What Orthopaedic Surgeons Should Know About Glucagon-like Peptide-1 Receptor Agonists in Joint Replacement Surgery

This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [9]. The protocol adhered to the Cochrane Handbook for Systematic Reviews of Interventions and was prospectively registered on PROSPERO (Registration number: CRD420251143459) [10]. Three databases—PubMed, Scopus, and Web of Science—were searched from their inception to June 16, 2025. To ensure comprehensive literature coverage, no restrictions were applied to publication date or study setting during the search. Searches were restricted to published, peer-reviewed comparative studies to synthesize outcomes from real-world clinical practice. Trial registries and grey literature were not searched, as the review’s objective was to analyze comparative effectiveness and safety data from completed studies with reported results, rather than study protocols or unpublished data. However, reference lists of eligible papers and relevant reviews were screened for additional records. The search strategy combined Medical Subject Headings (MeSH) and free-text keywords related to glucagon-like peptide-1 receptor agonists (GLP-1 RAs) and total joint arthroplasty (TJA). The full search strings for each database, which utilized the comprehensive strategy detailed in the main text, are provided in Supplementary Table S1.

Eligible studies were comparative studies published as full-text articles in peer-reviewed journals, comparing postoperative outcomes in adult patients undergoing primary THA, TKA, or TSA who received GLP-1 RAs versus a control group (no GLP-1 RAs or standard care). Studies were required to report at least one of the following outcomes with a minimum follow-up of 3 months: perioperative or postoperative complications (e.g., periprosthetic joint infection [PJI], surgical site infection [SSI], venous thromboembolism [VTE], myocardial infarction [MI], pneumonia, acute kidney injury [AKI], or deep vein thrombosis [DVT]), adverse drug events (e.g., gastrointestinal complications such as nausea, vomiting, diarrhea, or constipation), revision rates, length of stay (LOS), readmission rates, or emergency department (ED) visits. The intervention included any GLP-1 RA (e.g., semaglutide, liraglutide, dulaglutide, exenatide, lixisenatide, tirzepatide, or albiglutide). The operational definition of GLP-1 RA use differed across the included administrative database studies, primarily based on prescription fills within a specified window relative to surgery (e.g., “at the time of surgery,” “within 90 days before surgery,” or “within 1 year preoperatively”). The specific exposure definition for each included study was extracted and is presented in Table 1. For our narrative synthesis, these varying definitions were considered conceptually to reflect either preoperative exposure (documented use only before surgery) or perioperative exposure (use documented in a window spanning the immediate pre- and postoperative period). Exclusion criteria included single-arm studies, conference abstracts without results, secondary studies (e.g., reviews), studies involving revision arthroplasty or non-total joint arthroplasty (TJA) procedures, and studies lacking relevant outcome data. Studies with mixed TJA populations, excluding those with separable THA, TKA, or TSA data, were also excluded. Two authors independently screened records using the Rayyan web application, resolving disagreements by discussion with a third author.

Data were extracted independently by two authors using a predefined standardized spreadsheet. Extracted details included study design, sample size, data source, cohort period, patient demographics (age, sex, body mass index [BMI], comorbidities such as diabetes mellitus), intervention details (GLP-1 RA type and exposure definition), comparator, and outcomes (perioperative/postoperative complications, adverse drug events, revision rates, LOS, readmission rates, and ED visits). For continuous outcomes (e.g., LOS), means and standard deviations were recorded when available; medians with ranges were noted if means were unavailable. Discrepancies between authors were resolved by consensus with a third author.

The Newcastle–Ottawa Scale (NOS) [11] was used to assess the quality of included retrospective cohort studies across three domains: selection, comparability, and outcome (nine items, scored 1–9). Two authors independently evaluated each study, assigning scores based on cohort selection (e.g., representativeness, exposure ascertainment), comparability of groups (e.g., use of propensity score matching), and outcome assessment (e.g., follow-up duration, adequacy of outcome reporting). Disagreements were resolved by consultation with a third experienced author.

