Comparison of Several Prediction Equations Using Skinfold Thickness for Estimating Percentage Body Fat vs. Body Fat Percentage Determined by BIA in 6-8-Year-Old South African Children: The BC-IT Study.
Lynn Moeng-MahlanguMakama Andries MonyekiJohn Joseph ReillyHerculina Salome KrugerPublished in: International journal of environmental research and public health (2022)
Body composition measurement is useful for assessing percentage body fat (%BF) and medical diagnosis, monitoring disease progression and response to treatment, and is essential in assessing nutritional status, especially in children. However, finding accurate and precise techniques remains a challenge. The study compares %BF determined by bioelectrical impedance analysis (BIA) and calculated from available prediction equations based on skinfolds in young South African children. A cross-sectional study performed on 202 children (83 boys and 119 girls) aged 6-8 years. Height and weight, triceps and subscapular skinfolds were determined according to standard procedures. %BF was determined with BIA and three relevant available equations. SPSS analyzed the data using paired samples tests, linear regression, and Bland-Altman plots. Significant paired mean differences were found for BIA and Slaughter (t 201 = 33.896, p < 0.001), Wickramasinghe (t 201 = 4.217, p < 0.001), and Dezenberg (t 201 = 19.910, p < 0.001). For all of the equations, the standards for evaluating prediction errors (SEE) were above 5. The Bland-Altman plots show relatively large positive and negative deviations from the mean difference lines and trends of systematic under- and over-estimation of %BF across the %BF spectrum. All three equations demonstrated a smaller %BF than the %BF measured by BIA, but the difference was smallest with the Wickramasinghe equation. In comparison, a poor SEE was found in the three %BF predicted equations and %BF derived from BIA. As such, an age-specific %BF equation incorporating criterion methods of deuterium dilution techniques or 'gold-standard' methods is needed to refute these findings. However, in the absence of developed %BF equations or 'gold-standard' methods, the available prediction equations are still desirable.