Do Early Results of Proximal Humeral Allograft-Prosthetic Composite Reconstructions Persist at 5-year Followup?

Abstract

Background

Insufficiency of the rotator cuff is a major problem after resections of proximal humeral tumors and can limit shoulder motion despite preservation of the deltoid muscle and axillary nerve. Allograft-prosthetic composite reconstruction offers one method to reattach the rotator cuff tendons and has been successful in small studies with short followup. However, data are lacking with regard to implant durability, changes in Musculoskeletal Tumor Society (MSTS) scores over time, and delayed complications with extended followup.

Questions/purposes

(1) What is the cumulative incidence of allograft-prosthetic composite revision surgery 5 years after the procedure? (2) What are the early- and intermediate-term MSTS scores of allograft-prosthetic composite reconstruction of the shoulder? (3) What are the complications of allograft-prosthetic composite reconstruction?

Methods

Twenty-one patients underwent allograft-prosthetic composite reconstruction after tumor resection of the proximal humerus between 2000 and 2015. Six patients who were lost to followup were not included. All patients had malignant or aggressive benign tumors that could be treated with a wide intraarticular approach preserving the deltoid muscle, axillary nerve, and glenoid. Cumulative incidence of implant revision was calculated with death of the patient as a competing risk. Minimum followup was 24 months (with the exception of one patient who died at 22 months), and median followup was 97 months (range, 20-198 months). The upper extremity MSTS score was used to assess function. Various complications were identified from radiographs and charts.

Results

The cumulative risk of implant revision was 10.1% at 5 years (95% confidence interval [CI], 1.6%-28.0%). Mean MSTS scores were 86% (± SD 9%) at 1 year and 78% (± SD 13%) at 5 years (mean difference ± SD 9% ± 14%, p = 0.015). Mean active forward elevation was 101° (± SD 33°) at 1 year and 92° (± SD 34°) at 5 years (mean difference ± SD 8° ± 36°, p = 0.41). Notable adverse events included progressive radiographic superior subluxation > 1 cm after 12 months followup (12 of 21 patients), delayed union > 12 months (10 of 21 patients), resorption of the greater tuberosity (nine of 21 patients), and aseptic loosening (three of 21 patients).

Conclusions

At intermediate 5-year followup, allograft-prosthetic composite reconstruction of the proximal humerus has an acceptable overall MSTS score and a low incidence of implant revision, but loss of patients to followup and exclusion from the study likely make the results seem better than they actually are. The MSTS score deteriorates between 1 and 5 years. Decreased active forward elevation is not likely to be the sole reason for worsening MSTS scores. A variety of delayed complications including delayed union, resorption of the greater tuberosity, and superior subluxation occurs frequently and may contribute to overall scores. Future studies that compare allograft-prosthetic composites against other forms of reconstruction should attempt to control for possible selection bias and have sufficiently long followup to detect the deterioration of MSTS scores that occur with time.

Level of Evidence

Level IV, therapeutic study.

Introduction

Shoulder reconstruction after proximal humerus resection is challenging. Endoprosthetic replacement provides predictable pain relief and durable implant survival, but restoration of normal shoulder strength and active ROM have not been achieved with segmental endoprosthetic hemiarthroplasty [3, 6, 7, 9, 12]. Patients face difficulties raising the arm to a useful level for activities of daily living. The reasons for the inability to restore full function may include the extent of soft tissue resection, muscle denervation, capsular laxity, shoulder instability, and subluxation. Inadequate soft tissue attachment of the rotator cuff may be a critical factor.

Allograft-prosthetic composite reconstruction has been used in various anatomic sites to improve tendon and capsular attachments. Theoretically, this type of shoulder reconstruction would enable the tendons of the rotator cuff to be connected to the respective allograft insertions and the capsule to be reconstituted. Patients undergoing osteoarticular or allograft-prosthetic composite reconstructions of the proximal humerus seem to enjoy relatively good early function with respect to pain relief, active ROM, and Musculoskeletal Tumor Society (MSTS) scores [1, 4, 10, 13]. However, there are relatively few papers pertaining to the durability of the construct and changes in the MSTS or other functional outcome scores over time [1, 11, 13].

