Introduction
Glucagon-like peptide-1 receptor agonists (GLP-1As) were approved by the Food and Drug Administration in 2005 for the treatment of type 2 diabetes mellitus (DM). GLP-1As have demonstrated high glycemic efficacy while conferring physiologically induced weight loss through the reduction of appetite and energy intake [1–4]. Notably, semaglutide, commonly known as Ozempic, has rapidly increased in popularity because of its efficacy for weight loss [1,5–7].
Despite the growing use of GLP-1A since the early 2020s, their effects on surgical outcomes remains underexplored. Only recently have an increasing number of studies considered the relationship between GLP-1A use and postoperative outcomes [8–10]. Furthermore, animal studies have demonstrated that the effects of GLP-1As extend beyond tightening glycemic control, with evidence indicating accelerated epithelial regeneration and angiogenesis and improved wound healing by lowering oxidative stress and inhibiting platelet activation [10]. Emerging data from hip and knee arthroplasty indicate reduced rates of complications, readmissions, revisions, and mortality among GLP-1A users [11]. However, their roles in spine surgery—particularly in adult spinal deformity (ASD) and multilevel arthrodesis—have not been well explored.
ASD involves complex spinal misalignments, primarily in the thoracic and lumbar spine, and often requires high-risk surgery with complication and revision rates as high as 39% and 26%, respectively [12–16]. Patients typically need substantial preoperative optimization of comorbidities, bone health, and nutrition. Given the rapidly increasing interest in the use of GLP-1As for the management of type 2 DM and weight reduction, healthcare providers and surgeons must understand their potential association with outcomes of multilevel spinal fusion for ASD correction [17].
This study aimed to examine 1- and 2-year postoperative outcomes in patients with ASD who underwent deformity correction and spinal fusion and compare outcomes of GLP-1A users with those of nonusers.
Materials and Methods
TriNetX database
This retrospective, multicenter cohort study used the TriNetX Global Collaborative Network, which compiles deidentified electronic health records from 146 healthcare organizations worldwide, primarily large academic centers. The database includes over 160 million patient records. For this study, data of patients with ASD who underwent spinal arthrodesis, with or without GLP-1A use, were extracted. Of the 146 institutions participating in TriNetX, only 49 contributed relevant patient data. All data are deidentified per HIPAA guidelines (§164.514[a]); therefore, this study qualified for institutional review board exemption [18].
Study population and outcomes
On August 8, 2025, the TriNetX Global Database was queried to generate two cohorts: a GLP-1A group, consisting of patients who had been on GLP-1A therapy for at least 1 year before surgery, and a non-GLP-1A group of patients had no history of GLP-1A use. Medications were identified using RxNorm codes for GLP-1As semaglutide (1991302), liraglutide (475968), lixisenatide (1440051), dulaglutide (1551291), or tirzepatide (2601723), a dual glucose-dependent insulinotropic polypeptide/glucagon-like peptide-1 (GIP/GLP-1) agonist. Procedures were identified using International Classification of Diseases, 10th Revision (ICD-10), and Current Procedural Terminology (CPT) codes, focusing on posterior arthrodesis (CPT 22800, CPT 22802, and CPT 22804), anterior arthrodesis (CPT 22808 and CPT 22810), three-column osteotomy (3CO) (CPT 22206 and CPT 22207), and vertebral column resection (VCR) (CPT 22818). A subgroup analysis was performed on 3CO and VCR cases. Patients aged ≥18 years were included. Notably, the CPT codes—excluding 3CO—did not specify the location of the deformity.
Postoperative outcomes were assessed at 1- and 2-year intervals (Supplement 1). Hardware failure was defined as any instance of mechanical failure, such as implant breakdown, hardware displacement, and loosening. Pseudoarthrosis, defined by ICD-10 as bone healing failure after fracture or fusion, is typically diagnosed within 1 year postoperatively using X-ray or computed tomography imaging [19]. However, the TriNetX platform does not specify diagnostic criteria. All procedural and diagnostic codes used are listed in Supplement 1.
Propensity-based matching
The cohorts were 1:1 greedy propensity-score matched based on patient demographics and comorbidities (Tables 1, 2). Body mass index (BMI) was encoded as a categorical variable. Following propensity-score matching, the main analysis included 396 patients in both cohorts, whereas the 3CO subgroup generated 49 patients (Fig. 1).
