INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

BIOELECTRIC-IMPEDANCE ANALYSIS (BIA) was first introduced as a method of estimating body composition ~10 years ago (14). Since then, use of the method has become widespread in both clinical and epidemiological research. Despite considerable experimentation and theoretical advances, BIA remains something of a "black box" method of in vivo body-composition analysis. Technical and theoretical aspects of the BIA method have been reviewed in detail elsewhere (1). In brief, BIA measures the impedance, or "opposition," of the body to the flow of a low-amplitude, high-frequency alternating electric current. Impedance is measured in BIA by using a tetrapolar-bridge approach with separate current (source) and voltage (detector) electrodes placed on the skin surface at standard anatomic locations. The current is introduced between the distally located source electrodes, and the voltage drop caused by impedance is measured between the dectector electrodes, which are typically situated ~5-20 cm proximal to their paired source electrodes. The phase shift caused by the capacitive effects of cell membranes and other dielectric materials is also measured and is used to partition total impedance into resistance (R) and reactance components in most BIA devices, although some directly determine R and reactance separately. Both R and reactance are frequency and temperature dependent. In the human body, >90% of the measured impedance is composed of R. For this reason, as well as others, most BIA applications use R, rather than impedance, to predict body composition (1).

The electrical charge of the current is conducted by free electrolytes in the body fluids, and the R of any specific body composition component is proportional to the concentrations of fluids and mobility of ions in that component. Thus R is low for blood, urine, and muscle but high for adipose tissue, bone, and air, which contain little or no fluid or electrolyte ions. Because the current tends to follow the path of least resistance, measured R correlates most strongly with total body water (TBW), and correlations decrease for other body composition components, depending on the amount of water in these components. In vivo measurements of impedance, however, reflect the joint conductive properties of all the materials within the total body volume (V) or within the V of the arm, leg, or trunk when body segments are measured separately. As a result, it is not clear to what extent in vivo measurements of impedance directly reflect the V of specific components with high conductivities or are influenced by other component V with lower conductivities (1).

In most applications, BIA is used to predict TBW and fat-free mass (FFM). These components of body composition do not occupy discrete spaces within the body V but are distributed in varying concentrations in body tissues. Consequently, it is difficult, if not impossible, to model these components as specific conductive paths within the total body V. Muscle, adipose tissue, and bone are components that constitute discrete, anatomic volumes (Vm, Vat, and Vb, respectively). Although the composition of these components, in particular their fluid concentration, may vary within and between individuals, they can be defined as discrete conductive paths, or resistors, in an equivalent-circuit model as shown in Fig. 1.

View larger version (31K): [in this window] [in a new window]   Fig. 1.   Parallel tissue-resistor model. This model considers a limb to represent concentric cylinders of subcutaneous adipose tissue (SAT), muscle, and bone that form a circuit of resistors in parallel when conducting an electric current. Am, cross-sectional muscle area; Aat, adipose tissue area; Ab, bone area; m, resistivity () of muscle; at,  of adipose tissue; b,  of bone.

Several investigators have reported that bioelectric impedance overestimates FFM in samples of obese people (8, 9, 11, 12, 19, 20). These studies show that residuals for the prediction of FFM by BIA are correlated significantly with percent body fat. Different hypotheses have been advanced to explain this effect. One is that hydration and/or fluid distribution in either or both the nonadipose and adipose tissues is altered in obesity and that the criterion methods used for calibrating prediction equations do not account for these changes. Although some studies have shown that BIA is sensitive to alterations in hydration and fluid distribution, there have been few, if any, good tests of the hypothesis that this explains specifically the overestimation of FFM in obese subjects (6).

A second hypothesis is that the tendency of BIA equations to overestimate FFM in obese subjects is caused by differences in body geometry. Some studies have shown that measures of whole body R are dominated by the arms and the legs and depend strongly on variation in the cross-sectional areas of the distal extremities (3, 7). Lukaski (13) demonstrated empirically that the use of proximal, rather than distal, electrode placements on the limbs reduced significantly the correlation between percent body fat and FFM residual scores. Despite this finding, there is no clear theory that explains an association between body geometry and the tendency to overestimate FFM in obese subjects.

