Anthropometric indices of obesity: validity in stunted populations

Abstract

Ratio indices of body mass distribution have been used to estimate obesity while controlling for allometric scaling (JCK Wells & CG Victora. Int J Obes 2005;29:–9). A good index should be uncorrelated with its denominator; this often requires exponentiation of the denominator (e.g., weight/height2). The stability of the derived exponents across populations is not known. We obtained subscapular (SUBS) and triceps (TRI) skinfolds, height (HT) weight (WT), fat mass (FM) and fat free mass (FFM) in a population of Guatemalan adults (height 1.63 ± 0.06 m (868 males); 1.51 ± 0.06 m (1047 females)). We derived sex-specific exponents for the indices SUBS/TRIP, TRI/SUBSP, FM/WTP, FM/FFMP and WT/HTP, through log-log regression (table). For all indices the exponents differed from 1, were uncorrelated with the denominator, were different for men and for women, and different between males in our sample and a sample of Brazilian men. The derived indices were not more strongly correlated with adiposity than were first-order ratio measures. Complex indices may lack generalizability, especially across populations of differing body proportions, and hence need to be compared cautiously. Table 1. Log-log regression coefficients in three populations