A Comparison of Revision Rates for Osteoarthritis of Primary Reverse Total Shoulder Arthroplasty to Primary Anatomic Shoulder Arthroplasty with a Cemented All-polyethylene Glenoid: Analysis from the Australian Orthopaedic Association National Joint Replacement Registry

Abstract

Background 

There has been decreased use of anatomic total shoulder arthroplasty (aTSA) because reverse TSA (rTSA) is increasingly being used for the same indications. Although short-term studies generally have not found survivorship differences between these implant designs, these studies are often small and their follow-up is limited to the short term. Likewise, the degree to which patient characteristics (such as gender, age, and American Society of Anesthesiologists [ASA] score) may or may not be associated with survivorship differences calls for larger and longer-term studies than is often possible in single-center designs. Large national registry studies may be able to help answer these questions.

Questions/purposes 

By analyzing a large Australian registry series of primary aTSAs with cemented all-polyethylene glenoids and rTSA for osteoarthritis (OA), we asked: (1) Is the revision risk for OA higher for aTSA with all-polyethylene glenoids or for rTSA, adjusting for patient characteristics such as age, gender, ASA score, and BMI? (2) Is the patient’s gender associated with differences in the revision risk after controlling for the potentially confounding factors of age, ASA score, and BMI?

Methods 

In this comparative, observational registry study performed between January 1, 2015, and December 31, 2019, all primary aTSAs with all-polyethylene glenoids and rTSA for OA as determined by the treating surgeon and reported to our national registry formed two groups for analysis. The study period was set to time-match for the collection of ASA score and BMI in 2012 and 2015, respectively. Our registry enrolls more than 97% of all shoulder arthroplasties undertaken in Australia. There were 29,294 primary shoulder arthroplasties; 1592 hemiarthroplasties, 1876 resurfacing and stemless shoulders, 269 stemmed, and 11,674 reverse shoulder arthroplasties were excluded for other diagnoses. A total of 1210 metal-backed glenoids in stemmed aTSA for OA were excluded. A total of 3795 primary aTSAs with all-polyethylene glenoids and 8878 primary rTSAs for OA were compared. An aTSA with an all-polyethylene glenoid and rTSA were more likely to be performed in women (56% and 61% of patients, respectively). The mean age was 69 ± 8 years for aTSA with all-polyethylene glenoids and 74 ± 8 years for rTSA. One aTSA for OA was performed in a patient with an unknown glenoid type. The ASA score (n = 12,438) and BMI (n = 11,233) were also recorded. The maximum follow-up was 5 years for both groups, and the mean follow-up was 2.6 ± 1.4 years for aTSA with all-polyethylene glenoids and 2.1 ± 1.4 years for rTSA. The endpoint was time to revision (all causes), and the cumulative percent revision was determined using Kaplan-Meier estimates of survivorship (time to revision) and HRs from Cox proportional hazard models that were adjusted for age, gender, ASA score, and BMI category.

Results 

Overall, there were no differences in the 4-year cumulative percent revision between the groups; the 4-year cumulative percent revision was 3.5% for aTSA with all-polyethylene glenoids (95% CI 2.9%-4.2%) and 3.0% for rTSA (95% CI 2.6%-3.5%). There was an increased risk of revision of rTSA compared with aTSA using all-polyethylene glenoids in the first 3 months (HR 2.17 [95% CI 1.25-3.70]; p = 0.006, adjusted for age, gender, ASA score, and BMI). After that time, there was no difference in the rate of revision, with the same adjustments. In the first 3 months, men undergoing rTSA had a higher rate of revision than men with aTSA using all-polyethylene glenoids (HR 4.0 [95% CI 1.72-9.09]; p = 0.001, adjusted for age, BMI, and ASA). There was no difference between men in the two groups after that time. Women with aTSA using all-polyethylene glenoids were at a greater risk of revision than women with rTSA from 3 months onward (HR 2.77 [95% CI 1.55-4.92]; p < 0.001, adjusted for age, BMI, and ASA), with no difference before that time.

Conclusion 

Given the absence of survivorship differences at 4 years between rTSA and aTSA, but in light of the differences in the revision risk between men and women, surgeons might select an aTSA with an all-polyethylene glenoid to treat OA, despite the current popularity of rTSA. However, there are survivorship differences between genders. Future studies should evaluate whether our comparative findings are replicated in men and women undergoing aTSA with all-polyethylene glenoids and rTSA for primary diagnoses such as rheumatoid arthritis or post-traumatic arthritis, and whether there are functional differences between the two implant designs when used for OA.

