Comparison of body composition measures assessed by bioelectrical impedance analysis versus dual-energy X-ray absorptiometry in UK Biobank

Background Bioelectrical impedance analysis (BIA) and dual-energy X-ray absorptiometry (DXA) are used commonly for assessing body composition. This study aimed to evaluate the agreement between BIA and DXA measures to assess the validity of BIA measures in UK Biobank. Methods UK Biobank participants with body fat mass (FM) and fat free mass (FFM) estimates derived from BIA and DXA measures performed on the same date were included. BIA and DXA-derived estimates were compared using Pearson correlation coefficients. Bland-Altman analysis was performed to quantify the differences and its distribution. Multivariable linear regression was used to identify predictors for the difference between the two measures. Finally, prediction models were developed to calibrate BIA measures against DXA measures. Results 34437 participants (female 51.4%, mean age 55.2 years) were included. BIA and DXA measurements were highly correlated (correlation coefficient 0.96 for FM and 0.97 for FFM), with similar values in males and females and across body mass index subgroups. BIA underestimates FM by 1.84 kg (23.77 vs. 25.61), but overestimated FFM by 2.56 kg (52.49 vs. 49.93). The BIA-DXA difference was associated with individual FM, FFM, BMI and waist circumference. Prediction models showed overall good model performance in the training and testing data. Conclusion We found good agreement between BIA-and DXA-derived body composition measures at a population level in UK Biobank. However, BIA-DXA difference existed at individual level, and was associated with anthropometric measures. Future studies may consider using prediction models to calibrate BIA measures.