## Abstract

Dual-energy X-ray absorptiometry (DEXA) provides a measure of lean soft tissue (LST). LST hydration, often assumed to be constant, is relevant to several aspects of DEXA body composition estimates. The aims of this study were to develop a theoretical model of LST total body water (TBW) content and to examine hydration effects with empirically derived model coefficients and then to experimentally test the model's prediction that, in healthy adults, LST hydration is not constant but varies as a function of extra- and intracellular water distribution (E/I). The initial phase involved TBW/LST model development and application with empirically derived model coefficients. Model predictions were then tested in a cross-sectional study of 215 healthy adults. LST was measured by DEXA, extracellular water (ECW) by NaBr dilution, intracellular water (ICW) by whole body ^{40}K counting, and TBW by ^{2}H_{2}O dilution. TBW estimates, calculated as ECW + ICW, were highly correlated with (*r* = 0.97, SEE = 2.1 kg, *P* < 0.001) and showed no significant bias compared with TBW measured by ^{2}H_{2}O. Model-predicted TBW/LST was almost identical to experimentally derived values (means ± SD) in the total group (0.767 vs. 0.764 ± 0.028). LST hydration was significantly correlated with E/I (total group, *r* = 0.30, SEE = 0.027, *P* < 0.001). Although E/I increased with age (men, *r* = 0.48; women, *r* = 0.37; both *P* < 0.001), the association between TBW/LST and age was nonsignificant. Hydration of the DEXA-derived LST compartment is thus not constant but varies predictably with ECW and ICW distribution. This observation has implications for the accuracy of body fat measurements by DEXA and the use of TBW as a means of checking DEXA system calibration.

- whole body counting
- body composition
- water distribution

an important advance in body composition research is the availability of dual-energy X-ray absorptiometry (DEXA) for partitioning body mass into three components: fat, lean soft tissue (LST), and bone mineral (Mo) (20). Total body water (TBW) is present entirely within the LST compartment, and one assumption of DEXA measurements is that the water content of LST or the similar, larger fat-free mass (FFM) compartment (i.e., FFM = LST + Mo) is relatively stable [e.g., TBW/FFM (means ± SD) = 0.730 ± 0.029, coefficient of variation = 4.0%] across subjects (25, 26). This concept is also embodied in the classical approach for estimating FFM as TBW divided by the hydration factor 0.73 (22).

The assumed stability, or “constancy,” of FFM and LST hydration is important when body composition models are being developed (7) for confirming the calibration of DEXA scanners (20) and when hydration-induced errors in DEXA estimates of fat mass are considered (25).

Moore et al. (16) first highlighted the existence of two main components at the cellular level of body composition that are included in FFM: body cell mass (BCM) and extracellular fluid (ECF). In a follow-up study, Moore and Boyden (15) hypothesized that the hydration of FFM would vary systematically as a function of changes in the proportions of body mass as ECF and BCM. The ECF compartment has a high extracellular water (ECW) content (i.e., ECW/ECF = ∼0.98; 2% nonaqueous solids as minerals, electrolytes, and protein), whereas the intracellular water (ICW) content of BCM (i.e., ICW/BCM) is lower, ∼0.70 (1, 23, 26). Moore and Boyden's classic model (15) predicts greater lean tissue hydration with increasing ECF/BCM or ECW/ICW (E/I). Although widely cited, a critical test of Moore and Boyden's model predictions is lacking. Importantly, their model predicts that lean tissue hydration is not “constant” but varies systematically with fluid distribution (e.g., mean TBW/FFM = 0.73 with range from 0.69 to 0.74), even in healthy adults.

In a recent study, Wang et al. (26) also reported a theoretical model based on empirical coefficients from Reference Man that predicts FFM hydration varies as a function of E/I This equation revealed that relative changes in water distribution have a small effect on TBW/FFM. The mean magnitude of TBW/FFM ranges from a low value of 0.72 at E/I = 0.8 to a high value of 0.74 at E/I = 1.2 in healthy adults. These observations highlight the importance of considering FFM hydration in body composition methodology aimed at estimating body fat content.

