Total shoulder arthroplasty provides effective, reliable treatment of end-stage inflammatory and degenerative glenohumeral arthrosis. The National Center for Health Statistics estimated that, in 2004, >13,000 total shoulder arthroplasties were performed1, and population-based studies have projected a continued growth in the number of total shoulder arthroplasties2. One persistent problem has been glenoid component loosening with cement fixation, with loosening rates ranging from 2% to 10% with use of clinical criteria and from 30% to 60% with use of radiographic criteria3-8. Bone-ingrowth components were designed with the hope of achieving more stable fixation to the bone and a corresponding increase in implant survival9-15.
On the basis of prior studies, it has been hypothesized that the clinical results of a bone-ingrowth glenoid component should be comparable with the published results of cemented and other uncemented glenoid components. The purpose of this study was to examine the clinical findings, the radiographic results, and the survivorship of one type of bone-ingrowth glenoid component as a part of a total shoulder arthroplasty system.
Between 1989 and 1994, 435 primary total shoulder arthroplasties were performed by the senior author (R.H.C.) at our institution for the treatment of symptomatic glenohumeral arthritis. One hundred and sixteen consecutive primary ingrowth total shoulder arthroplasties were performed during this time. The study group included eighty-three arthroplasties, after thirty-three shoulders were excluded. The inclusion criteria were the diagnosis of osteoarthritis or posttraumatic arthritis, a patient of average to above-average physical size as the ingrowth component was somewhat larger than standard polyethylene components, a patient who had minimal or no osteopenia to ensure sound initial fixation between the ingrowth component and the bone, and a patient who had the desire to remain active so that he or she would have the potential to obtain the proposed benefits of durable, stable bone-ingrowth fixation over time—avoiding problems with cement fragmentation and component loosening. The exclusion criteria included a diagnosis of rheumatoid arthritis, rotator cuff tear arthropathy, a previous rotator cuff repair, or a rotator cuff tear at the time of shoulder arthroplasty. In addition, a patient was excluded if, after a thorough discussion with the surgeon, the patient wished to proceed with the more traditional cemented polyethylene component. The study was approved by our institutional review board.
The average age of the patients who underwent total shoulder arthroplasty was sixty-eight years (range, forty-one to eighty-seven years). There were forty-seven men and thirty-one women; four men and one woman underwent bilateral total shoulder arthroplasty. Forty-five right shoulders and thirty-eight left shoulders were replaced. The indications for total shoulder arthroplasty were osteoarthritis in seventy-four shoulders, traumatic arthritis in five shoulders, and osteonecrosis in four shoulders. One of the eighty-three shoulders had undergone a previous arthroscopic débridement procedure and had been found to have degenerative joint disease. The average length of clinical follow-up was 9.5 years for all eighty-three shoulders included in the study. Follow-up was less than two years for four shoulders that were revised for component fixation problems (revision surgery also served as the end point for data collection). The duration of follow-up was between two and five years for seven shoulders, between five and ten years for twenty shoulders, between ten and fifteen years for forty-eight shoulders, and greater than fifteen years for four shoulders.
Components and Surgical Technique
The ingrowth glenoid component consists of four parts: a titanium-alloy metal tray with three columns projecting from its surface that are designed to penetrate into the bone of the glenoid neck, two titanium-alloy cortical bone screws, and an ultra-high-density polyethylene insert. The porous coating is limited to the portion of the metal tray abutting the subchondral bone plate of the glenoid. This surface does not extend onto the columns that penetrate into the glenoid neck as the surface applied to the subchondral bone of the glenoid was quite large in area, fitted exactly to the bone. Avoiding the extension of the porous coating into the columns facilitates component removal should that become necessary. This porous coating is 1 mm deep; the average pore diameter is 250 µm, and the average porosity is 30%9.
