All patients who were to have a shoulder arthroplasty between December 1992 and December 1996 were evaluated for inclusion in this study. The indication for all operations was pain in the shoulder due to osteoarthritis that had been unresponsive to medical treatment and that interfered with activities of daily living. The criteria for inclusion in the study were a diagnosis of osteoarthritis, an intact rotator cuff, and a concentric glenoid. A diagnosis other than osteoarthritis was a criterion for exclusion.
The integrity of the rotator cuff was determined intraoperatively with visual inspection and palpation. Radiographic templates were used preoperatively to estimate the concentricity of the glenoid, and the final determination was made with use of a trial implant at the time of the operation. We used a combination of the patient's history and the findings on physical and radiographic examination to make the diagnosis of osteoarthritis16. A history indicative of osteoarthritis included an absence of major trauma to the shoulder, no previous operation involving the shoulder, and no other known cause of secondary degenerative joint disease. Physical findings indicative of osteoarthritis included painful, limited glenohumeral motion, and radiographic findings included narrowing of the joint space, periarticular sclerosis, periarticular osteophytes, and the absence of any other pathological condition. In order to increase the uniformity of the study group, the criteria for inclusion required that, at the time of the operation, the glenoid demonstrate degeneration of the cartilage of the articular surface and a concentric osseous surface with no flattening or bone loss. Also, uneven bone loss was determined on preoperative plain axillary radiographs, computerized tomographic scanning, or magnetic resonance imaging at the time of the operation. Plain radiographs were made for all patients; magnetic resonance images, for fourteen; and computerized tomographic scans, for two.
We conducted a power analysis to determine the sample size necessary to demonstrate a difference resulting from the use of a glenoid component in shoulder replacement operations. Amstutz et al.1 assessed the difference between patients who had received two different types of glenoid components. The effect size of the type of glenoid component was calculated to be 0.30. On the basis of that effect size, an effect size of 0.60 was estimated for this study. Calculation of the necessary sample size, with use of the effect size of 0.60, demonstrated that, with a one-tailed or unidirectional analysis at a desired power of 0.80, approximately thirty-five subjects per group would be required8.
One hundred and thirty-four consecutive shoulders in 124 patients were evaluated for inclusion in this study. Seventy-one patients (seventy-seven shoulders) met one criterion for exclusion and therefore were not enrolled in the study. Of the patients who were excluded, twenty-six had sustained acute or chronic trauma to the shoulder, eleven had rheumatoid arthritis, ten had osteoarthritis with uneven bone loss, nine had avascular necrosis, six had osteoarthritis following an episode of instability, four had a tear of a rotator cuff tendon that was identified at the time of the operation, three had cuff-tear arthropathy, and two had a non-union of the proximal part of the humerus. Fifty-three patients (fifty-seven shoulders) who met the criteria for inclusion were asked to participate in the study, and fifty-one patients (fifty-five shoulders) agreed. The two patients who refused to participate had had a previous total shoulder arthroplasty on the contralateral side and requested a similar procedure. Four patients were lost to follow-up. This left fifty-one shoulders in forty-seven patients for analysis.
Confidentiality and Consent
In order to maintain patient confidentiality throughout the investigation, each patient was assigned a unique identification number. The patients' names and identification numbers were kept separately, so that the analysis of the data was performed without any knowledge of the patient's identity.
All patients provided written informed consent; they stated that they comprehended the purpose of the study as well as the potential risks and benefits of the operation. The patients were informed that they had the right to alternative forms of operative and nonoperative care other than that offered in the study. They also were advised that they had the right not to participate in the study even if they selected the type of care offered in the study. The patient's understanding of these rights was documented in writing. The study was approved by our institutional review board.
Preoperative Assessment
In order to allow comparison of our findings with those that have previously been reported in the literature, sufficient data were collected to allow scoring according to the University of California at Los Angeles shoulder scale9 and the American Shoulder and Elbow Surgeons shoulder index22. The University of California at Los Angeles shoulder scale is a 35-point scale, with a maximum of 10 points assigned for pain; 10 points, for function; and 5 points each, for motion, strength, and patient satisfaction. The American Shoulder and Elbow Surgeons shoulder index is calculated with use of scores that the patient assigns for pain on a 10-point visual analog scale, with 0 representing no pain and 10 representing unbearable pain, and with use of scores that the patient assigns for ten self-assessed activities of daily living. The pain score on the visual analog scale is subtracted from ten and the resulting number is multiplied by five, for a maximum of 50 points. The ten activities of daily living are scored from 0 to 3 points, for a maximum of 30 points. This number then is multiplied by five-thirds, for a possible total of 50 points; when this score is added to the pain score, 100 points are possible. All patients completed self-assessment questionnaires to document their levels of satisfaction and function.