Level of Evidence 

Level III, therapeutic study.

Introduction

Anatomic total shoulder arthroplasty (aTSA) and reverse TSA (rTSA) were initially designed to manage different end-stage shoulder conditions [^{6, 11, 14, 24, 31}]. With time, the indications for rTSA have expanded to include a wider range of pathologic findings [^{5, 13, 20, 22}]. In Australia, there has been an increase in the use of reverse shoulder arthroplasty; for osteoarthritis (OA), it is now the more frequently used shoulder arthroplasty option of those two designs [²]. The differences in survivorship for rTSA and aTSA for the indication of OA, if any, are not known [^{20, 22, 26, 29}].

In addition, a number of factors may be associated with poorer survivorship after aTSA or rTSA. These have included glenoid type [^{4, 10, 12, 25}] and patient characteristics including age, gender [^{8, 34}], ASA score [¹⁸], and BMI [³²]. However, to control for confounding among those variables, larger studies are needed than is usually possible in the context of the commonly performed single-center designs [^{21, 27, 28, 30}]. For this reason, we explored these themes using a large national registry.

We therefore asked: (1) Is the revision risk for OA higher for aTSA with all-polyethylene glenoids or for rTSA, adjusting for patient characteristics such as age, gender, ASA score, and BMI? (2) Is the patient’s gender associated with differences in the revision risk after controlling for the potentially confounding factors of age, ASA score, and BMI?

Patients and Methods

Study Design

This was a comparative, observational registry study. The Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR) began collecting data on September 1, 1999, and includes data on almost 100% of the hip and knee arthroplasty procedures performed in Australia since 2002. Data collection was expanded in April 2004 to include shoulder arthroplasty procedures, and the registry has documented 97.1% of shoulder arthroplasty procedures Australia-wide since November 2007 [¹⁵]. These data are externally validated against patient-level data provided by all Australian state and territory health departments. A sequential, multilevel matching process is used to identify any missing data, which are subsequently obtained by follow-up with the relevant hospital. Each month, in addition to internal validation and data quality checks, all primary procedures are linked to any subsequent revision involving the same patient, joint, and side. Data are also matched biannually to the Australian National Death Index data to identify patients who have died. The AOANJRR began collecting the American Society of Anesthesiologists (ASA) score in 2012 and BMI in 2015.

Patients

The study period was January 1, 2015 (to time-match to the collection of patient ASA score [beginning in 2012] and BMI [beginning in 2015]) to December 31, 2019. There were 29,294 primary shoulder arthroplasties, 3468 of which were excluded per the study criteria (hemiarthroplasty, resurfacing, or stemless arthroplasty). In addition, 269 stemmed anatomic and 11,674 reverse shoulder arthroplasties undertaken for primary diagnoses other than OA were excluded. To remove the known confounding effect of other glenoid types from the study [²⁵], we excluded 805 modular metal-backed and 405 non-modular metal-backed glenoids of stemmed aTSA for OA (Fig. 1). Two different cohorts were identified to compare the survivorship of rTSA for OA with that of aTSA for OA. These included all primary procedures undertaken for OA using either aTSA with all-polyethylene glenoids (stemmed) or rTSA.

Fig. 1: This flowchart shows the patients who were included in our study.

The study population consisted of 3795 primary aTSAs with all-polyethylene glenoids and 8878 primary rTSAs (Table 1). The mean age was 69 ± 8 years in the aTSA with all-polyethylene glenoid group and 74 ± 8 years in the rTSA group for OA. Women comprised 56% of the aTSA with all-polyethylene glenoid group and 61% of the rTSA group (Table 1).

