we fixed the LARS ligament around the glenoid component by suturing through the holes of the rim. Meticulous suturing of the soft-tissue attachments to the LARS ligament was performed to provide additional stability. There might be a concern that the addition of LARS could increase the risk of infection. However, neither this study nor our previous report about proximal humerus replacement showed a higher infection rate when compared with historical literature [1‐3]. Finally, the patients followed strict immobilization of the shoulder in an abduction brace for 4 weeks to allow scar formation around the shoulder. After taking these measures, no more cases of dislocation were observed during the follow-up in this cohort. Recently, we also improved the design of the engaging mechanism by using a long screw connecting the proximal humeral component, the glenoid component and the scapula (Fig. 5).

PHR with hemiarthroplasty is the most common method for segmental defects of the proximal humerus with a wide range of postoperative MSTS-93 scores [1, 2]. Teunis et al. performed a systematic review of the reconstructive techniques after proximal humerus resection, which showed that the MSTS scores of prosthetic reconstruction ranged from 61 to 77% and the implant survival ranged from 38 to 100% [2]. Another systematic review by Dubina et al. demonstrated the mean MSTS score of megaprosthesis as 72% with a reoperation rate of 10% [1]. We previously reported the outcomes of proximal humeral replacement with/without synthetic mesh augmentation in 2015 [3]. In that study, the average operative time and intraoperative blood loss were 158 min and 339 ml in the mesh group. The mean MSTS-93, ASES, ROM of flexion, and ROM of abduction were 80%, 85, 77° and 68°, respectively. In this study, the results showed that this new prosthesis did not significantly increase the operative time (151.5 min) or intraoperative hemorrhage (410 ml), indicating equivalent perioperative safety of the arthrodesis prosthesis. The mean MSTS-93 (83%) and ASES scores (79.4), as well as the ranges of flexion (71.3°) and abduction (61.3°) of this new prosthesis, were also comparable with those of our previous study and those of other studies [1‐3, 21].

Most of the available prostheses with total shoulder arthroplasty for oncological cases were reverse total shoulder arthroplasty (rTSA) according to the literature [1, 6‐8, 27, 28, 35]. The reported average MSTS-93 score (72.3 to 96.6%) and ROM of the shoulder (flexion: 71° to 96°, abduction: 62° to 88°) were excellent for cases of rTSA but varied across different studies [1, 8, 27, 28]. Griffiths et al. reported that in a cohort of rTSA using a Bayley Walker prosthesis, the mean MSTS score was 72.3%, and the mean ROM was 71° forward flexion and 62° abduction [35]. Grosel et al. reported that the mean range of forward flexion for rTSA patients was 85°, and the mean ASES score was 59 [6]. In a recent study of 22 patients who underwent resection of a proximal humeral tumor and replacement with a modular rTSA while preserving an innervated deltoid muscle, Trovarelli et al. reported a mean MSTS-93 score of 96.6%, a mean ASES score of 81, a mean abduction angle of 103°, and a mean forward flexion angle of 117° [27]. The results from Trovarelli et al. were much better than those of other retrospective studies [6, 7, 35], indicating the essential role of deltoid muscle in functional restoration after rTSA. Moreover, rTSA was reported to have a higher rate of complications than other arthroplasty options [1, 27, 35]. The risks of dislocation and shoulder instability could be as high as 30% in oncological cases [27, 29, 35]. Compared with rTSA, our arthrodesis prosthesis could provide better stability of the glenohumeral joint and similar upper limb function but would rely much less on the preservation of the deltoid and axillary nerve. Although this new prosthesis had a detachment rate of 20%, the overall risk of prosthetic complications was similar to that of rTSA. As a result, the arthrodesis prosthesis was an alternative choice to rTSA for cases without preservation of the axillary nerve.

Glenohumeral arthrodesis with allografts or autografts has been an alternative to prosthetic replacement in patients requiring resection of the rotator cuff, deltoid, or axillary nerve. The mean MSTS-93 scores were satisfactory, ranging from 73 to 88%, and the mean forward flexion and abduction could reach up to 80° according to the literature [1, 11, 13, 17, 33]. However, the procedures of arthrodesis are complex and tedious, and complications are common. The risks of nonunion ranged from 4.7 to 62.5% [11, 13, 17, 26, 34, 36], while the risks of infection ranged from 7 to 21% [1, 12, 34, 36]. Moreover, the risks of graft fracture ranged from 10 to 67%, and the risks of hardware failure could be up to 20% [11, 12, 32‐34, 36, 37]. By applying the new arthrodesis prosthesis, we created a stable joint immediately, and permanent arthrodesis could be achieved by osseointegration at the implant-bone interface (Fig. 4). It achieved an MSTS-93 score and ROM similar to those of arthrodesis with bone grafts while avoiding the complicated procedures and risks of nonunion, infection, fracture and hardware failure, as reported in the literature [1, 11‐13, 17, 26, 32‐34, 36, 37].

There were several major limitations in this study. First, this was a retrospective study that had the intrinsic weakness of selective and recalling bias. Second, as a preliminary observation of this newly designed prosthesis, we did not establish a strict criterion for selection of the prostheses at the beginning. Third, the sample size was small, and the follow-up duration was relatively short. Although the shortest follow-up duration for the surviving patients exceeded 24 months in this study, further recruitment of patients and longer follow-up durations are needed to evaluate the risk of mechanical complications and long-term function. Finally, although there were concerns about mechanical failure because the lever arm was high with respect to upper limb weight, the osseointegration at the interface could be strong enough to endure the stress. Further biomechanical analysis is needed to justify the mechanical strength of the integrated bone-implant interface.