The aim of the present study was twofold: to extend Wang's FFM hydration model to the DEXA-relevant LST compartment so that predictions can be made of E/I effects on TBW/LST, and to then experimentally test the hypothesis in healthy adults that LST hydration varies as a function of water distribution.

## EXPERIMENTAL PROCEDURES

### Subjects

Subjects were recruited via newspaper advertisements and from the local community. Healthy adults >18 yr of age with body mass index (BMI) <35 kg/m^{2} were included in this study. Subjects with BMI >35 kg/m^{2} were excluded due to DEXA technical measurement concerns in very obese subjects. Health in the evaluated subjects was established by physical examination and screening blood studies. These subjects were part of an earlier study aimed at developing skeletal muscle mass measurement methods (29). A total of 218 subjects were available for analysis. The study was approved by the Institutional Review Board of St. Luke's-Roosevelt Hospital, and all subjects signed an informed consent before participation.

### Protocol

A TBW/LST theoretical model fitted with empirical coefficients was first developed as a means of evaluating the factors that potentially influence DEXA LST hydration. The model's predictions were then evaluated using experimental data collected earlier in a large cohort of healthy adults.

Each subject had LST, bromide and deuterium dilution space, and total body potassium (TBK, mmol) measurement on the same day. The bromide dilution space and TBK were used to estimate ECW (i.e., ECW by Br^{−} dilution, in kg), ICW [i.e., ICW = TBK/152, in kg (4, 13, 22, 23)], E/I (i.e., ECW/ICW), and TBW (i.e., TBW = ECW by Br^{−} + ICW by TBK, in kg). The deuterium dilution space was also used to independently estimate TBW and to derive hydration as TBW/LST. The water distribution estimates (i.e., E/I) are thus independent of the DEXA LST hydration estimates.

### Model Development

On the cellular body composition level (Fig. 1, *right*), FFM can be divided into three components: BCM, ECF, and extracellular solids (ECS) (27). The ECS compartment consists of organic ECS (i.e., ECS protein) with three types of protein and inorganic ECS or Mo, which includes primarily calcium hydroxyapatite, and therefore consists of Mo and some LST components in the DEXA model. LST measured by DEXA can thus be expressed as the sum of three cellular level compartments: (1) Similarly, TBW can be expressed as the sum of ICW and ECW (2) On the basis of *Eqs. 1* and *2*, LST hydration can be expressed as (3) This formula provides an expression of LST hydration on the cellular body composition level. Body cell mass and ECS can, on the basis of previous models, be expressed as BCM = ICW/a and ECF = ECW/b, where a and b are the fractions of BCM and ECF as water, respectively (Fig. 1, *right*). These hydration ratios are the basis of Moore and Boyden's model (15) linking FFM hydration to fluid distribution as ECF and BCM. Moore and Boyden assumed that b > a and that, accordingly, TBW/FFM value would increase as a function of b/a (15).

In addition, the ratio of ECS protein to Mo is assumed relatively constant at 0.732 (28). Extracellular solids protein can be expressed as a function of TBW: ECS protein = 0.732 × Mo = 0.732/1.732 × ECS = 0.423 × c × TBW = 0.423 × c × (ICW + ECW), where c is the ratio of ECS to TBW. *Equation 3* can thus be converted into (4) In *Eq. 4*, ICW and ECW are interrelated compartments of body water, and ECW can be expressed as a function of ICW, ECW = (E/I) × ICW, where E/I is the ratio of ECW to ICW. The equation above can be converted and simplified to (5) The proposed TBW/LST model indicates that LST hydration is a function of four determinants [i.e., TBW/LST = *f*(a, b, c, E/I)], and each of the determinants may vary within an assumed range for young adults: a, from 0.69 to 0.71; b, from 0.97 to 0.99; c, from 0.12 to 0.16; and E/I from 0.58 to 1.36 (26). Factors a, b, and E/I are in direct proportion, and c is in inverse proportion, to TBW/LST magnitude. One can thus estimate the biological range of LST hydration if the four determinants take their extreme values. When a is 0.69, b is 0.97, c is 0.16, and E/I is 0.58, TBW/LST may reach its low value according to *Eq. 5* When a is 0.71, b is 0.99, c is 0.12, and E/I is 1.36, TBW/LST may reach its high value The predicted variation range of LST hydration for healthy young adults is thus approximately from 0.73 to 0.82.

