Total shoulder arthroplasty has become an established treatment option for degenerative disorders of the shoulder joint. Significant improvements in function, quality of life, and pain relief have been described1-5. On the basis of these parameters, the results seem to be superior to those of hemiarthroplasty1,6-8. However, the implantation of a glenoid component may lead to loosening in the medium and long term, and is a major concern after total shoulder replacement9-13.
Various implant concepts such as cemented or uncemented humeral and glenoid components have been developed. Unsatisfactory medium and long-term results with high failure rates have been described for uncemented glenoid components in the past14-17, and revision rates between 0% and 23% have been reported14,15,18,19. Many studies have focused on radiographic analysis of cemented glenoid components. The authors of a study of a large cohort of 972 total shoulder replacements performed with a cemented glenoid component noted good overall survivorship of 95% at ten years and 92% at fifteen years (87% for the Cofield-1 and 94% for the Neer-II prosthesis)19. Two recent multicenter studies demonstrated excellent survivorship of cemented keeled glenoid components with revision as the end point but less satisfactory radiographic results: with radiographic loosening as the end point, glenoid survivorship was only 51.5% at ten years13 and 34% at fifteen years12. Other groups have found better radiographic results, possibly because of different examination protocols20,21.
The aim of this study was to examine the clinical and radiographic outcomes of patients treated at a single center with a cemented third-generation total shoulder replacement and followed for a minimum of ten years postoperatively.
Between June 1995 and October 2000, 125 shoulder arthroplasties were performed at our institution and were included in a prospectively recorded database. Total shoulder arthroplasty was performed in sixty cases (fifty-six patients). Of the remaining sixty-five shoulders, forty-three underwent a hemiarthroplasty, nineteen were treated with a fracture hemiprosthesis, and three underwent a reverse shoulder arthroplasty. The decision to perform a total shoulder arthroplasty was made by the surgeon in cases without any articular cartilage remaining on the glenoid.
Inclusion criteria for this study were (1) implantation of the same third-generation cemented implant, (2) a diagnoses of primary osteoarthritis, (3) a minimum duration of follow-up of ten years, and (4) complete clinical and radiographic examination.
Overall, forty-six total shoulder replacements in forty-three patients fulfilled the inclusion criteria. Seven patients died during the follow-up period, leaving thirty-six patients (thirty-nine shoulders). All patients were included in the Kaplan-Meier survival analysis. The seven patients who died were excluded from the clinical and radiographic analysis because the follow-up was less than ten years.
There were twenty-six women and seventeen men in the cohort. Twenty-three patients were right-handed. The dominant shoulder was treated in twenty-one cases. The patients’ mean age at the time of arthroplasty was sixty-four years (range, forty-three to seventy-nine years). The mean duration of follow-up was eleven years (range, ten to fifteen years).
All arthroplasties were performed by or under the supervision of the same surgeon.
Clinical Protocol
Clinical evaluation included the Constant score22 (adjusted for age and sex) and active shoulder motion. Sagittal plane shoulder flexion, coronal plane abduction, and external rotation with the elbow at the side were recorded in degrees, whereas internal rotation was graded according to the posterior spinal region that could be reached by the thumb. Pain, activity, mobility, and strength were graded with use of the Constant score (pain: 0 points [severe pain] to 15 points [no pain]; activity: 0 points [no activity] to 20 points [full activity]; mobility: 0 points [no mobility] to 40 points [full mobility]; and strength: 0 points [0 kg] to 25 points [12.5 kg]).
The patients independently rated their subjective assessment of the outcome as “very satisfied,” “satisfied,” “somewhat disappointed,” or “very disappointed.”
Radiographic Protocol
Standardized true anteroposterior and axillary radiographs of the shoulders were obtained preoperatively and postoperatively. For the axillary view, the patient sat on a chair and leaned sideward (to the side of the affected shoulder) with the arm raised upward. The x-ray beam was directed perpendicular to the cassette and the shoulder joint. The humeral and glenoid bone stock was evaluated with magnetic resonance imaging (MRI) or computed tomography (CT) of the shoulder in all cases preoperatively. The morphology of the glenoid was recorded according to the classification described by Walch et al.23. The glenoid was type A2 in eleven cases, type B1 in sixteen, type B2 in ten, and type C in two.
