Diagnostic accuracy of anthropometric indices for obesity screening among Asian adolescents.

INTRODUCTION Weight-and-height-based anthropometric indices have long been used for obesity screening among adolescents.However, the ability of their age-and-sex-specific reference values in classifying adolescent as "obese" in different populations was not fully established. Our study aimed to validate the existing international (BMI-for-age charts from WHO, CDC, IOTF) and local cut-offs [percent weight for height (PWH)] for obesity against body fat percentage, as assessed by 4 skinfolds measurement. MATERIALS AND METHODS A cross-sectional sample of 6991 adolescents aged 12 to 18 years was measured. All anthropometric measurements were compliant with the internationally accepted protocol. Obesity was defined as percentage body fat greater than or equal to 95 percentile, specific to age and sex. The validity of the existing classification criteria in detecting obesity was evaluated by comparing their respective diagnostic accuracy. RESULTS Both prevalence of obesity and diagnostic accuracy indices varied by the classification criteria. While all criteria generated very high specificity rates with the lowest being 95%, their sensitivity rates were low ranging from 43% to 71%. Youden's index suggested that CDC and WHO criteria had optimal sensitivity and specificity. ROC analysis showed that overall performance could be improved by refining the existing cut-offs. CONCLUSIONS Clinical validity of weight-and-height-based classification systems for obesity screening in Asian adolescents is poorer than expected, and this could be improved by refining the existing cut-offs.