Ability of body mass index to predict abnormal waist circumference: receiving operating characteristics analysis

Body mass index (BMI) and waist circumference (WC) are the most used anthropometric measures to identify obesity. While BMI is considered to be a simple and accurate estimate of general adiposity, WC is an alternative surrogate measure of visceral obesity. However, WC is subject to significant inter-examiner variation. The aim of the present study was to correlate BMI and WC measures in a group of Brazilian adults to determine the most accurate BMI values for predicting abnormal WC.

BMI and WC were measured in 1184 volunteers (45.6 ± 17.3 yrs; 69% female) using standard procedures. Abnormal WC was defined as ≥88 cm in women and ≥102 cm in men using the traditional criteria, and ≥80 cm in women and ≥90 cm in men using the new criteria. Statistical analysis involved the calculation of Pearson's correlation coefficients and receiver operating characteristic (ROC) curves.

BMI was strongly correlated with WC (women: r = 0.87, p < 0.0001, area under ROC curve = 0.93 ± 0.1; men: r = 0.89, p < 0.0001, area under ROC curve = 0.94 ± 0.01). The most accurate BMI cutoff point for abnormal WC was 27.1 kg/m2 for men and 26.8 kg/m2 for women using the traditional WC criteria, and 24.7 kg/m2 for men and 24.9 kg/m2 for women using the new WC criteria.

Based on the strong correlation found with WC, BMI can be used as the primary anthropometric measure to estimate adiposity, since both obese and most overweight subjects will have abnormal WC.