Statistical analysis
The outcomes of interest assessed at 1-year and 2-year postoperative periods and analyses were generated within the TriNetX platform. Baseline and matched attributes of the cohorts were compared based on descriptive characteristics and presented as means with standard deviation for continuous variables and percentages for categorical variables. Fisher’s exact test was conducted to highlight differences in categorical variables between cohorts, given the small sample size, whereas independent sample t-tests were employed for continuous variables. Odds ratio (OR) and 95% confidence intervals (CIs) were calculated to evaluate differences in postoperative outcomes and complications. Significance was defined as p-value <0.05.
To assess residual confounding, standardized mean differences were calculated for the matching variables, whereas E-values were used to determine the strength of confounders needed to explain the observed associations. Subclass analyses stratified GLP-1 agents as traditional or dual GIP/GLP-1 agonists, and outcomes were compared with matched controls. Heterogeneity was assessed by comparing effect estimates, and sensitivity analyses were conducted by sequentially excluding subclasses. A parallel analysis focused on hardware-related complications to assess device-specific effects independent of systemic drug benefits.
Results
Demographics
Of the 26,684 patients who underwent multilevel deformity correction with arthrodesis and osteotomy, 417 were prescribed a GLP-1A within 1 year. Following propensity-score matching, 396 patients remained in each cohort for 1- and 2-year outcomes; 215 non-GLP-1 users were excluded for falling outside the 20-year window. In the GLP-1A cohort, 61.9% were female, 81.1% were White, 11.6% were Black, the mean age was 61.7±11.6 years, and the BMI was 35.6±6.6 kg/m2. In the non-GLP-1 cohort, 61.6% were female, 81.8% were white, 13.1% were Black, the mean age was 63.9±11.3 years, and the BMI was 32.7±5.7 kg/m2. Of the 2,329 patients who underwent 3CO, 54 were prescribed GLP-1A. After matching, 49 patients remained in each cohort. In the GLP-1A cohort, 65.3% were female, 79.6% were White, 20.4% were Black, the mean age was 60.9±10.2 years, and the BMI was 34.7±5.6 kg/m2. In the non-GLP-1 group, 61.2% were female, 85.7% were White, 20.4% were Black, the mean age was 60.3±15.5 years, and the BMI was 32.0±6.1 kg/m2 (Tables 1, 2).
1-year outcomes
Compared with the non-GLP-1A group cohort, the GLP-1A cohort preoperatively had significantly lower odds of pneumonia (OR, 0.356; 95% CI, 0.189–0.669; p=0.001), pseudoarthrosis (OR, 0.39; 95% CI, 0.282–0.682; p<0.001), surgical site infections (OR, 0.448; 95% CI, 0.261–0.769; p=0.003), wound dehiscence (OR, 0.394; 95% CI, 0.212–0.731; p=0.002), sepsis (OR, 0.315; 95% CI, 0.172–0.576; p<0.001), thromboembolic events (OR, 0.570; 95% CI, 0.363–0.895; p=0.014), hardware failure (OR, 0.527; 95% CI, 0.324–0.855; p=0.009), readmission (OR, 0.166; 95% CI, 0.121–0.226; p<0.001), mortality (OR, 0.379; 95% CI, 0.259–0.555; p<0.001), and cardiac complications (OR, 0.504; 95% CI, 0.547–0.999; p=0.046) after multilevel deformity correction (Table 3).
2-year outcomes
At this time interval, the GLP-1A cohort preoperatively demonstrated significantly lower odds of pneumonia (OR, 0.483; 95% CI, 0.283–0.824; p=0.007), pseudoarthrosis (OR, 0.481; 95% CI, 0.322–0.718; p<0.001), surgical site infection (OR, 0.389; 95% CI, 0.231–0.654; p<0.001), sepsis (OR, 0.379; 95% CI, 0.223–0.543; p<0.001), wound dehiscence (OR, 10.385; 95% CI, 0.212–0.702; p=0.001), thromboembolic event (OR, 0.506; 95% CI, 0.330–0.777; p=0.002), hardware failure (OR, 0.480; 95% CI, 0.314–0.735; p=0.001), readmission (OR, 0.185; 95% CI, 0.135–0.251; p<0.001), cardiac complications (OR, 0.496; 95% CI, 0.267–0.922; p=0.024), and mortality (OR, 0.404; 95% CI, 0.284–0.575; p<0.001). Notably, the association between GLP-1A use and hardware failure demonstrated an increase in significance over time, from the 1-year to the 2-year follow-up (Table 3).
