Which metric of relative weight best captures body fatness in children?

To evaluate the relative merits of BMI (kilograms per meter squared) and age- and gender-adjusted BMI, age- and gender-specific z score of BMI, and age- and gender-specific percentiles of BMI as surrogate measures of body fatness among a sample of youth.

The sample comprised 596 children and adolescents 5 to 18.7 years old and was 40% male and 55% white. Height and weight were measured by trained research staff. DXA was used to determine body fat mass. BMI, age- and gender-specific percentile of BMI, and age- and gender-specific z scores of BMI were computed, and these metrics were compared with measured body fatness.

The BMI values in the sample ranged from 12.9 to 55.0 kg/m(2), with a mean of 24.9 kg/m(2). The Spearman correlations with percentage body fat were similar for all of the BMI metrics (r = 0.82 to 0.88). Linear regression models with age- and gender-specific percentiles of BMI explained significantly less of the variance (65%) than models with log-transformed BMI (81%) or age- and gender-specific z scores of BMI (75% to 79%). z scores were the most accurate at classifying children who were overfat (sensitivity = 0.84, specificity = 0.96 for z score > or = 1). However, using a BMI > or = 85th percentile or a BMI > or = 20 kg/m(2) was also accurate at classifying youth.

The BMI metrics had similar correlations with body fatness, but age- and gender-specific percentiles of BMI were the least accurate proxy measure of body fatness. However, a BMI z score > or = 1, BMI percentile > or = 85, and BMI> or = 20 kg/m(2) are all useful for identifying children who may be overfat.