Association of Insulin Resistance with Dysglycemia in Elder Koreans: Age- and Sex-Specific Cutoff Values

Abstract

Background/Objectives: Dysglycemia including pre-diabetes mellitus (Pre-DM) and type 2 diabetes mellitus (T2DM) is associated with insulin resistance. This study aimed to support personalized early diagnosis of dysglycemia by proposing optimal, sex- and age-specific cutoff values for Homeostatic Model Assessment of Insulin Resistance (HOMA-IR) and Homeostatic Model Assessment of Beta-Cell Function (HOMA-β) in Koreans aged ≥65 years. Methods: This study analyzed 3862 older Koreans from the 8th Korea National Health and Nutrition Examination Survey data (2019–2021), excluding those with prior diabetes or medication. The participants were classified into normal and dysglycemia groups, based on fasting plasma glucose (FPG) and glycated hemoglobin (HbA1c). Sex- and age-specific optimal cutoff values were determined using Youden’s Index (YI) and area under the curve (AUC). Results: For T2DM, the optimal HOMA-IR cutoff was 2.25 for men and 2.03 for women, with strong discriminative performance (AUCs: 0.828 and 0.823, respectively). Stratifying cutoff values further by sex and age improved the diagnostic accuracy (AUC > 0.83 in most subgroups), underscoring the value of tailored thresholds. For pre-DM, the HOMA-IR cutoff was 1.73 in men and 1.85 in women (AUCs: 0.682 and 0.665, respectively). Age- and sex-specific cutoffs modestly improved AUCs, particularly in men (up to 0.7), although the improvement was less consistent among women. HOMA-β showed no significant association with dysglycemia, and no meaningful cutoff values were identified. Conclusions: HOMA-IR is a promising marker for the early identification of dysglycemia in older adults when interpreted through a personalized lens. Applying sex- and age-specific cutoff values enhances diagnostic precision and supports a more individualized approach to metabolic risk assessment. Further longitudinal studies are warranted to validate these personalized thresholds and to optimize early detection strategies in diverse populations.