Skip to main content
. 2013 Jan 30;97(3):480–486. doi: 10.3945/ajcn.112.047787

TABLE 4.

Results of receiver operating characteristic curve analyses for the utility of visceral adipose tissue (cm2), BMI (kg/m2), waist circumference (cm), total fat mass (kg), and percentage of body fat in the prediction of the upper quintile of the continuous metabolic syndrome score1

AUC (95% CI) Threshold Sensitivity Specificity
White women
 Visceral adipose tissue 0.785 (0.743, 0.823)a 141.0 0.674 0.805
 BMI 0.721 (0.676, 0.763)b 31.1 0.616 0.732
 Waist circumference 0.781 (0.739, 0.820)a 95.1 0.663 0.776
 Total fat mass 0.701 (0.655, 0.744)b 34.3 0.593 0.752
 Percentage of body fat 0.631 (0.583, 0.677)c 40.3 0.605 0.624
African American women
 Visceral adipose tissue 0.696 (0.641, 0.746)a,b 97.1 0.694 0.667
 BMI 0.665 (0.609, 0.717)a 30.2 0.742 0.526
 Waist circumference 0.726 (0.673, 0.775)b 95.6 0.710 0.715
 Total fat mass 0.663 (0.607, 0.715)a 31.1 0.726 0.558
 Percentage of body fat 0.605 (0.548, 0.659)c 40.2 0.532 0.691
White men
 Visceral adipose tissue 0.734 (0.689, 0.777)a 140.2 0.778 0.609
 BMI 0.646 (0.597, 0.693)b 29.1 0.778 0.474
 Waist circumference 0.680 (0.633, 0.726)b 95.6 0.901 0.388
 Total fat mass 0.644 (0.595, 0.691)c 23.3 0.815 0.462
 Percentage of body fat 0.616 (0.567, 0.663)d 26.0 0.827 0.462
African American men
 Visceral adipose tissue 0.789 (0.696, 0.865)a 81.9 0.900 0.575
 BMI 0.814 (0.724, 0.885)a 31.3 0.750 0.775
 Waist circumference 0.799 (0.707, 0.872)a 101.5 0.750 0.775
 Total fat mass 0.791 (0.698, 0.866)a 25.5 0.750 0.800
 Percentage of body fat 0.788 (0.694, 0.863)a 26.6 0.700 0.825
Total sample
 Visceral adipose tissue 0.734 (0.708, 0.758)a 111.5 0.751 0.610
 BMI 0.689 (0.663, 0.715)b 31.1 0.614 0.666
 Waist circumference 0.726 (0.700, 0.750)a 95.1 0.767 0.616
 Total fat mass 0.668 (0.641, 0.694)c 31.1 0.590 0.670
 Percentage of body fat 0.586 (0.558, 0.613)d 26.6 0.900 0.253
1

Groups with different superscript letters were significantly different from each other within sex-by-race groups on the basis of pairwise comparisons by using the method of DeLong et al (31).