The technique of placing this glenoid component differs from the techniques used in this time period to place components that rely on methylmethacrylate bone cement for fixation. A series of instruments is used to prepare the glenoid surface and the glenoid neck to create an exact fit between the component and the scapula. Briefly, a small centering hole is placed in the glenoid surface. A facing reamer is positioned, and 1 to 1.5 mm of the subchondral bone plate is removed. Removal creates a subchondral bone surface that exactly matches the shape of the porous surface on the metal tray and includes an admixture of cortical and cancellous bone. Templates and drills are used to create three columns extending into the glenoid neck. The metal tray of the component is impacted into position and is further secured to the scapula with two cortical screws that penetrate the upper and lower columns of the implant and achieve purchase in the cortical bone at the base of the glenoid neck. The polyethylene insert is then impacted into position on the metal tray9. The chromium-cobalt humeral component is composed of one part that is available with two head sizes and four stem diameters. The ingrowth coating of the humeral component is similar to that of the glenoid component in structure and is limited to the undersurface of the prosthetic humeral head. Just as for the glenoid component, a humeral resection guide, drill, and reamers were developed to create an exact fit between the humeral component and bone.
A deltopectoral approach was used in eighty shoulders, and an anteromedial approach, carefully releasing and repairing the deltoid origin from the clavicle and anterior acromion, was used in three shoulders16. The subscapularis tendon and anterior capsule was incised at the medial edge of the lesser tuberosity in seventy-seven shoulders, the subscapularis was taken from bone in three shoulders, and a z-plasty lengthening of the subscapularis tendon was performed in three shoulders. Bone-grafting was required for a limited amount of glenoid bone erosion or cyst formation in six shoulders.
Clinical Assessment
At our institution, clinical assessment of all patients who undergo shoulder arthroplasty is recorded preoperatively and at the time of each patient contact through use of a standard shoulder analysis data form6. The latest clinical follow-up assessment was by examination for fifty-seven shoulders and by a standard self-administered shoulder questionnaire for twenty-six shoulders. In a previous study by Smith et al.17 on patient and physician assessment of shoulder function after arthroplasty, a high level of agreement was demonstrated between patients and physicians using this questionnaire to measure pain, motion, and functional parameters.
Shoulder pain was graded on a scale from 1 to 5, with 1 indicating no pain; 2, slight pain; 3, pain after unusual activity; 4, moderate pain; and 5, severe pain. Range of motion in active elevation and external rotation was recorded in degrees. Range of internal rotation was measured on the basis of the highest posterior vertebral segment the patient could reach with the thumb. The results were classified according to a modification of the Neer result-rating system6. The result was considered excellent if the patient was satisfied with the result and had no or slight pain, active elevation to =140°, and external rotation to =45°. The result was satisfactory if the patient was satisfied with the procedure and had no, slight, or moderate pain only with vigorous activity; active elevation to =90°; and external rotation to =20°. If any of these criteria were not met or if the patient had undergone additional surgery, the result was unsatisfactory.
The radiographs reviewed included those made in the preoperative period, in the early (one to two-month) postoperative period, and at a minimum of two years. The mean radiographic follow-up was 7.1 years. Three projections were used for radiographic analysis: an axillary radiograph and two true anteroposterior (internal rotation and external rotation) radiographs of the glenohumeral joint. All shoulders in this study had radiographs of excellent quality to assess the implant-bone interface changes (Fig. 1). Components were divided into radiographic zones to measure periprosthetic lucency; the glenoid had five zones and the humerus had eight zones. Radiolucent lines were evaluated according to their presence or absence, location, and thickness. Radiographs were also examined for evidence of polyethylene loss with metal wear of the tray. Three observers compared the early postoperative and final follow-up radiographs and reached a consensus as to whether the glenoid or humeral components had shifted in position. To combine the data on lucent lines, their extent and thickness, and the data on a shift of the component position, we selected a set of changes that orthopaedic surgeons would, in our opinion, usually agree were worrisome for component loosening and could be associated with clinical problems. We designated these combinations as representing components that were radiographically "at risk" for clinically symptomatic loosening. A glenoid component was defined to be at risk if observers identified migration or tilt of the component or if a complete lucent line was present and some part of it was =1.5 mm in width8,18. A humeral component was defined to be at risk if observers identified tilt or subsidence of the component or if a lucent line of =2 mm in width was seen in three or more zones9,18.
Statistical Methods
For ordinal and continuous measurements, a paired t test was performed to compare preoperative and postoperative changes. Postoperative assessments for pain and range of motion were made at the time of the last follow-up. In patients undergoing a revision arthroplasty, the last clinical and radiographic information prior to revision was used. For the implant survival with revision only and revision and/or radiographic failure as the end points, shoulders were followed from the date of the primary shoulder arthroplasty. Survival was estimated with the Kaplan-Meier method, reporting the estimated rate and the 95% confidence interval. Associations between revision and risk factors were assessed with the Cox proportional survival method, adjusting for correlated data (two shoulders in the patients with a bilateral total shoulder arthroplasty). The alpha level was set at 0.05 for significance.
