A computerized search was performed to identify all patients who had undergone a total shoulder arthroplasty with a cemented Neer-II metal-backed component (Kirschner Medical, Fairlawn, New Jersey) at our institution between 1985 and 1991. In order to be included in the study, patients had to have a diagnosis of osteoarthritis and the surgery had to have been performed by one of three surgeons with a specific focus on upper extremity reconstructive surgery. Patients with a rotator cuff tear, a history of shoulder trauma, or a diagnosis other than osteoarthritis were excluded from the study. One hundred total shoulder replacements in ninety-four patients met these inclusion criteria. Five shoulders with less than two years of follow-up were excluded from the analysis of the clinical results. They included one shoulder in a patient who died less than two years after surgery, one in a patient with incomplete follow-up data, and three in patients who were lost to follow-up. Thus, ninety-five total shoulder replacements in eighty-nine patients who had been followed for a minimum of two years (mean, 10.8 years) or until revision were included in the clinical analysis in this retrospective review. All 100 total shoulder replacements were included in the Kaplan-Meier survival analysis, with a reoperation as the end point.
Patients with all other diagnoses and other glenoid components were excluded from the study. In addition, patients who required structural bone-grafting of the glenoid were excluded from the study. During this time period, a total of 274 total shoulder arthroplasties were performed for the diagnosis of osteoarthritis at our institution. The decision to use the Neer metal-backed component was made on the basis of the preference of the surgeon. The surgeon chose the metal-backed cemented component on the basis of patient size. Small patients received an all-polyethylene component. Larger patients received the cemented metal-backed component because the keel and metal rim are larger than the all-polyethylene component. There were no other factors in the decision, as it was not clear there was any advantage of one component over the other.
The mean age (and standard deviation) of the patients was 68 ± 8 years. The study group included sixty-one men and thirty-three women. Of the 100 shoulders, three had undergone a previous procedure. These included an open débridement for arthritis (one shoulder) and distal clavicular excision (one shoulder). A third patient previously underwent a biceps tenodesis and rotator cuff repair. At the time of shoulder arthroplasty, the rotator cuff was intact.
For the 100 shoulders that met the inclusion criteria, the length of clinical follow-up was a mean (and standard deviation) of 10.4 ± 5.1 years (range, one to 19.6 years).
Patients were identified with the use of a computerized database that contained the files of all patients who had had a joint arthroplasty at our institution. All patients who had undergone a joint arthroplasty at our institution were asked to return for an examination, interview, and radiographic evaluation at regular (five-year) follow-up intervals. Those who were unable to return for evaluation were sent a validated questionnaire to evaluate their functional status and satisfaction6. In addition, patients were requested to have a local orthopaedic surgeon send the results of a clinical examination, complete the questionnaire, and send recent radiographs. The most recent follow-up evaluation was by questionnaire for seventy-one shoulders and occurred in the clinic of the operating surgeon for twenty-four shoulders.
Operative Data
A deltopectoral approach was used in eighty-six shoulders, and an anteromedial approach was used in fourteen. A tuberosity reconstruction was not performed in any shoulder. All patients received a Neer-II metal-backed, keeled glenoid component inserted with cement. The glenoid vault was prepared by hand. The cement was pressurized manually. Forty Neer humeral components and sixty Richards humeral components (Smith and Nephew, Memphis, Tennessee) were used. The humeral components were press-fit in ninety-six shoulders and were cemented in the other four shoulders. The humeral component was implanted in a mean of 30° of retroversion. The humeral component was monoblock in both systems. The forearm was used as a guide to determine the amount of retroversion for the Neer humeral components. An intramedullary guide and an extramedullary rod based off the forearm were used for the Richards humeral components. Bone graft from the humeral head was placed in cysts in the proximal aspect of the humerus in four shoulders. Cancellous bone from the head was placed in contained, small glenoid cysts in six shoulders. The three surgeons performed forty-eight, forty-one, and eleven operations, respectively.
Clinical Review
The clinical assessment of all patients who had undergone shoulder surgery at our institution is recorded with use of a standard shoulder-analysis sheet6. Pain was graded on a numerical scale as previously described by Neer et al.7 and Cofield8, with 1 point indicating no pain; 2 points, slight pain; 3 points, pain after unusual activities; 4 points, moderate pain; and 5 points, severe pain. Range-of-motion assessment was performed by the operating surgeon without the use of a goniometer. This included active elevation in the scapular plane, passive external rotation with the arm at the side, and active internal rotation graded according to the posterior spinal level reached with the thumb.
