Reverse Prostheses in Arthropathies With Cuff Tear: Are Survivorship and Function Maintained Over Time?

Abstract

Background

The use of reverse shoulder arthroplasty has considerably increased since first introduced in 1985. Despite demonstrating early improvement of function and pain, there is limited information regarding the durability and longer-term outcomes of this prosthesis.

Questions/purposes

We determined complication rates, functional scores over time, survivorship, and whether radiographs would develop signs of loosening.

Patients and Methods

We retrospectively reviewed 527 reverse shoulder arthroplasties performed in 506 patients between 1985 and 2003. Clinical and radiographic assessment was performed in 464 patients with a minimum followup of 2 years and 148 patients with a minimum followup of 5 years (mean, 7.5 years; range, 5–17 years). Cumulative survival curves were established with end points being prosthesis revision and Constant-Murley score of less than 30 points.

Results

Eighty-nine of 489 had at least one complication for a total of 107 complications. Survivorship free of revision was 89% at 10 years with a marked break occurring at 2 and 9 years. Survivorship to a Constant-Murley score of less than 30 was 72% at 10 years with a marked break observed at 8 years. We observed progressive radiographic changes after 5 years and an increasing frequency of large notches with long-term followup.

Conclusions

Although the need for revision of reverse shoulder arthroplasty was relatively low at 10 years, Constant-Murley score and radiographic changes deteriorated with time. These findings are concerning regarding the longevity of the reverse shoulder arthroplasty, and therefore caution must be exercised when recommending reverse shoulder arthroplasty, especially in younger patients.

Level of Evidence

Level IV, therapeutic study. See Guidelines for Authors for a complete description of levels of evidence.

Introduction

Nonconstrained shoulder arthroplasty reportedly provides high improvement of function and ROM for patients with shoulder arthritis [18]; however, function is substantially compromised in the presence of rotator cuff deficiency [1, 29]. The absence of a rotator cuff causes superior migration of the humeral head and an increased risk of glenoid loosening through a rocking horse effect [7]. The use of hemiarthroplasty avoids the problem of glenoid loosening, but four of 18 needed revision in the study of Arntz et al. [1], and five of 21 achieved less than 90° of active flexion after operation in the study of Sanchez-Sotelo et al. [22]. Progressive superior migration of the head also causes wear of the acromial arch and reportedly is associated with worsening function over time [22]. An alternative treatment is the constrained prosthesis with a fixed and lateralized center of rotation, but this is associated with complication rates ranging from 9% to 28% and loosening rates of 5% to 20% at 6.8 years [14, 15, 20].

The reverse shoulder prosthesis, initially proposed by Grammont for cuff tear arthropathy (CTA), reportedly improves pain and mobility at 2 years [2]. In one study, implant survival at 5 years was reported as 91% [25]. However, as a result of its semiconstrained nature, there remain concerns regarding an increased risk of wear and loosening [11, 25].

Distribution of this prosthesis was initially limited; however, many new models of reverse shoulder arthroplasty have been introduced and only limited followup has been reported (Table 1). Sirveaux et al. [25] reported one of the earliest series of reverse shoulder arthroplasty performed for CTA and found 91.3% implant survival rate at 5 years and a Constant-Murley score improvement of 42.3 points after an average of 44 months. Guery et al. [11] found 91% implant survival at a minimum followup of 5 years and substantially better survival rate in those patients with arthropathy associated with a massive cuff tear (MCT) than other indications and therefore suggested the best indication for a reverse shoulder arthroplasty was CTA. Wall et al. [27] reported other indications, especially revision surgery, were associated with a higher risk of complications, including the need for revision. Rittmeister and Kerschbaumer [21] reported three of eight shoulders developed glenoid loosening in a small series of patients with rheumatoid arthritis. From these studies, we can expect a high survival rate at 5 years in patients with CTA, but the mean followup used to determine survival curves was moderate (44 [25] to 69 months [11]) and the number of patients assessed after 5 years’ followup was low. Moreover, we do not know whether the survival rate and functional score persist after 5 years and whether there are progressive radiographic changes including signs of loosening.

Table 1.