For the eighty-three shoulders included in the study, the mean pain scores decreased from 4.7 to 2.0 (p < 0.001). Mean active abduction improved from 102° (range, 15° to 180°) to 135° (range, 9° to 180°) (p < 0.001). Mean external rotation increased from 27° (range, -20° to 90°) to 56° (range, 0 to 90°) (p < 0.001). Mean internal rotation increased from the L5 vertebral level (range, inability to reach the abdomen to the ability to reach T8) to the L1 vertebral level (range, inability to reach the abdomen to the ability to reach T5) (p < 0.001). According to the modified Neer ratings, the result was excellent for twenty-nine shoulders, satisfactory for twenty-one, and unsatisfactory for thirty shoulders; no response relative to patient satisfaction was provided for three shoulders. The thirty unsatisfactory results included twenty-six revisions for various reasons and four shoulders with movement deficiencies. Specifically, they did not have active elevation to =90° and external rotation to =20°.
Clinical parameters assessed preoperatively and at the time of the latest follow-up for the fifty-seven shoulders that had not undergone a revision were compared. In this group, the mean pain rating decreased from 4.7 to 1.9. Mean abduction improved from 102° to 140°, and mean external rotation improved from 29° to 60°. The mean range of internal rotation improved from the L5 to the L1 vertebral level. All improvements were significant (p < 0.001). Thirty-eight patients reported that the shoulder was much better; twelve, that it was better; five, that it was the same; none, that it was worse; and two had no response. According to the Neer ratings in this group, the result was excellent in twenty-nine shoulders, satisfactory in twenty-one, unsatisfactory in four shoulders, and no response in regard to satisfaction in this segment of questioning was provided by three patients. The patients who had no response in regard to satisfaction had an otherwise complete clinical and radiographic follow-up.
Complications and Revisions
An acute complication occurred in one patient who had sustained an intraoperative humeral shaft fracture. The fracture was fixed with cables at the time of surgery. Twenty-one other patients had delayed clinical or radiographic complications. There was rotator cuff stretching or tearing in seven shoulders, moderate humeral subluxation in ten shoulders, and dislocation in four shoulders. There were no infections in the group.
Twenty-six shoulders underwent a revision procedure that included removal or replacement of one or both components. The issues present at the time of the revision are illustrated in the Venn diagram in Figure 2. At the time of the revision, eighteen shoulders had component issues. Polyethylene wear was seen in fifteen shoulders; glenoid metal wear, in twelve; glenoid component loosening, in nine; and humeral component loosening, in six shoulders. The patients who had metal wear of the glenoid component had complete wear-through of the polyethylene component, leading to the humeral head wearing directly on the metal glenoid component in all twelve shoulders. Ten of the shoulders that had loosening of one of the components had substantial polyethylene wear and osteolysis, which was a major contributor to loosening of the component. With the numbers studied, we could not identify any patient, disease, or surgical characteristics associated with the need for revision surgery.
Radiographic Results
On radiographic analysis of all eighty-three shoulders, no lucent line was seen around thirty-three glenoid components. Lucent lines incompletely surrounded the glenoid component in sixteen shoulders, and such lines completely surrounded the glenoid component in thirty-four shoulders. Seventeen glenoid components had lucent lines of =1 mm. Thirty-three glenoid components had lucent lines in which some segment was =1.5 mm in width; the lines were incomplete in one component and were complete in thirty-two components. The distribution of radiolucent lines of =1.5 mm is depicted in Figure 3-A. Radiographically, polyethylene wear was determined by visualization of obvious narrowing of the radiolucent gap between the humeral head and the metal portion of the glenoid component. It must be noted that this represents only the components with severe polyethylene wear, and the actual polyethylene wear could be much higher. Thirty glenoid components demonstrated obvious polyethylene wear, and twenty-one had evidence of metal wear. Of the forty-four shoulders that had greater than ten years of clinical follow-up, twenty-one glenoid components had complete lucent lines of which some segment was =1.5 mm wide, and were considered at risk. Twelve had failure by metal wear. In this group of patients with more than ten years of follow-up without a revision, four shoulders ultimately underwent revision at an average of 14.4 years postoperatively. All four had radiographic evidence of failure by metal wear at the time of revision.
