Use of waist circumference to predict insulin resistance: retrospective study

H Wahrenberg, K Hertel, BM Leijonhufvud, LG Persson… - Bmj, 2005 - bmj.com
H Wahrenberg, K Hertel, BM Leijonhufvud, LG Persson, E Toft, P Arner
Bmj, 2005bmj.com
We analysed a sample of 2746 healthy volunteers (798 male) from retrospectively collected
data. Ages ranged from 18 years to 72 years, body mass index (kg/m2) from 18 to 60, and
waist circumferences from 65 cm to 150 cm (see table A on bmj. com for further data). We
determined height, weight, waist circumference (midway between the lateral lower ribs and
the iliac crest), and hip circumference. Results from analyses of venous plasma for glucose,
insulin, lipids, and leptin concentrations were used. We used homoeostasis model …
We analysed a sample of 2746 healthy volunteers (798 male) from retrospectively collected data. Ages ranged from 18 years to 72 years, body mass index (kg/m2) from 18 to 60, and waist circumferences from 65 cm to 150 cm (see table A on bmj. com for further data). We determined height, weight, waist circumference (midway between the lateral lower ribs and the iliac crest), and hip circumference. Results from analyses of venous plasma for glucose, insulin, lipids, and leptin concentrations were used. We used homoeostasis model assessment (HOMA index) as a measure of insulin sensitivity (plasma glucose (mol/l)× plasma insulin (mU/l)/22.5)—an established test in epidemiological studies. 1 We defined insulin resistance as a HOMA score> 3.99, on the basis of a definition for a white population. 2
We used multivariate regression models to assess the predictive power of the variables (see bmj. com). We used receiver operating characteristics (ROC) curve analysis to select an appropriate cut-off for variables. In the multiple regression model, waist circumference was the strongest regressor of the five significant covariates (standardised partial regression coefficients: waist circumference β1= 0.37; log-plasma triglycerides β2= 0.23; systolic blood pressure β3= 0.10, high density lipoprotein cholesterol β4=− 0.09; and body mass index β5= 0.15 (P< 0.001)). The areas under the ROC curves were 0.8915 (standard error 0.008) for men and 0.8644 (0.007) for women, respectively, indicating a very good discriminating power. On the basis of the ROC curves, we set the optimal cut-off for detecting insulin resistance at 100 cm for waist circumference in both sexes. The table shows the number of true and false positives and negatives in both sexes (see also the figure on bmj. com). Sensitivities and specificities were between 94-98% and 61-63% respectively in both sexes. The positive predictive values in our sample were 61% in men and 42% in women (these figures depend on the prevalence of insulin resistance in the actual sample). The negative predictive value was 98% in both sexes. With a cut-off of 88 cm in women (the level cited in guidelines) the specificity dropped to 49%. 3
bmj.com