Our previously published study (26) reported three factors that are assumed constant for model development purposes in healthy adults: a = 0.70, b = 0.98, and c = 0.14. As E/I is the only factor that varies widely within subjects over time and between subjects, the cellular level TBW/LST model (*Eq. 5*) can be simplified as (6) *Equation 6* indicates that changes in water distribution have an effect on the TBW/LST ratio. Factor E/I is in direct proportion to TBW/LST. For example, when E/I increases from 0.80 to 1.20, a range commonly observed in healthy adults, TBW/LST increases from 0.765 to 0.790. The curvilinear TBW/LST function generated by *Eq. 6* is shown in Fig. 2. The simplified model and Moore and Boyden's earlier model (15) thus predicts that TBW/LST increases as a function of E/I.

### Body Composition

Measurements were made with all subjects clothed in a hospital gown. The subject's body weight was then measured in kilograms to the nearest 10 g with a Weight-Tronix electronic scale (Scale Electronics Development, New York, NY). Standing height was measured in millimeters without shoes to the nearest 10 mm with a wall-mounted Holtain stadiometer (Crosswell, Wales, UK).

### Whole Body ^{40}K Counting

A 4π whole body counter was used to detect the natural 1.46 MeV γ-ray of ^{40}K as previously reported (18). The γ-ray of ^{40}K counts accumulated over 9 min was adjusted for body size on the basis of a ^{42}K calibration formula reported by Pierson et al. (19). TBK was calculated in millimoles as ^{40}K/0.000118 (3). The coefficient of variation (CV) for repeated counting of a ^{40}K phantom in our laboratory is 2.4% (8). ICW was calculated as ICW (kg) = TBK (mmol)/152 (22).

### DEXA

DEXA was used to measure the mass of fat, LST, and Mo in kilograms. The scan was completed with a whole body pencil beam DPX system (version 3.6 software; Lunar Radiation, Madison, WI). Detailed methods for image acquisition and analysis are described elsewhere (9).

### Bromide Dilution

Each subject was asked to ingest 5 g of a 4 mol/l solution of NaBr. The NaBr concentration in plasma was then measured by high-performance liquid chromatography before and 3 h after tracer administration. The corrected bromide space was measured with a CV of 1.4%. The dilution space was then converted into ECW (in kg) by correcting for the weight fraction of water in plasma (0.94), the Gibbs-Donnan effect (0.95), and the penetration of bromide into the intracellular space of erythrocytes (0.90) (22).

### Deuterium Dilution

Upon arrival at the testing laboratory, each subject was asked to void completely to provide a background urine. The subject then ingested 10 g of ^{2}H_{2}O (ICON, Summit, NJ), and the dose was rinsed three times with 10 ml of regular water. Three hours after the dose, 7 ml of blood were taken in a heparinized tube for later analysis by infrared spectrophotometry (Avatar 360; Nicolet, Madison, WI). The ^{2}H_{2}O dilution space, measured with a CV of 1.2%, was converted into TBW mass (in kg) assuming a water density of 0.99371 g/cm^{3} at 36°C (11, 22).

### Statistical Methods

Differences in baseline characteristics and body composition between the men and women were descriptively examined using *t*-tests, with *P* < 0.05 considered statistically significant.

The main study hypothesis was tested by examining the association between water distribution (E/I) and LST hydration (TBW/LST) by use of simple linear regression analysis, with *P* < 0.05 considered statistically significant.

