Receiver Operating Characteristic Curve Analysis of BMI and Percentage Body Fat in Type 2 Diabetics of Punjab

K. D. SINGH, j. K. DHILLON, ANUPREET ARORA* AND B. S. GILL**
Departments of *Physiology and **Medicine,
Govt. Medical College,
Patiala – 147 001

( Received on January 9, 2003 )

click here for full article in pdf

Abstract

The present study attempted to establish appropriate cut off levels of Body Mass Index (BMI) for defining overweight as a risk for the development of type 2 diabetes considering percentage body fat (BF) as standard. A total of 300 patients of known type 2 diabetes participated in the study (150 males and 150 females, all ³ 40 years of age). Clinical examination was done. Anthropometric measurements as BMI, Waist Circumference (WC) and Waist-hip ratio (WHR) were calculated. Percentage BF was calculated using skinfold thickness method from the equation of Durnin and Womersley. Mean BMI for males was 24.97 (SD 4.3) kg/m2 and for females was 27.56 (SD 5.14) kg/m2. Mean percentage BF for males was 28.19 (SD 0.74) and for females was 38.22 (SD 5.29). A comparison of BF and BMI data with various ethnic groups revealed conspicuous differences. Receiver operating characteristic (ROC) curve analysis showed a low sensitivity of conventional cut off value of BMI (25kg/m2) in identifying subjects with overweight as compared to the cut off values based on percentage BF (males >25, females >30). This results in substantial misclassification. Based on the ROC curve, a lower cut off value of BMI 22.3kg/m2, displayed the optimal sensitivity and specificity, and less misclassification in identification of type 2 diabetics with high percentage BF. BF: BMI was calculated and was found to be higher in females.