Between March 2002 and October 2005, 278 shoulder arthroplasties were performed at our institution and were included in a prospective database. Fifty-eight patients were managed with a cementless humeral surface replacement arthroplasty. From this cohort, fifteen patients (eighteen shoulders) were managed with cementless humeral surface replacement arthroplasty for the treatment of humeral head osteonecrosis and were retrospectively included in the present study. All clinical and radiographic data from these patients were analyzed retrospectively. The inclusion criteria were (1) cementless humeral surface replacement arthroplasty for the treatment of posttraumatic or nontraumatic osteonecrosis of the humeral head that initially had been treated either operatively or nonoperatively, (2) a minimum duration of follow-up of two years, (3) an intact rotator cuff, and (4) sufficient humeral bone stock (minimum, 60% of the humeral head). The exclusion criteria were (1) rheumatoid arthritis, (2) infection of the affected shoulder or shoulders, (3) neural injuries, and (4) bone defects of the glenoid. The patients included ten women and five men. The dominant shoulder was involved in ten cases. The mean age at the time of the arthroplasty was forty-eight years (range, thirty-five to eighty-two years). One patient (an eighty-two-year-old woman with posttraumatic osteonecrosis) died six months after surgery; no information on the short-term outcome was available in this case. Thus, fourteen patients with seventeen cementless humeral surface replacement arthroplasties were included in the present study. The mean duration of follow-up was three years (range, twenty-four to sixty-eight months).
There were nine cases of nontraumatic osteonecrosis in six patients and eight cases of posttraumatic osteonecrosis of the humeral head in eight patients. The etiology of the nontraumatic osteonecrosis was Hodgkin lymphoma in three patients (four arthroplasties), acute myeloid leukemia in one patient (two arthroplasties), non-Hodgkin lymphoma in one patient (one arthroplasty), and surgical removal of a pituitary adenoma in one patient (two arthroplasties). All patients were managed with corticosteroids before the arthroplasty because of the above-mentioned conditions, and none of the affected shoulders were treated with a surgical procedure before the arthroplasty. Three male patients had bilateral nontraumatic osteonecrosis (see Appendix). Other than the presence of osteonecrosis, the gross appearance of the proximal part of the humerus was normal in all cases.
The initial cause of the injury in patients with posttraumatic osteonecrosis was a simple fall in four cases, a sports-related injury in three, and a traffic accident in one. The initial fracture patterns were classified according to the system described by Neer20. There were five three-part fractures (two of which involved the lesser tuberosity and three of which involved the greater tuberosity), two four-part fractures, and one subcapital fracture (a surgical neck fracture). Two three-part fractures were primarily treated nonoperatively. The other six fractures were primarily treated operatively with open reduction and internal fixation. Of the six patients who initially were managed with open reduction and internal fixation, four required one operation and two required two operations. Malunion of the tuberosities was found in five of the eight patients, and the gross appearance of the proximal part of the humerus was normal in only three cases.
Humeral head necrosis was graded with use of radiographs, according to the system of Cruess9, as stage 1 (preradiographic stage), stage 2 (mottled sclerosis), stage 3 (subchondral fracturing [crescent sign]), stage 4 (collapse of subchondral bone and loss of sphericity), or stage 5 (degenerative changes of the glenoid). In the nontraumatic group, osteonecrosis was graded as stage 2 in one shoulder, stage 3 in seven, and stage 4 in one. In the posttraumatic group, osteonecrosis was graded as stage 3 in three shoulders and as stage 4 in five.
Preoperative and postoperative radiographs of the shoulders were made in the true anteroposterior and axillary projections for all patients (Figs. 1-A through 2-D).
Implant loosening on the immediate postoperative and most recent follow-up radiographs was compared by analyzing implant inclination, the distance above the greater tuberosity, the vertical humeral head height, and the perpendiculars to the observed humeral implant diameter running to the humeral shaft axis or the lateral cortex of the greater tuberosity for measurable migration of the implant in relation to the proximal part of the humerus as described by Rydholm and Sjögren21 (Fig. 3). Changes in implant position can be detected with this method by comparing the immediate postoperative and most recent follow-up radiographs. As stated by Rydholm and Sjögren, a difference of >3 mm or 5° of inclination was graded as a change in implant position. The true diameter of the implant was divided through the observed diameter (Fig. 3) to control for and to calculate the magnification effects21.
