Reverse shoulder arthroplasty vs. hemiarthroplasty for the treatment of osteoporotic proximal humeral fractures in elderly patients: A systematic review and meta‑analysis update.

The present meta-analysis was conducted to compare the safety and effectiveness of reverse shoulder arthroplasty (RSA) and hemiarthroplasty (HA) in the treatment of osteoporotic proximal humeral fractures in elderly patients. The Embase, Pubmed Central, Cumulative Index to Nursing and Allied Health Literature, ProQuest Dissertations and Theses, Cochrane Library and Chinese Biomedical databases were searched between January 2009 and January 2022 to identify relevant studies. According to the search strategy, a total of 210 associated studies were retrieved and 16 were finally included. Review Manager 5.4 software was used for the data analysis. This study indicated that patients in the RSA group had significantly improved treatment outcomes compared with patients in the HA group, as assessed by Constant-Murley Shoulder Outcome Score (95% CI, 1.69-3.76; P<0.001), American Shoulder and Elbow Surgeons score (95% CI, 11.81-24.88; P<0.001) and shoulder range of motion (ROM; 95% CI, 3.41-9.07; P<0.001). However, the HA group was superior to the RSA group in terms of the Oxford Shoulder score (95% CI, 2.89-11.11; P<0.001). There was no significant statistical difference between the two groups in terms of the Disabilities of the Arm, Shoulder and Hand score and complications. Overall, for the treatment of osteoporotic proximal humeral fractures in the elderly, the RSA group had improved postoperative ROM and functional scores compared with the HA group, without significant difference in the incidence of complications. However, HA remains a safe and reliable treatment option. Copyright: © Han et al.

Introduction

Proximal humeral fractures are the third most common fractures after proximal femur and distal radial fractures, accounting for 4-5% of all fractures (1), especially in the elderly population. With an increase in the incidence of osteoporosis in the aging society, the incidence of proximal humeral fractures is increasing year by year and the degree of fracture is often more serious, mostly occurring as a three-part or four-part fracture based on the Neer fracture classification system (2,3). At present, humeral locking plates and intramedullary nails have achieved good clinical efficacy, but there are more complications, such as increased operation times and intraoperative blood loss, as well as humeral head necrosis and screw perforation, when compared with shoulder arthroplasty (4). With the improvement of surgical techniques and implant materials, hemiarthroplasty (HA) and reverse shoulder arthroplasty (RSA) have gradually gained attention and made some progress in clinical practice (5). Conventional HA can solve the problems of internal fixation failure and humeral head necrosis of complex proximal humerus fractures. However, HA has certain requirements for the union of tubercle fractures and the integrity of the scapular glenoid (6). HA cannot deal with complicated rotator cuff injuries and glenoid cavity fractures (6), while RSA has higher requirements on the deltoid muscle function. There is still a lot of debate over which type of shoulder arthroplasty is more suitable for elderly comminuted proximal humerus fractures (7). Therefore, the present study aimed to systematically compare the differences between them by means of a meta-analysis to provide theoretical guidance for clinical practice.

Materials and methods

General information

The present study included domestic and overseas clinical controlled studies that were published between January 2009 and January 2022. The patients were diagnosed with complex proximal humerus fractures (Neer three-part and four-part fractures) based on the patient's medical history (8), and physical and imaging examinations. All patients required shoulder arthroplasty. Non-case controlled studies, case reports, literature reviews, letters, duplicate reports, studies that included cases with a mean age of <60 years or those that did not provide sufficient relevant data were excluded. Intervention measures were HA replacement and RSA. Outcome indicators were Constant-Murley Shoulder Outcome Score, Disabilities of Arm, Shoulder and Hand (DASH) score, American Shoulder and Elbow Surgeons (ASES) score, Oxford Shoulder score, shoulder range of motion (ROM) and complications (9). The Preferred Reporting Items for Systematic Reviews and Meta-Analyses checklist was followed to perform the present meta-analysis (10).

