Precision of measuring body fat distribution in adolescent African American girls from the 'healthy growth study'

• Published in: American journal of human biology Vol. 8; no. 3; pp. 325 - 329
• Main Authors: Mueller, William H., Taylor, Wendell C., Chan, Wenyaw, Sangi-Haghpeykar, Haleh, Snider, Sharon A., Hsu, Hai-An
• Format: Journal Article
• Language: English
• Published: New York Wiley Subscription Services, Inc., A Wiley Company 1996
• ISSN: 1042-0533; 1520-6300
• DOI: 10.1002/(SICI)1520-6300(1996)8:3<325::AID-AJHB3>3.0.CO;2-Z

Precision estimates are given for anthropometric assessment of body fat distribution in participants (n = 86) of the Healthy Growth Study (total n = 154). This five year longitudinal study explored the psychosocial and biologic influences on activity levels in urban adolescent African American girls. The basic anthropometric data include height, weight, four body and limb circumferences, and five skinfold measurements. It is proposed that ratio indices of body fat distribution are likely to have poorer precisions than the single variables which they comprise, and that ratios based on skinfolds may be particularly sensitive to this problem. The precision of the body mass index (BMI) and principal components of skinfold fatness and fat distribution are also considered. Precisions were greater than 0.95 (intraclass correlation from a random‐effects analysis of variance) for most anthropometric dimensions and indices, including the BMI and the first principal component of fatness. However, three of the five skinfolds had lower precisions (0.84–0.93). In contrast, the precisions of all indices of body fat distribution were 0.90 or less (range: 0.57–0.87 skinfold ratios; 0.83–0.90 circumference ratios). Of the ratios, conicity was most stable at 0.90 for repeated measures by both the same and different observers. A second principal component of central fat did not have precisions noticeably better than skinfold ratios (0.79–0.82). Indices of fat distribution may have lower reliabilities than the single variables that they comprise, because errors may be compounded when dividing one variable by another or in linear combinations of measurements other than the first principal component. This problem should be taken into account in clinical and epidemiologic investigations of body fat distribution. © 1996 Wiley‐Liss, Inc.