No lucent line was seen around fifty-two humeral components. Lucent lines incompletely surrounded the humeral component in twenty-three shoulders. Eight shoulders had complete lucent lines surrounding the humeral component, and nine components had lucent lines of =1 mm. Twenty-two components had lucent lines of which some segment was =2 mm in width. The distribution of radiolucent lines of =1.5 mm in width is depicted in Figure 3-B. Of the fifty-two shoulders that had greater than ten years of clinical follow-up, ten humeral components were considered at risk.
The components were evaluated for a change in position between the early radiograph and the radiograph made at the time of the most recent follow-up. The glenoid component had tilted or displaced medially on these sequential radiographs in nine shoulders (Fig. 4). The humeral component had shifted in position, either displacing distally into the humerus or tilting, in seven shoulders. When the data for radiographic lucent lines and component shifts in position were combined, thirty-three of the eighty-three glenoid components were judged to be at risk. Fifteen of the eighty-three humeral components were judged to be at risk. Twenty-one of the eighty-three were found to have radiographic evidence of failure by metal wear (Fig. 5). If any one of the three above risk factors is considered as a risk of failure, fifty shoulders were found to be at risk. No glenoid or humeral bead shedding was noted on the latest follow-up radiographs.
Similarly, radiographic data were collected for the subset of fifty-seven shoulders that were not revised. When the data for radiographic lucent lines and component shifts were combined, twenty-three of the fifty-seven glenoid components were judged to be at risk. Nine of the fifty-seven humeral components were judged to be at risk. Furthermore, nine of the fifty-seven had radiographic evidence of humerus-on-glenoid metal wear. If any one of the three above risk factors is considered to portend clinical failure, twenty-eight of these fifty-seven total shoulder replacements were found to be at risk of clinical failure.
Three patients had a subscapularis z-lengthening at the time of the index arthroplasty. One of these replacements was revised for glenoid wear and osteolysis. All three had subluxation. Of the two that were not revised, both had evidence of polyethylene wear, both had a glenoid component at risk, one had failure by metal wear, and one had a humeral component at risk.
Survival estimates were calculated for two separate clinical scenarios. Kaplan-Meier survival estimates were performed for the end point of revision only (Fig. 6-A). The one-year survival estimate was 94.0% (95% confidence interval, 89.0% to 99.2%). The five-year survival estimate was 86.7% (95% confidence interval, 79.7% to 94.3%). The ten-year survival estimate was 78.5% (95% confidence interval, 70.0% to 88.2%).
Kaplan-Meier survival estimates were also performed for the end point of revision and/or radiographic failure, including metal wear, a glenoid component at risk, and a humeral component at risk (Fig. 6-B). The one-year survival estimate was 92.8% (95% confidence interval, 87.4% to 98.5%). The five-year survival estimate was 79.9% (95% confidence interval, 71.6% to 89.3%). The ten-year survival estimate was 51.9% (95% confidence interval, 41.0% to 65.8%).
The use of a bone-ingrowth glenoid component in total shoulder arthroplasty has a long but rather limited history. The method of fabrication and the properties of a metal implant with a porous ingrowth surface were described in 1975 by Pilliar et al.13. In 1987, McElwain and English reported the early clinical results with twenty-one glenoid components of this type, and the results were quite favorable clinically10. Also, with greater than three years of follow-up, the components demonstrated no radiographic evidence of loosening. Weiss et al., in 1990, reviewed forty-two total shoulder arthroplasties with either the English-McNab ingrowth component or the Neer cemented prosthesis11. In that study, radiographic lucent lines developed around fifteen glenoid components, but only one of them was an ingrowth component. In the 1999 report by Wallace et al., twenty-six ingrowth components were placed without cement and thirty-two glenoid components were fixed with cement12. The cemented components were three times more likely to be classified as loose than were the ingrowth components; however, those authors recognized the potential for polyethylene and metal tray wear of uncemented glenoid components. In 1992, we reported on a new design of an uncemented bone-ingrowth glenoid component9. Clinical results were very favorable, with twenty-seven of thirty-two patients reporting little or no pain, and the average amount of clinical abduction was 145°. Although these parameters were quite favorable, three of the glenoid components had probable loosening on radiographic evaluation and eight shoulders had some degree of glenohumeral instability. Thus, overall, the clinical experience with the metal-backed ingrowth glenoid component was favorable, but there were issues of concern including polyethylene wear, metal wear, instability, and component loosening.
