Validity of Bioelectrical Impedance Analysis in the Estimation of Body Composition Among Lebanese Adolescents

Abstract

Introduction: With the increasing prevalence of adolescent obesity globally, and the established link between excess body fat and cardiometabolic risk, accurate assessment of body composition is crucial. The common, easy, rapid, non-invasive and cost effective tool used to assess body composition is bioelectrical impedance analysis (BIA). However, its validity and accuracy for diverse populations and ethnicities requires further investigations. Given the lack of studies on the validation of BIA amongst adolescents in the MENA region, the objectives of the present study are to 1) assess the validity of BIA for estimating body composition in comparison with the reference method of deuterium dilution (DDL) in Lebanese adolescents, and 2) develop population-specific BIA prediction equations for improved accuracy in estimating body fat in this population group.

Methods: This study included 135 participants, 66 males (mean age16.5 ± 0.96 years) and 69 females (mean age16.5 ± 1.02 years). TBW, FFM and BF% estimated from BIA device (BIA 101, Akern, Via Lisbona, Italy) were compared to TBW, FFM and BF% measured by DDL. The latter involved collecting saliva samples using cotton wool balls, with the first sample obtained before and the second three hours after the ingestion of 60 mL of deuterated water. Both samples were analyzed using Fourier transform infrared (FTIR) spectroscopy at the American University of Beirut. Paired t tests were performed to assess the differences in TBW, FFM and BF% estimates between the two methods, and the correlation between BIA and DDL estimates was examined using Pearson correlation coefficients. The Bland-Altman plot was generated to assess the agreement between the two methods. For model development, five-fold cross-validation was performed within the study dataset to evaluate model stability and predictive accuracy. The final models were derived from the full dataset using predictors consistently selected across folds.

Results: Differences between BIA and DDL for FFM, TBW, and %BF were not statistically significant in the total sample (mean differences = 0.358 kg, 0.301 kg, and −0.69%, respectively), and the same was observed by gender. However, Bland–Altman analysis revealed relatively wide limits of agreement, indicating considerable individual variability between methods. This was more evident at higher adiposity levels and among girls. New cross-validated prediction equations for TBW and FFM were developed, yielding an R² of 0.93. The RMSE was 2.16 kg for TBW and 2.93 kg for FFM. The final equations were expressed as follows: TBW (kg) = 7.876 + 0.167 × weight (kg) + 0.412 × RI − 2.375 × gender (male = 1, female = 2); FFM (kg) = 10.653 + 0.224 × weight (kg) + 0.559 × RI − 3.139 × gender (male = 1, female = 2). Bland Altman plots indicated good agreement between predicted and reference values for TBW and FFM derived from DDL.

Conclusion: Inbuilt BIA equations showed acceptable group-level agreement with the reference method but relatively wide limits of agreement, indicating reduced individual accuracy. The newly developed, cross-validated prediction equations demonstrated improved performance with lower error and better agreement. These findings support the use of population-specific BIA equations for more accurate body composition assessment in Lebanese adolescents. Future studies should externally validate these equations in independent samples from Lebanon and the Middle East, particularly in children and adolescents.