Active ranges of motion were measured, including elevation, abduction, external rotation in abduction, and internal rotation behind the back. Passive elevation and external rotation (with the arm adducted) were measured by the examiner with use of a handheld goniometer and were recorded to the nearest 5 degrees. Internal rotation was recorded as the highest vertebral level reached by the patient's extended thumb.
We recorded all measurements during the initial physical examination and during subsequent visits to the clinic. No attempt was made to increase the precision of the measurement with use of such techniques as blinding of the examiner, calculation of test-retest validity or of interobserver and intraobserver reliability, instruction of the patient before measurements were obtained, or warm-up or exercise before the evaluation. Elevation strength was measured with the arm elevated 90 degrees in the scapular plane and internally rotated. Resistance against the examiner was graded from 0 to 5, with 0 representing no strength and 5, normal strength. We did not use injections of anesthetic before examining the shoulder for strength, so we were unable to quantify how much of the loss of strength was due to pain.
Radiographic Findings
Routine radiographs included an anteroposterior radiograph and an axillary lateral radiograph. No fluoroscopic examinations were performed. Radiographic evidence of loosening of the glenoid component included the presence or progression of radiolucent lines and fracture or migration of the component. All glenoid components were non-metal-backed, had a keel, and were inserted with cement. Radiographic evidence of loosening of the press-fit humeral stem included the presence or progression of radiolucent lines, widening of the humeral canal, osteolysis, and migration of the stem. All humeral stems were inserted without cement. All patients received the same type of prosthesis (Global; DePuy, Warsaw, Indiana), and all operations were performed by or under the direct supervision of one of us (G. M. G.).
Randomization was accomplished with use of a random-numbers list generated with Microsoft Excel (Redmond, Washington) and kept by the operating-room nurse. After completion of the humeral osteotomy, anterior and inferior capsular release, mobilization of the subscapularis, and exposure of the glenoid, the circulating nurse reviewed the random-numbers list. Patients who had been assigned an odd number were subsequently managed with a hemiarthroplasty, and those who had been assigned an even number were managed with a total shoulder arthroplasty.
All patients were managed with the same postoperative regimen, including administration of antibiotics and physical therapy. An active range of motion was allowed beginning four days after the operation. The physical therapist instructed the patients in a home-exercise program, according to a protocol provided by the surgeon10.
The patients were evaluated at two weeks, six weeks, three months, six months, and one year after the operation and yearly thereafter. At each annual visit to the clinic, before the examination, the patients completed self-assessment forms to allow tabulation of the shoulder scores described earlier.
Statistical Methods
The initial screening of the data was accomplished with use of scatterplots, histograms, and frequency tables for all variables. Additional diagnostics were performed on any potential outliers with use of regression diagnostics and studentized residuals. Violations of linearity, homoscedasticity, and independence were assessed on the scatterplots12. For the descriptive analysis, the total number of subjects who fit the criteria for inclusion was summed and differentiated according to group (hemiarthroplasty or total shoulder arthroplasty). The mean age of the subjects and their gender distribution also were differentiated according to group. Mean scores for the two outcome instruments and their preoperative and postoperative subscores were determined.
Independent t tests were performed in order to determine whether there was a preoperative difference between the two randomly assigned groups with regard to gender, age, or total scores according to the American Shoulder and Elbow Surgeons and the University of California at Los Angeles systems. The appropriate Bonferroni adjustment of the significance level was performed to prevent the increased possibility of a type-I error. The alpha value was set at p = 0.0125 for each comparison.
Independent-samples t tests were calculated to evaluate postoperative differences, with regard to the American Shoulder and Elbow Surgeons shoulder index, the University of California at Los Angeles shoulder score, and the subscales for each of the two shoulder-scoring systems, between the patients who had had a hemiarthroplasty and those who had had a total shoulder arthroplasty. A total of nine comparisons were made; therefore, the significance level was adjusted to p = 0.006 for each comparison.