Functional outcome was graded according to a modified Neer result-rating system9,10. The result was considered excellent if the patient had no or slight pain, was satisfied with the procedure, and had at least 140° of active elevation and at least 45° of external rotation. The result was considered satisfactory if the patient had no, slight, or moderate pain with vigorous activities only, was satisfied with the procedure, and had at least 90° of active elevation and at least 20° of external rotation. The result was considered unsatisfactory if any of these criteria were not met or if the patient underwent an additional operation.
Radiographic Analysis
Radiographs were available for eighty-three shoulders with a total shoulder replacement that had been followed for a minimum of two years or until undergoing a revision (Figs. 1-A, 1-B, and 1-C). The mean duration for radiographic follow-up was 8.5 ± 4.7 years. The preoperative, initial postoperative, and most recent radiographs were evaluated by two shoulder surgeons who were blinded to the clinical results, and a consensus was reached. Three projections were used for radiographic analysis: an axillary radiograph, a 40° posterior oblique radiograph with external rotation of the humerus, and a 40° posterior oblique radiograph with internal rotation of the humerus. Radiographs were reviewed to determine the presence of glenoid erosion, glenohumeral subluxation, periprosthetic lucency, and component shift in position.
Glenoid and humeral periprosthetic lucency was rated as grade 0 if there was no line, grade 1 if the line was 1 mm and incomplete, grade 2 if the line was 1 mm thick and complete, grade 3 if the line was 1.5 mm and incomplete, grade 4 if the line was 1.5 mm and complete, or grade 5 if the line was 2 mm and completely surrounding the component9. Glenohumeral subluxation was evaluated with regard to the direction (anterior, posterior, superior, or inferior) and the amount of translation of the center of the prosthetic head relative to the center of the glenoid component. Translation was graded as none, mild (<25%), moderate (25% to 50%), or severe (>50%)9. Erosion of the glenoid was graded as none if no erosion was observed, mild if there was erosion of the glenoid cortical bone, moderate if there was medialization of the glenoid subchondral bone with hemispheric conforming deformation of the glenoid, or severe if there was hemispheric deformation of the glenoid and central bone loss within 5 mm of the coracoid base9. Glenoid component tilt or migration was rated as either present or absent. Humeral component subsidence and tilt were rated as either present or absent.
Statistical Methods
Implant survival subsequent to the date of primary shoulder arthroplasty was estimated with the Kaplan-Meier method. The 95% confidence intervals were estimated with bootstrap confidence intervals (BCa type), which adjust for correlated data (two shoulders in the patients with bilateral total shoulder arthroplasty). Since only five shoulders were revised, no risk factor assessment was done.
Descriptive results are presented as the number and percentage for categorical variables and as the mean and standard deviation for continuous variables. A paired t test for improvement was used to compare preoperative and postoperative abduction, internal rotation, external rotation, and pain measurements. A 95% confidence interval for the mean improvement is given in Table I. Improvements in abduction, external rotation, and pain measurements were compared for the variables of sex, preoperative subluxation (none or mild compared with moderate or severe), Neer score (unsatisfactory compared with satisfactory or excellent), postoperative subluxation (none or mild compared with moderate or severe), postoperative humeral lucency (none compared with any), and postoperative glenoid lucency (none compared with any) with use of a two-sample t test with unequal variances. A chi-square test was used to compare postoperative subluxation, postoperative glenoid lucency, and postoperative humeral lucency with an unsatisfactory Neer score (categorized as unsatisfactory compared with satisfactory or excellent). When a low expected cell count of less than five was observed, a Fisher exact test was used.
The alpha level was set at 0.05 for significance.
Source of Funding
There was no outside funding source for this study.
Complications and Revisions
Three patients had a neurapraxia of the musculocutaneous nerve postoperatively, and it resolved without surgical treatment. Five shoulders underwent revision surgery (Fig. 2). The estimated survival rate (and 95% confidence interval) for total shoulder arthroplasty components was 98% (95% to 100%) at five years, 97% (93% to 100%) at ten years, and 93% (87% to 99%) at fifteen years (Fig. 2). Two of the five shoulders that underwent revision had the reoperation within the first two years after the index arthroplasty. Of those two shoulders, one had instability within one year and had the humeral component revised for aseptic loosening and the other one underwent revision total shoulder arthroplasty for aseptic loosening of the glenoid and humeral components. Of the other three shoulders, one underwent revision for polyethylene wear eight years after the index arthroplasty because of pain. Another shoulder that had a nonunion of a humeral fracture distal to the humeral stem, which was due to an injury twelve years after the index operation, was revised with a long-stem total shoulder replacement. The remaining shoulder underwent removal of the glenoid component and revision to hemiarthroplasty thirteen years after total shoulder arthroplasty because of aseptic loosening of the humeral and glenoid components.
