Measuring Body Fat: Calibrating the Rulers. Intermethod Comparisons in 389 Normal Caucasian Subjects R. N. PIERSON,JR,J. WANG,S. B. HEYMSFIELD,ET AL American Journal of Physiology 261 (Endocrinology and Metabolism 24):E103-E108,1991

Abstract

In this study from Columbia University, 389 (156 men and 233 women) healthy white subjects ranging from 19 to 94 years, with body mass index ranging from 15 to 35 kg/m2, had their body fat measured using eight different methods. These methods included anthropometric assessments as defined by Durnin and Womersley¹ and Steinkamp et al,² underwater weighing, total body potassium, dual-photon absorptiometry, total body water, total body electrical conductivity, and bioelectric impedance analysis. All subjects were studied after an overnight fast. Statistical analysis of the data utilized repeated-measures analysis of variance to test the hypothesis that the eight methods yielded the same average percent of body fat. Pearson correlation coefficients were calculated for each pair of body fat measurements. Linear regression equations were calculated for each pair of methods and for each method using age as an independent variable. The reliability of the methods was calculated using the intraclass correlation coefficient; the coefficient of variation of each method was also calculated.

Table I summarizes the coefficients of variation and reliability intraclass correlations for the eight methods tested. It is important to note the low coefficients of variation in dual-photon absorptiometry and underwater weighing and the high coefficient of variation using total body potassium. Table II summarizes the average percent body fat measurements for men and women using the various methods. Of note is that the use of bioelectric impedance analysis and total body electrical conductivity consistently resulted in the lowest values for percent body fat, whereas the use of total body potassium and anthropometric measurements, as defined by Steinkamp et al,² consistently resulted in the highest values for percent body fat. Additional data presented by the authors demonstrate correlations between percent body fat, measured by the different methods, and the linear regression equations of percent fat with age. The investigators conclude their data presentation with a full-page table of intermethod linear regression equations for men and women. The purpose of this table is to serve as a type of Rosetta stone. It provides the slopes and intercepts (and standard estimate of the error and r ² values) to be used to translate results obtained using one method to those that would be expected using any of the other methods according to the general formula y = mx + b.