A stepwise regression analysis was performed to assess the potential relationships between gender, age, treatment group, preoperative American Shoulder and Elbow Surgeons shoulder index, and preoperative University of California at Los Angeles shoulder score and the corresponding postoperative scores. The alpha value was set at p = 0.05 for each analysis, with the highest predictor being entered into the equation first. The other predictors were entered only if they contributed significantly more to the equation than the first predictor alone. For all calculations, standard statistical software (SPSS version 9.0; Chicago, Illinois) was used to analyze the data.
Initial Screening of the Data
Initial screening and analysis of the data revealed one patient as a possible outlier. The studentized residuals for both postoperative scores for this patient were greater than 3.0. According to Kleinbaum et al.13, a studentized residual of greater than 3.0 is highly suspicious for the presence of an outlier. However, additional testing of residuals did not support the removal of this patient from any subsequent analysis of the data. The information on this patient therefore was included in all additional analyses.
Demographic Data
Twenty-four shoulders in twenty-two patients (thirteen men [two of whom had a bilateral procedure] and nine women) had a hemiarthroplasty, and twenty-seven shoulders in twenty-five patients (fifteen men and ten women [two of whom had a bilateral procedure]) had a total shoulder arthroplasty. The mean age (with standard deviation) of the patients who had a hemiarthroplasty was 64.6 ± 6.3 years (range, forty-five to seventy-eight years) compared with 65.3 ± 8.4 years (range, fifty to eighty-six years) for the patients who had a total shoulder arthroplasty. Of the four patients who had a bilateral procedure, two had a bilateral total shoulder arthroplasty, one had a bilateral hemiarthroplasty, and one had a total shoulder arthroplasty on one side and a hemiarthroplasty on the other. The mean duration of follow-up was thirty-four months (range, twenty-seven to seventy months) after the hemiarthroplasties, and it was thirty-six months (range, twenty-four to seventy-two months) after the total shoulder arthroplasties.
Preoperative Shoulder Scores and Range of Motion
Significant differences were not observed between the two groups with regard to the preoperative American Shoulder and Elbow Surgeons shoulder index or the University of California at Los Angeles shoulder score (p > 0.0125) (Table I). The preoperative range of motion for the patients who had a hemiarthroplasty averaged 89 degrees (range, 45 to 150 degrees) of elevation, 34 degrees (range, 0 to 70 degrees) of external rotation, and internal rotation to the gluteal fold (range, the greater trochanter to the third lumbar spinous process). The preoperative range of motion for the patients who had a total shoulder arthroplasty averaged 86 degrees (range, 45 to 140 degrees) of elevation, 36 degrees (range, 0 to 65 degrees) of external rotation, and internal rotation to the gluteal fold (range, the greater trochanter to the twelfth thoracic spinous process). These differences were not significant (p > 0.0125).
Operative Findings
The humeral stem was press-fit in all shoulders. The mean operative time was sixty-three minutes (range, forty-five to ninety minutes) for the hemiarthroplasties and ninety-eight minutes (range, sixty to 132 minutes) for the total shoulder arthroplasties (p = 0.01). The mean intraoperative blood loss was 150 milliliters (range, 100 to 250 milliliters) for the patients who had a hemiarthroplasty and 300 milliliters (range, 150 to 600 milliliters) for those who had a total shoulder arthroplasty (p = 0.01). No transfusions were given to any patient in this study.
Complications
There were no infections, dislocations, or neurological complications. The humeral head did not dissociate from the stem in any patient.
Postoperative Scores (Tables II and III)
Since multiple comparisons were used with the same data set, the appropriate Bonferroni adjustment to the original alpha level of 0.05 was made to prevent an increased risk of a type-I error. Independent t tests were used for all postoperative comparisons, with the significance level adjusted to p = 0.006 for each.
Pain: A reduction in pain, as demonstrated with both rating scales, was observed after both procedures. The difference between the preoperative and postoperative pain scores was significant (p < 0.0005) in both operative groups. Pain relief after the total shoulder arthroplasties was significantly greater than that after the hemiarthroplasties (p = 0.002).
