Which measure of body fat distribution is best for epidemiologic research?

Multivariate associations were sought between risk factor levels (total cholesterol, high density lipoprotein (HDL) cholesterol, triglycerides, glucose, and systolic and diastolic blood pressures) and two sets of anthropometric variables (four circumferences and six skinfolds) to select a set of anthropometric indicators of body fat distribution that correlate most highly with risk of disease. Subjects were men (n = 285) and women (n = 672) from a study of gallbladder disease in a Mexican American population in Starr County, Texas, 1985-1986. The canonical correlations showed that circumferences (0.49-0.61) and skinfolds (0.42-0.60) were equally well correlated to risk factor levels independently of sex and age. Weights from the canonical analyses suggest that measurements at or above the waist and on the lower limb (thigh) are most heavily loaded toward risk (waist = highest risk; thigh = lowest risk). The simplest and most reliable index of body fat distribution for both sexes is the ratio of waist to thigh circumferences. The more commonly used waist/hip ratio proved more valid in women, but not in men. Simple skinfold indices of body fat distribution were more poorly correlated to risk factor levels than the corresponding circumference ratios. In women, body mass index and waist circumference by themselves did as well as body fat distribution indices in explaining variation in risk factors, suggesting the involvement of visceral fat in the body fat/body fat distribution disease relation.