In this study, we asked the following: (1) What is the cumulative incidence of allograft-prosthetic composite revision surgery at 5 years? (2) What are the early- and intermediate-term MSTS scores of allograft-prosthetic composite reconstruction of the shoulder? (3) What are the complications of allograft-prosthetic composite reconstruction?

Patients and Methods

Between 2000 and 2015 we performed 27 allograft-prosthetic composite reconstructions of the proximal humerus. Of these 27 patients who underwent allograft-prosthetic composite reconstruction, six were lost to followup before the minimum 24-month followup and were excluded from the analysis. Of note, one patient died at 22 months, but this patient was included because patient death is a competing risk for the calculation of the cumulative incidence of implant revision surgery (see subsequently), and the patient was not lost to followup. The study cohort is comprised of the remaining 21 patients. The median followup was 97 months (range, 20–198 months). Four patients died of disease (range, 22–91 months). Of the 17 patients who were alive, three were not seen within the last 5 years.

The data were gathered retrospectively from medical records and radiographs by the individuals not involved in surgical care (MEB, JL). The mean age at diagnosis was 41 years (range, 20–80 years). There were 14 men and seven women. The study received approval from the institutional review board.

There were three benign tumors, 17 primary malignant sarcomas, and one solitary metastatic renal cell carcinoma. The specific pathologic diagnoses included six chondrosarcomas, seven osteosarcomas, four Ewing’s sarcomas, two giant cell tumors, one desmoplastic fibroma, and one renal cell carcinoma. Ten of 21 patients received preoperative chemotherapy, and seven patients were given postoperative chemotherapy. No patient received local radiation pre- or postoperatively.

During the period in question, we considered an allograft-prosthetic composite when patients had a primary malignant tumor or aggressive benign tumor of the proximal humerus. Patients with metastatic carcinoma and multiple myeloma were not candidates unless they had a solitary metastasis from renal cell cancer and no disease elsewhere. Furthermore, to be operative candidates, patients had to have tumors that could be resected with wide margins and an intraarticular approach. The surgery had to be compatible with preservation of the axillary nerve, a remnant of the rotator cuff tendons, and most of the deltoid muscle. Patients who had an extraarticular shoulder resection were not candidates for allograft-prosthetic composite reconstruction. During the study period, there were a total of 67 patients who met the eligibility criteria for the procedure. Of these, 40 patients underwent segmental endoprosthetic reconstruction rather than allograft-prosthetic composite reconstruction. The reason for the choice was difficult to discern from reviewing the medical records. There was no algorithm or protocol for assigning which reconstruction to use. The final decision for the type of reconstruction was made at the discretion of the attending surgeon. However, some preference was given to younger patients with smaller extraosseous tumors (see the Discussion).

The procedure was performed through an extended deltopectoral approach. A hemiarthroplasty was performed in all patients with native glenoid preservation. A long-stem prosthesis was cemented into both the allograft and the host bone. An important aspect of the cementation was to remove soft cement from the bone interface with a knife before final seating of the implant. In seven more recent patients, a supplemental plate was placed at the osteotomy. The choice of a supplemental plate was made at the attending surgeon’s discretion. Surgeons applied primary bone graft to the osteotomy in all patients. A variety of materials for bone grafting was used, including iliac crest bone graft (eight patients), iliac crest bone marrow (two patients), corticocancellous allograft bone chips (six patients), calcium phosphate artificial bone graft substitute (Vitoss [Stryker Corporation, Kalamazoo, MI, USA], four patients), and/or demineralized bone matrix (three different manufacturers, 11 patients). The capsule and rotator cuff were repaired with interrupted, nonabsorbable, braided polyester sutures. The sutures were placed in a horizontal mattress, pants-over-vest fashion, leaving the ends tagged and untied until all sutures were placed. For the rotator cuff, the sutures for the supraspinatus tendon were tied first with the shoulder partially abducted and flexed forward to tension the repair. The sutures for the subscapularis and infraspinatus/teres minor were tied with the shoulder in neutral alignment to balance internal and external rotation.

A shoulder immobilizer with a 10° abduction pad was used in 15 patients for 6 weeks postoperatively. A larger 75° abduction pad was used in six patients. The choice of the type of immobilizer was made at the discretion of the attending surgeon. Early active-assisted ROM was begun 3 weeks postoperatively. Physiotherapy to achieve muscle power in forward elevation was emphasized. External rotation was limited to 30° in the first 3 months.