The study included two independent estimates of TBW, one by deuterium dilution and the other by summing ECW by bromide dilution and ICW by TBK derived by whole body ^{40}K counting. We examined these two measures of TBW by first comparing their respective mean values by means of *t*-tests with *P* < 0.05 considered statistically significant. Potential bias between the two measures of TBW was then examined using a Bland-Altman plot (2).

Bland-Altman tests were done using the Microsoft Excel 2002 statistical analysis package, and multiple linear regressions were done using SAS 8.2 (SAS Institute, Cary, NC). Data are reported as means ± SD.

## RESULTS

### Subjects

Of the 218 subjects available for analysis, three were excluded in whom TBW calculated from TBK and Br^{−} measurements were >35–50% higher than those measured by ^{2}H_{2}O dilution. A total of 215 ethnically diverse subjects, 99 males and 116 females, were included in the analysis (Table 1). Age ranged from 19 to 82 yr for females and 18 to 82 yr for males.

### Comparison of TBW Measures

The body composition results are presented in Table 1. The relationship between TBW measured by deuterium dilution (38.9 ± 9.5 kg) and that calculated independently from the sum of ICW and ECW as measured by TBK and Br^{−} dilution (38.9 ± 9.6 kg) is presented in Fig. 3. The two TBW means did not differ significantly (*P* > 0.05) and were highly correlated (TBW by sum of ICW and ECW = 0.96 × TBW by ^{2}H_{2}O + 1.58; *r* = 0.97, SEE = 2.1 kg, *P* < 0.001). The slope and intercept of the relation did not differ significantly from 1 and 0, respectively. There was no constant or relative bias detected with a Bland-Altman analysis; the between-TBW method difference was not significantly correlated with the mean of the two methods (*y* = −0.0051*x* + 0.1283; *r* = 0.002, *P* = nonsignificant). These relationships were independent of sex.

### Hydration Effects

#### Age and sex.

The men had more ECW and TBK than did the women (Table 1, both *P* < 0.001) and lower E/I ratio (0.728 ± 0.091 vs. 0.894 ± 0.126, *P* < 0.001). The E/I increased with age in both men and women (*r* = 0.48 and 0.37, both *P* < 0.001).

Men had significantly more TBW, LST, and FFM than did women (all *P* < 0.001). The measured TBW/LST from deuterium dilution and DEXA methodologies was lower in men than in women (0.756 ± 0.025 vs. 0.771 ± 0.029, *P* < 0.001) by ∼2%. The measured TBW/FFM was also significantly lower, ∼1.5%, in men than women (0.718 ± 0.023 vs. 0.729 ± 0.026, *P* < 0.01). The correlations between age and LST or FFM hydration were not statistically significant (TBW/LST vs. age, *r* = 0.045 for the men and *r* = 0.067 for the women; TBW/FFM vs. age, *r* = 0.045 for the men and *r* = 0.080 for the women; all *P* values > 0.05).

#### Predicted vs. measured TBW/LST.

LST hydration predicted by the simplified model (i.e., *Eq. 6*) was 0.767 for men and women combined. The measured TBW/LST was 0.764 ± 0.028, similar to the model-predicted TBW/LST value. The model-predicted LST hydrations for healthy men (0.760) and women (0.772) were also similar to the measured values (0.756 ± 0.025 for men and 0.771 ± 0.029 for women).

#### TBW/LST vs. E/I.

There was a significant correlation between LST hydration and E/I (*r* = 0.30, *P* < 0.001; Fig. 4*A*). Fitting a curvilinear function to the data minimally improved the association between TBW/LST and E/I. The regression line developed for the measured data shown in Fig. 4*A* tracks closely to the model prediction line based on *Eq. 6* shown in Fig. 4*B*.

