Document Type



Centralized obesity has been associated with increased risk of non-insulin dependent diabetes and cardiovascular disease. Paramount to a sensitive index of body fat distribution is that it contain a measure of lower limb fat (Ashwell et al. 1978; 1982; Mueller and Stallones 1981). However, many epidemiological studies of body fat distribution, which have used skinfold measurements, have been limited to estimating centralized obesity from the triceps and subscapular or other conventional upper body sites. The purpose of this study was to evaluate the sensitivities, specificities, and positive predictive values of skinfold indices of body fat distribution when only sites on the upper body are available. We were able to do this in a large population-based data set, the Canadian YMCA-LIFE study, which in­cluded adults 25 to 64 years of age and skinfold sites from upper and lower anatomical regions of the body.Sensitivities, specificities, and positive predictive values did not vary systematically with age group, sex or obesity level. Sensitivities (mean = 70%) and positive predictive rates (mean = 65%) were moderate for the most common two site index (triceps/triceps + subscapular) and were not notably improved with the addition of the suprailiac site. Simple percent extremity fat indices (e.g. triceps/(triceps + subscapular) X 100) were as effective in discriminating body fat distribution groups as an index involving the same variables in the form of a vector of log transformed measurements. Substituting lower limb fat (medial calf) for arm fat (triceps) in simple percent indices, provided important additional information (mean sen­sitivity = 77%, mean positive predictive rate = 70%).