A third hypothesis, which has not been tested, is that the increased Vat in obese subjects directly affects the measurement of R. This hypothesis has a theoretical basis in the parallel tissue-resistor model as described by Rush et al. (17), which predicts that adipose tissue, if sufficiently large, will affect directly the measurement of bioelectric R.

The present study attempts to test the third hypothesis by using a unique data set for body composition from whole body magnetic resonance imaging (MRI) scans, and also segmental and whole body measurements of bioelectric R and anthropometry, in 86 overweight and obese men and women.

	SUBJECTS AND METHODS

Top Abstract Introduction Methods Results Discussion References

The study group consisted of 40 male and 46 female volunteers in an exercise-diet weight-loss program at the School of Physical and Health Education, Queen's University, Kingston, CA (16). The mean age of participants was 38.5 ± 10.2 yr. All participants had body mass indexes >27 kg/m², were weight stable (±2 kg) for 6 mo before entry, and were taking no medications known to affect body composition. The data set analyzed represents baseline data collected before initiation of the weight-loss protocol. The study was conducted in accordance with the ethical guidelines of Queen's University, and all participants gave informed consent.

Anthropometry. All body measurements were taken by using standardized procedures as described in the Anthropometric Standardization Reference Manual (10). Weight was measured to the nearest 0.1 kg on a beam-balance scale. Stature, sitting height, and acromiale height were measured to the nearest 0.1 cm with a wall-mounted stadiometer. Arm length, defined as the distance from the anterior-lateral edge of the acromion to the distal end of the third phalange, was measured by using flexible steel tape on the right side to the nearest 0.1 cm. The elbow and fingers were fully extended, and the arm was abducted slightly from the side. Leg length was derived as the difference between stature and sitting height. Trunk length was calculated as acromiale height minus leg length.

Bioelectric R. Bioelectric R was measured by using a model 101B BIA analyzer (RJL Systems, Detroit, MI) with an operating frequency of 50 kHz at 800 µA. Whole body measurements were taken by using standard electrode locations on the hand and foot on the right side when the participant was fasting and had voided the bladder (14). The R of the right arm and leg, as well as R of the trunk, were taken by using the electrode placements described in Chumlea et al. (5). In brief, the R of the arm was measured with one electrode pair placed in the conventional locations on the posterior surface of the hand and wrist. The other pair of electrodes was placed with the detector electrode on a line from acromial process to the axillary fold, and the source electrode was ~5 cm medial. The R of the leg was measured with one electrode pair in standard location on the anterior surface of the foot and the other at the level of the gluteal crease. The R of the trunk was measured from the anterior surface of the thigh at the gluteal crease to the sternal notch. All R measurements were taken with the participant supine and with the arms abducted slightly, but not touching the sides, and with the legs separated so that there was no contact between the thighs.

Whole body MRI. The protocol used to acquire the MRI data for whole body and body-segment tissue V is described in detail elsewhere (16). In brief, images were taken with a Siemens 1.5-T whole body scanner (Erlangen, Germany). A series of 41 10-mm-thick axial images were made at 50-mm intervals from the tips of the extended fingers to the feet by using a T1-weighted, spin-echo sequence with a 210-ms repetition time and 15-ms echo time. Images for the abdomen were obtained by using a rectangular field of view (192 × 256 pixels) and a one-half Fourier transformation. The use of these parameters reduced the image-acquisition time for the abdominal images to ~26 s, during which the participants held their breath to minimize respiratory noise. The total time to acquire all 41 images was ~25 min.

Analytic models and methods. Body composition is measured by using MRI on what Wang et al. (21) have referred to as the tissue-system level of organization. Thus it is important to keep in mind that the Vat, Vm, and Vb that are measured from MRI scans do not correspond directly to fat and fat-free soft-tissue and bone mineral masses on the molecular level, as measured by using techniques such as dual-energy X-ray absorptiometry. This distinction is important for the approach taken to analysis in the present study, because the MRI tissue V fit better in the theoretical, geometric model of parallel conductors described in Fig. 1.

View larger version (14K): [in this window] [in a new window]   Fig. 2.   Equivalent-circuit model. Parallel tissue-resistor model in terms of a simple electric circuit. V, voltage; I, current; Rm, muscle resistor; Rat, adipose tissue resistor; Rb, bone resistor; I1, I2, and I3, currents through each resistor in circuit.