The volumes of the necrotic areas of the humeral heads were calculated with use of the volumetric measurement method with radiographs in two planes as described by Steinberg et al.22. This technique originally was used to calculate necrotic areas of the femoral head; however, it was easily adapted for the evaluation of anteroposterior and axillary radiographs of the humeral head in the present study.
With use of this technique, the humeral head was traced from the anteroposterior and axillary radiographs onto tracing paper by one observer (P.R.). The volumes of the entire humeral head and the necrotic area were calculated with use of an image-processing program (ImageJ, version 1.38x; National Institutes of Health, Bethesda, Maryland) (Fig. 4). The proportion of necrosis was calculated on both views separately, and then these two values were multiplied together.
All patients were evaluated preoperatively with use of the Constant score23,24, adjusted for age and sex. Additionally, active range of motion was recorded for shoulder flexion, abduction, and rotation, with the hanging arm in a neutral position and the elbow flexed to 90°. Shoulder flexion, abduction, and external rotation were recorded in degrees, whereas internal rotation was graded according to the part of the spine that could be reached by the thumb. The Constant score was used to grade pain (with 0 points indicating severe pain and 15 points indicating no pain), activity (with 0 points indicating no activity and 20 points indicating full activity), mobility (with 0 points indicating no mobility and 40 points indicating full mobility), and power (with 0 points indicating 0 kp [0 N] and 25 points indicating 12.5 kp [122.6 N]).
The morphology of the glenoid was classified according to the system of Walch et al.25. An A1 glenoid was found in fourteen cases, and an A2 glenoid was found in three cases.
Postoperatively, all patients were routinely evaluated after six and twelve months and at the time of the most recent follow-up examination with the same standardized clinical and radiographic protocol. The subjective result after surgery was rated by the patient as "very satisfied," "satisfied," "somewhat disappointed," or "very disappointed."
Operative Technique and Implants
In all shoulders, a deltopectoral approach was used as described by Neer et al.26. In no case was a rotator cuff tear found. After detachment of the subscapularis tendon and a capsular release, the joint was exposed. In all cases, the intraoperative joint status corresponded with the radiographic findings. The biceps tendon was dissected close to its glenoid attachment and was tenodesed in the bicipital groove in all cases. The humeral head was reamed with hemispherical reamers of decreasing size. The cancellous bone removed during reaming was used to fill in the necrotic zones. The uncemented resurfacing implant was placed with use of a press-fit technique. In all cases, primary stability of the implant as tested manually was achieved despite the presence of the humeral head necrosis. If movement was detected at the implant-bone interface, a stemmed implant was used. Over the past decade, this has been necessary in four other cases, not included in the present series, with <60% of the bone remaining in the humeral head.
In the first eight shoulders, a Copeland Shoulder (Biomet Europe, Dordrecht, The Netherlands) was used; in the subsequent ten shoulders, the EPOCA RH Cup (Argomedical, Cham, Switzerland) was used. All implants were coated with hydroxyapatite on the inner surface. All implant designs offered a spherical joint surface. For the Copeland Shoulder, primary stability in the humeral epiphysis was achieved with use of a centered peg, whereas for the EPOCA Cup, primary stability was obtained with use of a conical crown-shaped ring.
After placement of the implants, the subscapularis tendon was repaired with use of three to five nonabsorbable tendon-to-tendon sutures. Because of an internal rotation contracture, five shoulders in the posttraumatic cohort and one shoulder in the nontraumatic cohort were treated with a subscapularis release according to the method of Walch et al.27; this procedure was performed when a minimum of 0° of external rotation was not attained after provisional coaptation of the muscle.
Drains were removed on the first day after surgery. To protect the reconstructed subscapularis tendon, the arm was placed in internal rotation in a shoulder abduction pillow for four weeks. Postoperatively, the shoulder was mobilized passively by a physiotherapist for six weeks to 60° of flexion and abduction and 0° of external rotation. Patients were asked to support these movements actively. Free range of motion was allowed six weeks after surgery.
Statistical Analysis
The paired t test was used to compare the preoperative and postoperative Constant scores and subscores as well as to compare the parameters of implant loosening between the immediate postoperative and most recent follow-up radiographs. Two-factorial repeated-measurement analysis of variance was used to assess differences between the nontraumatic and posttraumatic groups. The level of significance was set at p < 0.05.
Source of Funding
The noncommercial research fund of the "Deutsche Arthrose-Hilfe e.V." supports clinical investigations using the "shoulder arthroplasty register" of the Orthopaedic Department of the University of Heidelberg.