Main series of reverse prostheses and their mean followup

Study	Year	Mean retrospective period (months)	Number of cases	Etiology
Rittmeister and Kerschbaumer [21]	2001	54	7	RA
Sirveaux et al. [25]	2004	45	80	CTA
Vanhove and Beugnies [26]	2004	31	32	CTA
Werner et al. [28]	2005	38	58	Varied
Seebauer et al. [23]	2005	18.2	57	CTA
Frankle et al. [6]	2005	33	60	CTA
Guery et al. [11]	2006	69.6	80	Varied
Boileau et al. [3]	2006	40	45	Varied
Levy et al. [17]	2007	44	19	Revisions
Bufquin et al. [4]	2007	22	43	Fracture
Gohlke and Rolf [8]	2007	31.5	34	Revisions
Wall et al. [27]	2007	39.9	191	Varied

RA = rheumatoid arthritis; CTA = cuff tear arthropathy.

We, therefore, determined complication rates, functional scores over time, survivorship, and whether radiographs would develop signs of loosening in patients with CTA who had reverse shoulder arthroplasties.

Patients and Methods

We retrospectively reviewed all 506 patients who underwent 527 reverse shoulder arthroplasties for CTA, MCT, or primary osteoarthritis with a deficient cuff from January 1, 1985, to December 31, 2003. The study was multicenter-organized by the French Orthopaedic Society. There were 383 women and 123 men, a gender ratio of approximately 3:1. Mean age at the time of surgery was 73 years (range, 40–90 years). The dominant side was involved in 76% of cases. In 21 cases, both shoulders were involved. One hundred thirty-four shoulders had undergone previous surgery for rotator cuff problems. At the time of the review, 12 patients died before 2 years’ followup; all had the prosthesis in place at the time of death. Five were lost to followup before 2 years. This left 489 patients with 509 prostheses. Of these 489 patients, 25 had prosthetic removal or conversion to hemiarthroplasty of their prostheses. This left 464 patients with 484 prostheses with a minimum followup of 2 years. Minimum followup was 2 years (mean, 4.5 years; range, 2–17 months) and mean age at review was 76.1 years (range, 50–103 years); 138 patients (148 shoulders) had a minimum followup of 5 years (mean, 7.5 years; range, 5–17 months), 65 (69 shoulders) had a minimum followup of 7 years, and 40 (41 shoulders) had a minimum followup of 9 years.

The patients were divided into three groups according to etiology and the five-grade classification of Hamada et al. [12]. That classification is based on radiographic features including narrowing of the subacromial space and degenerative changes of the glenohumeral joint. In our study, we had (1) 254 shoulders in a CTA group (Hamada et al. Grade 4 or 5); (2) 229 shoulders in a MCT group (Hamada et al. Grade 1, 2, or 3); and (3) 44 in a primary osteoarthritis group defined as glenohumeral arthritis with a deficient cuff (rupture of at least two tendons or fatty infiltration greater than Grade 2 according to the classification of Goutallier et al. [9]). That classification is a four-stage system based on the amount of fatty infiltration of rotator cuff muscles observed on CT scan.

Preoperatively, the Constant-Murley score [5] was 24.4 points (range, 0–68 points) and the relative Constant-Murley score [5] was 34.2 points (range, 0%–93%). The active anterior elevation was 70.5° (range, 0°–180°), the active external rotation was 8° (range, −45° to 80°), and the active external rotation in elevation was 23.5° (range, −20° to 90°).

The surgical approach was anterosuperior in 301 cases, deltopectoral in 215 cases, and a transacromial approach according to Baulot et al. [2] in 11 cases. The implant was a Delta^TM prosthesis (DePuy, Warsaw, IN) in 461 cases and an Aequalis Reversed^TM (Tornier, Inc, Edina, MN) in 66 cases. All patients were operated on in a beach chair position. If the subscapularis was intact, it was detached in its superior two-thirds. If the long head of the biceps was present, we performed either tenotomy or a tenodesis. We exposed the glenoid by initial release of the anterior capsule, then the inferior capsule, and, if necessary, the posterior and inferior parts of the capsule. Many of these procedures were the first performed by the various surgeons, and as a result, the vertical position of the glenoid was not standardized and variable. The humeral retroversion was often unknown. Glenoid implants were uncemented. The first model of Delta^TM (before 1995) had a glenosphere with a threaded locking mechanism and after 1995 had a glenosphere with a Morse taper. Humeral implants were cemented in 351 cases.

