Patient Selection
In a previous investigation7, radiographs and computed tomography scans were used to classify the glenoid morphology in 211 patients (216 shoulders) treated with a primary reverse shoulder arthroplasty from October 2004 through December 2007. Patients included in the study presented for a primary reverse shoulder arthroplasty after failing to respond to nonoperative measures. Our current investigation is an institutional review board-approved study to examine the clinical outcomes of the patients in the previous study. A cohort of 141 patients (143 shoulders) with minimum of two years of follow-up (average, 30.9 months; range, twenty-four to fifty-one months) was included in the current series (Table I). The average age of the patients at the time of surgery was 71.7 ± 8.3 years (range, forty-two to eighty-seven years). Of the 143 shoulders, fifty (35%) were in men and ninety-three (65%) were in women. Surgery was performed on eighty-five right shoulders (59%) and fifty-eight left shoulders (41%). Reverse shoulder arthroplasty was performed for an underlying diagnosis of primary rotator cuff deficiency (massive, irreparable rotator cuff tear and no previous surgery) in seventy-eight shoulders (55%), failed rotator cuff surgery in sixty-three shoulders (44%), and fracture sequelae in two shoulders (1%). All surgical procedures were performed by the senior author (M.A.F.).
Anatomic Evaluation
Details of the anatomic measurements for the group have been described previously7. Briefly, each patient underwent a preoperative radiographic evaluation (anteroposterior; anteroposterior tangential to the face of the glenoid with the arm in neutral rotation, i.e., a true anteroposterior or Grashey view of the glenohumeral joint; axillary; and outlet or scapular Y views) and two-dimensional computed tomography scans (with a slice thickness of =1 mm). The computed tomography scans were further used to create three-dimensional reconstruction models: the original computed tomography scan images, after isolating the scapula from the other structures, were converted into a stereolithography file by the program Mimics (Materialise, Leuven, Belgium) and then were imported into a computer-aided design program (SolidWorks, Concord, Massachusetts). Glenoid morphology was classified as either normal (eighty-seven shoulders; 61%) or abnormal (fifty-six shoulders; 39%) by three independent observers on the basis of (1) plain radiographs alone, (2) a combination of plain radiographs and two-dimensional computed tomography scans, and (3) the three-dimensional computed tomography reconstruction models (Fig. 1). Abnormal glenoids (those with acquired osseous defects) were then further subdivided by the site of erosion into posterior (twenty-five shoulders; 17%), superior (sixteen; 11%), global (eleven; 8%), and anterior (four; 3%). Anatomic measurements of the scapulae (see Appendix) were made with use of three-dimensional reconstruction images of each and included the glenoid height, width, version, and inclination; the distance from the coracoid base to the glenoid surface; and the distance from the acromion base to the glenoid surface (Fig. 2). Additionally, the bone available for purchase by the baseplate was determined by measuring the standard centerline perpendicular to the glenoid surface (defined as the line that originated in the center of the glenoid, perpendicular to the articular surface, and exited on the anterior aspect of the scapular neck)19,20 and the alternative scapular spine centerline (defined as the point that originated in the center of the glenoid and then was aligned with the scapular spine but was not necessarily perpendicular to the glenoid surface). The glenoid area available for peripheral screw placement was measured in the quadrants and columns with use of a previously described technique7.
Surgical Technique
Insertion of the reverse shoulder arthroplasty glenoid baseplate (Reverse Shoulder Prosthesis; DJO Surgical, Austin, Texas) in the eighty-seven shoulders with normal glenoid morphology proceeded in a standard fashion2,21. In these shoulders, the ideal position of the central screw followed the path of the standard glenoid centerline (Fig. 3, A) as described by Matsen et al.19 and Bicos et al.20. This placement results in the baseplate being oriented perpendicular to the glenoid surface and the central screw exiting anteriorly on the scapular body.
The surgical technique was modified in the fifty-six shoulders with acquired glenoid bone defects (as determined by preoperative radiographic evaluation and intraoperative inspection). In all fifty-six shoulders with acquired glenoid bone defects in the current series, a different orientation of the central screw was used to increase the amount of bone captured. This alternative centerline originated from the center of the glenoid surface, passed along the axis of the scapular spine, and passed through the junction of the scapular spine as it joined the body of the scapula7 (Fig. 3, B and C).