Search strategy

The Embase (https://www.embase.com), Pubmed Central (https://pubmed.ncbi.nlm.nih.gov/), ProQuest Dissertations and Theses (http://pqdtopen.proquest.com/), Cochrane Library (https://www.cochranelibrary.com/) and Chinese Biomedical databases (www.sinomed.ac.cn) were searched. The directories of periodical and references were manually retrieved and the gray literature (such as chapters in unpublished academic papers, monographs, non-publicly published documents, government documents, dissertations, conference documents, scientific and technological reports, technical archives) was retrieved. Meanwhile, no language constrains were applied to extend the search to all relevant content, and the papers were thereafter translated if necessary. The key words were ‘hemiarthroplasty’, ‘HA’, ‘reverse shoulder arthroplasty’, ‘RSA’ and ‘proximal humeral fractures’, and the search strategy was (reverse shoulder arthroplasty OR RSA) AND (hemiarthroplasty OR HA) AND (proximal humeral fractures).

Quality assessment of the literature

The included studies were independently analyzed by two physicians, and if disagreements occurred, they were resolved through discussion or handed over to a third senior physician to jointly determine the quality of the literature in strict accordance with the Cochrane risk of bias assessment criteria (11): i) Whether the experimental design adopts the random principle; ii) whether participants, performers and measurers use the double-blind principle; iii) whether the experimental data are complete and credible; iv) whether the allocation concealment method is adopted; v) whether the experiment adopts a selective data reporting method; and vi) other bias factors. According to the Newcastle-Ottawa scale (NOS) (12), the quality of the literature that met the inclusion criteria was evaluated.

Statistical analysis

Meta-analysis of the extracted data was performed using Review manager 5.4 software provided by the Cochrane Collaboration (https://training.cochrane.org/online-learning/core-software/revman). Dichotomous variables are expressed as the odds ratio and 95% CI. Continuous variables are expressed as the mean difference or standard mean difference and 95% CI. Heterogeneity among studies was tested by I2 statistics, with an I2 value of >50% indicating significant inconsistency. The reasons for the heterogeneity were analyzed and a random-effect model was used at this time, otherwise, a fixed-effect model was used. Sensitivity analyses were performed by removing certain studies and reanalyzing the data, and funnel plots were made to assess publication bias. P<0.05 was considered to indicate a statistically significant difference.

Results

Essential features of the included studies

A total of 210 related publications were retrieved according to the aforementioned search strategy. By reading the title and abstract, 181 publications that were not controlled studies, or were repeated publications and irrelevant to the research purpose were excluded, and so 29 suitable publications passed initial screening. Following full text reading and screening in accordance with the inclusion and exclusion criteria, 16 studies were finally included. Baseline patient information, such as age and course of disease were compared among the included studies, which was comparable (P>0.05). The literature screening process and results are shown in Fig. 1 and the basic characteristics of the included literature studies are shown in Table I.

Figure 1

Flow diagram of the literature search.

Table I

General characteristics of the included studies.

