Anatomic versus reverse shoulder arthroplasty: a mid-term follow-up comparison

Abstract

Background

Anatomic total shoulder arthroplasty improves pain and function with a reported reoperation rate of approximately 1% per year. With improved glenoid fixation, reverse shoulder arthroplasty implants may outperform anatomic total shoulder arthroplasty. We evaluate the functional outcomes and reoperation rate of anatomic total shoulder arthroplasty versus reverse shoulder arthroplasty at a minimum eight-year follow-up or revision.

Methods

Between 2005 and 2010, 187 shoulders (137 anatomic total shoulder arthroplasty, 50 reverse shoulder arthroplasty) were retrospectively reviewed at a mean of 8.8 years. The mean age at surgery was 67 years. Females were more commonly treated with reverse shoulder arthroplasty. Both groups had similar body mass index and comorbidities. Outcome measures evaluated included abduction, forward elevation, external rotation, internal rotation, Simple Shoulder Test, Constant score, American Shoulder and Elbow Score, University of California Los Angeles Shoulder score, and Shoulder Pain and Disability Index.

Results

At follow-up, anatomic total shoulder arthroplasty demonstrated greater overhead range of motion and external rotation. All patient-reported outcomes remained similar between groups. Reverse shoulder arthroplasty patients were more likely to rate shoulders as much better or better after surgery (90% versus 67%, p = 0.004). Complications were observed in 24% of anatomic total shoulder arthroplasties and 8% of reverse shoulder arthroplasties (p = 0.02). Reoperation was more common in anatomic total shoulder arthroplasties (23% versus 4%, p = 0.003).

Discussion

At mid-to-long-term follow-up, reverse shoulder arthroplasties demonstrated significantly fewer complications and reoperations than anatomic total shoulder arthroplasties. Despite similar patient-reported outcomes, reverse shoulder arthroplasty patients were more likely to be satisfied with their shoulder.

Introduction

Anatomic total shoulder arthroplasty (TSA) has been documented to provide reliable improvements in pain and range of motion (ROM) for end-stage glenohumeral arthritis.^1–6 When compared to reverse shoulder arthroplasty (RSA), internal and external rotation is greater following TSA.^7,8 However, similar patient-reported outcomes (PROs) are expected regardless of implant type.^7–9 Long-term outcomes following RSA remain limited, with formal FDA clearance of this prosthesis coming over 30 years after the introduction of TSA.^10–12 At early and midterm follow-up, RSA continues to demonstrate similar clinical outcomes and revision rates as TSA despite expanding indications.^13–15

Concern remains regarding glenoid component loosening and rotator cuff failure over time in TSA.^16,17 Schoch et al. ¹⁸ reported an average rate of reoperation of 1% per year for any cause following TSA. Because of high rates of radiographic glenoid loosening and concern for implant longevity, some surgeons believe that RSA implant longevity will outperform TSA due to improved glenoid fixation. These concerns have led to the use of RSA even in the setting of an intact rotator cuff. ¹⁹ However, concern still remains due to early reports of high postoperative complications following RSA.^20,21 The purpose of this study is to evaluate the functional outcomes, complications, and reoperations of TSA versus RSA at a minimum eight-year follow-up.

Methods

Following IRB approval, a retrospective review of a multicenter shoulder arthroplasty database was performed. Six institutions were included. The database was queried for all primary shoulder arthroplasties (TSA and RSA) performed between January 2005 and March 2010 using a single implant system (Equinoxe, Exactech, Gainesville, FL). This implant system uses a medialized glenosphere with lateralized on-lay humerus and a 145° neck shaft angle for the RSA. Humeral stems were cemented in 14% of TSA and 20% of RSA. All glenoid components were standard designs without augmentation, as these were not released until 2011. Shoulders were followed for a minimum of eight years or until revision. Shoulders with acute fractures were not included. One hundred and eighty-seven shoulders were identified in 73 males and 114 females. The mean age at arthroplasty was 67 (range, 41–83) years. One hundred and thirty-seven shoulders were treated with TSA and 50 with RSA components. Shoulders were evaluated at an average follow-up of 8.8 years (range, 2–10). Nineteen shoulders that underwent revision prior to eight years were included in the analysis with follow-up data immediately preceding revision surgery. No shoulders with clinical follow-up less than eight years without revision were included. Follow-up was slightly longer in the TSA group (8.8 (1.6) versus 8.5 (0.9) years, p = 0.05).