A quantitative meta-analysis was deemed inappropriate due to significant clinical and methodological heterogeneity across studies, including variations in patient populations, GLP-1 RA agents, outcome definitions, and follow-up duration. Crucially, the substantial risk of patient cohort overlap across studies using the same large administrative databases (e.g., TriNetX, PearlDiver) precluded the reliable pooling of data. Therefore, a narrative synthesis was conducted. To ensure transparency and mitigate the risk of double-counting patients from overlapping data sources, we implemented the following approach: (1) the primary database and cohort period for each study were clearly documented in Table 1; (2) when multiple studies utilized the same database with overlapping timeframes, this overlap was explicitly acknowledged in the narrative results, and their quantitative findings were not combined; and (3) the synthesis prioritizes comparing the direction and consistency of findings across independent studies, rather than relying on the aggregated sample size, which may be inflated by overlap. Findings were organized by procedure (THA, TKA, TSA) and then by outcome category (e.g., healthcare utilization, implant-related complications, systemic adverse events). For each outcome, results are presented descriptively, highlighting the direction of the effect (e.g., reduced, increased, neutral) and the approximate magnitude (e.g., odds ratios, risk differences) as reported in the primary studies. The aggregated total sample size is provided for context, but it is interpreted with caution due to acknowledged overlap.

The literature search identified 1,632 records. After removing 582 duplicates, 1,050 unique records were screened, of which 1,025 were excluded. The full texts of the remaining 25 articles were reviewed, and 15 studies [12–26] ultimately met the inclusion criteria. The PRISMA flow diagram detailing the study selection process is presented in Fig. 1.

The 15 included retrospective cohort studies [12–26] reported on an aggregate of 318,143 patients (114,365 THA, 125,505 TKA, 78,273 TSA), with 56,132 patients prescribed a GLP-1 RA. This total likely overestimates the number of unique patients due to overlap across large U.S. administrative databases (e.g., TriNetX, PearlDiver). A critical point of heterogeneity was the operational definition of GLP-1 RA exposure, which varied significantly across studies. (Table 1) Definitions ranged from broad historical prescriptions (e.g., any prescription within 1 year before surgery) to strict perioperative criteria (e.g., active prescription at the time of surgery, or for a minimum period spanning 3 months before and after the procedure). Thirteen studies used propensity score matching (ratios 1:1 to 10:1), one used inverse probability weighting [20], and one used an unmatched cohort design [14]. All studies were U.S.-based and published between 2023 and 2025.

Most studies included THA/TKA patients (n = 11); four focused on TSA [15, 20, 23, 26]. Follow-up averaged 18 months, typically reporting 90-day and/or 2-year outcomes. Most studies assessed various GLP-1 RAs, with three focusing exclusively on semaglutide [16, 24, 26]. The mean baseline age (64.6 years) and BMI (35 kg/m²) were similar across groups. (Table 1).

Comorbidity burden, assessed via the Charlson Comorbidity Index (CCI) or Elixhauser Comorbidity Index (ECI), and preoperative HbA1c (range: 6.3–6.5) were well-balanced. Osteoarthritis was the primary surgical indication. A total of 108,107 patients had diabetes, of whom 22,495 (21%) received a GLP-1 RA. Six studies enrolled exclusively diabetic patients [12, 15, 17, 20, 21, 23], one enrolled only non-diabetic patients [25], and the others enrolled mixed cohorts. Patients in both groups commonly used other anti-diabetic medications (e.g., Metformin, Insulin). (Tables 2 & 3).

Generally, studies [12–26] have shown good performance across the selection, comparability, and outcome domains. Aside from Baum et al. [14], which showed moderate quality, the overall body of evidence was predominantly of high quality, with scores ranging from eight to nine. It is important to note that while the NOS assesses specific methodological criteria, a high score does not eliminate the fundamental limitations inherent to observational database studies, such as residual confounding or exposure misclassification. Points were reduced in three studies (Baum et al. [14], Buddhiraju et al. [22], Seddio et al. [26]) due to short-term follow-up, while in one study (Baum et al. [14]), it was due to the omission of propensity score matching. (Table S2).