First author/s	Type of study	Country (city)	Publication year	RSA/HA	Cases, n	Mean age, yeas	Sex (M/F)	Outcome	NOS score	(Refs.)
Baudi et al	Retrospective	Italy (Modena)	2014	RSA	25	77.3	-/-	A,B,C,E	7	(14)
				HA	28	71.4	-/-
Bonnevialle et al	Retrospective	France (Toulouse)	2016	RSA	41	78.0±5.0	4/37	B,E,F	7	(15)
				HA	57	67.0±10.1	18/39
Boyer et al	Prospective	France (Besançon)	2017	RSA	65	-	-/-	B,F	7	(16)
				HA	69	-	-/-
Boyle et al	Retrospective	New Zealand (Auckland)	2012	RSA	55	79.6	4/51	D	6	(17)
				HA	313	71.9	69/244
Critchley et al	Retrospective	Australia (Adelaide)	2020	RSA	3,049	75.4	484/2565	F	6	(18)
				HA	2,897	71.2	623/2274
Cuff and Pupello	Prospective	America (Florida)	2013	RSA	24	74.4	-	C,EF	8	(19)
				HA	23
Alentorn-Geli et al	Prospective	Spain (Barcelona)	2014	RSA	20	78.6	4/16	A	7	(20)
				HA	12	82.6	4/8
Sebastiá-Forcada et al	Prospective	Spain (Alicante)	2014	RSA	31	74.7	4/27	B,E,F	9	(21)
				HA	30	73.3	5/25
Gallinet et al	Retrospective	France (Besançon)	2009	RSA	19	74.0	4/15	B,E	6	(22)
				HA	21	74.0	3/18
Garrigues et al	Retrospective	America (North Carolina)	2012	RSA	11	80.5	-/-	C,E	8	(23)
				HA	12	69.3	-/-
Cvetanovich et al	Retrospective	America (Chicago)	2016	RSA	125	-	-/-	F	6	(24)
				HA	404	-	-/-
Jonsson et al	Random	Sweden (Gothenburg)	2021	RSA	41	79.5	8/76	F	9	(25)
				HA	43
van der Merwe et al	Prospective	New Zealand (Auckland)	2017	RSA	218	78.2	22/196	D	6	(26)
				HA	427	71.6	97/330
Schairer et al	Retrospective	America (New York)	2015	RSA	420	-	-/-	F	6	(27)
				HA	1,124	-	-/-
Solomon et al	Retrospective	America (Cleveland)	2016	RSA	16	77.0	3/13	E	8	(28)
				HA	8	77.0	1/7
Young et al	Retrospective	New Zealand (Auckland)	2010	RSA	10	77.2	0/10	D,E	9	(29)
				HA	10	75.5	2/8

HA, hemiarthroplasty; RSA, reverse shoulder arthroplasty; M, male; F, female; NOS, Newcastle-Ottawa Scale; A, Constant-Murley Shoulder Outcome Score; B, Disabilities of the Arm, Shoulder and Hand score; C, American Shoulder and Elbow Surgeons score; D, Oxford Shoulder score; E, Shoulder range of motion; F, complications.

The present study included 1 randomized controlled trial, 5 prospective studies and 10 retrospective studies. To the best of our knowledge, the meta-analysis of retrospective, randomized controlled and prospective studies together will inevitably lead to certain biases in the results (13). However, when the literature was searched according to the search strategy, the number of randomized controlled studies and sample size was relatively small. Therefore, retrospective and prospective studies were included in the meta-analysis to increase the sample size. Consequently, NOS score was considered the most suitable for evaluating all the studies and the score was divided into low, medium and high quality grades, namely <5, 5-7 and 8-9 points (12). Among the literature, 3 papers scored nine points, 3 scored eight points, 4 scored seven points and 6 scored six points. Although the number of included studies was limited and there was a certain bias, the overall quality was moderate (Fig. 1 and Table I) (14-29).

Outcomes Constant-Murley Shoulder Outcome Score

In the present study, the Constant-Murley Shoulder Outcome Score was chosen as one of the outcomes of treatment. The Constant-Murley Shoulder Outcome Score is one of the most important scores for evaluating shoulder joint function (30); it mainly includes pain, daily activities, strength and shoulder joint ROM (31). A total of 6 studies reported the postoperative Constant-Murley Shoulder Outcome Score for RSA and HA in the treatment of elderly osteoporotic proximal humerus fractures, with scores divided into four subgroups according to pain, activity, strength and ROM. Due to the large heterogeneity (I²>50%) between the results and subgroups of each study, a random-effects model was used for the meta-analysis. The results showed that during the treatment of elderly osteoporotic proximal humeral fractures with joint replacement, the postoperative Constant-Murley Shoulder Outcome Score in the RSA group was improved compared with that in the HA group (95% CI, 1.69-3.76; P<0.001) and the difference was statistically significant. In terms of activity, the RSA group score was significantly improved compared with that of the HA group (95% CI, 2.25-4.70; P<0.001) and similarly, the RSA group outperformed the HA group for ROM (95% CI, 5.50-8.51; P<0.001). However, in the two subgroups of pain (95% CI, -0.17-4.58; P=0.07) and strength (95% CI, -1.10-2.32; P=0.48), the scores of the RSA and HA group were comparable (Fig. 2).

Figure 2

Meta-analysis of Constant-Murley Shoulder Outcome Score between RSA and HA in the treatment of elderly osteoporotic proximal humerus fractures. RSA, reverse shoulder arthroplasty; HA, hemiarthroplasty.