Pain
Total shoulder arthroplasty was associated with a mean improvement in the pain score of 2.7 ± 1.2 points (p < 0.001). The score had improved from a mean of 4.6 ± 0.5 points preoperatively to 1.9 ± 1.1 points postoperatively (Table I). No significant association was detected between improved pain relief and either sex or the presence of moderate or severe preoperative subluxation (p > 0.05). The mean difference with regard to the improvement in the pain score for patients with no or mild postoperative subluxation compared with those with moderate or severe subluxation was -0.2 point (95% confidence interval, -0.9 to 0.5 points) (p = 0.54). No significant association was detected between improvement in the pain score and humeral lucency or glenoid lucency (see Appendix).
Range of Motion
Active abduction improved by a mean of 54° ± 51°, from a mean of 92° ± 35° preoperatively to a mean of 146° ± 36° postoperatively (p < 0.001). No significant association was detected between an improvement in active abduction and sex or preoperative subluxation. In a comparison of preoperative and postoperative radiographs, patients who demonstrated subluxation on a postoperative radiograph had significantly less improvement in abduction (27° ± 53°) than did patients without subluxation (65° ± 46°) (p = 0.007). No significant association was detected between an improvement in active abduction and humeral lucency or glenoid lucency (see Appendix).
External rotation improved by a mean of 34° ± 37° (p < 0.001), from a mean of 26° ± 24° to a mean of 60° ± 30° (p < 0.001). No significant association was detected between the improvement in external rotation and sex or preoperative subluxation. In addition, no significant association was detected between an improvement in external rotation and postoperative subluxation, humeral lucency, or glenoid lucency (see Appendix).
On the average, active internal rotation improved from the L5 to the L2 level. Patients who had postoperative subluxation had a mean active internal rotation to L4, while patients who had not had postoperative subluxation had active internal rotation to L1.
Results According to the Neer Rating System
Forty-seven (49%) of ninety-five shoulders had an excellent result; twenty-seven (28%), a satisfactory result; and twenty-one (22%), an unsatisfactory result. The most common contributing reasons for an unsatisfactory result were lack of external rotation (fourteen shoulders), lack of active abduction (nine), pain (eleven), and revision surgery (five). No association was detected between an unsatisfactory result (according to the Neer score) and sex (six [21%] of twenty-nine women compared with twelve [22%] of fifty-four men; p = 0.87) or preoperative joint subluxation (two of ten shoulders with subluxation compared with fourteen [26%] of fifty-four without subluxation; p = 1.0). In addition, we were unable to detect a significant association between an unsatisfactory result and the presence on the most recent images of any glenoid lucency (fourteen [20%] of sixty-nine shoulders with a glenoid lucency compared with four of thirteen shoulders without a lucency; p = 0.47) or the presence of any humeral lucency (seven [19%] of thirty-seven shoulders with a humeral lucency compared with eight [40%] of twenty shoulders without lucency; p = 0.60). A weak association was detected between an unsatisfactory result and postoperative subluxation (eight [40%] of twenty shoulders with moderate or severe postoperative subluxation compared with ten [16%] of sixty-one shoulders with no or mild postoperative subluxation; p = 0.06).
Radiographic Analysis
On the most recent postoperative radiographs, mild superior subluxation was seen in sixteen shoulders; moderate superior subluxation, in seven; and severe superior subluxation, in four. Mild posterior subluxation was seen in three shoulders, and moderate posterior subluxation was seen in one shoulder. There was mild anterior subluxation in four shoulders. Two shoulders had mild superior subluxation combined with mild anterior subluxation, one had moderate superior subluxation combined with mild anterior subluxation, and one had mild superior subluxation combined with moderate anterior subluxation. We were unable to detect a significant association between the presence of postoperative subluxation and sex.
On the immediate postoperative radiographs, glenoid periprosthetic lucency was grade 1 (1 mm thick and incomplete) in thirty-three shoulders, grade 2 (1 mm thick and complete) in two shoulders, and grade 3 (1.5 mm thick and incomplete) in one shoulder. At the time of the most recent follow-up, glenoid periprosthetic lucency was present in sixty-nine shoulders. The lucency was grade 1 in thirty-two shoulders, grade 2 in eleven shoulders, grade 3 in seven shoulders, grade 4 (1.5 mm thick and complete) in four shoulders, and grade 5 (2 mm thick and completely surrounding the component) in fifteen shoulders. From the immediate postoperative to the most recent radiographs, twenty-eight remained the same grade, fifty-three had progressed, and two had improved. In the fifty-three shoulders that progressed over time, twenty-four had progressed one grade; six, two grades; five, three grades; twelve, four grades; and six, five grades. A glenoid component shift in position was present in twenty-nine shoulders. We were unable to detect a significant association between the development of glenoid lucency and sex or the presence of preoperative glenoid erosion.
