2.1 Standard protocol approvals, registrations, and patient consent

This study was conducted and reported in accordance with the Cochrane handbook, the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses)²¹ and the AMSTAR (Assessing the Methodological Quality of Systematic Reviews)²² Guidelines. The review protocol was registered in PROSPERO with the registration unique identifying number (UIN) of 42024608608. This article does not contain any studies with human participants or animals performed by any of the authors. For this type of study, formal consent is not required.

2.2 Search strategy

Two independent investigators systematically searched the PubMed, EMBASE and Cochrane Library electronic databases to conduct a systematic literature search for articles published from inception until May 1, 2024, without restrictions on language or time of publication. We searched PubMed and the Cochrane Library via medical subject headings (MeSH) and the and reverse shoulder replacement. To identify articles that may have been missed, we also manually searched reference lists for previous systematic reviews and meta-analyses (Supplementary Data Table 1).

2.3 Selection criteria

The selection criteria for this study followed the PICOS principles (population, intervention, comparison, outcome, and study design):

Inclusion criteria.(1)Population: total shoulder arthroplasty patients and reverse shoulder arthroplasty patients(2)Intervention: reverse shoulder arthroplasty(3)Control: total shoulder arthroplasty(4)Results: external rotation (ER)≥10°, active abduction (ABD)≥100°, active forwards flexion (FF)≥120°, internal rotation (IR) score≥4, simple shoulder test (SST)≥8, American Shoulder and Elbow Surgeons (ASES)≥70, forwards elevation≥120°, University of California at Los Angeles (UCLA) score≥30, Constant score≥65, shoulder pain and disability index (SPADI) score≥10, failure, fracture, loosening, pain, infection, dislocation, haematoma, revision, and pulmonary embolism(5)Study design: prospective cohort study, retrospective cohort study

Exclusion criteria.(1)Case reports, reviews, expert opinions, reviews, meta-analyses, duplicate literature, and conference reports.(2)Not enough data for comparison.

2.4 Data extraction

Researchers used an Excel software (version 2408 Build 16.0.17928.20114) designed to extract the table data in advance to extract the data manually. All disagreements among researchers were resolved through discussion or third-party adjudication. In our study, all patients were divided into two groups: the first group underwent total shoulder arthroplasty, and the second group underwent reverse shoulder arthroplasty. The researchers independently extracted the following characteristics: first author, year of publication, type of study design, geographic region, observation duration, sample size, mean age, percentage of female participants, disease type, outcome, and duration of follow-up.

2.5 Quality assessment

Each eligible study was independently assessed by the investigators via the Newcastle–Ottawa Scale (NOS),²³ which includes three quality parameters, namely, patient representation, exposure and outcome determination, and follow-up adequacy, with an overall score of 9 for each study, with scores ≥8 indicating high quality (low risk of bias).

2.6 Data analysis

All analyses were conducted via STATA software (version 12.0; Stata, University Station, Texas, USA). Owing to expected interstudy heterogeneity, random effects models were used to calculate the OR and 95 % CI for efficacy outcomes and complications in patients who underwent both procedures. The Cochrane Q test and I² test were used to evaluate interstudy heterogeneity. Heterogeneity was considered statistically significant when I² ≥ 50 % or P < 0.05.²⁴ The symmetry of the funnel plot was visually assessed in conjunction with Egger's test to assess potential publication bias.²⁵ In addition, we adjusted the risk estimates using Duvall & Tweedie's trim-and-fill to assess the potential impact of publication bias.²⁶ Sensitivity analysis was performed to evaluate the stability of the results. All of the statistical tests were bidirectional analyses, and P < 0.05 was considered statistically significant.

3.1 Literature search

Of the 17,228 articles identified in the database, 15,580 remained after duplicates were removed, and 1062 remained after primary screening of keywords in Endnote X9 (BLD 12062). By reading the titles and abstracts, 1020 articles were removed, and the remaining 41 articles were evaluated. After the full texts were read in detail, 14 studies that did not meet the inclusion criteria were excluded. Finally, a total of 27 articles^{7,12,20,27–50} met the inclusion criteria of our meta-analysis and were included in the analysis (Fig. 1).

