Traumatology and Orthopаedics of Kazakhstan · Oct 28, 2026

The Role of Artificial Intelligence in Osteoporosis Risk Assessment in Type 2 Diabetes Mellitus: Pharmacological Dimensions and Personalized Therapy Perspectives

By Sakhipova G, Kolbayeva A, Zhamaliyeva LM et al.

Skeletal fragility in type 2 diabetes mellitus (T2DM) is clinically important because fracture risk may remain elevated even when bone mineral density (BMD) is normal or increased. This discrepancy complicates screening, particularly when conventional fracture-risk prediction tools do not account for diabetes-specific factors, propensity to fall, or measures of bone quality. Glucose-lowering therapy may also affect skeletal outcomes, with effects that vary by drug class, while methods for using artificial intelligence (AI) to predict individual risk are advancing rapidly. Objective: To summarize the evidence on fracture risk in adults with T2DM, the limitations of current risk-assessment tools, the effects of glucose-lowering therapy on skeletal outcomes, and the potential role of AI in personalized fracture prevention. Methods. The search covered four predefined thematic areas: “diabetes and osteoporosis or fractures,” “traditional fracture-risk assessment and trabecular bone score,” “glucose-lowering therapy and skeletal outcomes,” and “artificial intelligence/machine learning and fracture prediction.” The main synthesis covered publications from 2020 to 2026. Twenty records from a selected, bibliographically verified reference set were reviewed; 19 were included in the qualitative analysis. These figures do not represent the full volume of the initial database search results. One landmark earlier cohort study was used for contextual discussion only. Results. Current evidence confirms an increased incidence of fractures in T2DM and indicates that BMD and standard calculated risk estimates may not fully capture skeletal vulnerability. The most consistent adverse skeletal signal has been observed with thiazolidinediones. Findings for incretin-based therapies are generally neutral or favorable, although they vary according to treatment duration and study design. AI models have demonstrated promising discriminatory performance in certain populations; however, heterogeneity in outcome definitions and limited external validation do not yet support claims of universal superiority over conventional tools. Conclusion. Fracture prevention in T2DM should be based on an integrated assessment of skeletal, metabolic, and fall-related factors. Clinical implementation of AI requires transparent reporting, calibration, and external validation.

doi.org/10.52889/1684-9280-2026-77-5-jto071 Open access Interventions