Survival and Complications
There was one reoperation in this series; specifically, a superficial wound infection was débrided eighteen days after the arthroplasty. There was no communication with the shoulder joint, and implant revision was not necessary. After this second operation, the patient received intravenous antibiotic therapy for two weeks. One patient in the nontraumatic cohort with a history of radiation therapy had development of neuralgic shoulder amyotrophy, which was detected forty-four months after the arthroplasty. This distinct clinical entity is associated with pain, multifocal paresis, and atrophy of the upper extremity; is more common in men (male:female ratio, 3:2); and occurs spontaneously28. There was no clinical or sonographic evidence of a rotator cuff tear in this patient. The patient had a Constant score of 53 points (58%) and was able to elevate the arm to 100° at the time of the most recent follow-up examination.
Clinical Results and Patient Satisfaction
The mean Constant score for the entire cohort (including the nontraumatic and posttraumatic groups) improved from 31 points (range, 8 to 57 points) preoperatively to 62 points (range, 27 to 83 points) postoperatively (p < 0.0001), and, adjusted by age and sex, from 34% (range, 5% to 63%) to 75% (range, 39% to 114%) (p < 0.0001). Significant differences also were found in terms of pain relief, power, activity, mobility, shoulder flexion, abduction, and external and internal rotation (p < 0.05). The findings of the preoperative and postoperative clinical examinations are shown in Table I.
The Constant scores and subscores for the nontraumatic and posttraumatic groups are shown in Tables II and III. Preoperatively, the Constant score and subscores (except for those for pain and flexion) were higher in the nontraumatic cohort. Significant postoperative differences between the nontraumatic and posttraumatic groups were detected in terms of the Constant score, abduction, and power (Table IV). Postoperatively, shoulder flexion, external and internal rotation, activity, mobility, and pain relief were higher in the nontraumatic cohort, but not significantly so.
At the time of the most recent follow-up examination, fourteen patients (fifteen shoulders) were very satisfied (eleven shoulders) or satisfied (four shoulders) with the procedure. The patient with neuralgic shoulder amyotrophy was somewhat disappointed with the result of surgery; another patient in the posttraumatic cohort was also somewhat disappointed because of continuous moderate pain.
Imaging Results
The mean portion of the humeral head occupied by necrotic zones was 18.6% (range, 8.9% to 30.9%).
In all cases, the necrotic zones were located at the articular surface between the apex and the lower third of the humeral head and were fully covered by the humeral surface replacement.
At no time during the follow-up period did we observe any obvious radiographic signs of implant loosening. When the immediate postoperative and most recent follow-up radiographs were compared on the basis of the parameters described in Figure 3, no significant differences were found.
No signs of secondary glenoid wear or humeral head dislocation were found on the most recent follow-up radiographs in two planes.
Humeral head necrosis can be a disabling disease. The loss of sphericity of the humeral head leads to shoulder joint incongruence, pain, and dysfunction. Conservative treatment options15 can be effective for patients with Cruess stage-1 and 2 osteonecrosis. Furthermore, joint-preserving surgical options such as arthroscopic débridement, core decompression, and vascularized bone-grafting have been described12-14,29,30. Shoulder arthroplasty may be considered if the above-mentioned treatment options have failed. The results of conventional hemiarthroplasty and total shoulder arthroplasty for traumatic and nontraumatic osteonecrosis of the humeral head have been documented. However, there have been only a few prospective studies7,31, and most published studies on this topic have had a retrospective design2,3,5,9. Good results of both hemiarthroplasty and total shoulder arthroplasty in cases of nontraumatic and posttraumatic osteonecrosis have been described. For nontraumatic osteonecrosis, a significant improvement in function and a decrease in pain have been found; the rate of patient satisfaction after surgery has ranged from 80% to 90%5,16,32. For posttraumatic osteonecrosis, the results of shoulder arthroplasty seem to be comparable or only slightly inferior2,5,31.
Good results have been reported for humeral head resurfacing as a bone-preserving treatment option to restore joint congruency in cases of primary and secondary osteoarthritis and rheumatoid arthritis18,33-35. There is only little information regarding cementless humeral surface replacement arthroplasty in cases of osteonecrosis18,19. One reason for the limited use of this treatment option in cases of osteonecrosis may be concern about the lack of sufficient osseous stability for the implant. However, necrosis of as much as 40% or 45% of the humeral head has been reported to leave sufficient bone to provide primary stability in patients managed with cementless humeral surface replacement arthroplasty35,36.
