Abstract
For developing race-specific anthropometry-based total body water (TBW) equations, we measured TBW using bioelectrical impedance analysis (TBW(BIA)) in 2,943 healthy Korean adults. Among them, 2,223 were used as a reference group. Two equations (TBW(K1) and TBW(K2)) were developed based on age, sex, height, and body weight. The adjusted R2 was 0.908 for TBW(K1) and 0.910 for TBW(K2). The remaining 720 subjects were used for the validation of our results. Watson (TBW(W)) and Hume-Weyers (TBW(H)) formulas were also used. In men, TBW(BIA) showed the highest correlation with TBW(H), followed by TBW(K1), TBW(K2) and TBW(W). TBW(K1) and TBW(K2) showed the lower root mean square errors (RMSE) and mean prediction errors (ME) than TBW(W) and TBW(H). On the Bland-Altman plot, the correlations between the differences and means were smaller for TBW(K2) than for TBW(K1). On the contrary, TBW(BIA) showed the highest correlation with TBW(W), followed by TBW(K2), TBW(K1), and TBW(H) in females. RMSE was smallest in TBW(W), followed by TBW(K2), TBW(K1) and TBW(H). ME was closest to zero for TBW(K2), followed by TBW(K1), TBW(W) and TBW(H). The correlation coefficients between the means and differences were highest in TBW(W), and lowest in TBW(K2). In conclusion, TBW(K2) provides better accuracy with a smaller bias than the TBW(W) or TBW(H) in males. TBW(K2) shows a similar accuracy, but with a smaller bias than TBW(W) in females.