Radiolucent Lines, Tilt, and Subsidence of the Components
Radiolucency around the cemented glenoid component was evaluated by two surgeons who specialized in shoulder surgery. They assessed the radiographs and reached a consensus. None of the reviewers were involved in the arthroplasties. The radiolucent line score was calculated according to the classification described by Molé et al.24. For this scoring system, the position of radiolucent lines was established with use of six zones: zones 1, 5, and 6 corresponded to the upper, lower, and middle parts of the tray, and zones 2, 3, and 4 were around the periphery of the keel. The thickness of the lines was classified as grade 1 (<1 mm), grade 2 (between 1 and 2 mm), or grade 3 (>2 mm). Scores were calculated by using the six zones and three grades for every case. The accumulation of all grades in each zone results in a maximum score of 18 points. A glenoid component is considered loose when the score is >12 points. This scoring system is based on anteroposterior radiographs only (Fig. 1-A).
Because radiolucent lines around the glenoid components are also frequently detected on axillary radiographs, we analyzed them in a manner comparable with the score of Molé et al.24. On the axillary views, the radiolucent lines were in six defined zones. Zones 1, 5, and 6 were on the anterior, posterior, and middle segments of the tray, and zones 2, 3, and 4 were around the periphery of the keel (Fig. 1-B). The thickness of the radiolucent lines was measured as described above. A glenoid component was considered not at risk when the score was 0 to 6 points, at risk when it was 7 to 12 points, and loose when it was >12 points on at least one view.
Every glenoid component was analyzed for tilt or subsidence by comparing the radiographic findings immediately after the operation with those on the most recent anteroposterior and axillary views. On the anteroposterior view, the tilt was measured by drawing one line along the superior and inferior rims of the glenoid component (the anteroposterior glenoid line) and another line along the lateral base of the coracoid process (the coracoid line) from the superior glenoid rim perpendicular to the bottom margin of the radiograph (Fig. 2). The anteroposterior radiograph is obtained with the patient in a standardized standing position so that the inferior border of the radiograph is parallel to the floor and the lateral base of the coracoid does not change with rotation of the scapula. This technique was described as reproducible by Boileau and Walch25 and later by Habermeyer et al.26, although there is no information about possible changes in the position and tilt of the scapula that might occur with variations of patient posture and stance. The angle between the two lines was measured3.
Tilting of the glenoid component on the axillary view was analyzed as follows. One line was drawn along the medial and lateral rims of the glenoid component (the axial glenoid line) and another line was drawn running through the anterior and posterior cortices of the scapular blade (Fig. 3). The angle between these two lines was measured. Tilting of the glenoid component by ≥5° was defined as loosening. Any tilting or subsidence automatically resulted in a score of 18 points as described in other studies12,13.
Radiographic assessment of the humeral component was performed with the method described by Sperling et al.27 (Fig. 4). The component was divided into eight zones and was defined as being at risk if a radiolucent line of >2 mm wide was present in three or more zones.
Cranial migration of the humerus was analyzed according to the method of Torchia et al.28; it was defined as mild if the center of the prosthetic humeral head had translated less than one quarter of its diameter relative to the center of the glenoid component, as moderate if the head had translated between one-quarter and one-half of this diameter, and as severe if the translation of the head was more than half of this diameter.
Operative Technique and Implants
The same unconstrained cemented third-generation anatomical implant (Aequalis Total Shoulder; Tornier, Edina, Minnesota) was used in all cases. The modular humeral stem with a matte surface finish, eccentric head, and variable head inclination was combined with a keeled ultra-high molecular-weight polyethylene glenoid component. A flat-back glenoid component was used in all cases. All humeral and glenoid components were cemented with high-viscosity bone cement (Biomet, Warsaw, Indiana).
