GLP‑1 receptor agonist therapy and 2‑year structural outcomes after arthroscopic rotator cuff repair in type 2 diabetes Authors: Mao Yi Yang, Li Wei Hung Published: 2026 ABSTRACT AbstractBackgroundThe peri‑operative management of glucagon‑like peptide‑1 receptor agonists (GLP‑1RAs) in orthopaedic surgery is controversial, and evidence regarding their influence on tendon healing after rotator cuff repair (RCR) is limited. Previous studies have mainly addressed incident rotator cuff tear or primary RCR, and one analysis reported no increase in reoperation among GLP‑1RA users. Whether pre‑index GLP‑1RA use affects tendon healing after arthroscopic RCR in patients with type 2 diabetes mellitus (T2DM) remains unclear.MethodsUsing the TriNetX Global Collaborative Network, a federated electronic medical record network comprising over 170 healthcare organizations, we identified adults with T2DM undergoing arthroscopic RCR (CPT 29827). In clinical practice, arthroscopic RCR is predominantly used for repairable small‑to‑large tears, whereas massive or irreparable tears are often managed with alternative procedures. GLP-1RA new users (prescriptions ≤ 6 months before index) were compared with an active-comparator diabetes therapy using 1:1 propensity score matching to balance age, sex, BMI, HbA1c, and key comorbidities. The primary outcome was 2-year revision surgery; the secondary outcome was a retear-proxy event (ICD-10 M75.1 plus MRI). Risks, risk ratios (RRs), and Cox hazard ratios (HRs) were estimated.ResultsAfter matching, 957 patients remained in each cohort with standardized mean differences < 0.1 for all covariates. Two‑year revision occurred in 10 GLP‑1RA users (1.0%) and 11 comparators (1.1%; RR 0.91, 95% confidence interval [CI] 0.39–2.13; HR 0.91, 95% CI 0.39–2.15). The retear‑proxy event occurred in 70 (7.3%) versus 82 (8.6%) patients (RR 0.85, 95% CI 0.63–1.16; HR 0.85, 95% CI 0.62–1.17). Kaplan–Meier curves and log‑rank tests showed no significant differences for either endpoint.ConclusionsIn T2DM patients undergoing arthroscopic RCR, pre-index GLP-1RA use was not associated with a detectable increase in 2-year risk of revision surgery or retear-proxy events compared with active-comparator non-GLP-1 therapy. These findings are consistent with the short-term structural safety of GLP-1RAs after RCR but should be interpreted with caution given the low number of revision events and the inherent limitations of the database.Level of evidenceLevel III.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12891-026-10059-w.Keywords: Glucagon‑like peptide‑1 receptor agonist, Rotator cuff repair, Revision surgery, Tendon healing, Type 2 diabetes, Federated healthcare database Introduction Glucagon‑like peptide‑1 receptor agonists (GLP‑1RAs) are increasingly prescribed for glycaemic control, weight reduction, and cardiovascular risk reduction in patients with type 2 diabetes mellitus (T2DM) and obesity [1]. As their use expands, orthopaedic surgeons and anaesthesiologists are increasingly confronted with decisions regarding the peri‑operative management of GLP‑1RAs [2], particularly in procedures where soft‑tissue or tendon healing is critical. Emerging evidence suggests that GLP‑1RAs may influence musculoskeletal tissues and tissue healing. Population‑level studies have linked GLP‑1RA use to varying risks of incident atraumatic rotator cuff tears and rotator cuff repair (RCR), especially when compared with sodium‑glucose cotransporter‑2 inhibitors (SGLT2is) [3]. Experimental work indicates that GLP‑1 receptor activation may modulate tendon biology; for example, exendin‑4 promotes tenogenic differentiation of mesenchymal stem cells [4], and liraglutide has been shown in animal models to attenuate inflammation and endoplasmic reticulum stress at the rotator cuff enthesis, enhancing healing via the GLP‑1R–AMPK/SIRT1 pathway [5]. More recently, a database study reported that GLP‑1RA use was not associated with increased reoperation after RCR, suggesting a neutral peri‑operative safety profile in this context [6]. However, important gaps remain. Previous studies have often focused on the risk of incident rotator cuff tear or primary RCR in general diabetic populations, rather than tendon integrity following RCR. Moreover, existing analyses have typically relied on single‑payer administrative claims without detailed laboratory data, and have not consistently used