Both groups were similar in regards to body mass index, comorbidities, and treatment with prior corticosteroid injections (see Table 1). Osteoarthritis was the most common diagnosis for shoulders undergoing TSA (88%). A greater distribution of preoperative diagnoses was seen for shoulders undergoing RSA, with cuff tear arthroplasty being the most common indication for surgery (70%). A full list of diagnoses is included in Table 1. Females and patients with prior surgical procedures were more commonly treated with RSA.

Table 1.

Demographic information.

	TSA (137)	RSA (50)	p value
Age a	65.7 (41–82; 7.8)	71.7 (53–83; 6.2)	<0.001
Sex (M/F)	60/75 (2 not recorded)	13/37	0.02
Height (in.) a	66 (58–75, 4.1)	63 (55–75, 4.2)	<0.001
Weight (lb) a	186 (106–360, 49)	163 (99–270, 36)	0.004
BMI a	29.8 (18–59, 6.7)	28.7 (19–53, 6.4)	0.4
Prior surgery	9 (7%)	19 (38%)	<0.001
Injection	27 (20%)	13 (26%)	0.4
Follow-up (months) a	106 (24–155, 20)	102 (59–121, 10.2)	0.05
Preop diagnosis			<0.001
Osteoarthritis	121	6
Osteonecrosis	3	0
Rotator cuff tear	0	4
Cuff tear arthropathy	0	35
Inflammatory arthritis	4	2
Posttraumatic arthritis	7	2
Post capsulorraphy arthropathy	1	0
Tumor	1	1
Comorbidities
Hypertension	30%	23%	0.5
Coronary artery disease	7%	8%	1.0
Diabetes	7%	4%	1.0
Nicotine abuse	4%	0%	0.6

BMI: body mass index; RSA: reverse shoulder arthroplasty; TSA: anatomic total shoulder arthroplasty.

^aResults presented as mean (range, standard deviation).

All shoulders were followed routinely as part of the multi-institutional database protocol. Surgeons or clinical research assistants evaluated each patient prior to surgery and at routine follow-up visits including annual visits after one year postoperatively. ROM, strength, and PROs were obtained at each visit in accordance with the database protocol. Abduction, forward elevation, and external rotation were measured in degrees. Internal rotation was assessed as the highest vertebral level reached by the thumb and recorded according to the scale previously described by Flurin et al. ⁹ In this scale, abdomen or less = 0, hip = 1, buttocks = 2, sacrum = 3, L5–L4 = 4, L3–L1 = 5, T12–T8 = 6, and T7 or higher = 7. PROs obtained included the following: the Simple Shoulder Test (SST), Constant score, American Shoulder and Elbow Score (ASES), University of California Los Angeles (UCLA) shoulder score, and the Shoulder Pain and Disability Index. Subjective postoperative satisfaction was assessed as much better, better, unchanged, or worse compared to before surgery. The database was also reviewed for physician-reported complications and reoperation.

Postoperative radiographs (Grashey and axillary lateral) were evaluated by the performing surgeon using a standardized follow-up form. Humeral lucencies were graded according to Sanchez-Sotelo et al.^22,23 Humeral loosening was considered when lines of 2 mm or more were identified in three or more contiguous zones, or the component was noted to have shifted in position. ²⁴ Glenoid component lucencies for TSA were graded according to the Lazarus scale. ²⁵ Notching was assessed according to the Sirveaux scale. ²⁶

Statistical analysis

Statistical analysis was performed using SPSS (Version 24, Armonk, NY: IBM Corp). Descriptive statistics are described as mean (range) or mean (standard deviation) for continuous measures and number (percentage) for discrete variables. A Student’s t-test was used to compare continuous variables. A Wilcoxon-rank-sum test was used to assess ordinal variables. A Chi-squared test or Fisher’s exact test was used to compare nominal data depending on sample size. The alpha level for all tests was set at 0.05 for statistical significance.