3CO subgroup outcomes
At 1 year following the procedure, the GLP-1A group cohort demonstrated lower incidences of pseudoarthrosis (p=0.247) but higher readmission rates (p=0.236), although these differences did not meet the significance threshold. Rates of pneumonia, surgical site infections, wound dehiscence, hardware failures, thromboembolic events, cardiac events, sepsis, and mortality were comparable between the groups. At 2 years, pseudoarthrosis and hardware failure rates remained lower in those taking GLP-1As, although not significant (p=0.371 and p=0.218, respectively) (Table 4).
Sensitivity/subclass analyses
Subclass analyses showed consistent outcomes across therapeutic categories, with no significant heterogeneity between traditional GLP-1 and dual GIP/GLP-1 agonists (p=0.543 and p=0.492). Both subclasses demonstrated protective effects with overlapping CIs; dual agonists had slightly greater relative risk reduction (62%–72% vs. 58%–72%). Results remained stable across subclasses, confirming that conclusions were not dependent on subclasses. Sensitivity analysis isolating mechanical failures (rod fracture, screw loosening, cage subsidence, and junctional failure) confirmed that GLP-1 agonists provide biomechanical protection beyond systemic effects (OR, 0.52; 95% CI, 0.32–0.83; p=0.009). Standardized mean differences were <0.10, indicating minimal confounding, and stratified Mantel-Haenszel ratios aligned with unpooled estimates. E-values for readmission and hardware failure were 9.9 and 3.1, respectively, indicating that an unmeasured confounder would be needed to increase the odds approximately 10-fold to negate findings.
Discussion
This study revealed a significant association between preoperative GLP-1A use and reduced odds of pseudoarthrosis, hardware failure, readmissions, and mortality at 1 and 2 years in patients undergoing multilevel deformity correction for ASD. These benefits were also seen in the 3CO subgroup despite its smaller sample size, proposing clinical consistency across both arthrodesis and osteotomy procedures. In this study, the authors hypothesize that these trends resulted from the pharmacokinetic effects of GLP-1As on bone metabolism. Animal studies have shown that these medications suppress osteoclast differentiation and resorption, thereby limiting bone degradation. In addition, liraglutide-treated diabetic mice exhibited reduced osteoclastic activity and protective effects on bone [20,21]. GLP-1As also enhance bone formation by upregulating osteoblast-related genes through mesenchymal stem cell turnover, which may promote more robust fusion [20,22].
The literature on GLP-1A use in spinal surgery is limited, as no studies have specifically examined patients with ASD undergoing deformity correction. Using the TriNetX Database, Ghali et al. [23] studied patients with DM taking GLP-1As who underwent lumbar fusions and found higher rates of all-cause anemia, renal failure, opioid use, emergency visits, and wound complications within 90 days postoperatively. However, these patients had lower rates of pseudoarthrosis, and no significant differences were noted in readmissions, adjacent segment disease, hardware failure, or postlaminectomy syndrome. The authors theorized that this difference could be attributed to the protective rather than restorative effects of GLP-1As on the bone matrix. Although the latter findings align with our overall theme of long-term equipoise in GLP-1A effects, conclusions about 90-day outcomes could not be drawn, as the present study’s earliest postoperative interval was 1 year following arthrodesis. Notably, the present study showed a trend toward nonsignificance for pseudoarthrosis at 2 years, possibly due to increased early surveillance and detection of surgical failure, whereas delayed fusions may manifest and be corrected over time.
This study has two findings of interest: the reduced rates of readmission and mortality observed among patients taking GLP-1As. As previously mentioned, surgical treatment for ASD is associated with a high risk of complications, including death. Zuckerman et al. [24] investigated 6,000 patients who underwent deformity correction and found that PNA, DVT, and unplanned intubation were independently predictive of mortality within 1 year of surgery. A high incidence of PE has also been identified as a major cause of death among this patient population [25]. The present study demonstrated a significant decrease in the odds of thromboembolic events (p=0.001), which may be a result of GLP-1As’ inhibition of platelet aggregation and thrombosis [26]. The significant reduction in readmission rates may reflect the reduced need for higher-level care in these patients. Our results underscore the medical effects of GLP-1As in patients undergoing ASD surgery and their association with improved downstream outcomes.