Table 1. - Summary of primary TSA by type of procedure (primary diagnosis of OA)

Variable	Primary TSA (n = 3795)	Primary reverse TSA (n = 8878)	Total (n = 12,673)
Follow-up in years, mean ± SD	2.6 ± 1.4	2.1 ± 1.4	2.3 ± 1.4
Age in years, mean ± SD	69 ± 8	74 ± 8	72 ± 8
% of procedures revised (n)	3 (103)	2 (202)	2 (305)
Gender, % (n)
Men	44 (1661)	39 (3462)	40 (5123)
Women	56 (2134)	61 (5416)	60 (7550)
Age group, % (n)
< 55 years	5 (183)	1.3 (112)	2 (295)
55-64 years	24 (897)	10 (899)	14 (1796)
65-74 years	47 (1790)	41 (3592)	43 (5382)
≥ 75 years	24 (925)	48 (4275)	41 (5200)
ASAa grade, % (n)
1	7 (254)	3 (269)	4.2 (523)
2	56 (2068)	44 (3787)	47 (5855)
3	37 (1368)	51 (4405)	46 (5773)
4	0.9 (33)	3 (251)	2 (284)
5	0.1 (2)	0 (1)	0 (3)
BMIb class, % (n)
Underweight	0.3 (11)	0.6 (46)	0.5 (57)
Normal	14 (463)	15 (1214)	15 (1677)
Pre-obese	32 (1084)	35 (2741)	34 (3825)
Obese Class 1	31 (1039)	28 (2191)	29 (3230)
Obese Class 2	14 (480)	14 (1065)	14 (1545)
Obese Class 3	8 (270)	8 (629)	8 (899)

^aExcludes 235 procedures in patients with an unknown ASA class.

^bExcludes 1440 procedures in patients with unknown BMI.

The diagnosis of OA was selected by the treating surgeon at enrollment. This was a hierarchal, categorical selection from a list of 11 choices. Other diagnostic options included post-traumatic arthritis, rotator cuff arthropathy, rheumatoid arthritis or other inflammatory arthritis, or avascular necrosis. There was one aTSA for OA in a patient with an unknown glenoid type. Additional patient characteristics included the ASA score (excluding 235 of 12,438 unknown) and BMI (excluding 1440 procedures of 11,233 unknown) (Table 1).

Primary Study Outcome

The primary outcome measure for this study was the time to revision as calculated by the unadjusted cumulative percent revision. Comparisons of these revision rates used age-, gender-, ASA-, and BMI-adjusted HRs.

Ethical Approval

The AOANJRR is approved by the Commonwealth of Australia as a federal quality assurance activity under section 124X of the Health Insurance Act, 1973. All AOANJRR studies are conducted in accordance with the ethical principles of research (the Helsinki Declaration II).

Statistical Analysis

Kaplan-Meier estimates of survivorship were used to report the time to the first revision, with censoring at the time of death and closure of the dataset at the end of December 2019. The cumulative percent revision with 95% CI was calculated using unadjusted pointwise Greenwood estimates. The cumulative percent revision is displayed until the number at risk for the group reaches 40, unless the initial number for the group is less than 100, in which case the cumulative percent revision is reported until 10% of the initial number at risk remains.

Age-, gender-, ASA-, and BMI-adjusted HRs were calculated from Cox proportional hazard models to compare the rate of revision between groups. The assumption of proportional hazards was checked analytically for each model. If the interaction between the predictor and the log of time was statistically significant in the standard Cox model, then a time-varying model was estimated. Timepoints were selected based on the greatest change in hazard, weighted by a function of events. Timepoints were iteratively chosen until the assumption of proportionality was met, and HRs were calculated for each selected time period. For the current study, if no time period was specified, the HR was calculated over the entire follow-up period. All tests were two-tailed at 5% levels of significance. The statistical analysis was performed using SAS software, version 9.4 (SAS Institute Inc.).

Results

Adjusted Revision Risk: aTSA versus rTSA

Overall, there were no differences in the 4-year cumulative percent revision between the groups; the 4-year cumulative percent revision was 3.5% for aTSA with all-polyethylene glenoids (95% CI 2.9%-4.2%) and 3.0% for rTSA (95% CI 2.6%-3.5%) (Supplementary Table 1; Supplemental Digital Content 1, https://links.lww.com/CORR/A601). In the first 3 months, rTSA had a higher rate of revision than aTSA with all-polyethylene glenoids (HR 2.17 [95% CI 1.25-3.70]; p = 0.006) after adjusting for age, gender, ASA score, and BMI (Fig. 2). After 3 months, there was no difference between rTSA and aTSA with all-polyethylene glenoids (Fig. 2).

Fig. 2: This graph shows the cumulative percent revision of primary TSA by the type of primary procedure (primary diagnosis of OA).