Results

Prior to index arthroplasty, both TSA and RSA groups demonstrated similar overhead ROM. Active internal and external rotation was greater in the RSA group prior to surgery. Despite the differences in internal and external rotation, both groups demonstrated similar preoperative PROs (see Table 2 for full details). At final follow-up, TSA shoulders demonstrated greater ROM overhead and externally (see Table 2). However, all PROs remained similar between groups at follow-up, except UCLA scores, which were better in the RSA group (28.9 versus 26.3, p = 0.03). When evaluating improvements preoperatively to postoperatively, the TSA group demonstrated significantly greater improvements in ER (31° versus 4°, p < 0.001) and IR (1.7 versus 0.3, p = 0.001) compared to the RSA group. Gains in forward elevation were similar between TSA and RSA (37° versus 33°, p = 0.9). Improvements in all PROs were similar between groups (see Table 3). Subjective satisfaction was greater in shoulders treated with RSA. Ninety percent of RSA patients rated their shoulder as much better or better than before surgery, compared to 67% of TSA patients. Additionally, 19% of TSA patients rated their shoulder as worse, compared to 2% of RSA patients.

Table 2.

Pre- and postoperative outcomes by implant type.

	TSA (137)	RSA (50)	p value
Preopa
Pain	6.3 (2–10, 2.0)	6.3 (1–10, 2.0)	0.9
Abduction	81 (20–160, 25.2)	75 (30–130, 32.1)	0.3
Forward elevation	98 (30–170, 29.3)	95 (15–180, 46.6)	0.7
IR	2.9 (0–7, 1.4)	4.3 (1–7, 1.8)	<0.001
ER	11 (−35 to 65, 20.3)	24 (−30 to 85, 23.3)	0.001
SST	3.5 (0–9, 2.7)	2.8 (0–8, 1.9)	0.4
Constant	38.8 (9–66, 13.0)	35.6 (12–69, 15.0)	0.3
ASES	35.6 (2–70, 15.3)	34.9 (12–65,12.6)	0.8
UCLA	14.1 (6–22, 4.0)	13.9 (8–22, 3.7)	0.9
SPADI	82.5 (34–130, 20.7)	81.4 (45–115, 17.9)	0.8
Postopa
Pain	2.8 (0–10, 3.1)	1.7 (0–9, 2.3)	0.07
Abduction	113 (15–180, 31.5)	105 (55–160, 23.2)	0.08
Forward elevation	135 (0–180, 34.7)	128 (45–175, 27.2)	0.2
IR	4.6 (0–7, 1.4)	4.6 (1–7, 1.8)	0.9
ER	42 (–20 to 90, 9.3)	28 (−30 to 80, 23.1)	<0.001
SST	8.7 (0–123.4)	9.1 (0–12, 2.8)	0.9
Constant	63.5 (12–86, 16.5)	64.3 (24–88, 12.7)	0.7
ASES	69.8 (8–100, 26.5)	74.8 (22–100, 20.0)	0.2
UCLA	26.3 (8–35, 8.0)	28.9 (13–34, 5.0)	0.03
SPADI	35.5 (0–123, 32.0)	33.2 (0–94, 27.1)	0.6
Satisfaction
Much better	67	34	0.004
Better	28	11
Unchanged	19	4
Worse	23	1

ASES: American Shoulder and Elbow Surgeons; ER: external rotation; IR: internal rotation; RSA: reverse shoulder arthroplasty; SPADI: Shoulder Pain and Disability Index; SST: Simple Shoulder Test; TSA: anatomic total shoulder arthroplasty; UCLA: University of California Los Angeles.

^aResults reported as mean (range, standard deviation).

Table 3.

Improvement in outcomes by implant type.