Additional observations regarding the biomechanical properties of GLP-1As on articular surfaces may help in further illuminating the aforementioned findings. Previous studies have demonstrated that this class of medications promotes cartilage synthesis and decreases macrophage-mediated inflammatory pathways, contributing to enhanced osseous and periosteous integrity [20,27]. These findings align with the results of Vatsia et al. [27] and Zheng et al. [20] and are further supported the findings of Agrawal et al. [18], who reported reduced pseudoarthrosis rates in GLP-1A users undergoing single-level lumbar fusion using similar TriNetX data. Although the database lacks specific diagnostic criteria for pseudoarthrosis, both the present study and the study by Agrawal et al. [18] define this variable using the same ICD-10 parameters, supporting the consistency of findings. Overall, the effects of GLP-1As on deformity correction outcomes are encouraging, demonstrating a reduction in postoperative complications.
Several orthopedic studies have explored the effects of GLP-1A use on total joint arthroplasty outcomes. Magruder et al. [28,29] conducted two retrospective cohort studies and showed that semaglutide use before total hip and knee arthroplasty was linked to lower rates of 90-day readmissions and reduced prosthetic joint infections and sepsis but higher risks of MI and pneumonia. Furthermore, semaglutide was found to reduce infectious processes through improved glucose control, reduced oxidative stress, and modulation of immune cells [30]. Although these reductions in readmission and infection rates align with our findings, only one GLP-1A was studied, and previous clinical studies have shown slightly different mechanisms among GLP-1As available on the market [5,7]. Therefore, assessing the generalizability of these studies poses a challenge.
Several hypotheses may explain the significant reduction in wound dehiscence seen in this analysis of GLP-1A use. A recent experimental study suggested that GLP-1As aided surgical wound healing in diabetics through mechanisms beyond glycemic control, including reduced inflammation, enhanced angiogenesis, and tissue regeneration [31]. Animal studies further show that GLP-1As accelerate wound reepithelialization by promoting proangiogenic responses, often impaired in uncontrolled diabetes [32–35]. Additionally, the findings of reduced rates of cardiac complications in GLP-1A users align with the results of a previous study indicating longitudinal cardiovascular risk reduction benefits [36,37].
Finally, GLP-1As have documented adverse effects. GLP-1As cause delayed gastric emptying. Although it improves satiety, it also carries the risk of aspiration pneumonia, specifically during the immediate perioperative period [38–40]. Although this study revealed lower rates of pneumonia in patients taking GLP-1A, this was likely due to an infectious process as opposed to aspiration. Therefore, providers should carefully weigh the risks and benefits of surgical treatment in this patient population to minimize postoperative complications.
This study has several limitations, primarily related to its retrospective design and use of de-identified data. The lack of access to individual patient health records in TriNetX may have led to unrecognized discrepancies in demographics, comorbidities, and diagnoses. The relatively small size of the GLP-1A cohort, particularly in the 3CO subgroup, reflects both the niche population studied and the recent adoption of GLP-1As since the early 2020s [17]. Outcome measurements and cohort characteristics rely on ICD-10 codes, which may limit accuracy and fail to differentiate primary from revision surgeries and potentially influence results based on the sequence of surgeries. Moreover, the clinical follow-up intervals were 1 and 2 years owing to the study’s retrospective cohort design, and long-term clinical follow-up could have demonstrated the results in a more succinct, longitudinal fashion. This study did not stratify the location of deformity or fused levels owing to CPT code limitations and the need to preserve statistical power. Key predictors of fusion and mechanical failure—such as surgeon experience, pelvic incidence–lumbar lordosis mismatch, implant characteristics, lumbar lordosis correction, and lordosis distribution index—were unavailable in deidentified databases such as TriNetX. Future studies incorporating these fundamental variables could add substantial clinical weight to our findings.
Conclusions
Despite the limitations of this study, the results reveal that preoperative GLP-1A use was associated with lower odds of pseudoarthrosis, hardware failure, readmissions, and mortality. These results, alongside existing orthopedic literature, merit careful consideration by spine surgeons treating patients on GLP-1A therapy. Thus, thorough preoperative evaluation in this population is necessary to optimize outcomes and minimize complications. Our findings also warrant discussion as to whether GLP-1As could serve as an adjunctive therapy for eligible patients to improve surgical recovery and reduce disease burden. Additional prospective data should be gathered to assess the pharmacokinetic window of the benefit of GLP-1As for patients who underwent surgery to deliver proper postoperative care. Given the continued paucity of data in the spine community, more studies are needed to enhance the power of the present findings and evaluate the longitudinal mechanisms underlying these findings.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Full text via PMC
Download Citation
Supplement