From the chart review and from prospectively collected data sheets, we analyzed ROM with particular attention to active forward elevation and the MSTS score for the upper extremity [5] at 1 year and 5 years postoperatively (mean, 64 months; range, 60-71 months). Although a goniometer was used for ROM at the most recent followup visits, we could not ascertain from the clinic notes whether a goniometer was used in all patients. Data sheets for MSTS scores were collected during clinic visits, and the data were transferred to a longitudinally maintained departmental surgical database. Missing data were excluded from statistical analysis. Steps to mitigate bias in outcomes assessment included analysis of radiographs without knowledge of MSTS scores, completion of MSTS functional outcome score sheets by clinicians other than attending surgeons, and abstraction of data by the authors not involved in any of the operations (MEB, JL).

The assessment of surgical complications for the third research question of this study was made by review of both charts and radiographs. For most complications that were noted in progress notes or reports in the medical chart, radiographic verification with specific criteria for each complication was performed (see subsequently). Revision of the implant for the purposes of this study was defined as removal of the allograft-prosthetic composite and replacement with another construct. Bone grafting and other procedures for delayed union were not counted as revisions. For the purposes of this study, we define delayed complications as those occurring after 12 months from the time of the index operation.

Review of radiographs was necessary to establish whether complications had occurred. For the purposes of this study, we defined delayed superior subluxation as progressive upward migration of the humeral prosthetic head > 1 cm relative to the glenoid on followup AP radiographs taken at a median of 92 months (range, 20-198 months) and analyzed by two of the authors (MEB, JL) who were not involved in patient care. The displacement of the prosthetic head was measured from the bottom of the glenoid to the bottom of the prosthetic humeral head, and the final followup radiograph was compared with the initial postoperative radiograph to establish whether progressive superior migration had occurred (Fig. 1A-B). We defined delayed bone healing as the lack of bridging callus at the bone-allograft junction on two orthogonal radiographs 1 year postoperatively. Patients were considered to have nonunion if there was no bridging callus on two orthogonal radiographs at the time of last followup or at the time of revision surgery for implant removal. Aseptic loosening was deemed to be present if there was a complete radiolucent line around the host bone-cement interface, displacement of the stem relative to host bone, or fracture of the cement mantle in the host bone. Final radiographs were screened for the presence of periprosthetic fractures, dislocation of the glenohumeral joint, and resorption of the greater tuberosity. The latter phenomenon was deemed to be present if there was loss of the cortical bone of the greater tuberosity on any radiographic view when comparing followup radiographs against immediate postoperative radiographs. Resorption of the greater tuberosity was distinguished from loss of bone as a result of local recurrence, which affected two patients.

Fig. 1 A-B.

Superior glenohumeral subluxation gradually develops over time. Superior subluxation was defined as progressive migration > 1 cm of the humeral head relative to the glenoid. The distance “d” was measured between the inferior glenoid (“b”) and the inferior prosthetic humeral head (“a”) on AP radiographs. (A) In this example, the early postoperative radiograph showed a distance “d” of 0.6 cm, (B) whereas the late radiograph at 5 years showed a distance “d” of 4.0 cm (migration = 3.4 cm). Resorption of the greater tuberosity is present at late followup. The patient underwent revision to a segmental reverse shoulder arthroplasty for painful subluxation and stiffness.

Statistical Analyses

The majority of statistical calculations was performed with SPSS® version 23 (IBM Corp, Armonk, NY, USA). We used the paired t-test to compare the means of dependent continuous variables and the Student’s t-test for independent samples. We analyzed the relationship between the ordinal subscores at 1 year and 5 years followup with the nonparametric Wilcoxon signed-rank test for two related samples. The chi-square and Fisher’s exact tests were used to query the possible association between categorical variables.

To quantify the risk of implant revision, we used the cumulative incidence function because the Kaplan–Meier survival-based estimates count patients who die without implant revision as simple censoring events rather than as competing risks. The graph and table were produced by R version 3.4.3 (R Foundation for Statistical Computing, Vienna, Austria). The data were prepared using SAS 9.4 (SAS, Cary, NC, USA). A p value < 0.05 was accepted as statistically significant.