Improvement versus preop
	TSA (137)	RSA (50)	p value
Pain	3.5 (−3 to 10, 3.2)	4.6 (−1 to 9, 2.7)	0.3
Abduction	32 (−80 to 115, 37.8)	30 (−30 to 115, 36.5)	0.9
Forward elevation	37 (−110 to 150, 39.3)	33 (−50 to 110, 37.6)	0.9
IR	1.7 (−4 to 6, 2.0)	0.3 (−5 to 5, 2.5)	0.001
ER	31 (−25 to 115, 25.2)	4 (−80 to 70, 31.4)	<0.001
SST	5.2 (−5 to 12, 3.1)	6.3 (1 to 11, 2.5)	0.1
Constant	24.7 (−28 to 72, 18.7)	28.7 (−7 to 58, 14.1)	0.7
ASES	34.2 (−35 to 95, 27.7)	39.9 (0 to 76, 19.2)	0.3
UCLA	12.2 (−4 to 27, 7.9)	15 (2 to 24, 5.3)	0.6
SPADI	47.0 (−37 to 111, 34.2)	48.2 (8 to 95, 23.9)	0.6

ASES: American Shoulder and Elbow Surgeons; ER: external rotation; IR: internal rotation; RSA: reverse shoulder arthroplasty; SPADI: Shoulder Pain and Disability Index; SST: Simple Shoulder Test; TSA: anatomic total shoulder arthroplasty; UCLA: University of California Los Angeles.

Results presented as mean (range, standard deviation).

The presence of any humeral line (yes/no) was similar between TSA and RSA (23% versus 26%, p = 0.8). The average number of zones with a lucent line was 0.4 (range, 0–8) in the TSA group and 0.5 (range, 0–3) in the RSA group (p = 0.6). The most common zone to show a humeral lucent line was zone 8, followed by zone 7, for both groups. Anatomic glenoid radiolucent lines were present in 47% of shoulders. Lucencies were graded as 1 (7), 2 (9), 3 (11), 4 (3), and 5 (7). Notching was present in 19% of RSA. These were graded as 1 (2), 2 (3), and 3 (3).

Complications were observed in 33 patients (24%) treated with TSA and 4 (8%) with RSA (p = 0.02). The most common complication following TSA was rotator cuff tearing (14), followed by radiographic aseptic glenoid loosening (11) and infection (6). One patient also developed aseptic humeral loosening and another combined humeral/glenoid loosening. The mean time to glenoid component loosening in the TSA group was 8.1 years postoperatively (range, 6.6–9.8). Postoperative complications following RSA included instability (1), periprosthetic fracture (1), aseptic glenoid loosening (1), and infection (1). Reoperation was more common in the TSA group (31 shoulders, 23%) compared to the RSA group (2 shoulders, 4%; p = 0.003). Indications for reoperation in the TSA group included the following: rotator cuff tear (13), aseptic glenoid loosening (10), infection (6), humeral loosening (1), and combined humeral/glenoid component loosening (1). Indications for reoperation in the RSA group included infection (1) and glenoid component loosening (1).

Discussion

Early reports on RSA documented complication rates in excess of 19%.^20,21 However, with design advancement and improved techniques, complications for both RSA and TSA have decreased. ²⁷ In this study, RSA showed a significantly lower rate of reoperation and complications compared to TSA. This is extremely important as the rate of shoulder arthroplasty increases. Surgeons continue to compare RSA to the historical performance of TSA when considering which implant to use. Furthermore, at a minimum eight-year follow-up, RSA also demonstrated significantly greater patient satisfaction compared to TSA.

Long-term reports on RSA remain limited.^12,28 Gerber et al. ¹² reported on 22 RSAs performed with a Grammont design implant at a mean follow-up of 16 years (minimum, 15 years). Similar to this study, RSA demonstrated durable improvements in ROM and function. Gerber et al. reported an average Constant score of 58, which was stable over time. This is similar to our series where the mean Constant score was 62.6. However, Gerber did not include the six shoulders with clinical failure in the analysis of the Constant score. Postoperative ROM in this study was greater than that in the Gerber series (forward elevation 126 versus 101, abduction 104 versus 86, external rotation 26 versus 18). This may be due to the expanded indications for RSA in this series and changes in design features (lateralized humerus and 145° humeral neck shaft angle). It remains unclear if this group will lose overhead ROM with further follow-up as reported by Gerber et al. ¹² and Bacle et al. ²⁸