active‑comparator, new‑user designs to limit confounding by indication [7]. In addition, few studies have incorporated diagnostic and imaging‑based proxies of retear that capture clinically suspected failures not culminating in revision surgery. The present study aimed to evaluate whether pre‑operative GLP‑1RA use is associated with increased 2‑year risk of structural failure after arthroscopic RCR in patients with T2DM, using a global federated electronic medical record network and an active‑comparator, propensity‑matched design. We hypothesized that GLP‑1RA use would not be associated with higher rates of revision surgery or retear‑proxy events compared with non‑GLP‑1 diabetic therapy. Methods Data source and study design This retrospective cohort study used de-identified data from TriNetX (Cambridge, MA, USA), a global federated electronic medical record (EMR) network that aggregates encounter-level data from over 170 healthcare organisations across North America, Europe, and Asia. Participating sites include academic medical centres, community hospitals, and specialty practices, contributing diagnoses, procedures, medications, laboratory results, and demographic information in a harmonised format. Data remain behind institutional firewalls and are queried in real time; only aggregate counts and de-identified summaries are returned to the investigator, which reduces the risk of data breaches while enabling large-scale, multisite observational research based on time-to-event analyses. The study protocol was reviewed and approved by the Institutional Review Board of Shin Kong Wu Ho-Su Memorial Hospital (IRB number: 20260127R), and the requirement for individual informed consent was waived because only de-identified data were analysed. Clinical trial registration was not applicable. Study population Adults aged ≥ 18 years with a diagnosis of T2DM (ICD-10-CM E11) who underwent arthroscopic RCR (CPT code 29827) were eligible for inclusion. The index event was defined as the first occurrence of arthroscopic RCR within the network, with the study period spanning the past 20 years. In clinical practice, this cohort predominantly represents repairable small-to-large rotator cuff tears; massive or irreparable tears are typically managed with alternative procedures—such as debridement, superior capsular reconstruction, or reverse total shoulder arthroplasty—which were not captured under this primary repair code. To ensure sufficient data for analysis, patients were required to have at least 365 days of observable history prior to the index date and 730 days of potential follow-up thereafter. Exposure definition The GLP-1RA cohort consisted of T2DM patients who received a GLP-1RA (ATC A10BJ) within 6 months prior to index and had no GLP-1RA prescriptions 6–12 months before index, approximating a new-user design. The active-comparator cohort included T2DM patients with prescriptions for metformin (RxNorm 6809), SGLT2is (ATC A10BK), or dipeptidyl peptidase-4 inhibitors (DPP-4is; ATC A10BH) within 6 months prior to index, and no GLP-1RA use in the year before index. This design reduces confounding by indication by restricting comparisons to patients receiving contemporary diabetes therapies [7]. Exposure was defined on the basis of pre‑operative prescriptions; detailed information on peri‑operative interruption, dose adjustments, or post‑operative continuation beyond the index date was not consistently available across sites, so our analyses cannot distinguish sustained from transient GLP‑1RA use throughout the early tendon‑healing period [8]. Covariates Covariates were assessed during the 365 days preceding index and included demographic, clinical, and laboratory variables. Demographics comprised age at index and sex. Clinical comorbidities included body mass index (BMI), hypertension, hyperlipidaemia, chronic kidney disease (N18), ischaemic heart disease (I20–I25), age‑related osteoporosis (M81.0), and other cardiovascular and metabolic conditions. Glycaemic control was captured using haemoglobin A1c measurements when available. These covariates were selected based on prior literature on rotator cuff healing and surgical outcomes in diabetic and orthopaedic populations [9–14]. Outcomes The primary outcome was 2‑year revision shoulder surgery following index RCR. Revision was defined as any of the following procedures occurring 1–730 days after index: repeat arthroscopic