After the central hole was drilled along the spine centerline, depth gauge measurement confirmed a minimum screw length of 25 mm. The tap was then placed, and a cannulated glenoid reamer prepared the glenoid surface. If <80% of the underside of the glenoid baseplate was in contact with the reamed glenoid (Fig. 4) because of eccentric bone loss, which occurred in twenty-two of the fifty-six shoulders, either a bulk graft from the humeral head (twenty-one shoulders) or a femoral head allograft (one shoulder) was used. The graft was contoured and placed directly into the defect and was held in place by either Kirschner wires or a 3.5-mm screw. The construct was then reamed to conformity with the tap still in place. The tap was then removed and the baseplate inserted. The baseplate was tightened to provide compression of the graft to the native glenoid surface (Fig. 5). Five-millimeter locking screws were placed, and the temporary Kirschner wires were removed. The structural graft was then sculpted around the implant with a burr.
Glenosphere selection was based, first, on the presence of a glenoid defect and, second, on the ability of the components to provide stability without restricting motion. In the presence of a defect, a larger glenosphere was selected for 46% (twenty-six) of fifty-six shoulders to adequately cover the defect. In glenoids without bone loss, a smaller size (32-mm [—4-mm offset], which has a 6-mm center of rotation lateral to the glenoid surface, for female patients and 32-mm [neutral offset], which has a 10-mm center of rotation lateral to the glenoid surface, for male patients21) was routinely used. The humeral socket chosen for 89% (127) of the 143 shoulders was the standard (neutral) size. Next, the shoulder was reduced, and humeral and glenoid component articulation was verified. Finally, shoulder range of motion was tested. If the shoulder was unstable or there was a restriction in motion due to tight soft-tissue structures, both components would be changed to achieve the optimal combination.
A larger glenosphere with a hood was utilized in fifteen shoulders to overhang and protect the graft (Figs. 6 and 7). All operations were performed in a single stage regardless of glenoid morphology. All humeral components were cemented in 30° of retroversion. The system utilized in this study contains humeral sockets with different thicknesses: the neutral thickness for each diameter is the thinnest, and the +4-mm and +8-mm sockets represent an increase in the thickness of the polyethylene relative to the neutral socket of 4 or 8 mm, respectively. In the eighty-seven shoulders with normal glenoid morphology, a neutral socket was selected for 92% (eighty shoulders); a +4-mm socket, for 7% (six shoulders); and a +8-mm socket, in one shoulder. In the fifty-six shoulders with acquired glenoid bone defects, a neutral socket was placed in 84% (forty-seven shoulders); a +4-mm socket, in 13% (seven shoulders); and a +8-mm socket, in 3% (two shoulders). Socket height was chosen on the basis of soft-tissue tension during trial insertion of the components.
Clinical Evaluation
Patients were followed clinically for a minimum of two years. At each follow-up visit, data were routinely collected for the American Shoulder and Elbow Surgeons (ASES) scores22, the Simple Shoulder Test (SST)23, and a 10-point visual analog scale for pain and function. All range-of-motion measurements reported were active range of motion and were made preoperatively and postoperatively with use of a previously described method24.
Radiographic Evaluation
Postoperative radiographs were reviewed for all 143 shoulders. The initial radiographs were compared with interval radiographs through the patient's most current follow-up. The standard four radiographs of the shoulder, including an anteroposterior radiograph, an anteroposterior tangential radiograph with the arm at neutral rotation (a true anteroposterior or Grashey radiograph of the glenohumeral joint), an axillary radiograph, and an outlet or scapular Y-view radiograph, were compared to identify any evidence of component failure or notching, progressive radiolucent lines, or obvious baseplate or graft failure. The Grashey radiographs were evaluated at each postoperative visit, and they were compared with successive radiographs to determine what, if any, interval changes occurred. Finally, the first postoperative radiographs were compared with the most recent radiographs, and any radiolucencies at that time were noted.