Postoperatively, the shoulder was immobilized with use of a simple sling or a slight abduction pillow that held the arm in internal rotation for 3 weeks. Passive ROM was begun immediately. After 3 weeks, use of the sling was discontinued and the patient was allowed to use the hand on the involved side for simple activities of daily living.

Patients were seen within the first 3 months, between 6 and 12 months at 2 years, and then every 2 or 3 years. At each postoperative visit, clinical assessment was based on the Constant-Murley score and active ROM and we obtained an AP and glenoid profile view if possible under fluoroscopy to ensure an accurate view.

We had minimum 2-year followup radiographs in 464 of the 484 patients. One surgeon evaluated all 464 images for the presence of a notch according to the classification of Sirveaux et al. [25], the presence of abnormalities of the prosthesis (disassembly, humeral or glenoid unscrewing), fracture or lysis around the screws or the central peg, and the existence of lysis of the different humeral Zones 1 to 7 by a modification of the Gruen classification for the hip [10].

Survival curves were established with the Kaplan-Meier technique [13] and with 95% confidence intervals according to Rothman. We considered two end points: (1) implant removal or replacement for any reason; and (2) a Constant-Murley score of less than 30 points. The curve was cumulative. The patients who died with prostheses still in place were censored. The patients lost to followup before 2 years were considered failures. We compared pre- and postoperative scores with a Wilcoxon signed-rank test and determined differences between subgroups (before 5 years’ followup and after 5, 7, and 9 years’ followup) with a Mann-Whitney U test. We made all analyses with StatView^® (Abacus Concepts Inc, Berkeley, CA).

Results

Eighty-nine of the 489 patients had at least one complication for a total of 107 complications. Seventy-three had one complication, 14 had two complications, and two had three complications. Three complications were observed most frequently: (1) infection: 27 cases, 12 of which required the removal of the prosthesis; (2) glenoid complications (unscrewing, loosening): 27 cases, 13 of which required revision to a hemiarthroplasty; and (3) dislocation: 19 cases. The other complications were hematoma in 14 cases, humeral complication (fractures, unscrewing of metaphysis, loosening) in 11 cases, neurologic complication in six cases, and fracture of the scapula spine in three cases.

Of the 138 patients (148 shoulders) with a minimum 5-year followup, all parameters of the Constant-Murley score improved. Active motion was improved in forward elevation and external rotation in elevation but not in external rotation with the elbow at the side (Table 2). The 138 patients with greater than 5 years’ followup had Constant-Murley scores similar to (p = 0.38) those of the 326 patients with less than 5 years’ followup. The 65 patients with a minimum followup of 7 years had lower (p = 0.03) relative Constant-Murley scores than those with less than 5 years. The 40 patients with a minimum followup of 9 years had lower pain (p = 0.04), mobility (p = 0.006), relative Constant-Murley (p = 0.003), and Constant-Murley scores (p = 0.01) than those with less than 5 years (Table 3).

Table 2.

Clinical results of the 148 shoulders reviewed with a minimum followup of 5 years

Parameter	Preoperative	Postoperative	p Value
Constant-Murley score parameters (points)
Pain	3.3 ± 3.3	12.2 ± 3.7	< 0.001
Activity	6.2 ± 2.5	15.8 ± 4.0	< 0.001
Mobility	12.7 ± 6.5	26.0 ± 8.3	< 0.001
Strength	1.7 ± 3.2	7.5 ± 4.5	< 0.001
Absolute Constant-Murley score	23.9 ± 9.9	61.5 ± 16.9	< 0.001
Relative Constant-Murley score	33.1 ± 13.6	85.2 ± 23.9	< 0.001
ROM (°)
AAE	69.3 ± 34	128.6 ± 32.6	< 0.001
AER1	4.9 ± 17.6	10.6 ± 18.8	0.01
AER2	23.5 ± 23.3	42.1 ± 30.2	< 0.001

Values are expressed as mean ± SD; AAE = active anterior elevation; AER1 = active external rotation with the elbow at the side; AER2 = active external rotation at 90° of elevation.

Table 3.