RCR (CPT 29827), shoulder revision or reconstruction procedures, or total shoulder arthroplasty (CPT 23472, 1004201 and 1004224). The secondary outcome was a retear-proxy event, defined a priori as a new ICD-10-CM diagnosis code for rotator cuff tear (M75.1) followed by an upper-extremity magnetic resonance imaging (MRI) study within 2 years of index surgery. In this federated EMR environment, this combination of a new rotator cuff tear code plus advanced imaging was used as a pragmatic surrogate for clinically suspected structural failure, because routine surveillance imaging is not part of standard post-operative care; requiring both elements was intended to preferentially capture symptomatic events that prompted targeted MRI rather than incidental coding or background imaging. Propensity score matching Propensity scores for GLP-1RA exposure were estimated within the TriNetX platform using logistic regression. Covariates were selected a priori based on clinical relevance and prior literature linking them to rotator cuff healing and postoperative outcomes, including age, sex, body mass index, haemoglobin A1c, and key comorbidities (hypertension, hyperlipidaemia, chronic kidney disease, ischaemic heart disease, and osteoporosis). One-to-one greedy nearest-neighbour matching without replacement was performed using a caliper of 0.2 of the standard deviation of the logit of the propensity score. Covariate balance after matching was assessed using standardized mean differences (SMDs), with SMD ≤ 0.1 indicative of an acceptable balance. Statistical analysis Within the TriNetX platform, we calculated event counts and risks for each outcome and derived risk ratios (RRs) and 95% confidence intervals (CIs). Kaplan–Meier survival curves were generated for time to revision and the retear-proxy event, with groups compared via log-rank tests. Cox proportional hazards models were fitted to estimate hazard ratios (HRs) and 95% CIs for GLP-1RA exposure in the matched cohort. To ensure the robustness of our findings, separate multivariable Cox models were also performed in the unmatched sample, adjusting for the same baseline covariates. In sensitivity analyses, GLP-1RA users were separately compared with metformin-only and SGLT2i-only cohorts using similar propensity score–based matching and time-to-event methods for retear-proxy outcomes (Additional file 1: Supplementary Table S1). Proportional hazards assumptions were evaluated using standard diagnostics within the network, with no major violations detected. All tests were two-sided, and p-values < 0.05 were considered statistically significant. Results Cohort characteristics Before matching, 961 GLP‑1RA users and 3,355 active‑comparator patients met inclusion criteria. GLP‑1RA users were slightly younger and had different distributions of sex, BMI, hypertension, and ischaemic heart disease compared with comparators (SMDs up to approximately 0.27). After propensity score matching, 957 patients remained in each cohort, and baseline characteristics were well balanced, with SMDs < 0.1 for age, sex, BMI, hypertension, chronic kidney disease, ischaemic heart disease, osteoporosis, hyperlipidaemia, and haemoglobin A1c. The mean age at index was 64.4 years in both groups, and approximately half of patients were female (Table 1). Table 1Baseline characteristics of GLP-1RA users and active-comparator patients before and after propensity score matchingCharacteristicsBefore Propensity Score MatchingAfter Propensity Score MatchingGLP-1RA user (n = 961)Non GLP-1RA user (n = 3,355)Std diffGLP-1RA user (n = 957)Non GLP-1RA user (n = 957)Std diffDemographics Age, years, mean ± SD64.3 ± 8.966.7 ± 9.10.2764.4 ± 8.864.4 ± 9.00.005 Male, n (%)500 (52.0%)1,988 (59.3%)0.15500 (52.2%)490 (51.2%)0.021 Female, n (%)461 (48.0%)1,367 (40.7%)0.15457 (47.8%)467 (48.8%)0.021Comorbidities Body Mass Index, n (%)366 (38.1%)1,067 (31.8%)0.13364 (38.0%)341 (35.6%)0.05 Hypertension, n (%)749 (77.9%)2,754 (82.1%)0.1749 (78.3%)743 (77.6%)0.015 Hyperlipidemia, n (%)494 (51.4%)1,779 (53.0%)0.03494 (51.6%)480 (50.2%)0.029 Chronic kidney disease, n (%)193 (20.1%)651 (19.4%)0.02192 (20.1%)176 (18.4%)0.042 Ischemic heart disease, n (%)234 (24.3%)980 (29.2%)0.11234 (24.5%)214 (22.4%)0.049 Osteoporosis, n (%)38 (4.0%)179 (5.3%)0.0738 (4.0%)31 (3.2%)0.039Laboratory values Hemoglobin A1c, %, mean ± SD7.0 ± 1.46.9 ± 1.30.067.0 ± 1.47.0 ± 1.30.023GLP-1RA Glucagon‑like