Results

The cumulative incidence function was used to describe failure of the implant, as defined by revision surgery necessitating implant removal, with death of the patient as a competing risk. The cumulative incidence of implant revision in this competing risk analysis was 10.1% (95% confidence interval [CI], 1.6%-28.0%) at 5 years (Fig. 2). Four of 21 patients underwent revision operations and implant removal. Four patients died of disease without revision of the implant. The causes for revision were aseptic loosening with concomitant nonunion (two patients), severe subluxation (one patient, Fig. 1A-B), and infection (one patient). Three patients were revised to segmental reverse shoulder arthroplasty. The fourth patient had an infected implant, which was treated by allograft and prosthesis removal, followed by antibiotic spacer placement, and finally implantation of an endoprosthesis (without reverse shoulder arthroplasty). That patient remains free of disease and infection after 9 years. No patient underwent upper limb amputation.

Fig. 2.

Competing risk analysis with death of the patient as the competing risk showed that the cumulative incidence of revision of the implant was 10.1% (95% CI, 1.6%-28.0%) at 5 years and 25.4% (95% CI, 6.7%-49.9%) at 10 years. There were 13 patients remaining at 5 years and six patients remaining at 10 years. The cumulative incidence of implant failure is shown as a solid green line, and the 95% CI bands are dotted green. Death of patient as a competing event is shown as a dashed red line.

Mean total MSTS scores were 25.7 of 30 (86%) ± SD 2.87 at 1-year followup and 23.4 (78%) ± SD 4.53 at 5-year followup (mean difference ± SD 2.4 ± 3.52; p = 0.014; Table 1). The individual component scores showed no difference between 1 and 5 years (Table 1). As measured on clinical examination, the mean active forward elevation was 101° (± SD 33°) at 1 year and 92° (± SD 34°) at 5 years (mean difference ± SD 8° ± 36°, p = 0.41).

Table 1.

Mean MSTS scores at 1 and 5 years

The most common complication was delayed superior subluxation (> 1 cm, occurring after 12 months) in 12 patients (Fig. 1A-B) followed by delayed union in 10 patients, allograft resorption at the greater tuberosity in nine patients, aseptic loosening in three patients, local recurrence in two patients, periprosthetic fracture in one patient, and infection in one patient. Of the 10 patients with delayed healing, six eventually united at the osteotomy site (one with no further treatment, two with iliac crest bone graft, and three with vascularized bone graft); two patients had nonunion at the time of revision (at 26 months and 65 months); and two patients died of disease without bridging callus (at 22 months and 71 months). Primary bone graft to the osteotomy was used in all patients at the time of the index operation, but this consisted of a variety of different materials, including iliac crest bone and demineralized bone matrix (see Patients and Methods). Delayed union and nonunion were not associated with any specific form of bone graft in chi-square analysis (p > 0.05 for all forms of bone graft). Of the seven patients who had plating of the junction at the time of the initial operation, three had delayed union, whereas of the 14 patients without primary plating, seven developed delayed union (p = 0.12). Chemotherapy was not associated with delayed or nonunion (p = 0.14). All nine patients with resorption of the greater tuberosity had a concomitant superior subluxation, whereas this did not occur in the nine patients without greater tuberosity resorption (Table 2; p = 0.001). Allograft resorption from the greater tuberosity did not extend to the osteotomy and did not directly affect bone union. With the numbers we had, neither greater tuberosity resorption nor superior subluxation was associated with final active forward elevation or MSTS scores. We did not find an association between aseptic loosening and any patient parameter. Although all three patients with aseptic loosening had a delayed or nonunion, seven of 18 patients without aseptic loosening also had delayed or nonunion (p = 0.078).

Table 2.