Based on the available observational data, the use of perioperative GLP-1 RA was not associated with a consistent negative impact on short-term healthcare utilization metrics. For THA and TKA, the length of stay (LOS) was similar or slightly shorter among GLP-1 RA users compared to controls, with reported averages of 2.2 days versus 3.1 days in THA and 2.7 days versus 2.9 days in TKA [12, 18]. This aligns with the known pharmacodynamics of GLP-1 RAs, which are characterized by delayed gastric emptying but generally do not prolong recovery times in orthopaedic settings [27]. The 90-day readmission rates were also low and not consistently different among the groups [12, 22]. The use of emergency department (ED) services was inconsistent; however, some data suggested a possible benefit, with several studies favoring GLP-1 RA users, especially within the THA group [22, 25, 26]. This decrease may indicate improved glycemic control or enhancements in mobility and self-care due to weight loss, thereby reducing the necessity for acute postoperative care [5, 28, 29]. For TSA, data on healthcare utilization were limited. However, one study found significantly fewer 90-day ED visits (OR 0.37) and another one reported an increased odds of 90-day readmissions (OR 1.6) [23, 26].

One of the most notable associations in the included studies was a signal suggesting a potentially reduced risk of local complications, particularly surgical site infections (SSI) and periprosthetic joint infections (PJI), in patients undergoing THA and TKA who received treatment with GLP-1 RAs. Our narrative synthesis revealed that the apparent protective signal was most consistent and pronounced in studies that defined GLP-1 RA exposure in the immediate perioperative period (i.e., at the time of surgery or spanning weeks before and after), compared to those using longer-term preoperative definitions. In high-risk populations—such as those with morbid obesity or diabetes—several studies reported significantly lower rates of PJI/SSI [16, 18, 21, 24]. For example, in diabetic patients undergoing THA, the 2-year PJI rates were 1.6% compared to 2.9% [16], and similar patterns were noted in TKA groups [12, 21, 24]. This association may be mediated by the acute anti-inflammatory and glycemic-stabilizing effects of GLP-1 RAs during wound healing, rather than chronic metabolic benefits from long-term use [28–30]. Furthermore, the reduced incidence of hematoma in THA (0% vs. 1.3%) suggests a possible hemostatic advantage, which may be associated with decreased inflammation [18]. These findings are consistent with a recent meta-analysis suggesting that GLP-1 RA use was associated with reduced postoperative complications, including lower rates of periprosthetic joint infection (PJI), following total knee or hip arthroplasty [31].

The results regarding implant survival, which involve revision surgeries and dislocations, were infrequent and comparable across the different groups within the available short follow-up period. The revision rates ranged from 0.6% to 4.0%, showing no association with an increased early risk among GLP-1 RA users [22, 25]. Similarly, the dislocation rates were low, ranging from 0.8% to less than 2.3% [18, 25]. These findings may be reassuring for short-term implant stability, particularly given the hypothetical concerns about potential GLP-1 RA-related sarcopenia or alterations in bone density that may affect implant survivorship [32]. However, the predominantly short-term follow-up, averaging 18 months, is a critical limitation. Consequently, although the available data do not indicate an increased short-term revision risk, they are insufficient to conclude on long-term safety regarding implant fixation and survival [16, 24]. The neutral short-term revision risk observed aligns with recent meta-analyses indicating neutral fracture risks among diabetic populations on GLP-1 RAs [33]. Nevertheless, prospective studies with extended follow-up are necessary to definitively address this clinical gap.

