Predicting body composition by densitometry from simple anthropometric measurements

Abstract

New equations have been developed to predict body fat (percent BF) calculated from body density measured by underwater weighing from simple anthropometric measurements, using stepwise-multiple-regression analysis in 63 men and 84 women. Log10 sum of four skinfold thicknesses explained 80.1% (SE = 3.8) of variance of percent BF in men and 76.4% (SE = 4.6) in women. Alternative equations using limb lengths instead of height may be valuable for epidemiologic and clinical work, with particular advantages for the chair- or bed bound, for whom no previous predictive equations existed. Five equations combining triceps-skinfold thicknesses with other anthropometric measurements explained > 80% (men) and 77% (women) of variance. The most powerful prediction was from waist circumference and triceps-skinfold thickness, which explained 86.6% (SE = 3.2) of variance of percent BF in men and 79.0% (SE = 4.0) in women. Percent BF for men = 0.353 waist (cm) + 0.756 triceps (mm) + 0.235, age (y) = 26.4, for women = 0.232 waist (cm) + 0.657 triceps (mm) + 0.215 age (y) = 5.5. The equations were tested in a separately studied validation sample of 146 men and 238 women aged 18-83 y. Skinfold- thickness measurements continued to give good predictions of mean body density, but with significant bias at extremes of body fat and age. The most robust prediction with the least bias was from waist circumference adjusted for age. Percent BF for men = 0.567 waist (cm) + 0.101 age (y) = 31.8; and four women = 0.439 waist (cm) + 0.221 age (y) = 9.4.