The patients were positioned in the beach-chair position. A deltopectoral surgical approach was used as described by Neer et al.29. After tenotomy of the subscapularis tendon and capsular release, the joint was exposed. The biceps tendon was released close to its glenoid attachment and a soft-tissue tenodesis in the bicipital groove was performed in all cases. The intra-articular portion of the tendon was removed. The humeral head was resected in anatomical fashion under direct view. After preparation of the humeral intramedullary canal, the glenoid was reamed with spherical reamers of increasing size. If required, the surgeon attempted to correct the version of the glenoid by reaming, taking care to protect the subchondral bone. Curettage was performed to remove cancellous bone of the glenoid vault after the preparation of the keel slot. After use of jet lavage and sponge drying, the cement was placed into the keel slot with a syringe. The glenoid component was cemented under continuous pressure until the cement was cured. Cementing of the humerus included third-generation cementing techniques with a distal plug, jet lavage, retrograde filling, and continuous pressurization in a standardized fashion. After the placement of the implants, the subscapularis tendon was repaired with three, four, or five nonresorbable tendon-to-tendon sutures. A tear of the supraspinatus tendon was detected in two shoulders and was repaired with transosseous nonabsorbable sutures after implantation of the components. A drain was placed and was removed on the first day postoperatively.
To protect the reconstructed subscapularis tendon, the arm was placed in a sling with a 20° shoulder abduction pillow in 20° of internal rotation for four weeks. For the first six weeks postoperatively, the shoulders were mobilized with active-assisted exercises, with the aid of a physiotherapist, consisting of flexion and abduction of 60° and 0° of external rotation without aggressive stretching or strengthening. The pillow was removed and placed again after the exercises. Free range of motion was allowed thereafter.
Statistical Methods
The empirical distribution of continuous end points is reported by tabulation of the mean, standard deviation, minimum, and maximum. To detect variation over time the paired t test was used, assessing paired preoperative and postoperative Constant scores, pain, strength, activity, and mobility (all according to the Constant scores) as well as shoulder joint motion (flexion, abduction, and external and internal rotation). A two-tailed p value of <0.05 was considered significant. Survivorships of the implants were calculated with use of the Kaplan-Meier method, including 95% confidence intervals.
Source of Funding
There was no external funding source for this study.
Clinical Results
At the most recent follow-up examination (at a minimum of ten years after surgery), the patient was very satisfied with the outcome of the shoulder replacement in twenty-five cases, satisfied in eleven, and disappointed in three.
The mean Constant score improved from 27 points (range, 11 to 54 points) preoperatively to 61 points (range, 21 to 86 points) postoperatively (p < 0.0001). The score adjusted for age and sex increased from 35% (range, 15% to 80%) preoperatively to 79% (range, 23% to 122%) postoperatively (p < 0.0001). Significant differences were also found for pain relief, strength, activity, mobility, shoulder flexion, abduction, and external and internal rotation (p < 0.0001). The findings of the preoperative and postoperative clinical examination are presented in Table I.
Imaging Results
The mean radiolucent line score for the glenoid components was 8.8 points (range, 2 to 18 points; standard deviation [SD], 6.8 points) on the anteroposterior view and 8.8 points (range, 2 to 18 points; SD, 7 points) on the axial view. As seen on the anteroposterior view, twenty-two shoulders had a score of <7 points, four shoulders had a score of 7 to 12 points, and thirteen shoulders had a loose glenoid component (>12 points). As seen on the axillary view, twenty-five shoulders had a score of <7 points, two shoulders had a score of 7 to 12 points, and twelve shoulders had a loose glenoid component (>12 points). Altogether, twenty-one shoulders (54%) were not at risk for loosening, four (10%) were at risk, and fourteen (36%) had loosening. Of the fourteen loose glenoid components, thirteen were seen to have a tilt (only on the anteroposterior view in five, only on the axial view in five, and on both views in three; Figs. 5, 6, and 7) while one only had a radiolucent line score of >12 points without tilting as seen on the axial view. The distribution of radiolucent lines on the anteroposterior and axial views among the different types of glenoids according to the system of Walch et al.23 is summarized in the Appendix. The mean radiolucent line score for the B2 and C glenoids (n = 12) was significantly greater, on both views, than that for the A2 and B1 glenoids (n = 27) (11.2 versus 7.7 points on the anteroposterior view and 11.4 versus 7.5 points on the axial view; p < 0.05).
There was no significant difference in the clinical outcomes (Constant score and subscores, flexion, abduction, internal and external rotation, strength, and pain) between patients with and those without radiographic loosening of the glenoid component. Age at the time of surgery did not influence the occurrence of glenoid loosening during the follow-up period.
Cranial migration of the humeral component was detected in twenty-seven shoulders (69%). The migration was mild in ten cases (37%), moderate in twelve (44%), and severe in five (19%). There was a trend toward inferior outcomes in cases with severe migration but no significant differences between groups (p > 0.05) (see Appendix). There were no cases of humeral component loosening. According to the scoring system of Sperling et al.27, none of the humeral shafts was at risk.