In contrast, the results on TSA were more diverse, with one study showing a significant decrease in 90-day surgical site infection (SSI) odds (OR 0.25) [26], while another one found no significant differences or trends towards SSI [20]. The inconsistency in TSA findings may also relate to differing exposure definitions, further highlighting how variable methodology can obscure a clear signal. Periprosthetic fractures were rare and showed neutral outcomes across all procedures, suggesting that GLP-1 RA was not associated with harm to bone health [34]. This discrepancy may reflect fundamental differences between the procedures. Periprosthetic joint infection (PJI) in weight-bearing joints like the hip and knee is strongly linked to obesity and metabolic syndrome, conditions that GLP-1 RAs directly target. In contrast, the infection risk in non-weight-bearing shoulder arthroplasty is less clearly tied to these systemic metabolic factors, which may explain why a consistent benefit from GLP-1 RAs is not observed in TSA [4].

Systemic complications showed a wide range of outcomes, with notable inconsistencies across various studies and procedures. In THA and TKA, the evidence for systemic cardiopulmonary, renal, and thromboembolic events was conflicting, without a clear indication of protective or harmful effects. For instance, one study reported potential advantages, such as lower transfusion requirements (OR 0.53 for non-diabetics) and reduced blood loss anemia (OR 0.57) [25]. Conversely, other studies indicated mixed risks: one in TKA found higher odds of 90-day myocardial infarction (OR 1.49) and pneumonia (OR 1.67) [24]. At the same time, another reported lower rates of 90-day acute kidney injury (3.4% vs. 4.0%) [12]. The study reporting increased cardiopulmonary risks employed a strict perioperative exposure definition, suggesting the timing of drug use may influence risk. In contrast, gastrointestinal side effects were consistently more frequent among GLP-1 RA users, with increased rates of nausea/vomiting (18.2% compared to 6.0%) and diarrhea, aligning with the known effects of GLP-1 RAs on gastric motility [13, 27]. Considering this mixed profile, careful perioperative management to minimize gastrointestinal symptoms is essential, along with current anesthesiology guidelines [35].

In TSA, the findings were highly contradictory: one study indicated lower risks of 90-day venous thromboembolism (OR 0.36), cardiac events (OR 0.32), and pneumonia (0.25) [26], while another showed higher risks (DVT OR 3.0; MI OR 2.8; pneumonia OR 2.2) [23]. This variation likely stems from more than just differences in cohort characteristics, follow-up duration, or the use of anticoagulants [36]. The opposing perioperative effects of GLP-1 RAs may lead to different outcomes, and the systemic benefits of the medications, such as reduced inflammation and enhanced endothelial function, may explain the observed decreased risks [37]. However, in specific patient subgroups, changes in body weight and fluid balance induced by GLP-1 RA therapy could be less well tolerated—particularly in individuals with lower baseline BMI or pre-existing cardiac dysfunction—potentially contributing to transient physiological stress and explaining conflicting results [38]. Importantly, these biological effects may be linked to how studies define exposure windows. The study showing reduced risks considered use within 1 year before surgery, likely representing long-term stable users [26]. Conversely, the study reporting increased risks focused on use within 3 months before surgery, potentially capturing patients experiencing acute physiological changes [23]. This contrast underscores that the overall systemic effect may depend on the definition of drug use relative to surgery. Given this profound inconsistency, coupled with the limited number of TSA studies, the evidence is currently insufficient to support any reliable conclusion regarding the systemic safety or risk profile of GLP-1 RAs in shoulder arthroplasty. Isolated signals, such as a reduction in mortality (OR 0.077) [20], are intriguing but require validation in larger, prospective cohorts. The absence of data on gastrointestinal complications in TSA further limits a comprehensive assessment of the condition.

For the orthopaedic surgeons, the synthesis of current observational data suggests that, in patients undergoing total hip or knee arthroplasty, perioperative GLP-1 RA use has not been associated with a clear signal of increased short-term surgical risk in the available studies. A consistent finding across several studies is a potential association with a reduced risk of periprosthetic joint infection, particularly in subgroups with obesity or diabetes. However, this evidence is derived solely from retrospective cohorts with methodological heterogeneity, most notably in the definition of drug exposure. For total shoulder arthroplasty, the evidence is limited and contradictory, offering no interpretable signal. Therefore, while these data are hypothesis-generating and can inform the design of future studies, they are insufficient to guide definitive clinical practice. Clinical decisions regarding perioperative GLP-1 RA management should remain individualized and based on multidisciplinary assessment, considering the known gastrointestinal side-effect profile. Future research must prioritize prospective, controlled designs to establish causality. These studies should employ standardized, clinically relevant definitions of GLP-1 RA exposure (e.g., active perioperative use vs. long-term prescription) to clarify the timing-dependent effects suggested by this review. Additional priorities include extended follow-up to assess long-term implant survivorship, differentiation by procedure and specific GLP-1 RA agent, and the inclusion of patient-reported outcomes.