Statistical Analysis
Outcomes Assessment
A post hoc power analysis was performed on the basis of our primary outcome measure, the ASES score. This analysis revealed that a minimum sample size of thirty-seven was required to detect a 15-point difference in the ASES score25 (a = 0.05 and ß = 0.90). Preoperative and postoperative outcome measures as well as range of motion were compared for all patients and between groups on the basis of sex, glenoid morphology, use of bone graft, and glenosphere size. The normality of all variables was evaluated with use of the Kolmogorov-Smirnov test. Comparisons between the patients with normal morphology and those with acquired glenoid bone defects with regard to the scores obtained preoperatively and postoperatively were conducted with use of the nonparametric Mann-Whitney U test to determine whether there were baseline differences between the two groups. The differences between preoperative and postoperative values were then analyzed with use of a t test with a Levene test to evaluate equality of variance. All variables including outcome scores, glenosphere size, and the use of bone graft were compared on the basis of patient sex and glenoid morphology with use of either the t test or the Mann-Whitney U test for continuous variables and either the Fisher exact test or the chi-square test for categorical variables. Alpha was set at 0.05.
Morphologic Analysis
Previously collected data7 on the morphologic features of the glenoid, including height, width, inclination, version, acromial-glenoid distance, coracoid-glenoid distance, standard centerline distance, and spine centerline distance, as well as the peripheral screw placement area, were used for the current study. These values were compared, with use of the t test or Mann-Whitney U test, on the basis of glenoid morphology (normal or acquired glenoid bone defects) and the need for bone graft.
Surgical Technique Analysis
Operative technique was analyzed retrospectively to identify accommodations made for various types of glenoid erosion. Particular attention was given to the use of bone graft and the size of glenosphere selected. These factors were first compared between the normal and acquired glenoid osseous defect groups in a univariate analysis. Next, a multivariate regression analysis with use of forward stepwise likelihood ratio methods was performed to determine the independent effects of variables, including patient sex, glenoid morphology, patient age at the time of surgery, and previous rotator cuff surgery, on operative technique (selection of glenosphere size and use of bone graft).
Source of Funding
The Foundation for Orthopaedic Research and Education, a private foundation, received research support from DonJoy Orthopaedic Surgical (DJO Surgical). DJO Surgical manufactures the Reverse Shoulder Prosthesis under investigation. The senior author (M.A.F.) is the designing surgeon for the Reverse Shoulder Prosthesis.
Outcomes Assessment
All Patients
A summary of the preoperative and postoperative ASES and SST scores and the range of motion for all patients is presented in the Appendix. All mean outcomes for the entire group improved postoperatively (p < 0.0001). The mean total ASES score improved from 39.1 to 75.1, the mean SST score improved from 1.9 to 6.4, and the mean visual analog pain score improved from 6.1 to 1.5. Range-of-motion analysis revealed improvements in the mean values for forward flexion (from 67° to 140°), abduction (from 65° to 126°), external rotation (from 19.8° to 49°), and internal rotation (from a spinal level of S1 to L2). The outcome was rated as excellent by eighty patients, good by forty-one, satisfactory by nineteen, and unsatisfactory by one.
Patient Subsets
Preoperative and postoperative range of motion and functional scores (including ASES, SST, and visual analog scale scores) were also compared for specific patient subsets (Table II and Appendix). There were significant differences between male and female patients on the preoperative SST scores (p = 0.002) and differences that were approaching significance on the postoperative ASES function and SST scores; however, these findings disappeared when preoperative and postoperative differences were compared. No significant difference was detected when the patients were grouped by glenoid morphology or use of bone graft.
Postoperative Radiographs
No evidence of failure was noted with respect to the graft or component loosening in any shoulder. No radiolucent line was noted at the baseplate-bone interface. No progressive radiolucent lines were evident around the bone-cement interface on the humeral component. One shoulder with grade-2 notching of the scapular neck, according to the Nérot grading system26, was noted in a patient with normal glenoid morphology. This finding was seen on the one-year postoperative radiograph. No progression was noted through the thirty-six-month follow-up period, and thus the patient continued to be treated conservatively.
