Osteoarthritis (OA) is a common degenerative joint disease with limited long-term nonoperative treatments. Type 2 diabetes mellitus (T2DM) is a recognized risk factor for OA and may accelerate disease progression via chronic low-grade inflammation, insulin resistance, and metabolic dysfunction. Novel glucose-lowering agents such as SGLT2 inhibitors and GLP-1 RAs have pleiotropic metabolic and anti-inflammatory effects that could plausibly influence OA risk. This study compared real-world associations between initiation of SGLT2 inhibitors and GLP-1 RAs and subsequent OA-related outcomes in adults with T2DM.
The investigators performed a retrospective cohort study using the TriNetX US Collaborative Network, an electronic health record (EHR)–based research database drawing from 67 health care organizations. The study period for identifying new users of the index drugs was January 1, 2017, through December 31, 2019. Follow-up continued until outcome occurrence, death, or December 31, 2024.
Adults (≥18 years) with T2DM (ICD-10 E11) who were newly prescribed either an SGLT2 inhibitor or a GLP-1 RA were eligible. New users had no prescriptions for the index class or comparator in the 12 months preceding the index date. Patients with a prior diagnosis of OA (ICD-10 M15–M19) or prior use of the comparator medication were excluded. SGLT2 inhibitors included canagliflozin, dapagliflozin, empagliflozin, ertugliflozin, bexagliflozin, and sotagliflozin; GLP-1 RAs included semaglutide, albiglutide, dulaglutide, liraglutide, lixisenatide, and exenatide.
After applying inclusion and exclusion criteria, the initial cohorts comprised 586,195 SGLT2 inhibitor users and 545,583 GLP-1 RA users. A 1:1 propensity-score matching (PSM) procedure produced two matched cohorts of 452,445 patients each.
Covariates used in propensity-score estimation included demographic variables, HbA1c, comorbidities (essential hypertension, ischemic heart disease, chronic kidney disease, cerebrovascular disease, peripheral vascular disease, obesity), and concurrent medications (including metformin and insulin). Matching used a greedy nearest-neighbor algorithm with a caliper of 0.2 standard deviations of the logit of the propensity score. Covariate balance was assessed via standardized mean differences (SMD), with SMD < 0.1 considered acceptable. After matching, all SMDs were < 0.1, though some small residual differences remained for chronic kidney disease, cerebrovascular disease, peripheral vascular disease, and HbA1c.
The primary outcome was incident OA over five years, defined as ICD-10 codes M15–M19 recorded on at least two separate clinical encounters to increase diagnostic specificity. Secondary outcomes included hip OA (M16) and knee OA (M17), each defined similarly; major joint injections (CPT 20610, 20611) defined as at least two recorded injection events; and arthroplasty procedures—total hip arthroplasty (THA; CPT 27130) and total knee arthroplasty (TKA; CPT 27447).
Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazards models. Kaplan–Meier survival analyses and log-rank tests were used to compare time-to-event outcomes. Statistical significance was defined as P < 0.05. All analyses were performed within the TriNetX platform.
From the TriNetX dataset of >104 million individuals, 17,480,946 met initial visit criteria; 6,436,562 had used either SGLT2 inhibitors or GLP-1 RAs. After exclusions and matching, each study arm included 452,445 patients. Before matching, the SGLT2 inhibitor group was older and had a higher proportion of males; these and other baseline differences were balanced after PSM, with acceptable SMDs.
Over five years, initiation of SGLT2 inhibitors was associated with a lower risk of incident osteoarthritis compared with GLP-1 RAs (HR 0.921; 95% CI 0.895–0.946; P < 0.001).
When site-specific OA was examined, the SGLT2 inhibitor group had a significantly lower risk of knee osteoarthritis (HR 0.831; 95% CI 0.783–0.882; log-rank P < 0.001). The difference in hip OA incidence was smaller and not statistically significant (HR 0.931; 95% CI 0.843–1.025; log-rank P = 0.111).
There were no statistically significant differences between groups for major arthroplasty endpoints: THA (HR 1.168; 95% CI 0.865–1.577; P = 0.311) and TKA (HR 0.877; 95% CI 0.651–1.182; P = 0.389). However, users of SGLT2 inhibitors had fewer major joint injections compared with GLP-1 RA users (1.14% vs 1.31%; HR 0.915; 95% CI 0.882–0.951; P < 0.001).
In this large, multi-institutional EHR-based cohort of adults with T2DM, new initiation of SGLT2 inhibitors was associated with a modestly lower risk of incident osteoarthritis compared with initiation of GLP-1 RAs, with a more marked relative reduction for knee OA and a lower rate of major joint injections. The authors propose that metabolic, anti-inflammatory, and weight-loss effects of these agents could influence OA risk, but differences in mechanisms or residual confounding may explain the observed between-class differences.
Key limitations arise from the observational design and reliance on EHR-coded diagnoses and procedures. Although the study used PSM to reduce confounding, residual confounding is possible. Exposure was classified by initial prescription without reclassification for subsequent switching or add-on therapy. The definitions required two diagnostic encounters to enhance specificity, but misclassification remains a possibility. Data access restrictions prevent sharing of patient-level data; supporting definitions and aggregate results are provided in the manuscript and supporting files.
Among adults with T2DM in the TriNetX US Collaborative Network, initiation of SGLT2 inhibitors was associated with a lower incidence of OA overall and a notably lower incidence of knee OA, compared with initiation of GLP-1 RAs. SGLT2 inhibitor users also had fewer major joint injections. The authors recommend further research to confirm these associations and to explore underlying mechanisms.