Chi-square analysis of resorption of greater tuberosity and superior migration

Discussion

Restoration of shoulder function after proximal humeral resection continues to be a difficult challenge in orthopaedic oncology. For segmental endoprosthetic reconstruction with hemiarthroplasty, previous authors have noted that late subluxation of the joint and superior migration of the prosthetic head can cause subacromial impingement and pain [6, 11]. Active forward elevation is difficult to restore, and our previous analysis of 83 segmental endoprosthetic replacements of the proximal humerus showed a mean active forward flexion of only 42°, rarely reaching 90° [2]. The observation is similar to the experience of other authors [7, 12]. Although allograft-prosthetic composite reconstruction offers the theoretical advantage of reconstituting bone and tendinous structures, some authors have suggested that this form of reconstruction is not superior to endoprosthetic reconstruction in terms of MSTS scores [13]. What has not been emphasized in previous studies is the effect of time on MSTS scores and the prevalence of delayed complications. In this study, we found that implant survival and MSTS scores deteriorate with time. Notable complications, including subluxation, allograft resorption, delayed bone healing, and aseptic loosening, tended to occur after 12 months of followup.

The small number of patients is a major weakness of the study and affects many of the statistical tests. Although we were able to show a decrease in the overall MSTS score over time, the individual component scores were not different, and it is possible that we could not detect changes in the subcomponents as a result of insufficient power. Similarly, we did not find a statistical association between aseptic loosening and delayed union although all three patients with aseptic loosening had delayed union. Chemotherapy can affect bone healing, and it may also potentially be a factor that affects the overall MSTS score, but our study did not find an effect of chemotherapy on bone healing or MSTS score, possibly as a result of the small cohort of patients. Furthermore, the small sample size made it impractical to study the effect of specific agents and chemotherapy protocols. A second major weakness of the study is that our measures of functional outcome were limited to ROM and the MSTS score. A third major weakness is the inability to control for the consistency and quality of the surgery and soft tissue repair. This is operator-dependent and may contribute to some of the variation in strength, ROM, and bone healing. The technique of stem fixation may affect the rate of delayed healing, because it is theoretically possible that the cement can interfere with callus formation, especially at the bone junction and the intramedullary contribution to new bone formation. The variability in bone grafting techniques compromised the ability to discern which method may have been superior for bone healing.

Our study is subject to certain biases, which are important to recognize. Selection bias can certainly affect the results. Of the 67 patients who were candidates for the operation during the study period, only 27 received the operation. There was probably some bias among the attending surgeons toward patients who were more favorable candidates for allograft-composite reconstruction in terms of their being younger, healthier, and having smaller extraosseous tumors. The surgeons performing the operations did agree with the recommendations of previous authors who advocated the procedure for younger patients with primary tumors [11]. This would potentially make the overall results seem better in terms of revision rate, MSTS scores, and active motion than if the procedure had been more widely used. However, inclusion of all patients may not necessarily diminish the deterioration of MSTS scores over time or the development of delayed complications that we are reporting in this study. In fact, one might predict that the magnitude of these negative observations would be increased. Transfer bias in the study may be present because six patients were excluded for followup < 24 months, and three patients in the study were lost to followup after 24 months. It is possible that these patients sought treatment for complications elsewhere, and again, this would have resulted in worse overall results than we found. Finally, assessment bias may be present because the MSTS score and active ROM are both subject to clinician bias, making the absolute scores and ranges better than they really are. We did detect worsening of the MSTS score with time, but with the numbers we had, we could not demonstrate a decrease in active forward flexion over time. It is possible that a more objective unbiased method of motion measurement such as with an electronic motion detector might obviate this problem in the future.

At 5 years after surgery, the cumulative incidence of implant removal and revision to another construct was 10.1%. There was one early revision resulting from infection at 2 months. The other revisions occurred later, between 37 and 109 months after the index operation, and were associated with aseptic loosening, subluxation, and allograft resorption at the greater tuberosity. Estimates of proximal humeral implant survival after allograft-prosthetic composite reconstruction are sparse in the available evidence. In one series, the estimated Kaplan-Meier survival was 91% at 5 years, which is similar to our study [11]. Another study reported a long-term survival rate of 88% at 10 years [1], which was better than our results. The difference in implant survival may be related in part to the small number of patients, large CIs as well as the threshold for revising an implant with resorption of bone and subluxation. It is also noteworthy that the latter study [1] had only 10 patients with deltoid-sparing resections and included 13 patients with total deltoid resection. Deltoid removal might result in less implant stress and lower aseptic loosening risk. Currently, this is speculative, and further work is needed to test the idea critically. Furthermore, the degree to which deltoid function contributes to the daily use of the arm and the measured outcome scores is yet undefined. Although it is a goal of surgeons to restore active shoulder abduction, other factors figure prominently in patients’ satisfaction, including maintenance of hand dexterity, pain, and not having reoperations on the limb.

