Can Ultrasound Subcutaneous Fat Thickness be used to Estimate Percent Body Fat in Older Adults?

Abstract

Ultrasound imaging has been used to estimate body composition, which includes percent body fat (%BF), in young and middle-aged adults. However, it is unknown whether ultrasound imaging can predict %BF in older adults when using dual-energy x-ray absorptiometry (DXA) as the criterion method. PURPOSE: To develop prediction equations to estimate %BF in older men and women using ultrasound subcutaneous fat thickness. METHODS: Four-hundred and nineteen men (n=176) and women (n=243) participated in this study. The average age, BMI and %BF of participants were 70.6 ± 6.4 yrs, 23.5 ± 3.0 kg/m2 and 28.2 ± 7.4%. Participants were randomly separated into a model development group (n=260) and cross-validation group (n=159). B-mode ultrasound using a 5-MHz scanning head imaged subcutaneous fat thickness on the right side of the body at the anterior forearm, anterior and posterior upper arm, anterior trunk, posterior trunk, anterior and posterior thigh, and anterior and posterior lower leg. A whole body scan using DXA was used to determine %BF. IBM SPSS Statistics 24 was used to analyze the data. For multicollinearity assumption, variables with a variance inflation factor (VIF) > 10 were excluded from analysis. Stepwise multiple linear regression analysis was used to develop prediction equations. Bland-Altman plots were used to validate the prediction equations. RESULTS: Variables included in the stepwise linear regression analysis were height, sex, and all fat thicknesses. Age and weight were not significantly correlated with %BF (p<0.05). The prediction equation for %BF was the following: %BF = 27.075 + (3.284*Anterior Trunk) + (4.916*Posterior upper arm) + (2.166*Sex) - (0.099*height) where males = 1 and females = 2. The adjusted r-square of the prediction equation was 0.701 with a standard error of the estimate of 4.1%. Bland-Altman plots revealed a mean bias of 0.2266 ± 7.9772 (95% confidence intervals). A significant correlation (r = 0.245, p=0.002) between the difference in %BF (measured - predicted) and the average %BF [(Measured + predicted %BF)/2] suggests some systematic error in the prediction equation. CONCLUSIONS: Ultrasound imaging can be used to predict %BF in older adults; however, there is some systematic error in the prediction equation.