Comparison of the clinical and radiographic results of the population with followup of less than 5 years and the populations with followup of 5, 7, and 9 years

Parameter	< 5 years	> 5 years	> 7 years	> 9 years
Number of shoulders	331	148	69	41
Constant-Murley score parameters (mean value) (points)
Pain	12.83	12.18	11.74	11.39*
Activity	16.03	15.84	15.68	15.32
Mobility	26.82	26.03	25.19	23.44*
Strength	7.2	7.50	7.35	6.59
Absolute Constant-Murley score	62.86	61.53	59.96	56.76*
Relative Constant-Murley score	88.46	85.24	82.13*	78.00*
ROM (mean value) (°)
AAE	130.11	128.59	124.92	124.12
AER1	12.08	10.60	9.06	6.89
AER2	42.17	40.07	35.59	29*
Radiographic parameters
Humeral radiographic complications	9%	21%*	32%*	39%*
Glenoid radiographic complications	6.8%	24.6%*	26.8%*	31.8%*
Notch Grade 3 or 4	18%	35%*	47%*	50%*

* p < 0.05; AAE = active anterior elevation; AER1 = active external rotation with the elbow at the side; AER2 = active external rotation at 90° of elevation.

Using removal of the prosthesis or conversion to hemiarthroplasty as an end point, we observed a survival rate of 89% at 10 years with a break in the curve at approximately 2 years and then at 9 years (Fig. 1). Twelve prostheses were removed for infection in the first 3 years at a mean followup of 23 months. We converted 13 shoulders to hemiarthroplasty at a mean followup of 51 months but with a substantial peak before 2 years (10 of 13 cases). These cases were all the result of glenoid problems. The survival using a Constant-Murley score of less than 30 was 72% at 10 years and showed a regular gradient with a break at approximately 8 years (Fig. 2).

Fig. 1.

A graph shows the survival curve with removal of the reverse prosthesis or its conversion to hemiarthroplasty as the end point. The survival relating to removal of the reverse prosthesis or its conversion to hemiarthroplasty was 89% at 10 years.

Fig. 2.

A graph shows the survival curve with an absolute Constant-Murley score of less than 30 points as the end point. The survival using a Constant-Murley score of less than 30 points was 72% at 10 years and showed a regular gradient with a break at approximately 8 years.

Likewise, the percentage of radiographic humeral or glenoid abnormalities differed (p = 0.002) between the two populations before and after minimum 5-year followup: we observed more Stage 3 and 4 notches in the population with the longest followup. The percentage of Stage 3 and 4 notches increased from 35% for 5 years’ followup to 49% for 9 years’ minimum followup, whereas the percentage of Stage 1 and 2 notches and Stage 0 notches decreased slightly. The presence of notching, even Stage 3 or 4, did not correlate with the Constant-Murley score (Table 4).

Table 4.

Comparison of the number of notches and Constant-Murley scores with increasing followup time

Notch stage	< 5 years		> 7 years		> 9 years
	Number	Constant-Murley score	Number	Constant-Murley score	Number	Constant-Murley score
Stage 0	25	62	7	60.4	4	52.7
Stages 1 and 2	68	61.9	29	59.7	16	59.1
Stages 3 and 4*	48	61.7	32	61.4	20	57.7

* The number of Stage 3 or 4 notches increased over time but without significant influence on Constant-Murley score.

Discussion

The use of reverse shoulder arthroplasty has considerably increased since first introduced in 1985. Some authors suggest glenohumeral arthropathy associated with MCT is the best indication for reverse prosthesis [3, 11]. In these patients, the survivorship of the prosthesis is reportedly approximately 91% at 5 years [11, 25] and the functional results remain stable until 7 years’ followup. The longer-term durability, however, is unknown. We, therefore, determined complication rates, functional scores over time, survivorship, and whether radiographs would develop signs of loosening.

We recognize a number of limitations in our study. First, throughout the study period, we did not follow a standardized clinical and radiographic assessment. The quality of radiographs was variable and the analysis was sometimes difficult, especially for staging of notches. It should be emphasized, however, that the radiographs were analyzed by a surgeon with substantial experience in the classification of notching. Second, this was a multicenter study. The number of patients included in this study was relatively large with a low number lost to followup. The number of patients assessed with greater than 5 years’ followup was high. However, there was wide variability in the operative technique. In particular, the positioning of the glenoid component was not always as recommended by Nyffeler et al. [19]. Moreover, the evaluators were the operative surgeons. Third, most of the surgeons were at the beginning of their experience with reverse prostheses and so the data include a learning curve. However, each surgeon had a subspecialist interest in shoulder surgery. Lastly, at the time of review, many of the patients were elderly and sometimes demented. This may have influenced the reliability of the clinical assessment but would not modify the survivorship of the prosthesis apart from when the general status of health is the reason for not revising a true or potential loosening.