Morphology
Normal Compared with Abnormal Glenoid Morphology
Significant differences between these groups were seen with respect to the glenoid height, width, inclination, coracoid-to-glenoid distance, acromion-to-glenoid distance, centerline distance, spine centerline distance, and peripheral screw area (Table II).
Bone Graft
Glenoids with acquired osseous defects that were treated with bone-grafting were compared with abnormal glenoids that had not had grafting. Glenoids treated with bone-grafting had a greater amount of erosion preoperatively as depicted by an increase in inclination (mean, 19.9° compared with 13.5°; p = 0.0065) (Table II).
Surgical Technique
Bone Graft
Bone-grafting was performed in twenty-two of 143 shoulders. Twenty-two (39%) of the fifty-six abnormal glenoids received a bone graft, and no normal glenoid was treated with a graft (p = 0.016).
Glenosphere Size
The most commonly used glenosphere sizes in all patients were a 32-mm (—4-mm offset) glenosphere (59%), a 32-mm (neutral offset) glenosphere (17%), and a 36-mm (—4-mm offset) glenosphere (15%). The other sizes (thirteen shoulders) were excluded from the remaining analyses because of low sample sizes. Univariate analysis revealed that glenosphere size varied depending on three factors: glenoid morphology (p < 0.001), sex (p < 0.001), and bone-graft use (p = 0.016). Among the shoulders with normal morphology, the majority had implantation of either a 32-mm (—4-mm offset) glenosphere (77%) or a 32-mm (neutral offset) glenosphere (21%). Shoulders with acquired glenoid bone defects demonstrated a different distribution, with a 32-mm (—4-mm offset) glenosphere implanted in 47% and a 36-mm (—4-mm offset) glenosphere implanted in 39%, indicating a shift toward larger sized implants. In the presence of bone graft, a 32-mm (—4-mm offset) implant was used in 27% (six) of twenty-two shoulders and a 36-mm (—4-mm offset) glenosphere was used in 55% (twelve shoulders). Sex was also found to have significant effects on the size of glenosphere chosen. The most frequently used sizes in women were 32-mm (—4-mm offset) (80%) and 36-mm (—4-mm offset) (19%), but the most commonly used sizes in men were 32-mm (neutral offset) (55%) and 32-mm (—4-mm offset) (36%). Logistic regression accounting for glenoid morphology, bone-grafting, sex, prior rotator cuff surgery, and age at the time of surgery revealed that the four factors identified in univariate analysis remained significant predictors for different device sizes. Specifically, female sex (p < 0.001) and no use of bone graft (p = 0.021) were predictive for 32-mm (—4-mm offset) glenosphere placement, and male sex (p < 0.001) was predictive of 32-mm (neutral offset). Normal morphology was associated with placement of both 32-mm (—4-mm offset) and 32-mm (neutral offset) glenospheres, whereas the use of bone graft was strongly predictive of placement of a 36-mm (—4-mm offset) glenosphere.
Complications
There were five complications. Two shoulders had an acromial fracture, which was treated nonoperatively. Two shoulders had a deep infection; one was treated with a two-stage revision and the other with extensive irrigation and débridement along with single-stage revision surgery. The remaining complication was a periprosthetic fracture that occurred when the patient lifted a heavy tree. The shoulder was treated nonoperatively. Of the patients with complications, two rated the outcome as excellent; two, as good; and one, as satisfactory. All complications, with the exception of the periprosthetic fracture, were in the group with normal morphology.
The present study showed that, in patients who have a primary reverse shoulder arthroplasty, surgical techniques can be modified to accommodate glenoid bone loss in shoulders with acquired osseous defects of the glenoid. Additionally, it showed, in contrast to our hypothesis, that patients with acquired glenoid bone defects can have clinical outcomes comparable with those in patients with normal glenoid morphology. These results indicate that, with a modified surgical technique, reverse shoulder arthroplasty for patients with acquired glenoid bone defects can yield short-term clinical outcomes similar to those in patients with normal glenoid morphology.
