Skip to main content
ARS Home » Plains Area » Houston, Texas » Children's Nutrition Research Center » Research » Publications at this Location » Publication #198543
Title: CROSS-VALIDATION OF LIPOMETER ESTIMATES OF BODY COMPOSITION: THE EFFECT OF GENDER AND SKIN COLOR

Author
item MCFARLIN, B - UNIVERSITY OF HOUSTON
item JACKSON, A - UNIVERSITY OF HOUSTON
item Ellis, Kenneth
item CALLIE, M - UNIVERSITY OF HOUTSON
item TRUETT, L - UNIVERSITY OF HOUSTON
item TURPIN, I - BAYLOR COLLEGE MED
item Bray, Molly

Submitted to: Obesity Research
Publication Type: Abstract Only
Publication Acceptance Date: 5/20/2006
Publication Date: 10/20/2006
Citation: McFarlin, B.K., Jackson, A.S., Ellis, K.J., Callie, M., Truett, L., Turpin, I., Bray, M.S. 2006. Cross-validation of lipometer estimates of body composition: the effect of gender and skin color [abstract]. Obesity Research. 14:A142.

Interpretive Summary:

Technical Abstract: The Lipometer (v12.1e; Graz, Austria) uses light-emitting diodes (lambda=660 nm) and a photodetector to measure subcutaneous adipose tissue (SAT) and estimate percent body fat (%fat). Since, the Lipometer uses a light beam to measure SAT, it is possible that skin color may influence the results, creating a skin color bias. Thus, the purpose of this study was to cross-validate Lipometer estimates of %fat (LIPOFAT) against dual-energy X-ray absorptiometry %fat (DXAFAT) and examine the effect of gender and skin color on prediction accuracy. Female (n=61) and male (n=38) college students were recruited to participate from the Training Interventions and Genetics of Exercise Response (TIGER) Study. The self-report race/ethnicity of the sample was: African-American (23%), Hispanic (29%), Caucasian (35%), and other ethnic groups (13%). The Lipometer was used to measure SAT at 15 body sites. Skin color was measured on the outer right arm, forehead, and inner right arm with a calibrated luminometer. DXA was used to measure %fat. The data were analyzed using general linear models (GLM) with skin color measures log transformed to account for skewness. The mean LIPOFAT and DXAFAT measures were similar for females (29.7±7.1 and 30.9±8.2) and males (19.7±9.0 and 19.7±7.6). LIPOFAT was correlated with DXAFAT (R2 = 0.57, SEE = 6.3 %fat). GLM analysis showed that both sex and log skin color accounted for independent sources of DXAFAT variance. Adding these variables to the model accounted for an additional 7% of DXAFAT variance (R2 = 0.64, SEE = 5.9 %fat). The LIPOFAT SEE of 5.9 is higher than what we previously re-orted for skinfold estimates of DXAFAT (R2 = 0.87, SEE = 3.3 %fat). When BMI was added to LIPOFAT, this model accounted for an additional 26% of DXAFAT variance (R2 = 0.89, SEE = 3.2 %fat). These results show that the Lipometer, as calibrated, does not accurately estimate DXAFAT. Adding gender, skin color and BMI to the Lipometer prediction model significantly enhanced validity, but further validation research is needed.