During that time period, biomechanical studies demonstrated that use of a metal-backed glenoid component resulted in a slight improvement in stress transfer to the cortical bone of the scapula19 and improved fixation when supplemented with the use of screws20. Encouraged by the early clinical results and the biomechanical studies, we redesigned a bone-ingrowth glenoid component of the type reported in the present study. In 2000, we reported on sixty-two shoulders that had a primary ingrowth total shoulder arthroplasty with a mean follow-up of 4.6 years18. The clinical parameters were very favorable in terms of pain relief and improvement in motion. With use of the same radiographic criteria for components at risk reported in this study, only four of the sixty-two glenoid components and six of the sixty-two humeral components were judged radiographically to be at risk for clinical failure. However, some issues that were emerging included instability in three shoulders, high-density polyethylene wear and displacement in three shoulders, glenoid loosening in one shoulder, and humeral loosening in one shoulder18. In parallel with this clinical experience, biomechanical laboratory studies have described high stresses within the polyethylene of metal-backed glenoid components with the implication that these components will have inferior wear properties21,22. The biomechanical studies, combined with our data, indicate that an environment of increased stresses introduced by the metal backing increases the polyethylene wear rate and leads to subsequent clinical failure in some shoulders.
For the fifty-seven shoulders in the present study that did not have a revision, the clinical parameters including pain rating and movement were all significantly improved. However, a comparison of the earlier study of sixty-two total shoulder arthroplasties with a mean follow-up of 4.6 years18 and these fifty-seven shoulders with a mean follow-up of 9.5 years demonstrated that the proportion of components radiographically at risk for clinical failure increased dramatically. For the glenoid components, this rate increased from 6.5% to 36%. An increased rate of failure in the humeral component over time was also noted in the study, and it is thought to be associated with increased polyethylene wear and associated osteolysis.
Eighteen of the twenty-six revision procedures had component wear or loosening as a major issue leading to the revision. Of the fourteen patients who had clinical instability, twelve went on to have a revision. Six of these patients had concomitant component issues. On the basis of these observations, the problems of high-density polyethylene wear, subsequent metal wear, osteolysis, and component loosening do not stabilize with time. These complications continue to accrue, thus dramatically mitigating the positive outcomes that were reported in earlier studies15. On a positive note, the instrumentation that was developed for the precise placement of ingrowth components seems to have transferred quite favorably to current shoulder arthroplasty practice; however, given the experience to date with metal-backed ingrowth glenoid components, our current practice is to cement an all-polyethylene glenoid component using these instruments and other improved surgical techniques in most patients.
In our current practice, the use of this metal-backed bone-ingrowth component is limited to special situations. These situations include focal glenoid bone loss that precludes the use of a cemented component in such patients as those with scapular dysplasia or those undergoing revision surgeries with moderate central glenoid bone loss. A bone-ingrowth humeral component continues to be used, but, on the basis of this experience, it has been modified to include a larger number of stem sizes for an improved interference fit and a larger area of bone-ingrowth surface extending onto the proximal aspect of the stem of the component.
There are limitations in this study. It is a retrospective review of one surgeon's experience with one type of metal-backed glenoid component in the setting of total shoulder arthroplasty. The preplanned methods of clinical follow-up are heterogeneous. The strengths of this study include the consistent record collection and the standardized fashion in which the patients were treated. We included patients with the primary diagnoses of osteoarthritis and posttraumatic arthritis and excluded patients who might have been predisposed to other modes of failure including those with rheumatoid arthritis or a rotator cuff tear.
In conclusion, as previously seen in biomechanical studies, the metal-backed glenoid design may introduce a biomechanical environment of unfavorable increased wear characteristics21. This study indicates that total shoulder arthroplasty with use of a metal-backed bone-ingrowth glenoid component has an increased level of high-density polyethylene wear, subsequent metal wear, osteolysis, and component loosening. This failure mechanism does not stabilize but continues to accrue with time. These findings raise substantial concern for the use of this component, and perhaps other metal-backed bone-ingrowth glenoid components, other than for special situations. 