Objective

To develop and validate predictive equations to estimate the body composition of women with grade III obesity, using the body mass index (BMI) as a predictive variable.

Methods

This cross-sectional study involved 104 patients treated at the hospital of the Universidade Federal do Rio de Janeiro randomly divided into two groups, the Equation Group, used to generate regression equations, and the Validation Group, used to validate the equations. Body fat mass (BFM), body fat percentage (BFP), skeletal muscle mass (SMM), fat-free mass (FFM) and total body water content (TBW) were valuated employing the bioimpedance method (InBody® 230).

Results

Polynomial equations exhibited the best fit and a general trend of results normalized by height squared presenting higher coefficients of determination (r2) was noted, positively affecting equation validations. Only one exception was observed, since the body fat percentage index (BFPI) resulted in an even lower correlation with BMI. Only these variables exhibited low r2 (0.11 to 0.29), while r2 values ranged from 0.51 to 0.94 for the other results.

Conclusion

Except for the BFP and BFPI, body composition can be estimated by the application of predictive BMI-based models. The equations employed for the indices normalized by the square of height were better predictors, while the use of equations that do not employ this normalization should consider the caveat that individuals with extreme BMI values (40 to 76 kg/m2) present greater estimate deviations in relation to the measured values.