A method of constructing an index of obesity.

Abstract

Body density data obtained from underwater weighing, of 458 children and adults aged from 6 to 51 years, in combination with weight (W) and stature (S), have been used to estimate percent body fat (%BF). A model for an obesity index of the form: %BF = C(Wm/Sk) is proposed. The parameters of the model were derived using a non-linear least squares procedure to maximize the relationship between the index and %BF. The constant C was independent of age but varied by sex. The estimated values of m (1.2) and k (3.3) for the sample studied were ageand sex-independent and significantly different from integers. The estimates of %BF from the index were significantly correlated with those from densitometry (r ranging from +0.5 to +0.75). It is considered that any estimated index of body fat based on measures such as W and S is likely to be population-specific. Nevertheless, the proposed general model can be used to obtain the "best" estimates of the parameters C, m and k in different samples, particularly samples for which more direct fat-related variables are not available. This approach requires little training in measurement techniques, it is non-invasive and is applicable to the very obese. The population specificity of the parameters would make it difficult to compare populations using such an index, but the approach should be useful for inter-individual comparisons and analyses of serial data, if densitometric data are available for subsets of the samples studied. A simple quick method of assessing quantitatively the fat content of the body would be of great practical importance. The "direct" methods, including densitometry, gamma radiometry or dilution, can be used only in laboratories and can be applied to few subjects. Several studies suggest there are significant relationships between accepted methods of determining percent body fat (%BF) and simple techniques that could be widely applied (Katch and McArdle, 1973). These simple techniques include body surface area estimated from weight and stature (Burmeister, 1980), skinfold thicknesses (Florey, 1970; Parizkova, 1976) and body circumferences (Katch and McArdle, 1973; Singh, 1978). Also a combination xAssiut College of Medicine, Department of Anatomy, Assiut City, Egypt. 2Department of Experimental Statistics, LOREX Pharmaceuticals, 5200 Old Orchard Road, Skokie, Illinois 60077. 3Division of Human Biology, Department of Pediatrics, Wright State University School of Medicine, 1005 Xenia Avenue, Yellow Springs, Ohio 45387. Human Biology, September 1985, Vol. 57, No. 3, pp. 415-430. © Wayne State University Press, 1985 This content downloaded from 40.77.167.32 on Sun, 31 Jul 2016 04:24:54 UTC All use subject to http://about.jstor.org/terms 416 A.K. Abdel-Malek, D. Mukherjee and A. F. Roche of weight and stature was used by Mellits and Cheek (1970) to estimate total body water. Functions of weight (W) and stature (S) are used commonly as indices of obesity, but there is a lack of agreement as to which index is to be preferred. Some consider a single index, such as W/S2, should be applied to all groups while others claim a population-specific index should be used with a power of S that maximizes the correlation with W, but minimizes the correlation with S (Benn, 1971; Billewicz et al. 1962). The W-S indices reported can be classified into two common forms according to the power parameters applied to W and S. One of these has the form Wm/S (e.g., Ponderal Index where m = 0.33); the other has the form W/Sk (e.g., Quetelet's Index, or Body Mass Index, where k = 2). Both these examples are special cases of the proposed index (Wm/Sk). A new general family of W-S indices is proposed with the form: