Comparison of the ability of anthropometric indices to predict the risk of Diabetes Mellitus in South African males: SANHANES-1

SOURCE: International Journal of Environmental Research and Public Health
OUTPUT TYPE: Journal Article
TITLE AUTHOR(S): M.D.Sekgala, R.Sewpaul, M.Opperman, Z.J.Mchiza
DEPARTMENT: Human and Social Capabilities (HSC)
Print: HSRC Library: shelf number 9812226
HANDLE: 20.500.11910/19294

If you would like to obtain a copy of this Research Output, please contact Hanlie Baudin at


This study aimed to assess the sensitivity of body mass index (BMI) to predict the risk of diabetes mellitus (DM) and whether waist circumference (WC), waist-to-hip (WHR) and waist-to height (WHtR) ratios are better predictors of the risk of DM than BMI in South African men aged 20 years and older. Data from the first South African National Health and Nutrition Examination Survey (SANHANES-1) were used. Overall, 1405 men who had valid HbA1c outcomes were included. The sensitivity, specificity, and optimal cut-off points for predicting DM were determined using the receiver operating characteristic (ROC) curve analysis. A total of 34.6% percent of the study participants were overweight/obese, while 10.5%, 10.4%, 36.6% and 61.0% had HbA1c, WC, WHR and WHtR above the normal reference ranges, respectively. Based on age-adjusted logistic regression analysis, the highest likelihood of DM was observed for those participants who had increased WC and WHtR (odds ratios [OR] were 6.285 (95% CI: 4.136-9.550; p < 0.001) and 8.108 (95% CI:3.721-17.667; p < 0.001)). The ROC curve analyses for WC, WHR, and WHtR displayed excellent ability to predict the risk of DM, with their areas under the curve (AUC) being 80.4%, 80.2% and 80.8%, respectively. The overall cut-off points to predict the risk of DM for WC, WHR, and WHtR were 88.95 cm, 0.92, and >0.54, respectively. The ROC analysis for BMI, on the other hand, showed acceptable ability to predict the risk of DM (AUC = 75.6%), with its cut-off point being 24.64 kg/m2. Even after stratifying the data by two age groups, WHtR remained a superior index to predict DM, especially in the younger age group. To conclude, no significant differences were observed between the AUC for BMI the AUCs for other indices. However, the AUCs for these indices showed significant excellent ability as opposed to the significant acceptable ability of BMI to predict DM in adult South African men.