peptide‑1 receptor agonist, SD Standard deviation, Std diff Standardized mean difference, SMD < 0.1 indicates acceptable covariate balance Baseline characteristics of GLP-1RA users and active-comparator patients before and after propensity score matching GLP-1RA Glucagon‑like peptide‑1 receptor agonist, SD Standard deviation, Std diff Standardized mean difference, SMD < 0.1 indicates acceptable covariate balance Primary outcome: revision surgery Within 2 years of index RCR, revision shoulder surgery occurred in 10 of 957 GLP‑1RA users (1.0%) and 11 of 957 active‑comparator patients (1.1%). The absolute risk difference was − 0.1% (95% CI − 1.0% to 0.8%), and the RR was 0.91 (95% CI 0.39–2.13). In Kaplan–Meier analysis, the 2‑year revision‑free survival was 98.04% in the GLP‑1RA group and 97.33% in the comparator group, with no significant difference on log‑rank testing (p = 0.83)(Table 2). The Cox model in the matched cohort yielded an HR of 0.91 (95% CI 0.39–2.15; p = 0.74) for GLP-1RA use, indicating no statistically significant association with revision surgery; however, the wide confidence interval reflects the low event rate and limits precision in estimating small differences in risk. In the unmatched multivariable Cox model, the adjusted HR for GLP-1RA exposure was 1.14 (95% CI 0.55–2.37; p = 0.73), likewise consistent with no detectable association with revision (Table 3). Kaplan–Meier curves for revision-free survival demonstrated nearly overlapping 2-year survival probabilities in the two groups (98.04% vs. 97.33%; log-rank p = 0.83), with no visually appreciable divergence over time (Fig. 1). Table 2Structural outcomes at 2-year follow-up in matched cohortsGLP-1RA userNon GLP-1RA userGLP-1RA user %Non GLP-1RA user %Risk Ratio (95% CI)Hazard Ratio (95% CI)p-valuePrimary outcome Revision shoulder surgery10/95711/9571.01.10.91 (0.39–2.13)0.91 (0.39–2.15)0.74 2-yr revision-free survival (%)98.0497.33Log-rank p = 0.83Secondary outcome Retear-proxy events70/95782/9577.38.60.85 (0.63–1.16)0.85 (0.62–1.17)0.76 2-yr freedom from retear-proxy (%)86.3486.59Log-rank p = 0.33GLP-1RA Glucagon‑like peptide‑1 receptor agonist, RR Risk ratio, HR Hazard ratio, CI Confidence interval Structural outcomes at 2-year follow-up in matched cohorts GLP-1RA Glucagon‑like peptide‑1 receptor agonist, RR Risk ratio, HR Hazard ratio, CI Confidence interval Table 3Cox proportional hazards regression resultsVariableHazard Ratio (95% CI)p-valueMatched cohort Revision outcomeGLP-1RA vs. Non GLP-1RA user0.91 (0.39–2.15)0.74 Retear-proxy outcomeGLP-1RA vs. Non GLP-1RA user0.85 (0.62–1.17)0.76Unmatched multivariable model Revision outcomeGLP-1RA vs. Non GLP-1RA user1.14 (0.55–2.37)0.73Age at Index (per year)0.93 (0.90–0.96)< 0.001Male sex1.03 (0.54–1.98)0.92BMI (per unit)0.97 (0.91–1.03)0.28Hemoglobin A1c (per %)0.94 (0.73–1.22)0.66Chronic kidney disease1.03 (0.39–2.69)0.96Ischemic heart disease0.26 (0.08–0.85)0.025 Retear-proxy outcomeGLP-1RA vs. Non GLP-1RA user0.94 (0.72–1.23)0.64Age at Index (per year)0.96 (0.95–0.98)< 0.001Male sex0.98 (0.96–1.00)0.07BMI (per unit)0.96 (0.88–1.06)0.43Hemoglobin A1c (per %)0.94 (0.73–1.22)0.66Chronic kidney disease0.94 (0.69–1.27)0.69Ischemic heart disease0.98 (0.75–1.27)0.86GLP-1RA Glucagon‑like peptide‑1 receptor agonist, RR Risk ratio, HR Hazard ratio, CI Confidence interval Cox proportional hazards regression results GLP-1RA Glucagon‑like peptide‑1 receptor agonist, RR Risk ratio, HR Hazard ratio, CI Confidence interval Kaplan–Meier curves for revision-free survival after arthroscopic rotator cuff repair. Kaplan–Meier estimates of revision-free survival over 24 months in propensity score–matched T2DM patients undergoing arthroscopic rotator cuff repair, comparing pre-operative GLP-1RA users (blue) with active-comparator non-GLP-1 diabetes therapy users (red). Two-year revision-free survival was 98.04% versus 97.33%, with no significant difference between groups (log-rank p = 0.83; hazard ratio 0.91, 95% CI 0.39–2.15). Numbers at risk are displayed below the x-axis Secondary outcome: retear‑proxy events Retear‑proxy events occurred in 70 GLP‑1RA users (7.3%) and 82 active‑comparator patients (8.6%) during 2‑year follow‑up. The risk difference was − 1.3% (95% CI − 3.7% to 1.2%), and the RR was 0.85 (95% CI 0.63–1.16; p = 0.31). Kaplan–Meier curves showed similar 2-year freedom from retear-proxy events in both groups (86.34% vs. 86.59%; log-rank