In this study, patients demonstrated significant improvements in ROM and PRO regardless of the type of procedure performed. At a minimum follow-up of eight years, shoulders undergoing TSA demonstrated greater ROM compared to shoulders undergoing RSA; however, only external rotation exceeded the MCID as described by Simovitch et al. ²⁹ This is similar to Steen et al., who showed greater postoperative ROM in all directions following TSA compared to a study cohort of shoulders treated with RSA at a mean follow-up of 3.5 years. Despite the differences in motion, Steen et al. ³⁰ also showed similar ASES and SST between TSA and RSA. Similarly, Kiet et al. ³¹ evaluated a cohort of 47 TSA and 53 RSA at two-year follow-up and demonstrated greater ROM following TSA but no difference in ASES scores. Lower ROM following RSA may be related to a greater prevalence of rotator cuff dysfunction in the RSA population as well as the increased age in the RSA cohort. ⁸ This is supported by previous studies comparing TSA and RSA in shoulders with primary osteoarthritis and an intact rotator cuff tear that have failed to show differences in postoperative ROM.^30,32 Another possible cause for the lower ROM in the RSA group may be the significantly higher rate of prior operations (38% versus 7%, p < 0.001). Frank et al. ³³ have previously shown that shoulders with prior surgery had significantly less forward elevation compared to shoulders without prior surgery (133.8 versus 142.3, p < 0.001).

The clinical significance of these ROM improvements should be questioned, as both groups demonstrated similar PROs at all time points. Cox et al. ³⁴ evaluated a cohort of 19 patients who underwent TSA on one side and RSA on the other. Similar to our results, ROM was better following TSA, but no difference in PROs was appreciated. When asked which shoulder they preferred, 68% preferred the RSA side, suggesting that ROM limits may not have a significant functional impact. Despite lower ROM in the RSA group, subjective satisfaction in our cohort was significantly higher in the RSA group at long-term follow-up. Shoulders treated with RSA were more likely to rate their shoulder as much better or better than before surgery (90% versus 67%). This may be in part due to the higher rate of complications and reoperations seen in the TSA group.

Villacis et al. ³⁵ previously reported a higher complication in RSA shoulders compared to TSA based on the California Office of Statewide Health Planning and Development state discharge database. Within two years of index arthroplasty, RSA was shown to have a significantly higher rate of complications (21.8 versus 14.3%, p < 0.001). ³⁵ This finding is in contrast to our study, where complications were lower following RSA (8% versus 24%). ³⁵ The rate of complications following TSA was higher in this study compared to Villacis et al. ³⁵ (24% versus 14.3%). The rate of TSA complications reported by Villacis et al. ³⁵ is likely under-representative of the true complication rate, as complications were only captured when the patient was readmitted. Glenoid component loosening without revision and periprosthetic fractures are is most commonly treated non-operatively outside of the inpatient/emergency department setting and would not be captured using their state-wide database set. The lower rate of complications following RSA in this study (8% versus 21.8%) may be in part due to surgeon experience. All surgeons in this study were fellowship trained, whereas Villacis et al.’s study included all surgeons in the state regardless of sub-specialty training. Hammond et al. ³⁶ have previously demonstrated that low-volume surgeons have a higher rate of postoperative complications following shoulder arthroplasty compared to their high-volume peers. Similarly, Walch et al. ³⁷ have previously shown a decreasing rate of complications with increased surgeon experience. The difference in surgical expertise may in part explain the difference in complications between studies. Our results are in contrast to Kiet et al., ³¹ who reported a similar rate of complications at two years between TSA and RSA. The higher complication rate seen following TSA in this study may be explained by longer follow-up and the occurrence of glenoid component loosening over time. In this study, 8% of TSA had radiographic glenoid loosening, compared to 2% of RSA. Schoch et al. ³⁸ have previously shown that glenoid component loosening is associated with worse overhead ROM and PRO. Additionally, 10% of TSA had a documented rotator cuff tear after surgery, which may have affected their postoperative function. ³⁹ The combination of these factors may have contributed to the greater dissatisfaction in the TSA group (19% worse than before surgery versus 2% of RSA).

