Publications from the DSR have described inferior patient-reported outcomes and a high revision rate of resurfacing hemiarthroplasty compared with stemmed total shoulder arthroplasty. Total shoulder arthroplasty is now the standard treatment for patients with end-stage osteoarthritis and an intact rotator cuff.
Arthroplasty for glenohumeral osteoarthritis
In Denmark, resurfacing hemiarthroplasty was the most common arthroplasty for osteoarthritis until 2012. The reason for this is unknown but may be related to concerns about late complications with glenoid loosening and the promising early results of resurfacing hemiarthroplasty. Levy et al. observed similar Constant scores for total resurfacing arthroplasties (
n = 39) and resurfacing hemiarthroplasties (
n = 30). They hypothesized that resurfacing arthroplasty had the advantage of a bone-preserving design, short operation time, and an easy revision, should the need for revision arthroplasty arise, and they recommended resurfacing hemiarthroplasty for osteoarthritis except in patients with nonconcentric or saddle-shaped erosion of the glenoid [
5]. A systematic review published in 2009 supported their results and concluded that resurfacing hemiarthroplasty is a viable option for shoulder replacement, especially in young patients [
2].
The DSR observed less promising results with unpredictable and disappointing patient-reported outcomes as well as a high rate of revision especially in young patients. Furthermore, there were poor patient-reported outcomes of revision arthroplasty after failed resurfacing hemiarthroplasty belying the hypothesis of an easy revision [
14,
16]. For the past 5 years, the DSR has advocated not using resurfacing hemiarthroplasty for osteoarthritis.
The DSR has advocated not using resurfacing hemiarthroplasty for osteoarthritis
According to the DSR, the most common reason for revision after resurfacing hemiarthroplasty was glenoid wear [
16]. This indicates that the rates of revision could have been improved if a glenoid component had been used. However, glenoid exposure with an intact humeral head is difficult and associated with a high risk of neurological complications [
8], and total resurfacing arthroplasty has only been used in very few cases during the existence of the DSR.
Anatomical total shoulder arthroplasty is now the most common type of arthroplasty for osteoarthritis in Denmark. Data from the DSR and NARA have shown superior patient-reported outcomes 1 year postoperatively and low rates of revision for anatomical total shoulder arthroplasty compared with hemiarthroplasty. This supports the findings and conclusions of a systematic review [
18] and the guidelines of the American Academy of Orthopedic Surgeons [
4] recommending anatomical total shoulder arthroplasty for osteoarthritis in patients with an intact rotator cuff.
In the NARA dataset there was no difference in short-term survival rates between stemmed total shoulder arthroplasty and stemless total shoulder arthroplasty. Very few studies have compared the two arthroplasty types and none of them has been able to find any statistically significant differences in functional outcome scores [
1,
17,
19]. Thus, it seems that stemmed total shoulder arthroplasty and stemless total shoulder arthroplasty perform equally and that the choice of the humeral component can be based on the surgeon’s preference. However, this needs to be confirmed by a large randomized clinical trial with long-term follow-up.
Other national registries have examined how the method of fixation and bone morphology of the glenoid can influence the longevity of the arthroplasty. The New Zealand Joint Registry included 1596 total shoulder arthroplasties with 1065 (67%) cemented and 531 (33%) uncemented glenoid components. The revision rate was 4.4 times higher for uncemented glenoid components than for cemented components [
3]. A similar study was published by the Australian Joint Registry, in which Page et al. included 10,805 total shoulder arthroplasties for osteoarthritis with 7646 (71%) cemented and 3159 (29%) uncemented glenoid components. The 10-year cumulative revision rate was approximately 7% and 22%, respectively. The HR for revision was 4.77 for uncemented glenoid components, with the cemented component as the reference [
7].
Benefits of a national registry
One of the greatest advantages of registry studies is their ability to undertake comprehensive long-term follow-up, and because of the large numbers studied, differences in arthroplasty survival rates can be analyzed with sufficient statistical power. This is rarely possible in randomized clinical trials and longitudinal studies where the limited number of patients and the low number of revision arthroplasties make it difficult to find any statistically significant differences between arthroplasty types. Furthermore, because of the large number of patients studied, it is also possible to estimate HRs as a measure of the risk of revision and thereby compare subgroups of arthroplasty types with adjustment for gender, age, and other factors that might influence the risk of revision. Finally, registry studies can report rare events such as a specific reason for revision. This could include the risk of revision because of infection or the risk of loosening for anatomical total shoulder arthroplasty.
In registry studies there is no control over who is included and what type of arthroplasty is used for a specific indication, and because the patients are not randomly allocated, there is the possibility of different distributions of covariables that could influence the outcome. This makes registry studies unsuitable for comparing treatment effect between arthroplasty types. However, the patients in randomized clinical trials and longitudinal studies are often operated on by a few experienced surgeons and the results cannot always be reproduced by other surgeons. By contrast, registry studies report the outcome of arthroplasty surgery performed by all surgeons on a national or even multinational level. This adds high external validity to the registry studies. Furthermore, in randomized clinical trials it can be laborious and difficult to study differences in outcome between arthroplasty types in smaller subgroups such as in young patients.
Methodological considerations
We acknowledge the limitations inherent to registry studies. The decision to use a specific implant type might be based on factors that are not recognized by the registry: comorbidity; functional demands; glenoid wear; rotator cuff status; and whether the bone stock of the glenoid is considered too poor for a glenoid implant. This might lead to selection bias. Another limitation of the DSR is the lack of a preoperative WOOS score. Any differences in preoperative WOOS score between the arthroplasty types might influence the WOOS score at 1 year. Furthermore, not all patients returned a complete WOOS questionnaire, and any systematic differences in WOOS between responders and nonresponders would have an influence on the results and their interpretation. Finally, incorrect reporting may diminish the accuracy and reliability of the registry data.
The inclusion of bilateral procedures in the survival analyses violates the assumption of independency, but this is probably of minor practical relevance. Another concern is that when arthroplasty survival rates are reported, patients are censored and no longer at risk of revision if they die or emigrate. This violates the assumption of competing risk. The Cox regression model is, however, recommended when reporting HRs after arthroplasty surgery. Even so, statistical limitations are worth considering when the results are interpreted.