Is Gender Associated with an Increased Revision Risk?

There was an increased risk of revision of rTSA compared with aTSA with all-polyethylene glenoids in the first 3 months for men (HR 4.0 [95% CI 1.72-9.09]; p = 0.001) after adjusting for age, BMI, and ASA score (Supplementary Table 2; Supplemental Digital Content 2, https://links.lww.com/CORR/A602). After 3 months, rTSA and aTSA with all-polyethylene glenoids were not different (Fig. 3). Women with aTSA using all-polyethylene glenoids were at a greater risk of revision than women with rTSA from 3 months onward (HR 2.77 [95% CI 1.55-4.92]; p < 0.001) after adjusting for age, BMI, and ASA score (Fig. 4). Before 3 months, for women, there was no difference between rTSA and aTSA with all-polyethylene glenoids (Fig. 4).

Fig. 3: This graph shows the cumulative percent revision of primary TSA by the type of primary procedure for men (primary diagnosis of OA).

Fig. 4: This graph shows the cumulative percent revision of primary TSA by the type of primary procedure for women (primary diagnosis of OA).

Discussion

In recent years, there has been increased use of rTSA in Europe, North America, and Australia [^{2, 17}]. This has mostly been reflected by increasing use in patients who have OA [^{2, 17, 20, 22}]. Although small series have evaluated the results of rTSA [^{1, 3, 7}], to evaluate the interactions of potentially important confounding variables—such as age, gender, and BMI—on implant survivorship, registry-based research is important. Our study, which is based on the Australian arthroplasty registry and which controlled for those important confounding variables, found that aTSA with all-polyethylene glenoids had the same revision risk as rTSA after 3 months. However, the risk of revision in the first 3 months was higher when rTSA prostheses were used. In addition, men had a lower revision risk for aTSA with all-polyethylene glenoids in the first 3 months than women did, but women were more at risk of revision of aTSA with all-polyethylene glenoids after 3 months. Although this supports the contemporary use of stemmed aTSA for OA, caution should be exercised in women. We believe that future research should consider survivorship for differing primary diagnoses, given the differences in revision risk outlined in this study.

Limitations

There are limitations to this study. The relative incidence of rTSA has increased over time in Australia [²], which reflects evolving surgical indications. This can complicate the interpretation of a registry study such as ours because we cannot be certain about the situations in which providers are choosing one implant type over the other. However, because we limited the time period of this study, the findings should apply to contemporary practice in which more surgeons seemingly are choosing rTSA in situations where aTSA would have been selected not long ago. We also mitigated selection biases by using one type of stemmed aTSA (with an all-polyethylene glenoid) and adjusting for patient characteristics to limit ambiguity in diagnoses and variations in cohort characteristics. In addition, although we could analyze aTSA with all-polyethylene glenoids, there are other designs in use, including hybrid glenoids with uncemented central pegs. These have promising early results [⁹] but they are not currently reported in our registry separately. This is a limitation, but probably not a significant one because we expect that few of these devices were used during our study period.

Patient-reported outcome measures and radiologic outcomes were not considered in this study, and we did not adjust for specific patient factors such as comorbidities (although we did control for ASA class). Because patient-reported outcome scores were not considered, some patients may have poorly performing and painful implants that were not revised. A study with the design we used, which focuses on revision as the endpoint, may be a best-case scenario for the implants we evaluated because we did not quantify the number of poorly functioning or painful implants. However, unless the proportion of poorly functioning or painful implants differs between aTSA and rTSA, we do not expect this to change the overall direction of the effect. However, clinical studies of patient-reported outcome scores generally find aTSA and rTSA to be comparable in the short term [^{20, 22, 26, 29}]. In addition, the time to revision (for all causes) is well-understood and can be measured in large numbers across countries without being subject to local variations in access of care. It also can be compared with the results from other large registries.

