Lack of correspondence among measures identifying the obese

Correspondence among eight measures commonly used to identify the obese was investigated in 225 men and 212 women 18-59 years of age. Measures included weight, body mass index (BMI), relative weight, skinfold thicknesses at triceps and subscapular sites, the sum of four skinfolds, percentage body fat (% BF) determined hydrostatically, and total weight of fat. Intercorrelations within various measures of body mass and total fat are high (0.86-0.99), while correlations between the two skinfolds, and among measures of body mass, subcutaneous fatness, and total body fat are much lower (0.57-0.70). The amount of variation in %BF accounted for by correlations with the other individual measures of obesity (other than fat weight) ranges from 32% to 48%. When analyzed categorically, the upper quintiles of the various measures poorly identify the same individuals as the fattest. The median chance-adjusted correspondences were 0.57 and 0.64 for men and women, respectively, with single values as low as 0.40. Analyses of variance (ANOVA) indicate the misclassification occurring among upper quintiles is nonrandom with respect to mean levels of these obesity measures. We find a considerable lack of correspondence among measures commonly used to identify the obese. This lack of correspondence may lead to unanticipated dissimilarity among people variously classified as obese and to bias regarding risks of associated disease or mortality.