This systematic review synthesizes evidence from retrospective cohort studies, many of which used large, nationally representative U.S. databases (e.g., TriNetX, PearlDiver), providing a reasonably broad overview of GLP-1 RA use in total joint arthroplasty. The review also incorporates diverse patient populations, including those with diabetes, obesity, and non-diabetic individuals, offering insight into potential variations in GLP-1 RA effects across different subgroups. A primary methodological contribution of this review is its critical analysis of the GLP-1 RA exposure definition. By systematically extracting and presenting each study’s operational criteria (Table 1) and analyzing outcomes in relation to these differing windows, this synthesis moves beyond merely acknowledging heterogeneity. We demonstrate that variable definitions (e.g., “active at surgery” vs. “prescription within 1 year”) can lead to divergent conclusions, providing a transparent framework for interpreting conflicting evidence, particularly in the context of TSA. This establishes exposure timing as a key consideration for future research.

However, the findings must be interpreted within the context of notable limitations inherent in the available evidence. Primarily, all included studies are retrospective and observational, which means causality cannot be established. A major concern is confounding by indication. Patients prescribed GLP-1 RAs form a selected group who tend to receive more structured metabolic management, closer perioperative monitoring, and may be more health-conscious—factors that independently reduce surgical risk and are challenging to fully measure or adjust for. Although propensity matching addressed many known confounders, residual confounding from unmeasured variables (e.g., nutritional status, physical activity, socioeconomic factors) is likely to remain. Additionally, methodological limitations of database studies are relevant. Risks of exposure and outcome misclassification exist; a filled prescription does not confirm perioperative adherence, and coding inaccuracies can influence the reported complications. While our synthesis directly tackles the issue, the inconsistent definitions of GLP-1 RAs across primary studies continue to be a fundamental limitation of the evidence base. Most importantly, a fundamental constraint is that GLP-1 RA use was operationally defined as a pre-surgical prescription in some of the included studies. While our synthesis provides a framework for interpreting this heterogeneity, it cannot overcome the core limitation: we cannot be certain whether the observed associations are due to the drug’s direct perioperative effects or to other factors related to its prior use, such as more structured metabolic care. Consequently, this variability fundamentally obscures the ability to determine an optimal, evidence-based perioperative regimen for GLP-1 RAs in TJA. Third, in our synthesis, we interpreted findings from studies with specific methodological limitations, such as a lack of propensity score matching, with greater caution, as these studies received lower quality scores in our quality assessment. Our overall conclusions are therefore driven by the consistent patterns observed across studies with stronger methodological designs. Fourth, our synthesis faces particular challenges as the substantial overlap in data sources (e.g., multiple studies using PearlDiver) means that the combined sample size overestimates the number of unique patients and may inflate the apparent precision of the evidence. Significant clinical and methodological heterogeneity, including variations in follow-up duration (3 months to 2 years), outcome definitions, and matching strategies, prevented a quantitative meta-analysis, necessitating a narrative synthesis. Fifth, the focus on short- to medium-term outcomes and the exclusion of non-primary and revision procedures limit insights into long-term implant survivorship and safety in more complex surgical scenarios. Furthermore, concentrating solely on U.S. populations may restrict the global relevance of the findings. Finally, by limiting our search to English-language, peer-reviewed publications and excluding grey literature, we might have introduced publication and language bias, potentially overlooking relevant data published in other languages or in non-journal formats.

Not Applicable.

The authors declare no competing interests.

All data generated or analyzed during this study are included in the published article.