Background and objective: Although Asian Indians have been shown to have increased body fat compared to Europeans, there have been very few studies in Asian Indians validating the various methods available for body fat measurement. The aim of this study was to test the validity of body fat measured by two commercial impedance analyzers (leg-to-leg and hand-held) as well as by skinfolds with Dual Energy X-ray Absorptiometry (DEXA) as the reference method in a population based study in southern India.
Methods: Body fat percentage (BF%) was measured in 162 South Indian urban men (n=76) and women (n=86) randomly selected from the "Chennai Urban Rural Epidemiology Study" (CURES), an ongoing population based study of a representative population of Chennai. The mean age of the subjects was 45.1 +/- 9.0 years and the body mass index ranged from 16.4 - 34.4 kg/m2. Percentage body fat was measured using DEXA, segmental impedance (leg-to-leg: BF%IMP-LEG; and hand-held BF%IMP-HAND) using the manufacturer's software and skinfolds using the prediction equation from the literature (BF%SKFD).
Results: Body fat (%) determined by the leg-to-leg method (BF%IMP-LEG 35.10 +/- 7.26) and the skinfolds (BF%SKFD 35.77 +/- 6.06) did not differ significantly from the reference method DEXA (BF%DEXA 35.82 +/- 8.33), but the hand-held impedance method (BF%IMP-HAND 31.38 +/- 6.24) showed significant difference (p < 0.001). The bias for estimation of body fat (%) for the bioimpedance leg-to-leg, hand-held and skinfolds were 0.73 +/- 5.70, 4.45 +/- 4.83 and 0.06 +/- 5.86 respectively. All the three methods showed a fairly good correlation with DEXA (BF%IMP-LEG: r = 0.741, p<0.001; BF%IMP-HAND: r = 0.817, p< 0.001; BF%SKFD: r 0.710, p< 0.001).
Conclusion: The study shows that in urban south Indians, measurement of body fat by the leg-to-leg impedance and the skinfold method have better agreement (lower bias) with DEXA than the hand-held impedance. However, all three methods (skinfolds, the leg-to-leg bioelectric impedance and hand-held impedance) show a fairly good correlation with DEXA.