•  body mass index;
  • diabetes risk;
  • obesity


Aims  To examine the utility of body mass index (BMI), waist circumference (WC) and waist-to-height ratio (WHR) in assessing diabetes risk across different ethnic groups.

Methods  Cross-sectional analysis of data for eight ethnic groups from the 2003–2004 National Health and Nutrition Examination Survey and 2003–2004 Health Survey for England was performed. In 11 624 adults ≥ 20 years old, self-reported as US White, US Black, Mexican American, English White, English Black, Bangladeshi, Pakistani, Indian or Chinese the presence of diabetes, defined as self-report of doctor diagnosis or glycated haemoglobin (HbA1c) > 6.1%, was ascertained. Comparisons of proportions were made using χ2-tests. Receiver operating characteristic (ROC) curves were calculated for BMI, WC and WHR predicting diabetes.

Results  Other ethnic groups had a higher prevalence of diagnosed diabetes than English Whites. The crude prevalence of diabetes in English Whites of normal weight (BMI < 25 kg/m2) was 3.4%. Higher prevalences were seen in other ethnic groups (5.0–10.9%). Based on ROC curves, both WC and WHR had better discriminating ability for diabetes than BMI for both genders and some ethnic groups.

Conclusions  Ethnic differences exist in the crude prevalence of diabetes, even in those characterized as normal weight by BMI. Thus, clinicians need to exercise caution in interpreting diabetes risk associated with a normal BMI. The use of other anthropometric measures, such as WC or WHR, may improve risk determination across different ethnic groups. More research is needed to determine the thresholds for different anthropometric measures that improve diabetes risk determination.