Our understanding of reverse shoulder arthroplasty has advanced substantially in recent years. This has enabled patients to obtain outcomes beyond the limited goals once set for shoulders with rotator cuff deficiency accompanied by varying degrees of arthritis27. However, glenoid bone loss continues to present a challenge in shoulder arthroplasty. Much of the literature regarding glenoid bone loss has been in the setting of conventional, unconstrained total shoulder arthroplasty, for which various strategies for glenoid component reimplantation have been described8,9,11,12,28-32. Overall, the results have been consistently inferior to outcomes in primary total shoulder arthroplasty without glenoid bone loss, regardless of technique. Failure rates of up to ten times that of primary arthroplasty without glenoid bone loss have been reported9-12. The consequences of not accounting for glenoid bone loss during reverse shoulder arthroplasty implantation have been linked to compromised outcomes as well18,24,33. The majority of the literature on the topic of glenoid bone loss is in the setting of revision surgery. However, on analyzing our patients presenting for a primary reverse shoulder arthroplasty, we found that 37.5% of them had an abnormal glenoid because of bone loss7. Although the prevalence of patients with glenoid bone loss in the general population is largely unknown, our experience suggests that this problem may occur with greater frequency than has been previously recognized.
Regardless of the type of shoulder arthroplasty planned, inadequate bone stock to support a glenoid component has been discussed as a reason to abandon component placement altogether because of the likelihood of decreased functional outcomes and early failure2,12,34,35. However, techniques to accommodate for bone loss and allow glenoid component implantation during reverse shoulder arthroplasty have been reported5,16,17,32. Neyton et al.16 reviewed the cases of nine patients with a Grammont-type reverse shoulder arthroplasty who underwent glenoid bone-grafting for severe bone loss. At a minimum two-year follow-up period, their results showed good pain relief but low postoperative functional scores (a mean Constant score of 53 points), and six patients had radiographic evidence of inferior scapular notching. Elhassan et al.17 reported on twenty-one patients who had reconstruction of the glenoid with a femoral head allograft because of insufficient bone stock during revision shoulder arthroplasty. Three of the patients underwent a Grammont-type reverse shoulder arthroplasty. Those patients were noted to have improvements in the level of pain and forward flexion but an overall decrease in external rotation.
To our knowledge, the present study is the first to compare outcomes of reverse shoulder arthroplasty in patients with and without glenoid bone loss. Despite the challenges presented by bone loss, clinical results in all outcomes measured were not significantly or clinically different from those achieved by the group with normal glenoid morphology. This may be due to several factors inherent in the reverse shoulder arthroplasty design that increase the constraint of the shoulder relative to that of an anatomic total shoulder replacement and make stability and proper soft-tissue tension attainable. Greater surface contact area between the glenosphere and socket in conjunction with the reversal of the articulation neutralizes the destabilizing force of the deltoid. This neutralization enables the baseplate to be relatively anteverted7 in order to accommodate for glenoid bone loss. Despite this anteversion, no shoulder had a dislocation.
Additionally, we believe that in the setting of acquired glenoid bone defects, initial fixation of the baseplate is of paramount importance in the surgeon's ability to successfully implant a glenoid component. By implanting the baseplate in the scapular body along the spine centerline, it is possible to achieve stable fixation even when bone loss is present. No glenoid fixation failure was observed in either patient group during the follow-up period. This appears to indicate that the initial fixation of the central axis of the glenoid baseplate (in this case, a central screw) is critical and can allow for the addition of peripheral osseous support with a bone graft. In our current series, the selection of the alternative scapular spine centerline made it possible to achieve stable glenoid fixation in all shoulders with an abnormal glenoid regardless of the level and location of bone loss.