Close

Fig. 1 Study flow chart.

Export to PPT

3.2 Study characteristics

The baseline characteristics of the patients included in the study are shown in Table 1. All included studies were published between 2012 and 2023. Eighteen studies^{12,20,28,31,33,34,36,37,39,40,42–44,46–50} reported the follow-up time of each group; the average follow-up time of the aTSA group was 42.3 months, and the average follow-up time of the rTSA group was 36.3 months. This analysis involved studies from seven countries and regions, sixteen^{7,27,29,30,32,34,36–38,40–42,44,45,47,49,50} in the United States, two^12,43 in Italy, one³⁹ in France, one³¹ in Venice, one⁴⁶ in the United Kingdom and Australia, four^20,28,35,48 in the United States and Europe, and one³³ in the United States and France.

3.3 Quality assessment

The quality evaluation of the included studies was independently conducted by two investigators, and any disagreements were resolved by consulting a third researcher. We used the Newcastle–Ottawa Scale (NOS) tool to evaluate the validity and quality of the included studies.²³ Among the 27 included studies, twenty-five^{7,12,20,27–36,38,39,41–50} studies had a low risk of bias, and two^37,40 had a high risk of bias. NOS scores of 0–7 (high risk of bias) and 8–9 (low risk of bias) indicate low and high quality, respectively. The quality evaluation results are summarized in Table 2.

4.1 Primary outcomes

4.1.1

4.1.1 ABD ≥100°

Eight studies^{27,33,34,36,37,42,43,48} compared the efficacy of aTSA and rTSA for ABD ≥100° (n = 3544; aTSA: 1585; rTSA: 1959). aTSA demonstrated superior efficacy (SMD = 0.27, 95 % CI: 0.09–0.45) with substantial heterogeneity (I² = 80.4 %) (Fig. 2).

Close

Fig. 2 Forest plot for ABD ≥100° in patients with shoulder joint disease after total shoulder arthroplasty or reverse total shoulder arthroplasty.

Export to PPT

Subgroup analyses showed that when sample size exceeded 500, the underlying disease was OA, mean age was >70 years, follow-up was ≤36 months, and study quality was high, aTSA significantly outperformed rTSA. Heterogeneity was markedly reduced in subgroups stratified by sample size, disease type, mean age, and follow-up duration, suggesting these factors may be potential sources of heterogeneity (Table 3).

Table 3 Subgroup analysis of the association between aTSA and rTSA at active abduction ≥100°.

Variables		SMD	95 % CI	I2 (%)	No. studies	P for interaction
Sample size						<0.01
	≤500	0.05	−0.17 to 0.28	44.5	5
	>500	0.50	0.43 to 0.57	0	3
Disease type						<0.01
	OA	0.41	0.26 to 0.56	55.9	5
	Non-OA	−0.14	−0.36 to 0.08	0	2
Average age						0.59
	≤70	0.23	−0.08 to 0.54	76.8	5
	>70	0.34	0.09 to 0.59	87.6	3
Average follow-up period(months)						0.03
	≤36	0.50	0.39 to 0.61	0	3
	>36	0.18	−0.09 to 0.45	87.3	5

Sensitivity analyses indicated no material change in SMD when excluding any single study (range: 0.21 [95 % CI: −0.02–0.45] to 0.34 [95 % CI: 0.18–0.50]) (eFig. 1). The funnel plot appeared generally symmetrical (eFig. 2), and Egger's test (P = 0.063) suggested no significant publication bias. Trim-and-fill analysis identified no potentially missing studies (Supplementary Data Table 2).

4.1.2

4.1.2 ER ≥ 10°

Ten studies^{12,27,31-34,36,37,42,44,48} reported data on ER ≥ 10° for both aTSA and rTSA groups (n = 4001; aTSA: 1783; rTSA: 2218). Meta-analysis showed that aTSA achieved a significantly greater effect for ER ≥ 10° (SMD = 0.61; 95 % CI: 0.50–0.71) with low heterogeneity (I² = 44.8 %) (Fig. 3).

