Comparative Performance of Fasting Insulin Resistance Surrogates for Questionnaire- Defined Osteoarthritis in National Health and Nutrition Examination Survey 2011-2023

Background/Aims: Metabolic dysfunction may contribute to a metabolically linked osteoarthritis (OA) phenotype, but it remains unclear which fasting insulin-resistance surrogate is most reproducibly associated with a questionnaire-defined OA phenotype and whether any single surrogate improves cross-sectional classification beyond routine clinical variables. A comparative analysis of metabolic score for insulin resistance (METS-IR), triglyceride–glucose index (TyG), and triglyceride-to-high-density lipoprotein cholesterol ratio (TG/HDL-C) was performed for their associations with prevalent questionnaire-defined OA and their incremental classification performance.

Materials and Methods: A repeated cross-sectional analysis was conducted using National Health and Nutrition Examination Survey (NHANES) 2011-2018 as the primary/derivation window and NHANES 2021-2023 as the temporal replication/validation window. Questionnaire-defined OA was defined as self-reported clinician-diagnosed arthritis with osteoarthritis/ degenerative arthritis reported as the arthritis subtype. Comparators were participants reporting no arthritis; arthritis-positive participants with non-OA, unknown, or missing subtype information were excluded. Survey-weighted logistic regression evaluated associations in fasting analytic sets. A prespecified no high-sensitivity C-reactive protein (no-hs-CRP) demographic-clinical base model was evaluated for cross-sectional classification performance.

Results: The 2011-2018 fasting sample included 8061 participants with 1137 questionnaire-defined OA cases; the 2021-2023 fasting sample included 2570 participants with 530 cases. In the 2011-2018 primary model excluding body mass index from the association adjustment set, METS-IR showed the strongest per-standard deviation association with questionnaire- defined OA (odds ratio [OR] 1.43, 95% confidence interval [CI] 1.29- 1.59), followed by TyG (OR 1.18, 95% CI 1.05-1.33), and TG/HDL-C (OR 1.11, 95% CI 1.01-1.21). Temporal replication supported a durable METS-IR signal (OR 1.57, 95% CI 1.32-1.86). The base model performed well in derivation and temporal validation, and adding individual surrogates yielded negligible incremental classification performance.

Conclusion: Metabolic score for insulin resistance, an adiposity-linked metabolic composite, showed the most reproducible cross-sectional association with questionnaire-defined OA. However, fasting insulin-resistance surrogates added little classification performance beyond routine demographic and clinical variables.

Cite this article as: Hu Y, Zhang H, Wang M, Shi Z, Pan C. Comparative performance of fasting insulin resistance surrogates for questionnaire-defined osteoarthritis in national health and nutrition examination survey 2011-2023. ArchRheumatol. Published online August 5, 2026. doi: 10.5152/ArchRheumatol.2026.26401.