(1)

(2)

View larger version (28K): [in this window] [in a new window]   Fig. 3.   Theoretical effects of increasing adipose tissue and changes in tissue  on measured R and estimation of muscle mass for leg. , Increasing SAT volume relative to muscle, with no change in ; , increasing SAT/muscle with decreasing at caused by increased hydration; , increasing SAT/muscle with decreasing m tissue caused by increased hydration; , increasing SAT/muscle with both decreasing  of SAT and decreasing m. Length, 70 cm; muscle volume, 6 liters; m, 300  · cm;  of SAT, 3,000  · cm.

(3)

	RESULTS

Top Abstract Introduction Methods Results Discussion References

Descriptive statistics (means ± SD) for the anthropometric, bioelectric R, and MRI volume measurements are shown in Tables 1 and 2. The mean body mass index was 32.29 ± 3.90 kg/m² in the men and 34.20 ± 4.99 kg/m² in the women, verifying that the study participants were overweight or obese. The mean SAT volume was 41 and 42% of the total volume of the arm and the leg, respectively, in the men. The ratios of adipose tissue to Vm in the men averaged 0.77 for the arm and 0.83 for the leg. In the women, SAT volume averaged 57% of total V of the arm and 54% of total V in the leg. The ratios of Vat to Vm in the women averaged 1.47 for the arm and 1.35 for the leg. In the men, 43% of total trunk V was SAT and 12% was IAT. In the women, in contrast, 29% of trunk V was SAT and 10% was IAT. Thus adipose tissue constituted a substantial fraction of the V of each body segment. In the women, SAT V was substantially greater than Vm for the arm and the leg and was well into the range in which an effect on measured R would be expected under the parallel tissue-resistor model.

Tables 3 and 4 show results for the multiple-regression analyses of 1/R on the MRI tissue V, adjusted for segment length according to formula (2), for the arm, leg, trunk, and whole body. Vm had statistically significant associations with 1/R, with the exceptions of the arm in the men and the trunk in both the men and women. SAT V for the trunk were associated significantly with trunk R in both sexes. The association of leg SAT V with leg R approached statistical significance (P < 0.06) in the women. Leg SAT and total SAT were associated significantly (P < 0.05) with whole body R in the women. IAT was not associated significantly with either whole body or body-segment R in any of the regression models. Vb were not associated significantly with 1/R for any of the body segments or the whole body. The variance explained by the combined tissue V ranged from 5% for the arm to 50% for the trunk in the men and from 12% for the trunk to 29% for the arm in the women. The combined segment tissue V explained 39-57% of the variance in whole body R in the men and women, respectively, in model A; and the sums of the tissue V explained 27-42% of whole body R in model B. Correlations between the length-adjusted muscle and SAT V were small (r < 0.25). Thus these results are not distorted by high multicollinearity among the variables.

We further evaluated the magnitude of the effect of SAT on R by using the regression equations and data for the leg and the whole body in the women. The full regression equation for the leg in women was Using the data in Table 1, we calculated R, assuming L = 74.29 cm, Vm = 6.2 liters, Vb = 0.8 liters, and Vat = Vm. The calculated R = 255.7  was then used in the equation (300  · cm) × L²/R to calculate Vm = 6.48 liters, or ~5% greater than the actual value. We then recalculated R, assuming that Vat = 1.6 × Vm = 9.86 liters. The calculated R was 243.3 , or ~5% less, and resulted in an estimated Vm = 6.81 liters, or ~10% greater than the actual value. To further illustrate the effect, in Table 4, we estimated R values by using model B for the whole body, first assuming Vat = Vm and then Vat = 1.6 × Vm. The calculated R was 14.4% less when Vat was assumed to be 1.6 × Vm than when the V were assumed to be equal.

Finally, we evaluated the effect of a 14% reduction in measured R caused by the effect of adipose tissue when applied to a standard model for estimating FFM. One of Lukaski's equations for women (13) has been given as where R is whole body R. When the mean values reported in Table 1 for stature, R, and weight in the women are inserted into this equation, the resulting estimated FFM is 52.6 kg and percent body fat is ~43%. If we assume that R is reduced by the effect of adipose tissue and recalculate FFM by using an R value that is 14% greater, the resulting estimated FFM is 49.56 kg, corresponding to ~46% body fat.