## DISCUSSION

In the present report, we extended earlier work by Moore and Boyden (15) and Wang et al. (26), providing quantitative predictions of water distribution effects on LST and related FFM hydration. Our model projection, confirmed by experimental observations, is that the DEXA-measured LST component has a fractional water content that varies in healthy adults as a function of water distribution. Greater TBW/LST is associated with a larger E/I. Our models also suggest that other factors contribute to LST hydration, but these determinants, as outlined in *Eq. 5* (e.g., the fraction of BCM as water, a = 0.70), are relatively stable and do not have a large influence on LST water content in healthy adults.

Although we did not specifically examine the basis of E/I variation in the current study, at least four moderating factors are recognized: variation in the proportion of total body as each organ and tissue [e.g., E/I for skeletal muscle = 0.54; kidney = 1.39; adipose tissue = 1.81 (23)], variation in the cellularity of each organ and tissue [e.g., with age (17)], physiological variation (e.g., menstrual cycle and fluid intake effects, etc.), and pathological variation (16).

Our findings also demonstrate close concordance between TBW measured by deuterium dilution and TBW derived from the sum of ECW estimated by bromide dilution and ICW estimated from TBK. Albeit not a direct proof, close agreement between the two TBW measures provides support for the measurement methods and assumptions used to quantify each of the components: ECW, ICW, and TBW. Moreover, this finding lends support to our approach for examining determinants of LST hydration (i.e., TBW/LST) using independently measured ECW and ICW (i.e., E/I). ECW estimates are typically method dependent (6), and our findings suggest that the combination of bromide dilution and TBK provides an excellent means of characterizing TBW (i.e., as ECW + ICW) by an alternative approach to isotopic water dilution.

### Hydration and Age

We observed a greater E/I in older subjects, a finding that confirms earlier studies that used different methodologies (6). The mechanism of this age-related effect is unknown, but increasing amounts of adipose tissue and LST atrophy, both known to cause a relative enlargement of the ECS (24), may contribute. However, the relatively larger ECW in older subjects was not accompanied by a corresponding significant difference from younger subjects in LST hydration. Insights into the lack of statistical significance for TBW/LST vs. age can be gained by applying our developed model with hypothetical subjects differing in age. For example, at age 20 yr, women in our study had an E/I of ∼0.8 compared with an E/I of ∼1.0 at age 80 yr. The corresponding TBW/LST predicted by *Eq. 6* is ∼0.765 at 20 yr and ∼0.779 at 80 yr, a remarkably small (1.8%) difference, perhaps explaining our failure to detect a significant increase in LST hydration with greater age. With such a small model-predicted age-related hydration effect, one could make the practical assumption that age, at least during adulthood, is an insignificant contributor to between-subject variation in TBW/LST. On the other hand, the E/I in newborns and children up to the age of 3 yr is high (∼1–1.5), as is hydration of FFM [∼0.80 (5a)]. Similar hydration effects during chemical maturation can be anticipated for DEXA-measured LST.

### Hydration Stability

A central concept in body composition research is the existence of stable, or relatively stable, component associations such as the hydration of FFM = 0.73 (22). Although our earlier models (26) suggest that FFM hydration also varies with E/I, as noted the effect is relatively small in healthy adults (e.g., TBW/FFM increases by 2.7% when E/I increases across the physiological range from 0.80 to 1.20).

Similarly, the association between TBW and LST is also relatively stable, the influence of E/I being small, and, as noted earlier, we were unable to detect a significant association between age and TBW/LST in adult subjects. Our model, as formulated in *Eq. 6*, predicted a TBW/LST of 0.767 for men and women combined, not different from the measured value of 0.766 ± 0.027. Thus, as expected, the hydration of LST (∼0.77) is greater than the larger FFM compartment (∼0.73) that includes Mo.

Conceptually, LST is a more appropriate denominator for TBW than FFM, as nearly all of body water is in the LST compartment with none in fat and very little in Mo as measured by DEXA. The water content of LST may thus be a sensitive measure of soft tissue hydration in health and disease, a topic worthy of future investigation.