Similar to postoperative complications, a significantly higher rate of reoperation was documented following TSA (23% versus 4%, p = 0.003). At earlier follow-up, other studies have documented similar rates of reoperation between TSA and RSA two years postoperatively.^31,35 However, beyond two years of follow-up, the cumulative rate of rotator cuff tears and aseptic glenoid loosening continues to grow, which likely accounts for the differences in reoperation following TSA. ¹⁸ This likely explains the higher rate of complications in the TSA group in this midterm follow-up study where 77% of the revisions were performed secondary to rotator cuff tearing or aseptic glenoid loosening. Our study demonstrates a lower rate of reoperation in the RSA group compared to historical reports with longer term follow-up.^12,28 At a mean follow-up of 12 years, Bacle et al. ²⁸ reported a reoperation rate of 12%, and Gerber et al. ¹² reported a 55% reoperation rate at a mean follow-up of 16 years (minimum, 15 years). It remains unclear if the difference in follow-up time may account for difference in reoperation rate. However, it is important to note that both of these studies used Grammont-style designs and were performed during the learning curve of RSA. Additionally, the Grammont design has previously been shown to have higher rates of scapular notching, which may lead to higher rates of polyethylene wear and component loosening.^40–44

The study has multiple limitations. First, surgeries were performed across six different centers, which resulted in variable operative indications, surgical techniques, and rehabilitation protocols. Controlling for preoperative diagnosis was not possible, given the historical variation in indications for these two procedures. However, we feel the comparison is still relevant as postoperative restrictions are similar following both procedures. Future studies may be better able to compare this implant based on a diagnosis of primary osteoarthritis, but this indication is relatively new and not yet widely accepted. ¹⁹ Despite varying diagnoses and poorer postoperative ROM with RSA, improvements in all PROs remained similar between groups. The difference in revision rate may partially be explained by the fact that RSA remains an option for failed TSA. However, satisfaction remained significantly higher in the RSA group, which suggests that there was not a large percentage of RSA patients unsatisfied with their shoulder without further operative options. Second, preoperative advanced imaging was not routinely available for all shoulders. Thus, we were unable to evaluate the effect of preoperative glenoid bone loss on the higher rates of complications, which was commonly due to glenoid loosening following TSA. It is possible that posterior bone loss and/or subluxation may have contributed to the higher failure rates with TSA.^45–47 It remains possible that some patients treated with TSA may have been better treated with RSA as these risk factors have become more well known over the last decade. Lastly, clinical and radiographic follow-up was completed by the surgeon performing the index operation leading to the potential for self-evaluation bias. Use of a standardized follow-up evaluation is used in an attempt to decrease variability between evaluators.

Anatomic total shoulder arthroplasty produces greater postoperative ROM compared to RSA shoulders, which is likely related to underlying rotator cuff disease in patients treated with RSA. However, these differences did not appear to meaningfully impact daily function as demonstrated by similar PROs regardless of the procedure performed at mid- to long-term follow-up. In this small cohort of patients, complications and reoperations were significantly lower in the RSA group at similar follow-up. At mid-to-long-term follow-up, RSA can be expected to perform similarly to TSA with a lower risk of complications and reoperation.

At mid-to-long-term follow-up, both TSA and RSA reliably improve pain and function. However, RSA is more likely to result in patient satisfaction with fewer complications and lower rates of reoperation. When compared to TSA function, only the difference in postoperative external rotation exceeds the MCID. This likely has minimal effect on daily life, as all improvements in PROs were similar between groups. Surgeons should consider these differences when counseling patients who may be considered for both procedures.

Ethical Review and Patient Consent

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. This study was IRB approved (WIRB 1112376) and each patient gave written consent for their deidentified information to be used in this study.

ORCID iDs

Bradley S Schoch https://orcid.org/0000-0002-9355-5069

Joseph J King https://orcid.org/0000-0002-9201-9408