DASH score. A total of 5 studies provided detailed information on the DASH score. The random-effect model was also used for the meta-analysis due to the large heterogeneity among the results (I²=70%). After the meta-analysis, no significance was observed (95% CI, -7.34-2.02; P=0.27) for DASH score (Fig. 3).

Figure 3

Meta-analysis of Disabilities of the Arm, Shoulder and Hand score between RSA and HA in the treatment of elderly osteoporotic proximal humerus fractures. RSA, reverse shoulder arthroplasty; HA, hemiarthroplasty.

ASES score. A total of 5 studies paid close attention to the ASES score for the treatment of elderly osteoporotic proximal humerus fractures. As there was little heterogeneity among the results of the studies (I²<50%), a fixed-effects model was used for the meta-analysis. The mean ASES score of the RSA group was higher than that in the HA group, and the difference was statistically significant (95% CI, 11.81-24.88; P<0.001; Fig. 4).

Figure 4

Meta-analysis of American Shoulder and Elbow Surgeons score between RSA and HA in the treatment of elderly osteoporotic proximal humerus fractures. RSA, reverse shoulder arthroplasty; HA, hemiarthroplasty.

Oxford Shoulder score. A total of 3 publications compared the postoperative Oxford Shoulder score between the RSA and HA groups. Due to the large heterogeneity (I²>50%) among the results of each study, the random-effects model was used for the meta-analysis. The results showed that the mean Oxford Shoulder score of the RSA group was higher when treating elderly osteoporotic proximal humerus fractures, and the difference was statistically significant (95% CI, 2.89-11.11; P<0.001; Fig. 5).

Figure 5

Meta-analysis of Oxford Shoulder score between RSA and HA in the treatment of elderly osteoporotic proximal humerus fractures. RSA, reverse shoulder arthroplasty; HA, hemiarthroplasty.

Shoulder ROM. A total of 9 studies compared the postoperative shoulder ROM of RSA and HA in the treatment of elderly osteoporotic proximal humerus fractures, with the scores divided into four subgroups according to anterior elevation, abduction, external rotation and internal rotation. Due to the large heterogeneity (I²>50%) among the results, the random-effect model was used for the meta-analysis. The results showed that during the treatment of elderly osteoporotic proximal humerus fractures with shoulder arthroplasty, the shoulder ROM in the RSA group was improved compared with that in the HA group (95% CI, 3.41-9.07; P<0.001) with statistically significant differences, especially with regard to the anterior elevation (95% CI, 18.27-39.38; P<0.001) and abduction (95% CI, 24.24-41.26; P<0.001). The meta-analysis for the other two subgroups involving external (95% CI, -5.33-1.12; P=0.20) and internal (95% CI, -1.86-1.06; P=0.59) rotations did not find a significant difference between both methods (Fig. 6).

Figure 6

Meta-analysis of shoulder range of motion between RSA and HA in the treatment of elderly osteoporotic proximal humerus fractures. RSA, reverse shoulder arthroplasty; HA, hemiarthroplasty; ROM, range of motion.

Complications. There are several possible postoperative complications in the treatment of elderly osteoporotic proximal humerus fractures with shoulder arthroplasty. The 8 included studies compared the postoperative complications of RSA and HA and divided them into 13 complication subgroups, including arthrofibrosis, greater tuberosity osteolysis, malunion greater tuberosity, diaphyseal fracture of humerus, infection, heterotopic bone, loosening, secondary cuff rupture, deep venous thrombosis, instability/dislocation, complex regional pain syndrome, malposition and pulmonary embolism. Due to the large heterogeneity (I²>50%) among the results of each study, the random-effect model was used for the meta-analysis. The results showed that in the deep venous thrombosis subgroup, the HA group had a lower incidence rate than the RSA group (95% CI, 1.21-5.68; P=0.01). No significant differences were observed in the other subgroups or in terms of complications overall (95% CI, 0.59-1.20; P=0.35; Fig. 7).