DEXA systems produce variable body composition results across instruments and manufacturers (5), and a need exists for in vivo calibration methods. Although LST could be derived directly from measured TBW (e.g., LST = TBW/0.77), our finding of an E/I effect on LST hydration supports the use of a reference approach insensitive to water distribution, such as a four-compartment method (30). The traditional four-compartment model includes fat, TBW, minerals, and residual mass (i.e., protein and glycogen); LST could be calculated in this model as the difference between FFM and Mo. The four-compartment model would therefore be ideally suited to test DEXA system estimates of fat and LST and would not be limited by water distribution concerns. As LST was measured in the present study using a specific DEXA system, future studies using LST derived using a four-compartment method would provide potentially less system specific normative estimates of TBW/LST.

### Hydration and DEXA Accuracy

Earlier studies have examined the effects of hydration on DEXA system fat estimates (10, 25). DEXA fat estimation errors occur as a function of added fluid “R value” (ratio of soft tissue attenuation) and the fraction of added fluid (21). The R value depends on fluid elemental content and is similar, although not identical, for ECW (1.377 at 40 and 70 KeV) and ICW (1.382). If the accumulating fluid has the same R value as LST and there is no R value difference, no fat estimation errors occur no matter how much extra fluid is added. If the hydration fluid and lean R values do differ, then greater relative fluid accumulation is associated with larger fat estimation errors. The theoretical R value of LST is 1.345, slightly lower than for either ECW or ICW. Thus ECW, ICW, and LST have very similar, although not identical, R values.

Simulation experiments by Testolin et al. (25) and Pietrobelli et al. (21) reveal that, when ECW is added to LST at a level accounting for 20% of the combined total mass, DEXA provides an estimate two percentage points below that of “actual” percent fat. If we assume an initial LST hydration of 0.77, this addition of ECW would increase hydration to 0.81. This hydration level is at the upper physiological range of 0.82 as estimated from our model in this study. Thus, at the extreme range of normal, DEXA may, in theory, under- or overestimate fat by two percentage points.

Larger fat estimation errors may occur in the presence of pathological states, including local fluid accumulation as with ascites or pedal edema. On the other hand, some small between-subject variability in DEXA fat estimation accuracy can be anticipated, as hydration varies with fluid intake and other physiological processes throughout the day and with varying conditions such as prior exercise and season of the year.

### Study Limitations

There were several limitations of the present study. First, LST was measured using a specific DEXA system from one manufacturer, and the extent to which there is agreement across systems in the observed relations is unknown. Earlier reports from our Center (12) show small but significant between-system and -manufacturer measurement differences. A related methodological concern is a potential lack of independence between DEXA-measured LST and the evaluated fluid compartments. The minerals and electrolytes within ECW and ICW, and thus LST, are major determinants of relative attenuation of DEXA's two main photon peaks (20). Nevertheless, the close agreement between our developed theoretical models and experimental observations provides support for our major study conclusions.

The second concern is that the reference values used in developing our models are based in some cases on small subject samples and consensus on appropriate studies included (23). The reference estimates are often provided as single values without variability statistics, and estimates of population variation are usually unavailable. However, this “best data set available” is invaluable for developing models such as those reported in the current study.

### Conclusion

Our developed models combined with experimental observations indicate that LST hydration is not “constant,” as is often assumed, but varies predictably with E/I. This observation extends the earlier FFM hydration predictions of Moore and Boyden (15) and Wang et al. (26) and other experimental studies of hydration and aging (22). Our quantitative estimates emerging from model development and evaluation also provide a general idea of percent fat errors introduced into DEXA measurements due to between-subject hydration differences, and we established the appropriateness of applying TBW as a means of calibrating DEXA system LST measurements.

These collective observations provide useful information and insights into DEXA-based LST estimates that are becoming increasingly available in the clinical and research setting.

## GRANTS

This study was supported by National Institutes of Health Grants RR-00645 and DK-42618. M.-P. St-Onge is supported by a Canadian Institute of Health Research fellowship.

## Footnotes

The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “

*advertisement*” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

- Copyright © 2004 by American Physiological Society