Satisfaction: Satisfaction was measured only with the University of California at Los Angeles system. Both procedures led to a significant improvement in satisfaction compared with the preoperative rating (p < 0.0005). The patients who had had a total shoulder arthroplasty demonstrated more satisfaction after the operation than did those who had had a hemiarthroplasty, but, with the numbers available, this difference was not found to be significant (p = 0.082).
Function: Both procedures resulted in a significant improvement in function compared with the preoperative rating (p < 0.0005). The patients who had had a total shoulder arthroplasty rated their function higher than did those who had had a hemiarthroplasty, but, with the numbers available, this difference was not found to be significant (p = 0.097 and p = 0.723 for the University of California at Los Angeles shoulder score and the American Shoulder and Elbow Surgeons shoulder index, respectively).
Range of motion: Both procedures resulted in a significant improvement in the range of motion compared with the preoperative rating (p < 0.0005). With the numbers available, no difference between the two groups was observed with regard to the University of California at Los Angeles range-of-motion score (p = 0.614). The mean range of motion after the hemiarthroplasties was 127 degrees (range, 90 to 150 degrees) of elevation, 61 degrees (range, 30 to 80 degrees) of external rotation, and internal rotation to the fourth lumbar spinous process (range, the greater trochanter to the twelfth thoracic spinous process). The mean range of motion after the total shoulder arthroplasties was 128 degrees (range, 20 to 150 degrees) of elevation, 61 degrees (range, 0 to 90 degrees) of external rotation, and internal rotation to the twelfth thoracic spinous process (range, the buttocks to the seventh thoracic spinous process). Internal rotation was the only range of motion that differed significantly between the two groups (p = 0.003).
Strength: Both procedures resulted in a significant improvement in strength compared with the preoperative rating (p < 0.0005). With the numbers available, no significant difference in the postoperative strength scores was noted between the two groups of patients (p = 0.441).
Shoulder score: Both procedures resulted in a significant improvement in the total shoulder scores, according to both systems, compared with the preoperative ratings (p < 0.0005). However, with the numbers available, no significant difference between the groups was noted with regard to either the postoperative University of California at Los Angeles shoulder score (mean, 23.2 points [range, 10 to 31 points] after the hemiarthroplasties compared with 27.4 points [range, 9 to 34 points] after the total shoulder arthroplasties) or the American Shoulder and Elbow Surgeons shoulder index (mean, 65.2 points [range, 15 to 94 points] after the hemiarthroplasties compared with 77.3 points [range, 3 to 100 points] after the total shoulder arthroplasties).
Regression Analysis and Correlation of Variables
Two stepwise regression analyses were performed, with use of the postoperative total scores for each shoulder scale as the two dependent variables. The independent variables were either the preoperative American Shoulder and Elbow Surgeons shoulder index or the preoperative University of California at Los Angeles shoulder score as well as age, gender, and operative group. With use of the postoperative American Shoulder and Elbow Surgeons shoulder index as the dependent variable, the preoperative index was entered into the equation first, with an r2 value of 0.129 (r = 0.359); gender was entered into the equation next; and the type of operation was entered last. With the three predictor variables, the overall r2 value was 0.403 (r = 0.635). The variable age did not contribute additionally to the regression equation. For the second analysis, with use of the postoperative University of California at Los Angeles shoulder score as the dependent variable, the type of operation was entered into the equation first, with an r2 value of 0.135 (r = 0.367), and gender was entered second, increasing the overall r2 value to 0.215 (r = 0.464). Neither the preoperative shoulder score nor age contributed additionally to the equation.
Analysis of Unsatisfactory Results
A total of six patients (six shoulders) had an unsatisfactory result. Three of these patients did not have a reoperation. Of these three, two (one of whom had had a hemiarthroplasty and one of whom had had a total shoulder arthroplasty) had marked stiffness that was unresponsive to postoperative rehabilitation; both refused to have an additional operation. The third patient, who had had a total shoulder arthroplasty, had severe pain (a score of 8 of 10 on the visual analog scale). This patient had a stable shoulder with full motion and strength. No loosening of the implant was noted on plain radiographs, and no evidence of infection was found on an aspiration arthrogram. The source of the pain in this patient was unclear.
