Objective: There are relatively low rates of screening for overweight and obesity among children and adolescents in primary care. A simplified method for such screening is needed. We examined if weight-for-age percentile curves are sufficiently sensitive in identifying overweight and obesity in the pediatric age range.
Design: We used data from two distinct sources: 1) four consecutive cycles of the National Health and Nutrition Examination Surveys (NHANES) from the years 2005-2012, using participants aged 2-17.9 years for whom data on age, sex, weight and height were available (n=12,884), and 2) primary-care clinic measurements from Israel (n=15,152). Primary outcomes were the threshold values of weight-for-age percentiles which best discriminated between normal-weight, overweight and obesity status.
Results: Receiver operating characteristic analyses demonstrated that weight-for-age percentiles well discriminate between normal-weight and overweight, and between non-obese and obese, individuals (area under curve= 0.956 and 0.977 respectively, both p<0.001). Following Classification and Regression Trees analysis, the 90th and 75th weight-for-age percentiles were chosen as practical cutoffs for obesity and overweight, respectively. These cutoffs had high sensitivity and negative predictive value in identifying obese participants (94.3% and 98.6%, respectively, for the 90th percentile) and in identifying overweight participants (93.2% and 95.9%, respectively, for the 75th percentile). The sensitivities and specificities were found valid against data obtained from NHANES 2011-2012 and from the primary-care setting.
Conclusions: Weight-for-age percentiles can discriminate well between normal-weight, overweight and obese children and adolescents. Children found to be above the aforementioned percentiles should be evaluated further, according to available pediatric obesity guidelines.