In 2001, Levy and Copeland presented the clinical results of cementless humeral surface replacement arthroplasty in a study of four shoulders with osteonecrosis18. When the preoperative and postoperative values were compared, the mean overall Constant score increased from 9 to 58 points, the mean pain score increased from 0 to 12 points (indicating a decrease in pain), the mean activity score increased from 2 to 11 points, the mean mobility score increased from 7 to 31 points, and the mean power score increased from 0 to 4 points. Furthermore, an improvement in shoulder motion was described. Except for more external rotation in the cohort of Levy and Copeland, the results in our entire cohort (including all patients with nontraumatic and posttraumatic osteonecrosis) were comparable or slightly superior (Table I).
Thomas et al.19 reported on one case of osteonecrosis that was treated with cementless humeral surface replacement arthroplasty. No information was provided regarding the clinical outcome in this patient.
We found only two studies comparing the results of shoulder arthroplasty in patients with nontraumatic and posttraumatic osteonecrosis5,31. Hattrup and Cofield5 reviewed 127 hemiarthroplasties and total shoulder arthroplasties after a mean duration of follow-up of nine years and found a significant difference between the nontraumatic and posttraumatic cohorts in terms of the mean postoperative American Shoulder and Elbow Surgeons score (69 compared with 55 points; p = 0.017). There were also significant postoperative differences between the two cohorts in terms of mean shoulder flexion (138° compared with 107°), mean abduction (125° compared with 96°), and mean external rotation (66° compared with 49°). Preoperative scores were not described. A complication rate of 32% was reported.
Orfaly et al.31, in a prospective study, analyzed the outcome of twenty-one shoulder arthroplasties in nineteen patients. The authors found a significant decrease in pain and an increase in function in the entire series with use of a visual analog scale and the American Shoulder and Elbow Surgeons Shoulder Score Index. No significant differences were found between the nontraumatic and posttraumatic cohorts with regard to outcome. Mean forward elevation increased significantly from 88° to 123° in the nontraumatic cohort and from 90° to 129° in the posttraumatic cohort. Mean external rotation increased significantly from -7° to 34° in the nontraumatic cohort and from 3° to 19° in the posttraumatic cohort. These findings seem to be comparable with those in the present study.
There are some limitations in the current study as well as in the above-mentioned trials. First of all, the clinical and radiographic data of the present longitudinal study were analyzed retrospectively. Moreover, the number of cases was relatively small as the abnormality is rare and the inclusion criteria were very strict. Another limitation is the different initial treatment in the two cohorts (with no previous operations in the nontraumatic cohort and a 75% rate of previous operations in the posttraumatic cohort) and the resulting selection bias. In the posttraumatic cohort, six of eight shoulders were treated with open reduction and internal fixation, and various kinds of primary fracture patterns were included. Moreover, two different implants were used during the present study. Malunion of the tuberosities and internal rotation contractures of the shoulder were found in five of eight cases in the posttraumatic group and in only one case in the nontraumatic group. Furthermore, the mean stage of osteonecrosis was higher in the posttraumatic group. The differences that were found between the two cohorts could be influenced by the different mean stages of osteonecrosis; however, we were able to demonstrate that cementless humeral surface replacement arthroplasty is a viable treatment option in cases of both nontraumatic and posttraumatic osteonecrosis of the humeral head.
Low revision rates and good clinical outcomes have been described for cementless humeral surface replacement arthroplasty for other etiologies18,33-35,37. The results of the present study seem comparable with those findings. To date, only one reoperation has been necessary. Glenoid erosion leading to persistent pain and the need for surgical intervention was not found during the relatively short follow-up period (mean, three years).
In the present study, we found cementless humeral surface replacement arthroplasty to be a viable bone-preserving treatment option for patients who have stage-2, 3, or 4 traumatic or nontraumatic osteonecrosis of the humeral head, even in the presence of a substantial bone defect. Levy et al. stated that a minimum of 60% of the humeral head bone stock is necessary to achieve sufficient primary stability with a cementless resurfacing implant34. In the present study, we demonstrated that cementless humeral surface replacement arthroplasty is a viable treatment option even in cases of bone stock necrosis of up to 31%. To date, no evidence of loosening or insufficient fixation of the cementless implant has been found.
While good clinical results and a low complication rate were found in the short term, long-term investigations are necessary to confirm these observations.