Adjusted Revision Risk: aTSA versus rTSA

In the first 3 months postoperatively, rTSA had an increased risk of revision compared with aTSA for the same indication, but they were not different after that time. This suggests that contemporary stemmed aTSA with all-polyethylene glenoids remains a viable option for primary OA. Previous studies have not demonstrated a clinical difference in revision rates between aTSA and rTSA; Kiet et al. [²⁰] reported that aTSA and rTSA had equivalent outcomes at 2 years, and Flurin et al. [¹³] analyzed 1145 aTSAs and rTSAs and found the same clinical outcomes and complication rates for the two types of prostheses at 9 years. However, our study had a much larger number of procedures than those studies did, which may explain the differences between our work and theirs. We cannot explain the difference between aTSA and rTSA based on our study alone. We considered patient characteristics that have been associated with changes in clinical outcome and survivorship such as age [²¹], gender [²³], obesity [³²], and comorbidities [^{16, 18}], but we controlled for these to minimize any confounding difference. We also excluded metal-backed glenoids because of their increased revision risk [^{2, 4, 25}]. Other potential associations such as glenoid morphology [³³] were not evaluated in this study because of limited enrollment of these data. Future studies comparing aTSA and rTSA survivorship should evaluate other primary diagnoses such as rheumatoid arthritis or post-traumatic arthritis.

The cumulative percent revision at 4 years for rTSA in this study was 3%, and it was 3.5% for aTSA with all-polyethylene glenoids. Despite the rising use of rTSA in multiple studies [^{2, 17}], we found no revision risk advantage over aTSA with an all-polyethylene glenoid for OA after 3 months. This is consistent with the survivorship of aTSA of 90.2% at 10 years reported by a single institution [³⁰], 87% in the Norwegian Arthroplasty Registry [¹²], and 96% in the Nordic Arthroplasty Register during the same period [²⁸]. The revision rates were also less than those for the first revision of shoulder arthroplasty from the same jurisdiction in a registry study (first revision cumulative percent revision of primary non-rTSA at 4 years = 15%, first revision cumulative percent revision of primary rTSA at 4 years = 18%) [¹⁵]. By comparison, other studies have reported higher primary revision rates for rTSA that ranged from 9% at 10 years in a registry study [²¹] to 10.1% at 2 to 7 years of follow-up in a systematic review [³⁵]. However, one institution reported an overall reoperation rate of 2.37% for procedures undertaken during a 7-year period [¹⁹]. Nationally, despite these results, other factors such as availability, ease of use, surgeon experience, and commercial prominence may also determine the choice. Our study supports the continued use of aTSA with all-polyethylene glenoids for OA, despite national trends. If there is no survivorship difference between aTSA with all-polyethylene glenoids and rTSA for OA, future research should pursue other avenues such as differences in economic cost, patient-reported satisfaction, or function.

Is Gender Associated with an Increased Revision Risk?

Our study showed that the patient’s gender was associated with a difference in the risk of revision for aTSA with all-polyethylene glenoids (for women) and rTSA (for men). In a British registry study, gender (men) combined with younger age were risk factors for both reoperation and revision [⁸]. Their 5-year cumulative percent probability of revision was 11.3 for women and 11.6 for men among all patients undergoing primary shoulder arthroplasty [⁸]. Wong et al. [³⁴] found inferior patient-reported outcomes for women who underwent rTSA but no differences in revision rates. We found that the patient’s gender was associated with different survivorship when comparing aTSA using all-polyethylene glenoids and rTSA. Although the outcome among men was the same as among the cohort population, overall, the outcome for women was the opposite. Although studies have noted gender differences, these differences are typically related to worse survivorship for men than for women, as we have found [^{8, 23, 34}]. We suggest that rTSA should be considered in women with OA if other patient characteristics make aTSA less favorable, such as glenoid deformity; this survivorship study supports that choice. Further research might examine whether gender is associated with survivorship differences in other primary diagnoses when comparing aTSA with all-polyethylene glenoids and rTSA.

Conclusion

After adjusting for multiple patient characteristics, we found a higher revision risk for rTSA in the first 3 months than for aTSA with all-polyethylene glenoids, with no difference after this time. Although this remained the case for men, for women, the revision risk was increased after 3 months in those undergoing aTSA with all-polyethylene glenoids. We conclude that for patients undergoing primary shoulder arthroplasty for OA, surgeons may, despite the popularity of rTSA, select an aTSA with an all-polyethylene glenoid. However, there are survivorship differences between genders. Future studies should evaluate whether there is a difference in the survivorship of aTSA and rTSA between men and women with other primary diagnoses such as rheumatoid arthritis and post-traumatic arthritis and whether there are functional differences between the two implant designs when used for OA.