Like Sirveaux et al. [25], Frankle et al. [6], and Werner et al. [28] (Table 5), we analyzed the results of the reverse prosthesis in MCT with or without glenohumeral arthropathy. The study of Guery et al. [11] considered other indications, but these were fairly nonhomogeneous with a substantial majority still being CTA. However, the grouping together of CTA on the one hand and other etiologies on the other hand revealed differences in prosthesis survival in favor of CTA. Rheumatoid arthritis also appeared to be an at-risk etiology [21]. In addition, the study of Wall et al. [27] provided additional information about the greater risk of complications in cases of revision. That is why we chose to analyze only MCT with or without glenohumeral arthropathy, which remains today the best indication for this prosthesis. This means the results reported here are potentially the best we can expect.

Table 5.

Results of main series of reverse prosthesis for massive cuff tears with or without glenohumeral arthritis

Study	Mean followup (months)	Revision rate (ablation or conversion to hemiarthroplasty)	Score	Percentage of Stage 3 and 4 notches
Sirveaux et al. [25]	44.5	4%	67 (Constant-Murley)	24%
Werner et al. [28]	17	8.6%	72% (Relative Constant-Murley)	46%
Frankle et al. [6]	33	13%	68.2 (ASES)	0%
Favard et al.	91	5%	61 (Constant-Murley)	35%

ASES = American Shoulder and Elbow Surgeons.

Guery et al. [11] concentrated on the appearance of the survival curve, which presented a break between 1 and 3 years in relation to prosthesis revision. In fact, most of the revisions occurred during this period because of either an infection or rapid loosening. Our study confirmed these findings. The frequency of infection complicating reverse shoulder arthroplasty has been reported in other series [4, 8, 28, 29]. The problem with unscrewing of the glenoid has now been resolved after technical modifications to the prosthesis. Apart from these problems, the fixation of the glenoid component seems to be stable and long-lasting with minimal or no glenoid problems once the initial period of 3 to 4 years has passed.

Sirveaux et al. [25] found substantial improvement in the Constant-Murley score and noted a revision rate of 6.2%. Among our 148 patients with more than 5 years’ followup, the Constant-Murley score was higher than preoperatively. On the other hand, the score decreased after 9 years. In the study of Guery et al. [11], the survivorship to a Constant-Murley score of greater than 30 revealed a break after the sixth year. We found a similar break at approximately 8 years. In the study of Guery et al. [11], the survival curves to a low Constant-Murley score and to a low relative Constant-Murley score were similar, taking age into account, thus suggesting age was not the cause of deterioration. The cause of this deterioration is, therefore, currently unknown but could be related to the delayed consequence of radiographic signs of potential failure.

Levigne et al. [16] noted a notching rate of 62% and an increasing frequency of notches with time. In their study, the presence of a notch had no impact on the Constant-Murley score. The explanation provided for the occurrence of a notch of this nature was either a mechanical impingement between the polyethylene and the scapula pillar or a biologic response relating to debris from wear [28]. We found a substantial increase in Stage 3 and 4 notches with increasing followup period, although these had no influence on the functional results. Our results are in accordance with those of other studies [16, 28] but disagree with some reports that have identified the occurrence of a notch as a prognostic factor for function [24, 25].

In conclusion, the long-term results of reverse prostheses for CTA can be summarized as follows: a survival rate of 89% after 10 years with the end point being removal or conversion of the prosthesis but only 72% with the end point being a Constant-Murley score of less than 30; minimal functional deterioration up to 8 years; increasing concern regarding radiographic signs after 5 years; and rate of Stage 3 and 4 notches increasing with time but without any major influence on the Constant-Murley score. We concur with the recommendations made by Guery et al. [11] that the reverse prosthesis is an indication for CTA. Attention must be paid to preventing infections and to the initial positioning of the glenoid implant to minimize early revisions. On average, patient function deteriorates progressively after the eighth year so caution must be observed in relation to the indication, especially in younger patients.