Close

Fig. 3 Forest plot for ER ≥ 10° of patients with shoulder joint disease after total shoulder arthroplasty or reverse total shoulder arthroplasty.

Export to PPT

Sensitivity analysis indicated that the pooled SMD remained stable after excluding any single study (range: 0.57 [95 % CI: 0.46–0.69] to 0.64 [95 % CI: 0.55–0.73]) (Supplementary Data Fig. 3). The funnel plot was approximately symmetrical, suggesting no publication bias (Supplementary Data Fig. 4), which was supported by Egger's test (P = 0.128) (Supplementary Data Table 2). Trim-and-fill analysis detected no potentially missing studies.

4.1.3

4.1.3 SST ≥8

Eight studies^{31–34,36,37,42,48} compared the proportion of patients achieving an SST ≥8 between the aTSA and rTSA groups (n = 3692; aTSA: 1632; rTSA: 2060). The aTSA group showed a higher likelihood of achieving SST ≥8 (SMD = 0.16, 95 % CI: 0.08–0.24), with low heterogeneity (I² = 14.1 %) (Fig. 4).

Close

Fig. 4 Forest plot for SSTs ≥8 of patients with shoulder joint disease after total shoulder arthroplasty or reverse total shoulder arthroplasty.

Export to PPT

Sensitivity analysis demonstrated that the pooled SMD remained stable when any single study was excluded (range: 0.14 [95 % CI: 0.04–0.23] to 0.17 [95 % CI: 0.10–0.24]) (Supplementary Data Fig. 5). Funnel plot inspection indicated approximate symmetry, and Egger's test (P = 0.803) confirmed the absence of publication bias (Supplementary Data Fig. 6 and Supplementary Data Table 2). The trim-and-fill method also detected no potentially missing studies.

4.2 Revision

A total of 14 studies^{7,12,20,28–31,34,35,41,44,46,49,50} involving 78,522 cases (aTSA: 34,621; rTSA: 43,901) compared postoperative revision rates between aTSA and rTSA. The pooled analysis showed that aTSA was associated with a lower revision rate than rTSA (OR = 0.71, 95 % CI = 0.55–0.91), with substantial heterogeneity (I² = 87.2 %) (Fig. 5).

Close

Fig. 5 Forest plot for the revision of patients with shoulder joint disease after total shoulder arthroplasty or reverse total shoulder arthroplasty.

Export to PPT

Close

Fig. 6 Forest plot for infection of patients with shoulder joint disease after total shoulder arthroplasty or reverse total shoulder arthroplasty.

Export to PPT

Subgroup analyses indicated that the lower revision rate in the aTSA group persisted in prospective studies, non-USA cohorts, studies conducted from 2020 onward, and studies with sample sizes ≤1000. Heterogeneity was markedly reduced in subgroups stratified by study design, geographic region, observation period, mean age, disease type, and sample size, suggesting that these factors may contribute to the observed heterogeneity (Table 4).

Sensitivity analyses demonstrated that excluding any single study did not materially alter the pooled OR (range: lowest = 0.66, 95 % CI = 0.53–0.82; highest = 0.77, 95 % CI = 0.61–0.98) (Supplementary Data Fig. 7). Funnel plot inspection showed approximate symmetry (Supplementary Data Fig. 8), and Egger's test (P = 0.331) supported the absence of publication bias (Supplementary Data Table 2). No potentially missing studies were identified by the trim-and-fill method.

4.3 Infection

Twelve studies^{7,20,28,34,35,38,41,44–46,49,50} compared postoperative infection rates between the aTSA and rTSA groups (n = 114,386; aTSA: 60,750; rTSA: 53,636). The risk of postoperative infection was approximately 1.5-fold higher in the aTSA group than in the rTSA group (OR = 1.55, 95 % CI: 1.09–2.20), with substantial heterogeneity (I² = 78.2 %) (Fig. 6).

Subgroup analyses showed a significantly higher infection rate in the aTSA group when studies were retrospective, had observation periods before 2020, included patients with a mean age >70 years, involved osteoarthritis, or had sample sizes >1000. Heterogeneity was markedly reduced in subgroups stratified by study design, mean age, and sample size, suggesting these variables may be potential sources of heterogeneity (Table 5).