p = 0.33; Table 2; Fig. 2). In the matched cohort, GLP‑1RA use was not associated with the risk of retear‑proxy events (HR 0.85, 95% CI 0.62–1.17, p = 0.76). In the unmatched multivariable Cox model further adjusting for age, sex, BMI, haemoglobin A1c, chronic kidney disease, and ischaemic heart disease, GLP‑1RA exposure likewise showed no significant association with retear‑proxy events (HR 0.94, 95% CI 0.72–1.23, p = 0.64) (Table 3). Kaplan–Meier curves for retear-proxy-free survival after arthroscopic rotator cuff repair. Kaplan–Meier estimates of freedom from the retear-proxy event (new ICD-10 code M75.1 plus upper-extremity MRI within 2 years) in the same matched cohort, comparing pre-operative GLP-1RA users (blue) with active-comparator non-GLP-1 diabetes therapy users (red). Two-year retear-proxy-free survival was 86.34% versus 86.59%, and GLP-1RA use was not associated with retear-proxy risk (log-rank p = 0.33; hazard ratio 0.85, 95% CI 0.62–1.17). Numbers at risk are shown below the x-axis Discussion In this global, propensity-matched cohort of T2DM patients undergoing arthroscopic RCR, pre‑operative GLP‑1RA use within 6 months of surgery was not associated with a detectable increase in 2-year risk of revision shoulder surgery or retear-proxy events compared with active-comparator non-GLP-1 therapy. In both the propensity-matched and unmatched multivariable Cox models, hazard ratios for revision and retear-proxy outcomes remained close to 1.0, with confidence intervals crossing unity and p-values well above conventional thresholds for statistical significance. The point estimates for revision and retear-proxy outcomes did not suggest a large relative increase in risk, and Kaplan–Meier curves for revision-free and retear-proxy-free survival were nearly superimposable; however, the low number of revision events results in wide confidence intervals, so modest effects cannot be excluded. Taken together, these findings are consistent with the short-term structural safety of GLP-1RAs after RCR in T2DM patients, while underscoring the need for cautious interpretation and further prospective study. Comparison with previous literature Our results complement and extend prior work on GLP‑1RAs and rotator cuff pathology. A large real‑world study comparing SGLT2is and GLP‑1RAs reported lower risks of incident rotator cuff tear and primary RCR among SGLT2i users, suggesting potential tendon‑protective effects of SGLT2is, but did not specifically address post‑operative outcomes in RCR recipients [3]. Another recent database analysis found no increase in reoperation after RCR among GLP‑1RA users, supporting a neutral safety profile, but relied on a single administrative claims source and lacked laboratory data and imaging‑based proxies of retear [6]. Our findings are consistent with recent large-scale matched-cohort analyses by Seddio et al. [15] and Lauck et al. [16]. Seddio et al. reported that semaglutide use was associated with a lower risk of postoperative complications and structural retears in T2DM patients undergoing RCR. Similarly, Lauck et al. observed no significant increase in complications or reoperation rates among GLP-1RA users, further reinforcing the safety profile of these agents in the peri-operative setting. By focusing on T2DM patients within a global federated EMR network, applying an active‑comparator new‑user framework, incorporating detailed laboratory covariates, and evaluating both revision surgery and a retear‑proxy event, the present study provides an independent and methodologically distinct confirmation that GLP‑1RAs do not markedly worsen RCR structural outcomes. Experimental studies further support a non‑detrimental or even potentially beneficial role of GLP‑1RAs in tendon biology. Liraglutide has been shown to reduce inflammatory signaling and endoplasmic reticulum stress in rotator cuff enthesis models, improving histological healing and biomechanical strength via GLP‑1R–AMPK/SIRT1 activation [5]. Exendin‑4 promotes tenogenic differentiation of mesenchymal stem cells and upregulates tendon‑related matrix genes, suggesting that GLP‑1R agonism may not inherently impair tendon repair [4]. Our clinical findings align with these mechanistic data by failing to demonstrate any evidence of increased revision or retear‑proxy events among GLP‑1RA users after RCR. Notably, recent evidence suggests that certain medications within our