We found that MSTS functional outcome scores deteriorate with time. The observation has not been emphasized in previous studies [1, 10, 11, 13] because there are little data examining the effect of time on functional outcome. Several notable patients in this study initially achieved excellent function and forward elevation beyond 150°; however, they subsequently lost shoulder strength, and their functional scores declined. Although it may be tempting to theorize that the function decline in our patients might be attributable to the resorption of the greater tuberosity, we did not find statistical support for this idea. Some patients who developed resorption and superior subluxation were noted to maintain active ROM and MSTS scores despite the radiographic findings. It seems more likely that the cause of worsening shoulder function was more complicated and multifactorial. The final mean MSTS score in our study was comparable to other studies. In one study, the mean active forward flexion was 70° for patients without deltoid resection, and the mean MSTS score was 87% [1]. Two other studies reported mean MSTS scores of 72% and 79% at last followup [11, 13], similar to our final mean MSTS score of 78%. Although the overall results in these studies as well as our study demonstrate only a moderate degree of success in terms of restoring shoulder function, it is perhaps important to recognize that the results may be a reflection of not only surgical technique and skill, but also our current technology with regard to bioengineering. We currently do not have the means to control biologic remodeling of the allograft and repopulation of allograft with live host cells and tissue at critical areas, including the osteotomy junction, tendons, and cortical bone. Unfortunately, the retrieved allografts from the revision surgeries were not subject to rigorous biologic study to determine how the host cells and tissues were interacting with and incorporating into the allograft.

Noteworthy complications in our series included superior subluxation (12 of 21 patients), delayed union (10 of 21 patients), resorption of the greater tuberosity (nine of 21 patients), and aseptic loosening (three of 21 patients). Other authors have also noted a high complication rate. One study found the following: 14% for subluxation, 11% for delayed union, and 8% for aseptic loosening [1]. Another series reported 20% of patients with infection, 20% with fracture, and 30% with subluxation [13]. Superior migration of the humeral head suggests rotator cuff incompetence, particularly the supraspinatus. We theorize that there may be an inflammatory reaction, which might contribute to graft resorption, tendon disruption, and greater tuberosity resorption. Indeed, there was a statistical association between superior subluxation and resorption. It is perhaps notable that three of the four patients needing revision in our study underwent reverse shoulder arthroplasty as the salvage operation. We did not try to compare the results of allograft-prosthetic composite reconstruction with reverse shoulder arthroplasty because our cohort of reverse shoulder replacements has much shorter followup. Interestingly, a recent report described the use of a reverse total shoulder replacement together with an allograft-prosthetic composite for two patients [8]. We also did not attempt to compare the allograft-prosthetic composite with segmental endoprosthetic replacement with shoulder hemiarthroplasty, in part because of the issue of selection bias, which was alluded to previously. Patients of the latter group may have been less favorable candidates for allograft-prosthetic composites because the patients tended to be older with larger tumors and more extensive soft tissue resection. A fair comparison of the techniques would ideally attempt to control for such selection bias.

In summary, allograft-prosthetic composite reconstructions of the shoulder can partially restore active ROM. At intermediate 5-year followup, the incidence of implant revision is low, and the overall MSTS score is acceptable, but loss of patients to followup may make these results seem better than they really are. The MSTS score deteriorates between 1 and 5 years, and complications after 1 year are common. Gradual superior migration of the humeral head was observed in more than half of the patients, and this was associated with resorption of the greater tuberosity. The procedure is technically demanding and may not be appropriate for older patients or those who need extensive resection of soft tissues around the shoulder. Surgeons considering the operation should be aware of worsening MSTS scores over time and the development of delayed complications, both of which would make this construct less attractive with extended followup. Future studies that compare allograft-prosthetic composites against other forms of reconstruction should attempt to control for possible selection bias, and the studies should have sufficiently long followup to detect the deterioration of MSTS scores that occur with time.