Figure 7

Meta-analysis of complications between RSA and HA in the treatment of elderly osteoporotic proximal humerus fractures. RSA, reverse shoulder arthroplasty; HA, hemiarthroplasty; CRPS, complex regional pain syndrome.

Publication bias and sensitivity analysis

The Review Manager 5.4 statistical software provided by the Cochrane Collaboration was used to analyze the publication bias of the Constant-Murley Shoulder Outcome Score, DASH score, ASES score, Oxford Shoulder score, shoulder ROM and complications after the shoulder arthroplasty. The results showed that the funnel plots were basically symmetrical, indicating that there was no obvious publication bias (Fig. 8). The analysis showed that there was high heterogeneity in the Constant-Murley Shoulder Outcome Score subgroups (pain, strength and ROM), DASH score, Oxford shoulder score and shoulder ROM subgroups (anterior elevation and internal rotation). The heterogeneity was reduced to 0% after the study by Bonnevialle et al (15) was excluded from the meta-results of Constant-Murley Shoulder Outcome Score-ROM. Due to strict compliance with the search and inclusion criteria, the inclusion of certain results in the literature is limited, and it would be difficult to complete the meta-analysis if certain studies were excluded. After further analyzing the heterogeneous sources using the one-by-one elimination method, the results of the meta-analysis showed no directional changes, indicating that the stability of the research results was good (Figs. 9 and 10).

Figure 8

Funnel plots for evaluation of publication bias. (A) Funnel plot for Constant-Murley Shoulder Outcome Score; (B) funnel plot for DASH score; (C) funnel plot for ASES score; (D) funnel plot for Oxford Shoulder score; (E) funnel plot for shoulder ROM; (F) funnel plot for complications. ASES, American Shoulder and Elbow Surgeons; ROM, range of motion; DASH, Disabilities of the Arm, Shoulder and Hand; CRPS, complex regional pain syndrome.

Figure 9

Risk of bias graph. Each risk of bias item is presented as a percentage across all included studies and indicates the proportional level for each risk of bias item.

Figure 10

Risk of bias summary. Methodological quality of the included studies according to a risk of bias tool that assessed randomization (sequence generation and allocation concealment), blinding (participants, personnel and outcome assessors), completeness of outcome data, selection of outcomes reported and other sources of bias. The items were scored with ‘yes’, ‘no’ or ‘unclear’.

Discussion

Proximal humerus fractures are one of the common long bone fractures, accounting for more than one-fourth of shoulder fractures (32). The purpose of treatment for proximal humerus fractures is to relieve pain, restore the ROM of the shoulder joint and improve the quality of life of the patient. Most of the proximal humerus fractures in adolescents are undisplaced and simple fractures, which can be treated conservatively. However, proximal humeral fractures in the elderly are mostly comminuted fractures due to long-term bone loss (33).

Surgical options for proximal humerus fractures include open reduction and internal fixation, HA and RSA. A number of studies have shown that shoulder arthroplasty has fewer complications and improved postoperative functional scores than the open reduction and internal fixation of proximal humerus fractures (34,35); however, there is still some controversy about the superiority of HA and RSA. Although HA can effectively solve the problem of humeral fracture pain, the matching degree of the prosthesis is relatively high and there are certain requirements for the integrity of the soft tissue around the shoulder joint (36). For example, the incompleteness of the greater tuberosity of the humerus is one of the reasons for the operation failure. Active anti-osteoporosis treatment and functional exercise are still required after the surgery. The RSA avoids a high degree of dependence on the rotator cuff by shifting the relationship between the glenohumeral joints and shifting the center of rotation downward. Although the requirement for rotator cuff integrity is lower than that of HA, the integrity of the greater tuberosity still has a significant effect on shoulder function after RSA (37). Additionally, RSA also has various complications such as periprosthetic fractures and shoulder loosening or dislocation (38).