active-comparator group may independently influence postoperative outcomes. Feng et al. demonstrated in a randomized controlled trial that metformin significantly reduces the incidence of shoulder stiffness after arthroscopic RCR by modulating inflammatory pathways [17]. Additionally, a comprehensive review by Morya et al. highlighted the potential protective roles of SGLT2 inhibitors in rotator cuff disease among patients with diabetes, mediated through anti‑inflammatory and tenogenic mechanisms [18]. To address the potential confounding effects of these specific agents, we performed additional sensitivity analyses comparing GLP-1RA users directly against metformin-only and SGLT2i-only cohorts (see Additional file 1: Supplementary Table S1). These analyses showed no significant differences in retear-proxy outcomes between GLP-1RAs and metformin (p = 0.069; HR 0.72) or SGLT2is (p = 0.665; HR 1.12). Taken together, these results suggest that the observed lack of an apparent effect of GLP-1RAs on structural outcomes is robust and not primarily driven by the specific composition of the active-comparator group; in particular, the broadly similar retear-proxy risks in GLP-1RA versus metformin-only and SGLT2i-only users indicate that our overall “no strong detrimental effect” findings are not simply masked by a protective effect within the composite comparator cohort. Clinical implications Given the growing use of GLP‑1RAs and ongoing debate about peri‑operative continuation versus cessation, especially in procedures where delayed gastric emptying and aspiration are concerns [19], orthopaedic surgeons may worry whether GLP‑1RAs adversely affect tendon healing [20–22]. In T2DM patients undergoing arthroscopic RCR, our data suggest that continuing or initiating GLP‑1RA therapy around the time of surgery is unlikely to substantially increase the risk of structural failure requiring revision surgery or prompting advanced imaging for suspected retear within 2 years. Decisions regarding peri‑operative GLP‑1RA management should therefore prioritize systemic metabolic and anaesthetic considerations rather than concerns about rotator cuff healing alone, while concurrently accounting for each patient’s overall risk profile [23–27]. Strengths and limitations Strengths of this study include the use of a large, multi-institutional federated EMR network; restriction to T2DM patients; an active-comparator new-user design; propensity score matching with good covariate balance; and the incorporation of both revision surgery and a retear-proxy event as structural outcomes. Nevertheless, several limitations warrant caution. First, specific tear characteristics such as tear size, tendon retraction, and fatty infiltration were not granularly available in the database. Although the use of the arthroscopic repair CPT code likely enriches the cohort for small-to-large tears deemed repairable—while massive or chronic retracted tears with substantial fatty infiltration are more often managed with non-repair strategies (e.g. debridement, superior capsular reconstruction, or arthroplasty) and thus were not captured—considerable variation in tear morphology and tendon quality may still remain within this group. Residual confounding by indication is therefore possible if GLP-1RA users differ systematically from comparators in unmeasured structural factors, surgical technique, or rehabilitation patterns despite propensity score matching on BMI, HbA1c, and major comorbidities. Second, the overall 2-year revision rate was approximately 1%, which is lower than rates reported in most single-centre series and likely reflects several factors. These include fragmentation of care leading to under-ascertainment of revisions performed outside the contributing network, as well as inherent data missingness in federated EMR systems, where the absence of seamless linkage between independent healthcare organisations may result in under-reporting of diagnostic codes or imaging events. Although the network includes a mix of academic and community hospitals, we cannot track patients across non-participating health systems, and it is unknown whether the propensity to receive revision care within versus outside the network differs between GLP-1RA users and comparators; for example, GLP-1RA users may, on average, be more likely to receive care in specialised centres that are not uniformly represented, which could