Computational methods are significant determinants of the associations and definitions of insulin resistance using the homeostasis model assessment in women of reproductive age.

Background: Insulin resistance (IR) plays an important role in the pathogenesis of polycystic ovary syndrome (PCOS), but identification of insulin-resistant individuals is difficult. The homeostasis model assessment (HOMA), a surrogate marker of IR, is available in 2 computational models: HOMA1-IR (formula) and HOMA2-IR (computer program), which differ in incorporated physiological assumptions. This study evaluates the associations of the 2 models as markers of IR, the metabolic syndrome (MS), and PCOS.

Methods: Anthropometric, hormonal, and biochemical parameters were measured in 92 PCOS women and 110 controls. HOMA1 and HOMA2 were used to assess IR. Regression analyses were used to find the associations of the 2 models with different variables, MS, and PCOS.

Results: The cutoff levels for definition of IR were HOMA1-IR ≥2.9 and HOMA2-IR ≥1.7. Mean HOMA1-IR (2.79) and HOMA2-IR (1.42) differed substantially. The difference (HOMA1-IR - HOMA2-IR) was significantly correlated with insulin, fasting plasma glucose, triglycerides, HDL cholesterol, waist circumference, leptin, and adiponectin (all P < 0.05). HOMA1-IR and HOMA2-IR were significantly associated with MS (odds ratio 5.7 and 4.2, respectively) and PCOS (odds ratio 3.7 and 3.5, respectively).

Conclusions: HOMA computational methods significantly affect the associations and cutoff values used for definition of IR. The correlations of the difference in the computational methods corroborate differences in captured physiological mechanisms. As precise identification of IR in PCOS patients is of practical importance, practitioners and researchers should be aware of these differences in the HOMA computational methods.