Type 2 Diabetes Mellitus Modifies the Association Between Exercise Therapy and Injury Risk and Long-Term Outcomes in Patients with Knee Osteoarthritis: A Retrospective Cohort Study

Abstract

Introduction Structured exercise is a core treatment for knee osteoarthritis (KOA), but whether its associations with musculoskeletal injury, knee arthroplasty utilization, and opioid use differ by type 2 diabetes mellitus (T2DM) status remains uncertain. Methods This retrospective cohort study used electronic health records from 2014–2024. A 3-month landmark design classified patients as receiving structured exercise therapy or usual care. Propensity score matching yielded 1,329 pairs in the T2DM+KOA cohort and 3,441 pairs in the KOA-only cohort. Injury was analyzed using clustered Cox regression, total knee arthroplasty (TKA) utilization using Fine–Gray models, and opioid-use trajectories using group-based trajectory modeling. Results Over 3.4 years of follow-up, exercise therapy was not significantly associated with injury in the KOA-only cohort (adjusted HR, 0.93; 95% CI, 0.76–1.13; p = 0.462), but was associated with increased injury risk in the T2DM+KOA cohort (adjusted HR, 1.39; 95% CI, 1.11–1.74; p = 0.004). The Exercise Therapy × T2DM interaction was significant (p = 0.006). Exercise therapy was associated with lower TKA utilization in the KOA-only cohort (adjusted subdistribution HR, 0.79; 95% CI, 0.68–0.92; p = 0.003), but not in the T2DM+KOA cohort (adjusted subdistribution HR, 0.92; 95% CI, 0.76–1.12; p = 0.408). The interaction was significant for the Progressive/Worsening opioid-use trajectory (OR, 1.71; 95% CI, 1.12–2.61; p = 0.014). Among patients with T2DM receiving exercise therapy, exploratory spline analysis showed a nonlinear association between glycated hemoglobin (HbA1c) and injury risk (p for nonlinearity = 0.006). Conclusion Exercise therapy showed different associations with injury according to T2DM status, but causal treatment-effect modification cannot be inferred. It was associated with lower TKA utilization in the KOA-only cohort, whereas no significant association was observed in the T2DM+KOA cohort. The HbA1c finding was exploratory and requires external validation. Individualized assessment and monitoring warrant evaluation, particularly among patients with poorer glycemic control.