However, preoperative identification of the level and location of bone loss is an important part of surgical planning. Identification of glenoid bone defects on standard shoulder radiographs should prompt further imaging, including a computed tomography scan with three-dimensional reconstruction. These images can help to guide placement of the central screw (the central axis of the baseplate) into adequate bone for initial fixation. Both the standard glenoid centerline and spine centerline axis should be identified and the proper alignment selected for the baseplate to allow for maximal central screw fixation. Additionally, identification of segmental bone loss, which may prevent sufficient contact with the backside of the baseplate, can be anticipated preoperatively. We found that abnormal glenoids that were supplemented with bone graft had greater glenoid inclination (in the superior direction) than those that had not had grafting. Osseous contact with the backside of the baseplate can be improved with the addition of bulk autograft or allograft to fill eccentric voids in the glenoid.
In addition to initial fixation of the baseplate, proper soft-tissue tensioning may play an important role in improving functional outcomes following reverse shoulder arthroplasty. The selection of various glenosphere sizes in our series reflects alterations in the soft-tissue envelope secondary to either altered glenoid morphology and/or sex. All glenosphere selections were made on the basis of the surgeon's perception of soft-tissue tension to provide a stable shoulder with a range of motion that was unencumbered by the remaining soft-tissue structures. Male patients and patients with acquired glenoid bone defects were treated more frequently with larger glenospheres. Patients who received a larger glenosphere had substantial improvement in all planes of motion including forward flexion (mean, 65.5° to 141°), abduction (mean, 63° to 127.7°), external rotation (mean, 21.2° to 43.9°), and internal rotation (mean, S1 to T12). We suspect that the smaller glenospheres in female patients relate to their smaller stature and, therefore, were selected at the time of surgery to accommodate the diminished volume of the soft-tissue envelope. Currently, there is little information available to assess optimal soft-tissue tension, but we believe that correct soft-tissue tensioning contributes to a successful outcome after reverse shoulder arthroplasty.
One weakness of our study was the lack of a true control group. Patients who had a glenoid with normal morphology were compared with a series of patients with acquired glenoid bone defects that required treatment with a modified technique. There was no control group consisting of patients with glenoid bone defects implanted with a reverse shoulder replacement with use of the standard technique. Therefore, while the modifications in surgical technique assisted in our ability to implant the device, we cannot isolate the effect of surgical technique on clinical outcomes. It is difficult to say whether glenoid fixation would have been possible without modifying the surgical technique.
Additionally, it is important to note that while radiographic outcomes were encouraging, these were not fluoroscopic views and radiographic follow-up was short term. Grashey radiographs were evaluated at each postoperative visit, and they were compared with successive radiographs to determine what, if any, interval changes occurred. Finally, the first postoperative radiographs were compared with the most recent radiographs and any radiolucencies at that time were noted.
Another weakness of the study was that the results were based on a single surgeon using a single type of prosthesis at a unique tertiary-care center that has the longest experience with reverse shoulder arthroplasty in the United States. As such, our experience may not represent that of practices outside the realm of a subspecialty-based tertiary-care referral center or that of practitioners who either have less experience or use other reverse shoulder devices. Techniques utilized in the study were employed with an implant design containing a baseplate with a central compression screw and a center of rotation lateral to the glenoid. Alternative techniques may be necessary to accommodate the varying baseplate and glenosphere designs currently available. Finally, while the results are encouraging for these techniques at a minimum two-year follow-up, this time period is relatively short considering the expected lifespan of the prosthesis. Longer follow-up is necessary to evaluate implant survival, the ultimate fate of the graft, and whether the clinical outcomes seen in this study can be maintained.
Clinically important, acquired glenoid bone defects may not be uncommon and were present in 39% of the shoulders in this series of primary reverse shoulder arthroplasties. Osseous erosion may require alteration in operative technique such as the use of the spine centerline, bone-grafting, and larger glenospheres. Surgeons can anticipate and prepare for these altered techniques with preoperative identification of glenoid bone loss. The spine centerline allows for secure initial fixation of the glenoid baseplate in abnormal glenoids, regardless of the location or severity of bone loss. Larger glenosphere sizes can accommodate for bone loss and appropriately tension the soft-tissue envelope. These modifications may enable patients with acquired glenoid bone defects to obtain short-term functional outcomes similar to those of patients with normal morphology.
Note: The authors acknowledge Jason Busse, DC, MSc, PhD, and Ola Alak, BA, for their efforts in the preparation of this manuscript.