Sensitivity analysis indicated that exclusion of any single study did not materially alter the pooled OR (lowest: OR = 1.36, 95 % CI: 0.99–1.86; highest: OR = 1.70, 95 % CI: 1.18–2.43) (Supplementary Data Fig. 9). Funnel plot inspection revealed approximate symmetry (eFig. 10), consistent with Egger's test (P = 0.875) (Supplementary Data Table 2). No potentially missing studies were identified after trim-and-fill adjustment.

4.4 Secondary outcomes

A total of seven studies compared the aTSA and rTSA groups in terms of IR score ≥4 and fracture risk. After treatment, the aTSA group had a higher proportion of patients with an IR score ≥4 (SMD = 0.40, 95 % CI: 0.25–0.54; I² = 59.2 %) (Supplementary Data Fig. 11) and a higher risk of fracture (OR = 2.05, 95 % CI: 1.03–4.07; I² = 79.0 %) (Supplementary Data Fig. 12).

Six studies evaluated failure rates, postoperative loosening, and haematoma risk. Compared with rTSA, aTSA was associated with lower failure rates (OR = 0.16, 95 % CI: 0.03–0.83; I² = 96.6 %) (Supplementary Data Fig. 14) and a lower incidence of loosening (OR = 0.40, 95 % CI: 0.23–0.71; I² = 87.8 %) (Supplementary Data Fig. 16), but a higher incidence of haematoma (OR = 1.96, 95 % CI: 1.56–2.54; I² = 0) (Supplementary Data Fig. 17).

Five studies (n = 2779; aTSA: 1231; rTSA: 1548) reported the proportion of patients achieving active forward flexion (FF) ≥120°. The aTSA group had a slightly higher proportion than the rTSA group (SMD = 0.17, 95 % CI: 0.01–0.33; I² = 68.1 %) (Supplementary Data Fig. 13).

Four studies (n = 70,774; aTSA: 35,097; rTSA: 35,715) compared postoperative pulmonary embolism incidence. The aTSA group showed a higher risk than the rTSA group (OR = 1.69, 95 % CI: 1.26–2.28; I² = 0) (Supplementary Data Fig. 15).

4.5 Outcomes without statistical significance

However, we did not find significant differences in pain (OR = 1.12, 95 % CI: 0.83–1.51), dislocation (OR = 1.37, 95 % CI: 0.86–2.17), ASES ≥70 (SMD = 0.11, 95 % CI95 % CI: −0.02–0.23), forwards elevation≥120° (SMD = 0.14, 95 % CI95 % CI: −0.09–0.38), UCLA score≥30 (SMD = 0.04, 95 % CI95 % CI: −0.13–0.22), Constant score ≥65 (SMD = 0.10, 95 % CI95 % CI: −0.03–0.22) or SPADI score ≥10 (SMD = −0.06, 95 % CI95 % CI: −0.14–0.02) between the aTSA group and the rTSA group. There was almost no difference in the outcomes of the two surgical methods in terms of the above aspects (Table 6).

4.6 Sensitivity analysis and publication bias

As mentioned in previous statistical analyses, the efficacy evaluations and complication outcomes mentioned in 27 studies were assessed for potential publication bias. Therefore, we conducted bias tests for revision, ER ≥ 10°, etc. (Supplementary Data Figs. 2,4,6,8,10,18-31). The results are all shown in the supplementary material. There was no publication bias in any of the efficacy evaluations or complication outcomes (Supplementary Data Figs. 2,4,6,8,10,18-31). In addition, we used sensitivity analyses to assess the stability of the combined results of the efficacy evaluations and complications for each study (Supplementary Data Figs. 1, 3, 5, 7, 9, 32-45 and Supplementary Data Table 3).

5.1 Principal findings

The purpose of this meta-analysis was to evaluate the postoperative efficacy and complications of aTSA versus rTSA by summarizing data from 27 studies (a total of 195,413 patients). Our results revealed that in the aTSA group, the patients were significantly more likely to have ER ≥ 10°, long-term postoperative IR score≥4, SST≥8, and ABD≥100°, and the probability of failure, loosening, revision and other complications was lower. However, the rTSA group had lower rates of complications such as fractures, pain, infection, dislocation, haematoma, and pulmonary embolism. There were no statistically significant differences between aTSA and rTSA for ASES≥70, forwards elevation≥120°, UCLA score≥30, Constant score≥65, or SPADI score≥10.

