Anthropometric predictors of body fat in a large population of 9-year-old school-aged children

Objective: To develop and cross-validate predictive models for percentage body fat (%BF) from anthropometric measurements [including BMI z-score (zBMI) and calf circumference (CC)] excluding skinfold thickness. Methods: A descriptive study was carried out in 3,084 pre-pubertal children. Regression models and neural network were developed with %BF measured by Bioelectrical Impedance Analysis (BIA) as the dependent variables and age, sex and anthropometric measurements as independent predictors. Results: All %BF grade predictive models presented a good global accuracy (≥91.3%) for obesity discrimination. Both overfat/obese and obese prediction models presented respectively good sensitivity (78.6% and 71.0%), specificity (98.0% and 99.2%) and reliability for positive or negative test results (≥82% and ≥96%). For boys, the order of parameters, by relative weight in the predictive model, was zBMI, height, waist-circumference-to-height-ratio (WHtR) squared variable (_Q), age, weight, CC_Q and hip circumference (HC)_Q (adjusted r² = 0.847 and RMSE = 2.852); for girls it was zBMI, WHtR_Q, height, age, HC_Q and CC_Q (adjusted r² = 0.872 and RMSE = 2.171). Conclusion: %BF can be graded and predicted with relative accuracy from anthropometric measurements excluding skinfold thickness. Fitness and cross-validation results showed that our multivariable regression model performed better in this population than did some previously published models.