The glenoid component was not at risk on either radiographic view in any of the seven patients excluded from clinical and radiographic analysis because the duration of follow-up was less than ten years (mean, thirty-eight months; range, twelve to seventy-eight months). The mean radiolucent line score for these patient was 3.1 points (range, 0 to 6 points) on the anteroposterior view and 1.9 points (range, 0 to 4 points) on the axial view. No humeral component was at risk, and no patient had revision surgery.
Survivorship Analysis
Survivorship of the implant (humeral and glenoid components) with revision of the prosthesis as the end point was 100% after thirteen years. Survivorship with any revision as the end point was 98% (95% confidence interval [CI], 83.7% to 98.9%) after thirteen years. Survivorship of the glenoid component with a radiolucent line score >12 points or tilt as the end point was 89% (95% CI, 78.5% to 96.6%) after ten years, 53% (95% CI, 44.6% to 79.5%) after eleven years, and 48% (95% CI, 16.8% to 70.4%) after thirteen years (Fig. 8).
Complications
There were two complications in this cohort, one of which led to surgical revision. In this case, a nontraumatic anterior dislocation of the humeral head was detected at the follow-up examination one year postoperatively. Repair of the subscapularis tendon was not possible, and a pectoralis major tendon transfer was performed. At the most recent follow-up examination (two years postoperatively), the shoulder remained stable. This patient died three years after surgery. In another patient, a brachial plexopathy occurred postoperatively but resolved completely six weeks after the operation. No periprosthetic fractures, infections, or other complications directly related to the surgical procedure were noted.
This study demonstrated satisfactory long-term functional outcomes and high implant survivorship of a third-generation cemented total shoulder system. Although no glenoid component required revision, we found a high rate of radiographic loosening. The results of our study are comparable with those of two recently published multicenter studies12,13. Young et al. reported the results of 263 total shoulder replacements after a medium to long-term follow-up (at a mean of 124 months [range, sixty-one to 219 months]) in patients with osteoarthritis12. The same anatomical third-generation implant that we used in our cohort, including a keeled flat-back glenoid, was used in all of their patients. They reported significant improvements in the Constant score and clinical findings. The survivorship with glenoid component revision as the end point was 94.5% at ten years and 79.4% at fifteen years. The survivorship with radiographic loosening as the end point of 80.3% at ten years and 33.6% at fifteen years was comparable with our findings. Young et al. found highly significant associations between radiographic loosening and clinical data such as pain, activity, shoulder motion, and Constant score. In our cohort, we found no correlation between radiographic loosening and clinical data, possibly because of the smaller number of patients in our study. In our opinion, loosening of a glenoid component may lead to pain, although this is not necessarily the case. Young et al. also found that younger patient age and curettage of the glenoid vault correlated with loosening of the glenoid component. As is usual in a multicenter study, surgical techniques varied among the centers.
Another risk factor for glenoid loosening was described in a multicenter study by Walch et al.13. They analyzed 333 total shoulder replacements followed for a mean of ninety months (range, sixty-one to 152 months) and found that excessive reaming of the glenoid into the subchondral bone correlated with higher rates of radiographic loosening. In our cohort, we did not perform excessive reaming. In some cases, the surgeon tried to correct the glenoid version as accurately as possible without destroying the subchondral bone. The humeral components used by Walch et al. were equivalent to those in our investigation. However, they used a convex-back cemented glenoid component (Aequalis, Tornier) and reported survivorship rates of 99.7% at five years and 98.3% at ten years with revision of the glenoid component as end point. The results were less promising when radiographic loosening of the glenoid component was defined as the end point: the survival rate was 99.7% at five years but only 51.5% at ten years.
It has been reported that curettage of the keel slot as compared with bone compaction leads to a significantly higher rate of radiolucent lines immediately after surgery and glenoid component loosening during the follow-up period30. This could be an explanation for the poor radiographic survivorship in the present study. In this series, we performed minimal reaming to protect the subchondral bone but we did perform curettage in all cases. Several other factors that we did not study, such as the surgical and/or cementing technique, the postoperative rehabilitation protocol, or patient-related factors, may have had an influence on the clinical and radiographic outcomes. This could explain why some of our results differ from the findings of the French multicenter group12 although the same implant was used. Szabo et al. found fewer radiolucent lines with curved components than with flat-back glenoids after two years of follow-up31, although no difference was detected in the same prospectively followed series after ten years32. We are not able to highlight one component design over the other on the basis of our study.