Three male patients who had had a hemiarthroplasty had a reoperation for resurfacing of the glenoid at nineteen, thirty-nine, and forty-eight months after the index procedure. The indication for the operation in each patient was increasing pain and decreasing space between the humeral head and the glenoid as noted on plain radiographs (Figs. 1-A and 1-B). No evidence of loosening of the humeral stem was noted on plain radiographs, and no evidence of infection was found on a preoperative aspiration arthrogram or on culture of specimens obtained during the revision operation. The most recent shoulder score that had been recorded before the revision operation was used as the final follow-up score. Each patient had removal of the modular head, insertion of a glenoid component with cement, and insertion of a new humeral head. In all three patients, a smaller humeral head was needed because the surface of the glenoid component was in a more lateral position than the native glenoid. These three patients noted a decrease in pain (a visual analog score of 8, 8, and 7 preoperatively compared with 4, 3, and 2 postoperatively). However, these results must be considered preliminary, as the mean duration of follow-up after the revision operations was only eighteen months (twelve, seventeen, and twenty-four months).
Radiographic Analysis
Standard anteroposterior and axillary radiographs were made at the time of the initial visit, the first postoperative visit, and each yearly examination. Fluoroscopic examination was not performed.
Adequate postoperative radiographs were available for a comparison of the preoperative and postoperative positions of the component in all twenty-four shoulders that had had a hemiarthroplasty. Progressive narrowing of the joint space was noted in fourteen shoulders (ten of the fifteen shoulders in male patients and four of the nine shoulders in female patients). Seven shoulders (six in male patients and one in a female patient) had increasing pain associated with loss of joint space, and in five of them (all in male patients) the pain was so severe that a revision operation was discussed with the patient. As noted earlier, only three of these patients had a revision operation. No humeral stem had subsided more than two millimeters or was associated with a radiolucent line. No patient had an operation because of symptomatic loosening of the humeral stem.
Adequate postoperative radiographs were available for all twenty-seven shoulders that had had a total shoulder arthroplasty. A radiolucent line with a width of more than two millimeters (measured according to the method of O'Driscoll et al.20) was noted extending around the glenoid component in four shoulders. Loosening, defined as a shift in the position of the component, was noted in two shoulders; however, we did not perform a revision in either patient, as both were asymptomatic.
Hospital Charges
The duration of hospitalization was similar between the two groups. The increased hospital charges for the longer operative time (thirty-five minutes) that was required for the total shoulder arthroplasties averaged $400 (in 1999 dollars). Additional quantifiable charges incurred by these patients included the cost of the glenoid component ($600) and the cost of the bone cement ($177). Thus, the minimum increase in charges for each total shoulder arthroplasty was $1177. The mean charge for the revision hemiarthroplasties was $15,998. In total, the twenty-seven index total shoulder arthroplasties cost $31,779 more than the twenty-four index hemiarthroplasties. However, the total cost of the three revision operations was $47,994.
Osteoarthritis is a common indication for shoulder arthroplasty2,10,11. There is consensus for the use of hemiarthroplasty to treat certain conditions of the shoulder, including inadequate glenoid bone (rheumatoid arthritis), irreparable rotator cuff tears, and avascular necrosis of the humeral head with normal glenoid articular cartilage23. Replacement of the glenoid is indicated in shoulders with normal or repairable rotator cuff tendons, loss of articular cartilage, and incongruent osseous surfaces. However, orthopaedic surgeons commonly encounter shoulders with loss of cartilage, a good rotator cuff, and congruent surfaces. In this situation, the advantages of hemiarthroplasty are that it is a simpler procedure, it costs less, there are no complications related to a glenoid component, and the clinical outcome is good.
Some authors have reported that the results of hemiarthroplasty are inferior to those of total shoulder arthroplasty. Cofield et al.7 and Levine et al.15 found that hemiarthroplasty may be effective for the treatment of glenohumeral arthritis but that the results are not as good as those of total shoulder arthroplasty. Levine et al., in a retrospective review, reported that 74 percent (twenty-three) of thirty-one patients had a satisfactory result after hemiarthroplasty. However, when the results were evaluated according to the degree of wear of the glenoid component, the result was satisfactory in thirteen of fifteen patients who had a concentric glenoid but in only ten of sixteen who had a nonconcentric glenoid. The shoulders with a concentric glenoid that were reported on by Levine et al. can be considered equivalent to the shoulders in the current series. The mean American Shoulder and Elbow Surgeons shoulder index in the series of Levine et al. was 68 points, which is comparable with the shoulder index of 65 points in the current study. If the rating system of Neer et al.19 is applied to the twenty-four shoulders that had a hemiarthroplasty in the current study, twenty had a satisfactory result, which is comparable with the rate of thirteen of fifteen reported by Levine et al. From the same institution, Pollock et al.21 reported that 97 percent (sixty-five) of sixty-seven patients had an excellent or good result after total shoulder arthroplasty, which is comparable with the rate of satisfactory results of twenty-four of twenty-seven after the total shoulder arthroplasties in the current study. Thus, the results in the current series are similar to those reported in other series.