5.2 Comparison with other studies

aTSA improves the ability of patients to perform external rotation and internal rotation, which is consistent with the conclusions of the latest research,^51,52 indicating that rTSA has a greater degree of limitation on internal and external rotation, especially external rotation. This may be attributed to changes in shoulder joint mechanics: the deltoid muscle becomes the primary power source, whereas the role of the rotator cuff muscle group (particularly the rotator muscles) is diminished or eliminated. Simultaneously, the rTSA is designed to enhance shoulder joint stability, especially when the rotator cuff muscles, including the rotator muscles, are severely damaged. To achieve this enhancement, surgery often compromises the range of motion of the joint to some extent to ensure stability. Internal and external rotations are particularly susceptible to causing instability in the shoulder joint. Consequently, this biomechanical shift restricts the rotational activities of the shoulder joint, particularly internal and external rotations.

For the efficacy scores of the included studies, the commonly used scoring criteria included the Constant score, ASES, UCLA score, and SST. No significant differences were found in the ASES, Constant score, SPADI score, or UCLA score, which aligns with the findings of a recent study⁵³; this could be attributed to the fact that the scoring system is primarily designed based on patients' subjective perceptions, reflecting more their overall feelings regarding surgical outcomes rather than precisely measuring shoulder joint range of motion, specific muscle strength, fine joint function, or complex movement patterns.

rTSA significantly reduces the incidence of surgical complications, particularly pain, infection, and haematoma, which aligns with recent research findings.^54–56 It is possible that rTSA alters the mechanical structure of the shoulder joint, shifting its centre, thereby enabling the deltoid muscle to become the primary dynamic muscle of the joint. This design compensates for deficiencies in rotator cuff function, enhances the uniformity of load distribution in the shoulder joint, and mitigates the occurrence of postoperative joint pain. PostrTSA surgery reduces reliance on the rotator cuff muscle group, leading to accelerated early recovery and minimizing complications stemming from an extended rehabilitation process. Additionally, fewer muscle tears and reduced biomechanical stress further decrease the likelihood of postoperative pain, infection, and haematoma.

Furthermore, aTSA and rTSA should be stratified based on age and sex. In articles that report patient age, the average age for aTSA is approximately 67 years, whereas for rTSA, it is approximately 72 years. However, some articles fail to disclose patient age. Regarding patient demographics, approximately two-thirds of the articles reported a greater proportion of women, although a few large-sample studies lacked sex ratio data. Therefore, caution is warranted when interpreting study results stratified by age and sex.

5.3 Strengths

The strengths of this study include the following: (a) Strict adherence to PRISMA guidelines, a comprehensive literature search strategy, and the use of Egger's tests and funnel plots for publication bias assessment. (b) Previous reviews were based on a small number of studies and patients. In contrast, 27 studies with 195,413 patients were included in this meta-analysis, which will significantly improve the validity and reliability of our results. (c) The comprehensive results of postoperative Er, IR score, ABD, FF, SST score, ASES score, forwards elevation, UCLA score, Constant score, SPADI score, complication rate and reoperation rate were comprehensively analysed for the first time. (d) For possible publication bias, trim-and-fill techniques were applied to adjust the pooled estimates, and the results were consistent with the main analysis.

5.4 Limitations

However, this study has several limitations: (a) The quality of some of the included studies was relatively low or medium. Some of the articles included patients with complex preoperative conditions, and some articles used different types of joint prostheses, which may have affected our results to some extent. (b) The number of patients in some studies was relatively small, which may have increased heterogeneity and bias. The sensitivity analysis revealed that our results are stable. (c) As with other observational studies, there may be selection bias and unexplained confounding factors that bias the results, but this bias has little effect on our results. More prospective high-quality articles are needed in the future to further compare the efficacy and safety of aTSA and rTSA.