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

The results of the present study suggest that adipose tissue, when sufficiently large, may directly affect measured R. This effect on R may result in an overestimation of muscle, lean soft tissue, or FFM when equations are applied that are calibrated by using data for nonobese, lean individuals. These effects are normally small, but they increase rapidly when the Vat is greater than Vm, as is the case in many persons with moderate to severe obesity.

The present study confirms for the first time that, from a simple body-composition model, the primary conductor in the bioelectric impedance method is appendicular skeletal muscle. Leg Vm in men and arm and leg Vm in women were correlated significantly with the R for the arm and the leg, as well as the for whole body. At this time, there is no clear explanation why arm Vm was not significantly associated with arm or whole body R in the men. The lack of association was not attributable to the presence of one or more unusual data points or outliers.

SAT on the trunk was associated significantly with the measurement of trunk R in both sexes. These results, however, should be considered with caution, because the parallel-tissue resistor model lacks validity for this body segment. Nonetheless, these findings for the trunk could help explain why the specific measurement of trunk R is greater than is theoretically expected and why trunk phase angle (ratio of reactance to R) is correlated with percent body fat (2, 3, 7, 13). In addition, this may pose a major limitation to the segment-impedance approach, because unbiased specific  for the trunk cannot be estimated with confidence from measurements of trunk R. It is important to note that neither trunk Vm nor Vat significantly influences whole body R. This suggests that the composition of the trunk is of relatively little concern in the conventional BIA approach to estimating body composition by using whole body R. As noted previously, the electrical conductive characteristics of the limbs are reported to primarily determine the use of BIA in predicting body composition (3, 7). However, Scheltinga (18) has noted that measurements of the trunk should be taken without moving the source electrodes (i.e., leaving them in their distal locations). It is possible that the use of this method would have improved results for the trunk.

A considerable portion of the variability in measured R was not explained by volumetric composition. This suggests that other factors contribute importantly to bioelectric R over and above the relative volumes of more- or less-conductive tissues and their errors of measurement. These factors probably include physiological and structural variables, such as body temperature, tissue composition, and fluid distribution, as well as anisotropic effects of muscle fibers. Technical factors affecting the accuracy of the measurement of resistance also must be studied.

It is important to consider some of the limitations of the model used in the present study. The model considers measured R in terms of parallel R offered by the major tissue V. Each tissue V, however, is composed of more basic constituents at cellular or chemical levels that may or may not be modeled in terms of parallel R circuits. In addition, tissues have different structural characteristics that may affect R. For example, skeletal muscle fibers can produce anisotropic effects on R. Moreover, reactance was not considered in the equivalent-circuit model, although reactance is small relative to R in biological conductors (1).

It is not known how variability in the composition of different tissues affects their specific . For example, although the increase in Vat in obesity is primarily caused by the increased amounts of lipid deposited within the adipocytes, there is controversy as to whether the concentration of water in adipose tissue changes with the expansion of V (15). An increase in the concentration of water with increasing Vat would be expected to lower the of this tissue. In the parallel tissue-resistor model, an increased Vat with a reduced  would be predicted to have an even greater effect on total measured R. An expanded adipose tissue mass might also be associated with changes in the distribution of fluid and electrolytes within muscle, altering its . There is no way of testing this hypothesis with the present data; however, our model predicts that changes in muscle  would have much greater effect on measured R than observed in the present study, as illustrated in Fig. 3.

In summary, the results of the present study suggest that the measurement of bioelectric R may be affected by adipose tissue and that these effects may explain the slight overestimation of FFM, and subsequent underestimation of percent body fat, that has been observed in obese subjects. Insofar as the results of the present study have validity, it may not be possible to eliminate this effect of adipose tissue by using the current equipment. This effect, however, appears to be slight and occurs only for moderately to severely obese subjects. BIA can be used in most applications with confidence that the true conductive V is muscle and that estimates are relatively unbiased by adipose tissue. Thus, the BIA method may be used to predict Vm and muscle mass with high validity, although the precision of the estimates remains to be established. Further research needs to be conducted to determine the best approaches for reducing the slight bias produced by the effect of adipose tissue on measured R in obese women.