Robert S Thiebaud1, Takashi Abe2, Jeremy P Loenneke2, Eiji Fujita3, Takuya Akamine3. 1. Department of Kinesiology, Texas Wesleyan University, Fort Worth, TX 76105, USA. Electronic address: rthiebaud@txwes.edu. 2. Department of Health, Exercise Science, and Recreation Management, Kevser Ermin Applied Physiology Laboratory, School of Applied Sciences, The University of Mississippi, MS 38677, USA. 3. Department of Sports and Life Sciences, National Institute of Fitness and Sports in Kanoya, Kanoya, Kagoshima 891-2393, Japan.
Abstract
BACKGROUND & AIMS: B-mode ultrasound accurately measures both muscle mass, body density and percent body fat (% BF) in younger adults, but how well it can estimate % BF in middle-aged and older adults using DXA-derived %BF as the criterion is unclear. We sought to develop % BF prediction equations for middle-aged and older adults using ultrasound subcutaneous fat thickness (SFT). METHODS: A cross-sectional study of Japanese adults (n = 414, 50-79 years) where 276 subjects were randomly assigned to a model development group and the other 138 subjects were assigned to a cross-validation group. B-mode ultrasound measured SFT at nine sites. Dual energy X-ray absorptiometry (DXA) measured % BF, arm fat mass (FM) and leg FM. Stepwise multiple linear regression developed prediction equations from anthropometric data (body mass, height, waist and hip circumference) and ultrasound SFT sites. Bland-Altman plots assessed validity of the prediction equations to measure % BF in the cross-validation group. RESULTS: The best prediction equation for % BF was the following: [% BF = 15.709 + (1.753*anterior trunk SFT) + (5.626*Sex) + (3.635*posterior upper arm SFT) - (4.428*anterior lower leg SFT) - (0.170*height) + (0.264*waist) + (anterior thigh SFT*2.241); r2 = 0.809, standard error of the estimate (SEE) = 3.3 kg]. Arm FM and leg FM prediction equations had r2 values ranging from 0.690 to 0.812 and SEEs of 0.29 and 0.75 kg. A small mean bias was noted for estimating % BF (-0.14%), but large limits of agreement were found (-8.0-7.7%) and systematic error was noted in all of the equations (r = 0.275 to 0.515, p < 0.05). CONCLUSIONS: Despite high r2 values and a small mean bias found between predicted and DXA % BF, wide limits of agreement were found with some systematic error present. Therefore, these prediction equations for middle-aged and older adults may not be sufficiently accurate to use in a clinical setting.
BACKGROUND & AIMS: B-mode ultrasound accurately measures both muscle mass, body density and percent body fat (% BF) in younger adults, but how well it can estimate % BF in middle-aged and older adults using DXA-derived %BF as the criterion is unclear. We sought to develop % BF prediction equations for middle-aged and older adults using ultrasound subcutaneous fat thickness (SFT). METHODS: A cross-sectional study of Japanese adults (n = 414, 50-79 years) where 276 subjects were randomly assigned to a model development group and the other 138 subjects were assigned to a cross-validation group. B-mode ultrasound measured SFT at nine sites. Dual energy X-ray absorptiometry (DXA) measured % BF, arm fat mass (FM) and leg FM. Stepwise multiple linear regression developed prediction equations from anthropometric data (body mass, height, waist and hip circumference) and ultrasound SFT sites. Bland-Altman plots assessed validity of the prediction equations to measure % BF in the cross-validation group. RESULTS: The best prediction equation for % BF was the following: [% BF = 15.709 + (1.753*anterior trunk SFT) + (5.626*Sex) + (3.635*posterior upper arm SFT) - (4.428*anterior lower leg SFT) - (0.170*height) + (0.264*waist) + (anterior thigh SFT*2.241); r2 = 0.809, standard error of the estimate (SEE) = 3.3 kg]. Arm FM and leg FM prediction equations had r2 values ranging from 0.690 to 0.812 and SEEs of 0.29 and 0.75 kg. A small mean bias was noted for estimating % BF (-0.14%), but large limits of agreement were found (-8.0-7.7%) and systematic error was noted in all of the equations (r = 0.275 to 0.515, p < 0.05). CONCLUSIONS: Despite high r2 values and a small mean bias found between predicted and DXA % BF, wide limits of agreement were found with some systematic error present. Therefore, these prediction equations for middle-aged and older adults may not be sufficiently accurate to use in a clinical setting.
Authors: Erika Aparecida Silveira; Larissa Silva Barbosa; Ana Paula Santos Rodrigues; Matias Noll; Cesar De Oliveira Journal: Arch Public Health Date: 2020-07-18