In another study, Khan et al. used the same keeled, flat-back implant as we used and followed their patients for a minimum of ten years20. Significant improvements in the Constant score and American Shoulder and Elbow Surgeons score were found postoperatively. Twenty-four patients were evaluated radiographically with the method described by Lazarus et al.33. Two glenoid components were revised and only two were loose (17%). In nine (38%) of the cases, no radiolucent lines were found after this long follow-up period. In another study, of young and middle-aged patients followed for a mean of seven years after replacement with the Aequalis total shoulder prosthesis, 52% of the patients had no evidence of radiolucent lines and loosening of the glenoid component was not detected21.
Radiographs in two planes are recommended to analyze implant seating and potential complications. In many studies of total shoulder arthroplasty, the authors analyzed radiolucency exclusively on anteroposterior views3,12,13,20,21,33-35. We have demonstrated that loosening or tilting of the glenoid component may be found on the anteroposterior view only, on the axial view only, or on both views. Nyffeler et al.36 concluded that glenoid component orientation cannot be determined accurately on standard axillary lateral radiographs; CT scans were recommended to assess the correct orientation. Although axillary radiographs are not as accurate, we were able to detect, on these views, loosening that was not seen on the anteroposterior views. CT scans could not be used to evaluate our patients because CT scanning was not approved by the ethics committee of our university. For future studies, we recommend a radiographic examination protocol that includes analysis of radiolucent lines and tilting on both anteroposterior and axillary views.
Despite the poor radiographic results in our cohort, only one patient underwent revision surgery, because of a nontraumatic anterior shoulder dislocation, and significant functional improvement was maintained throughout the cohort over a follow-up period exceeding ten years.
A survivorship of 100% was also found in a study previously reported by our group3. The rate of radiographic loosening of the glenoid in that study was 9% after five years and 33% after nine years. In 86% of the ninety-six treated cases, radiolucent lines were detected in more than three zones.
In a study of 972 all-polyethylene cemented glenoid components, Fox et al. reported overall survival rates of 95% at ten years and 92% at fifteen years (87% for the Cofield-1 and 94% for the Neer-II prosthesis) with revision as the end point19. Good survivorship was found for the Cofield-2 keeled glenoid (99% at five years and 94% at ten years) and pegged glenoid (99% at five years). Although the results of the pegged and keeled implants were similar at five years, it must be pointed out that no long-term data are available for cemented pegged glenoid components, to our knowledge.
In contrast to the high rates of loosening of the glenoid component in our and other investigations, we found no clinical or radiographic signs of loosening of the humeral component. In other studies, the reported rates of loosening of the humeral components were also very low3,20,21, which emphasizes the fact that the durability and survivorship of the glenoid component are the key problem, especially in the long term. Although loosening of the humeral component was not found, we detected a high rate of cranial migration of that component. Moderate or severe cranial migration of the humeral component has been described as a sign of rotator cuff failure12,37,38. The rate of humeral migration in this study is higher than that in other reports. Young et al.12 found moderate or severe migration in 27% of their cases. We could not identify a significant correlation between humeral cranial migration and the occurrence of glenoid loosening, perhaps because of our relatively small number of patients. There was a trend for decreased Constant scores for patients with more severe cranial migration, but the difference was not significant. A limitation of this study is the relatively low number of patients compared with the numbers in some of the above-mentioned multicenter studies. On the other hand, all patients were treated in the same standardized fashion by or under the supervision of the same shoulder surgeon. The clinical and radiographic examination protocol was performed uniformly before and after the operation. Nevertheless, the small number of patients may limit the statistical analysis, particularly regarding failure to demonstrate significant correlations. With the implants, cementation, and surgical technique that were utilized in this group of patients with primary glenohumeral osteoarthritis, radiographic loosening of the glenoid component and signs of rotator cuff deficiency were very common at ten to fifteen years postoperatively.
Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any aspect of this work. None of the authors, or their institution(s), have had any financial relationship, in the thirty-six months prior to submission of this work, with any entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. Also, no author has had any other relationships, or has engaged in any other activities, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.