Estimations of body fat by anthropometry or bioelectrical impedance differ from those by dual-energy X-ray absorptiometry in prefrail community-dwelling older women

This study aimed to compare different methods to assess body fat (BF). We hypothesized that bioelectrical impedance analysis (BIA) or anthropometry may be used to estimate BF in prefrail older women, equivalently to dual-energy X-ray absorptiometry (DXA). The cross-sectional study included 72 prefrail community-dwelling older women (71.13 ± 4.65 years old; body mass index [BMI] 28.89 ± 4.23 kg/m2). The BF percentage (%BF) was estimated using anthropometry with the Durnin & Womersley (D&W) and Petroski's predictive equations, BIA with 2 Baumgartner predictive equations (BIA 1 and BIA 2), and DXA. All methods differed significantly from DXA according to assessments using repeated measures ANOVA and pairwise comparisons. The mean %BF varied between 39.99 ± 3.42% (D&W) and 43.93 ± 5.06% (DXA). Multiple regression analysis with age and BMI as covariates showed positive correlations (R2 = 0.91) in models with D&W equation and BMI, and with BIA 2 and BMI; however, BMI explained more of the model (71%) than the equations. Furthermore, Bland-Altman test revealed a proportional bias for D&W and for BIA 2, with underestimation of BF varying across different %BF values. Petroski's skinfold equation showed a positive correlation on linear regression (R2 = 0.74) and no proportional bias; however, Bland-Altman analysis revealed high limits of agreement (-13.6 to -0.05), thus compromising clinical application. To conclude, compared with DXA, all the equations tested showed a high disagreement and wide limits of agreement, restricting their use in clinical practice to estimate the BF in prefrail older women.