Use of anthropometric measures to predict insulin resistance.

ABSTRACT

OBJECTIVE: Obesity is closely related to insulin-resistance (IR) but it is evaluated differently in the diverse definitions of the metabolic syndrome. The objective of this study was to verify the utility of different anthropometric measures to predict IR and to evaluate the best cut-off points. SUBJECTS AND METHOD: We performed a cross-sectional study of the general population aged 40 to 70 years old (n=2,143); a simple random sample of 305 non-diabetic persons was obtained. Sociodemographic data, physical examination and routine biochemical analysis with insulinemia were obtained. IR was defined by a HOMA index (Homeostasis Model Assessment) ≥2.9. To obtain the best variables to predict IR, a forward stepwise logistic regression was performed. Subsequently, a logistic equation was constructed and its predictive capacity was compared with the different anthropometric variables by the area under the ROC (receiver-operating characteristic) curve (AUC). The best cut-off points were established according to the Youden index. RESULTS: Body mass index (BMI) and the waist/hip ratio ×100 were entered into the model, but age, sex, waist, hip and body surface were not. The logistic equation found p(RI)=1/1+exp{-[-14.295]-[0.234×IMC]-[0.07×(waist/hip×100)]} showed good adjustment, and the probability calculated on the basis of this equation showed the greatest AUC overall and in both sexes, followed in women by BMI and by waist measurement in men, but without significant differences. CONCLUSIONS: No significant differences were found between the BMI, waist or hip measurements and the logistic model to predict IR.