Humeral periprosthetic lucency was not present on the initial postoperative radiographs among any of the eighty-three shoulders. However, at the time of the most recent follow-up, lucencies were present in thirty-seven shoulders. The lucency was grade 1 in nineteen shoulders, grade 2 in two shoulders, grade 3 in nine shoulders, grade 4 in three shoulders, and grade 5 in four shoulders. We were unable to detect a significant association between the development of humeral lucency and sex. A shift or subsidence of the humeral component was present in fifteen shoulders. No significant association was detected between a shift and sex (seven [13%] of fifty-three shoulders in men had a shift compared with eight [28%] of twenty-nine shoulders in women; p = 0.14) nor between shift and age (two of seventeen shoulders in patients who were less than sixty years old had a shift compared with thirteen [20%] of sixty-five shoulders in patients who were at least sixty years old; p = 0.73). No association was detected between shift and length of follow-up (mean [and standard deviation] of 8.6 ± 3.6 years for shoulders with a shift and 8.4 ± 4.9 years for shoulders without a shift; p = 0.85).
In the past, little information has been available to guide clinical decision-making with regard to the use of cemented metal-backed glenoid components in total shoulder arthroplasty. In addition, there have been few long-term studies of shoulder arthroplasty in patients of all ages. A review of forty-one series with 1858 total shoulder arthroplasties revealed a mean follow-up of only 3.5 years10. Twenty-one reports had a minimum follow-up of two years, and only five studies had a mean follow-up of five years or longer10. To our knowledge, the present study has the longest duration of follow-up of any reported series of shoulders managed with cemented metal-backed glenoid components.
Overall, there was significant pain relief as well as improvement in active abduction and external rotation among patients with osteoarthritis treated with a total shoulder arthroplasty with a cemented metal-backed glenoid component. In addition, postoperative subluxation was associated with less improvement in active abduction and less internal rotation. The unsatisfactory results according to the modified Neer rating system were usually a result of limited range of motion, especially external rotation and abduction. Although the prosthesis survival rate was quite high, radiographic analysis revealed a high rate of glenoid periprosthetic lucency that is a concern.
One can compare the results of the Neer metal-backed glenoid component used in this study with the results of the Neer all-polyethylene component for the treatment of osteoarthritis. Torchia et al.11 reported on the long-term results of shoulder arthroplasty with a Neer all-polyethylene glenoid component. The same surgeon performed most of the procedures in that study and in the current report. Among the patients with osteoarthritis treated with an all-polyethylene glenoid component, at the most recent follow-up, the mean active elevation was 143° and the mean external rotation was 55°. Thirty (88%) of thirty-four glenoid components had periprosthetic loosening. The majority of glenoid lucent lines were noted on the initial postoperative radiographs, which may reflect surgical technique rather than failure of the implant. In the current study, the results are very similar, with a mean elevation of 146° and a mean external rotation of 60°. Sixty-nine (83%) of eighty-three shoulders had glenoid periprosthetic lucency.
The present study has several limitations, including the fact that it is a retrospective review with differences in the type of follow-up performed. Three different surgeons performed the shoulder arthroplasty in a heterogeneous group, including some shoulders with a rotator cuff tear. No strict criteria were used to determine when a metal-backed glenoid component should be used. Radiographic follow-up was not available for all patients. While the radiographs that were reviewed were blinded to the reviewers of the clinical outcome of the patients, one of the reviewers was a surgeon who contributed patients to the study. A goniometer was not used to measure range of motion, and all patients were not examined by the same surgeon. A limiting factor with regard to the outcome measurement is that, at the time that the patients underwent the procedure, specific functional outcome measurements, such as the Simple Shoulder Test, were not yet available12,13. Specific outcome measurement values were also not available at the time of the most recent follow-up for all patients.
The data from this study suggest that the clinical outcomes and rate of periprosthetic lucency with a metal-backed glenoid component are not better compared with a cemented all-polyethylene glenoid design. The high rate of periprosthetic glenoid lucency associated with the cemented metal-backed component is concerning and requires additional follow-up and investigation.