### INTRODUCTION

_{BIA}) in a large study population to develop an anthropometry-based TBW equation. Then to validate this equation, we analyzed the agreement between the TBW

_{BIA}and the TBW derived from anthropometry-based equations in another control group.

### MATERIALS AND METHODS

_{BIA}was validated on healthy subjects (2). The procedure was performed in 3 min or less and the TBW

_{BIA}was automatically calculated from the BIA with equations installed in the instrument's program.

_{W}=2.447-(0.09156×age)+(0.1074×Ht)+(0.3362×BW)

_{W}=-2.097+(0.1069×Ht)+(0.2466×BW)

_{H}=(0.194786×Ht)+(0.296785×BW)-14.012934

_{H}=(0.34454×Ht)+(0.183809×BW)-35.270121

### Statistical analysis

*p*<0.01. TBW

_{BIA}was used as a dependent variable. Sex, age, Ht and BW were used as independent variables. Polynomial terms for continuous variables and multiplicative interaction terms were considered in the model building process. Pearson's correlation coefficient (r) was used to find the relationship between two variables. To analyze the differences in TBW

_{BIA}and TBWs derived from anthropometry-based equations, one-way analysis of variance (ANOVA) was performed with using the Bonferroni method for the post-hoc test. To assess the agreement, Bland-Altman plots using the means and differences between TBW

_{BIA}and calculated TBW were used (5). To quantitate the degrees of bias, we compared the correlation coefficients of the respective differences and means. The closer the correlation coefficient of Bland-Altman plot was to zero, the less the bias. Root mean square error (RMSE) and mean prediction error (ME) were also used. ME was also an indication of bias, but not of accuracy. The RMSE value was used as a measure of the goodness-of-fit of an equation. If there were more than one equation to fit the data, the one with the smallest RMSE value had the highest precision. The equations used for ME and RMSE are as follows:

*p*value less than 0.05 was considered as statistically significant.

### RESULTS

### Development of anthropometry-based TBW equations

_{BIA}was 34.9±6.6 L. The simple (TBW

_{K1}) and complicated (TBW

_{K2}) TBW equations based on the anthropometric variables were developed by linear regression analysis (Table 1). The adjusted R

^{2}was 0.908 for TBW

_{K1}and 0.910 for TBW

_{K2}.

### Validation of newly developed TBW equations

_{BIA}was 33.6±6.2 L. In males, TBW

_{BIA}showed the highest correlation with TBW

_{H}(r=0.951), followed by TBW

_{K1}(r=0.945), TBW

_{K2}(r=0.945) and TBW

_{W}(r=0.937) (Table 2). There were no differences between the TBW

_{BIA}and TBW

_{K1}or TBW

_{K2}. However, TBW

_{W}and TBW

_{H}were significantly larger than the TBW

_{BIA}. There were significant differences between TBW

_{W}and TBW

_{K1}or TBW

_{K2}and between the TBW

_{H}and TBW

_{K1}. In females, TBW

_{BIA}showed the highest correlation with TBW

_{W}(r=0.902), followed by TBW

_{K2}(r=0.895), TBW

_{K1}(r=0.890), and TBW

_{H}(r=0.887). There were no differences between TBW

_{BIA}and TBW

_{W}, TBW

_{K1}or TBW

_{K2}. The TBW

_{H}was significantly larger than the others.

_{K1}and TBW

_{K2}showed the lower RMSE (1.58, 1.58, 2.14, and 2.08 for TBW

_{K1}TBW

_{K2}and TBW

_{W}TBW

_{H}, respectively) and ME (0.526, 0.547, 1.426, and 1.362 for TBW

_{K1}TBW

_{K2}TBW

_{W}and TBW

_{H}, respectively) than the TBW

_{W}and TBW

_{H}(Table 3). On the Bland-Altman plot, the correlations between the difference and means were smallest for the TBW

_{K2}(r= -0.192), followed by the TBW

_{K1}, TBW

_{W}, and TBW

_{H}(Fig. 1A, C, E, G). In females, the RMSEs were smallest for the TBW

_{W}, followed by the TBW

_{K2}, TBW

_{K1}, and TBW

_{H}(1.49, 1.50, 1.62, and 1.70 for the TBW

_{W}, TBW

_{K2}, TBW

_{K1}and TBW

_{H}, respectively). The ME was closest to zero for the TBW

_{K2}, followed by the TBW

_{K1}, TBW

_{W}and TBW

_{H}(0.554, 0.556, 0.593, and 0.988 for the TBW

_{K2}, TBW

_{K1}, TBW

_{W}and TBW

_{H}, respectively). The correlation coefficients between the means and differences were highest for the TBW

_{W}(r=-0.553), and lowest for the TBW

_{K2}(r=0.057) (Fig. 1B, D, F, and H).

### DISCUSSION

_{K1}and TBW

_{K2}) for Koreans using TBW

_{BIA}as a reference. Among them, TBW

_{K2}showed the highest precision and the smallest bias for males and a similar precision and the smallest bias for females compared to the TBWs derived from Watson or Hume-Weyers formulas.

_{K2}may be helpful for assessing the nutritional status and dialysis adequacy more exactly for the Korean healthy control population and the Korean patients with end-stage renal disease.

_{W}showed a lower RMSE value than the TBW

_{K2}in females. Therefore, TBW

_{W}might have a better accuracy than TBW

_{K2}, at least in females. However, TBW

_{W}showed a greater bias than TBW

_{K2}, as shown in Fig. 1B, H. TBW

_{K2}had a similar RMSE value and its ME was closer to zero than TBW

_{W}. Furthermore, it had the least bias in females. Therefore, TBW

_{K2}seemed to be more suitable for the estimation of the TBW in Korean females.

_{K2}) still showed weak correlation between the means and differences in the Bland-Altman plot. Thus, the TBW derived from TBW

_{K2}might underestimate the real TBW in men with large BW. Fourth, this study was limited to the healthy subjects. Therefore, it should be validated for patients with the volume disorders such as acute renal failure, liver cirrhosis with ascites, ESRD, congestive heart failure, and nephrotic syndrome.