lead to differential under-capture of revisions. In addition, there is often a high clinical threshold for revision surgery in T2DM patients due to peri-operative risks, and many structural failures remain asymptomatic or are managed non-operatively. Our retear-proxy rate (7.3–8.6%) likely provides a more sensitive reflection of structural integrity than the revision rate alone; however, MRI-based definitions do not guarantee structural failure, asymptomatic retears are not captured, and the proxy we have employed may still underestimate the true incidence of anatomic retears. Moreover, the database did not capture post-operative shoulder range of motion or validated patient-reported outcome (PRO) measures, so our findings pertain primarily to structural and procedural endpoints rather than functional recovery or patient-perceived shoulder status. Third, residual confounding remains possible despite propensity score matching and multivariable adjustment, particularly with respect to unmeasured aspects of tear severity, repair techniques, rehabilitation protocols, and medication adherence. Fourth, our exposure definition captured GLP-1RA prescriptions in the 6-month pre-index window but did not reliably capture whether patients continued or temporarily withheld GLP-1RA therapy during the early postoperative period when tendon-to-bone healing is most active. Consequently, our results are best interpreted as reflecting the structural outcomes associated with pre-operative GLP-1RA use, rather than continuous GLP-1RA administration across the entire 6-month healing window. Finally, follow-up was limited to 2 years; longer-term effects of GLP-1RAs on tendon integrity and shoulder function after RCR remain unknown. In addition, the overall number of revision events was low, resulting in wide confidence intervals around the hazard ratio for revision and limiting the power to detect small but potentially clinically relevant differences in revision risk. Our findings are therefore best interpreted as showing no evidence of a large detrimental effect of GLP-1RAs on revision risk, rather than proving equivalence between treatment groups. Conclusions In this propensity-matched, real-world cohort of T2DM patients undergoing arthroscopic RCR, GLP‑1RA use in the 6‑month pre‑operative window was not associated with a detectable increase in 2-year risk of revision shoulder surgery or retear-proxy events compared with active-comparator non-GLP-1 diabetes therapy. These findings are consistent with the short-term structural safety of GLP-1RAs following RCR and contribute additional real-world evidence to the evolving literature on GLP-1RA use in orthopaedic surgery. Future prospective studies with imaging-confirmed retears, functional and patient-reported outcomes, and longer follow-up are needed to more definitively characterize the long-term effects of GLP-1RAs on tendon healing. Supplementary Information Supplementary Material 1. Acknowledgements None. Authors’ contributions “ **Mao Yi, Yang** :Conceptualization, Methodology, Formal Analysis, Investigation, Data Curation, Writing – Original Draft Preparation, Writing – Review & Editing, Visualization.; **Li Wei, Hung** : Conceptualization, Supervision. All authors have read and agreed to the final version of the manuscript. “. Funding This work was supported by the Ministry of Science and Technology, Taiwan (ROC) (NSTC114-2622-E-341-001) and the Shin Kong Wu Ho-Su Memorial Hospital, Tawan (2021SKHARD017). The funding body had no role in the study design, data collection, data analysis, interpretation of data, or writing of the manuscript. Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate This study was conducted in accordance with the ethical principles of the Declaration of Helsinki. The protocol was reviewed and approved by the Institutional Review Board of Shin Kong Wu Ho-Su Memorial Hospital, Taipei, Taiwan (IRB number: 20260127R). The need for written informed consent from individual participants was waived by the Institutional Review Board because the analysis used only de‑identified data from a federated electronic medical record network and involved no direct patient contact or intervention. Consent for publication Not applicable, as this study used only de‑identified data and did not involve individual patient images or case reports requiring consent. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.