The greatest difference between the two study groups was between the patients' self-assessments of pain. Pain was assessed with both scoring systems. The University of California at Los Angeles shoulder score demonstrated superior satisfaction, function, range of motion, and strength after total shoulder arthroplasty, but these results were not found to be significant, with the numbers available. Neither system specifically measures range of motion in degrees, but when each range of motion was analyzed separately internal rotation behind the back (p = 0.003), but not elevation or external rotation, was found to be significantly better after total shoulder arthroplasty. The American Shoulder and Elbow Surgeons shoulder index was 12 points higher for the patients who had had a total shoulder arthroplasty than it was for those who had had a hemiarthroplasty, but this difference was not found to be significant, with the numbers available (p = 0.057).
Hospital charges were $1177 greater, on a per-case basis, for the total shoulder arthroplasties. The mean charge was $15,998 for the three revision operations in the group that had had a hemiarthroplasty. Thus, the charges for the index hemiarthroplasties were lower than those for the index total shoulder arthroplasties, but this must be balanced against the potential charges for a revision operation.
Weaknesses in the radiographic analysis included the small number of shoulders in the study groups, the lack of fluoroscopic examination, and the short duration of follow-up. No shoulder had meaningful loosening or subsidence of the humeral stem, findings that are consistent with those in other reports5,14,19, in which problems related to the humeral stem have been relatively rare. The rates of radiolucent lines around the glenoid component (four of twenty-seven) and loosening of the glenoid component (two of twenty-seven) were lower than those reported previously10,21,23. We cannot explain these decreased rates other than in relation to the just mentioned weaknesses in the radiographic evaluations.
Weaknesses of the study as a whole included the small number of patients and the short duration of follow-up. The power analysis that was performed before the beginning of the study was a gross estimate and suggested that thirty-five patients be included in each subgroup. Because of the time constraint that was integral to the study, that goal was not achieved. Although enough data were obtained to observe significant differences, it is possible that the sample size was inadequate to differentiate the effect of the type of operation. Readers who are accustomed to a standard significance level of p = 0.05 may be confused by our failure to identify p = 0.008 as significant (Table III). Because we compared multiple items in our analysis, we adjusted the level of significance accordingly.
We also are concerned about the large spread of the 95 percent confidence intervals (Table III). One interpretation of these data is that the scoring systems that we used do not measure the status of the patient with sufficient precision. Although a modification of the American Shoulder and Elbow Surgeons patient self-assessment questionnaire has been tested for responsiveness and reliability3, we still are concerned that the present scoring systems may not optimally measure the outcome of shoulder arthroplasty. We recognize that the study would have been strengthened by use of an independent evaluator who was unaware of the type of operation that had been performed. Nonetheless, the study was prospective and randomized, and the variables were controlled in an attempt to specifically evaluate the effect of glenoid replacement.
Compared with hemiarthroplasty, total shoulder arthroplasty offers the patient who has osteoarthritis of the shoulder and a concentric glenoid significant improvement in terms of pain relief and internal rotation and higher (but not significantly higher) scores in terms of satisfaction, function, and strength, as assessed by the patient. Both total shoulder scores were higher for the patients who had had a total shoulder arthroplasty, but the difference was not significant given the size of the study groups and the number of comparisons made between them. Total shoulder arthroplasty has the disadvantages of increased technical complexity, operating time, blood loss, and cost. Although no patient had a revision operation because of loosening of the glenoid component during the study period, the potential for this complication remains. Hemiarthroplasty has the advantages of decreased complexity, operating time, blood loss, and cost; however, three of twenty-four shoulders needed a revision for insertion of a glenoid component.