The purpose of the present meta-analysis was to compare the efficacy of shoulder arthroplasty in the treatment of elderly osteoporotic proximal humerus fractures, as assessed by the Constant-Murley Shoulder Outcome Score, DASH score, ASES score and Oxford Shoulder score, as well as shoulder ROM and complications, in order to comprehensively and carefully compare the differences in postoperative functional scores and complications between the RSA and HA groups. It was found that the RSA group had a significantly improved shoulder joint function Constant-Murley Shoulder Outcome Score (95% CI, 1.69-3.76; P<0.001), ASES score (95% CI, 11.81-24.88; P<0.001) and shoulder ROM (95% CI, 3.41-9.07; P<0.001], while the Oxford shoulder score [95% CI, 2.89-11.11), P<0.001] in the HA group was significantly improved compared with that in the RSA group. The present analysis did not find a significant difference in the DASH score and complications between the two groups.

Jain et al (39) suggested that healing of the greater tubercle of the shoulder joint is key to the recovery of ROM after shoulder arthroplasty, and the degree of dispersion of shoulder joint mobility after HA was greater than that of RSA. It was concluded that HA had improved postoperative clinical outcomes compared with RSA in cases without complications. Wang et al (40) suggested that RSA is a preferable treatment option compared with HA in the treatment of complex proximal humerus fractures. In particular, RSA has obvious advantages in postoperative functional score and ROM of anterior elevation, which is consistent with the results of the present study. Moreover, since the present meta-analysis performed a more systematic and detailed comparison on the ROM of shoulder joint abduction, internal/external rotation, etc., it was found that RSA was also superior to HA in abduction ROM, which may be related to the fact that RSA affects the deltoid muscle less and thus retains the complete abduction ability of the shoulder joint, while with the overfilled HA prosthesis, it is easy to put extra pressure on the rotator cuff, thus affecting part of the abduction function (41). Shukla et al (42) obtained similar results further confirming the outcomes of this meta-analysis. In addition, Wang et al (40) found that RSA also has obvious advantages in terms of postoperative complications. However, the present study found no significant difference in postoperative complications between the two groups. We hypothesize that this may be more common for RSA in cases of severe rotator cuff injury due to less soft tissue-induced instability, which increases the complication rates. As the biomechanical principles of RSA become more understood in the future, it is considered that the incidence of RSA complications will continue to decline. Similarly, the study by Kleim et al (43) also showed that postoperative function and score of RSA are improved compared with those of HA. RSA not only ensured the postoperative ROM of the shoulder joint but also greatly improved the quality of life of patients after surgery, which is consistent with the results of the present study. In summary, although there have been literature reports on associated topics in the past, the present study included a larger number of original literature studies and more qualified case samples. Therefore, it is suggested that the conclusions made on the present updated meta-analysis are more reliable to evaluate the difference between RSA and HA with respect to their therapeutic effects.

Shoulder arthroplasty has a good postoperative clinical effect in the treatment of elderly osteoporotic proximal humerus fractures. The present meta-analysis revealed that the postoperative functional score of most cases in the RSA group was greater than that in the HA group (44). We hypothesize that RSA has obvious advantages over HA in the treatment of elderly osteoporotic proximal humerus fractures. However, the present meta-analysis could not determine whether one surgical method was superior to the other in terms of postoperative complications. For example, one study showed that RSA has a risk of postoperative shoulder dislocation higher than that in HA (45). Considering that such fractures vary in the degree of severity due to the age of the patients and degree of osteoporosis among individual cases, the selection of specific surgical methods should still follow the principle of individualized treatment.

The present study has the following limitations, which need to be further improved upon: i) A total of 16 foreign literature studies were included in the meta-analysis evaluation system, of which only 1 randomized controlled trial was included with a low evidence level, while the combined analysis of retrospective studies, prospective studies and randomized controlled studies were bound to cause some bias in the results; ii) some of the literature included by using the NOS scale were of low quality; iii) among the outcome evaluation indexes, the same evaluation method included 9 publications at most and 3 at least, the heterogeneity between the groups also increased slightly and the final follow-up time for outcome measures varied between different studies; iv) clinical studies require informed consent to participate from the patients; and v) this systematic review has not been prospectively registered in an appropriate registry (such as the PROSPERO database of the National Institute for Health Research). When it comes to the selection of specific treatment plans and medical ethics, low literature quality evaluation may also result in inevitable bias and affect the reliability of meta-analysis conclusions. Therefore, the aforementioned conclusions need to be further verified by larger sample, randomized controlled studies.