Comparison of Methods to Assess Body Composition Changes during a Period of Weight Loss


  • Madlyn I. Frisard,

    1. Clinical Trials Division, Pennington Biomedical Research Center, Louisiana State University System, Baton Rouge, Louisiana
    Search for more papers by this author
  • Frank L. Greenway,

    1. Clinical Trials Division, Pennington Biomedical Research Center, Louisiana State University System, Baton Rouge, Louisiana
    Search for more papers by this author
  • James P. DeLany

    Corresponding author
    1. Clinical Trials Division, Pennington Biomedical Research Center, Louisiana State University System, Baton Rouge, Louisiana
      6400 Perkins Road, Baton Rouge, LA 70810. E-mail:
    Search for more papers by this author

  • The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

6400 Perkins Road, Baton Rouge, LA 70810. E-mail:


Objective: To assess the accuracy of body composition measurements by air displacement plethysmography and bioelectrical impedance analysis (BIA) compared with DXA during weight loss.

Research Methods and Procedures: Fifty-six healthy but overweight participants, 34 women and 22 men (age, 52 ± 8.6 years; weight, 92.2 ± 11.6 kg; BMI, 33.3 ± 2.9 kg/m2) were studied in an outpatient setting before and after 6 months of weight loss (weight loss, 5.6 ± 5.5 kg). Subjects were excluded if they had initiated a new drug therapy within 30 days of randomization, were in a weight loss program, or took a weight loss drug within 90 days of randomization. Subjects were randomly assigned either to a self-help program, consisting of two 20-minute sessions with a nutritionist and provision of printed materials and other self-help resources, or to attendance at meetings of a commercial program (Weight Watchers). Body composition was examined by each of the methods before and after weight loss.

Results: BIA (42.4 ± 5.8%) underestimated percentage fat, whereas the BodPod (Siri = 51.7 ± 6.9%; Brozek = 48.5 ± 6.5%) overestimated percentage fat compared with DXA (46.1 ± 7.9%) before weight loss. Correlation coefficients for detecting changes in body composition between DXA and the other methods were relatively high, with Brozek Δfat mass (FM; r2 = 0.63), Siri FM (r2 = 0.65), tetrapolar BIA percentage fat (r2 = 0.57), and Tanita FM (r2 = 0.61) being the highest.

Discussion: In conclusion, all of the methods were relatively accurate for assessing body composition compared with DXA, although there were biases. Furthermore, each of the methods was sensitive enough to detect changes with weight loss.


The rapidly increasing prevalence of obesity has attracted the attention of the World Health Organization, which has issued a statement on the matter (1). Obesity is related to a number of disease states including hypertension, hypercholesterolemia, and diabetes. Body fat is also associated with increased mortality (2). Weight loss reduces cardiovascular disease risk (3, 4, 5) and improves insulin sensitivity (6). Weight loss practices and techniques with an emphasis on fat loss have become widespread, and the need for accurate measurements for the assessment of body composition is important. There are a number of methods to assess body composition. While some are more accurate than others, the cost of more accurate methods is not always feasible. Therefore, it is important to assess the validity of less expensive techniques.

Earlier body composition methods used a two-compartment model composed of fat mass (FM)1 and fat-free mass (FFM). Hydrodensiometry was such a model and was considered the gold standard for body composition assessment. The emergence of DXA has provided an alternative. DXA is easier to administer than hydrodensiometry, a factor that is important in certain populations such as the elderly (7). More importantly, DXA uses a three-compartment model and provides information on bone in addition to soft tissue. This method also seems relatively unaffected by hydration status. The major disadvantage of this method is the cost of equipment. DXA correlates well with other measures of body fat, such as hydrodensiometry, total body water, and multicompartment models (8, 9, 10, 11, 12). Bone mineral density may affect the accuracy of the underwater weighing (UWW) measurement (13). DXA, however, removes this confounding factor from the measurement (14). DXA is also able to detect changes with weight loss (15).

With the emergence of the BodPod (Life Measurement, Inc., Concord, CA), a number of researchers began looking at air displacement plethysmography (ADP) as a possibly less expensive alternative to other methods. ADP seems to be precise and reliable (16, 17, 18, 19, 20, 21, 22). Compared with other methods, researchers have found a correlation between ADP and DXA, UWW, and a four-compartment model (18, 20, 21, 23, 24, 25). Others have found that ADP underestimates body density and, therefore, overestimates percentage fat, especially in lean subjects (16, 26, 27, 28).

Studies evaluating the accuracy and precision of bioelectrical impedance analysis (BIA) have been inconsistent. Kushner et al. (29) validated BIA against deuterium oxide dilution in 12 obese women and found that it was accurate in predicting total body water in both obese and non-obese men and women. However, other studies have not obtained similar accuracy (30, 31).

There is little research comparing body composition methods over a period of weight loss (15, 29). Weyers et al. (32) found that both DXA and ADP are sensitive enough to measure moderate changes in body composition with weight loss. The accuracy of inexpensive “field techniques” is important for the measurement and assessment of body composition changes with weight loss in large populations. The purpose of this study was to assess body composition by three different methods, DXA, BIA, and ADP, before and after a 6-month weight loss program. We used both conventional bioelectrical impedance machines as well as the Tanita body fat analyzer, and we will differentiate between the two by the terms tetrapolar BIA and the Tanita system. The Tanita has become quite popular in the public as a measure of body fat, so we thought it was important to test this method against other proven methods measuring weight loss. Although multicompartment models are the ideal reference standards to compare methods, we chose to use the DXA as our reference method. In a validation study (unpublished data), DXA was consistent with the chemical analysis of the percentage fat in ground meat alone (25.4 ± 1.1% vs. 25.1 ± 0.3%), and when known amounts of lard were added, DXA accurately measured the amount of fat added (454 vs. 456 ± 27 g; 908 vs. 894 ± 16 g; 1816 vs. 1803 ± 5 g). We also validated percentage fat by DXA to UWW in 154 adults (R2 = 0.93; unpublished data) and compared DXA to UWW for measuring change in both FM (R2 = 0.78) and FFM (R2 = 0.67) in 114 children over a 2-year period (33).

Research Methods and Procedures


Seventy-three subjects were initially enrolled in the weight loss program. Healthy men (n = 22) and women (n = 34) with a BMI of 27 to 40 kg/m2, 18 to 65 years of age, were included in the study. Subjects were excluded if they had initiated a new drug therapy within 30 days of randomization, were in a weight loss program, or took a weight loss drug within 90 days of randomization. Body composition was measured in the fasted state by each of the methods once on the same day. Follow-up measures were performed 6 months later in the same order as before weight loss.


The participants were weighed in street clothing without shoes. The measurement was taken to ±0.1 kg with an electronic scale (Detecto, Webb City, MO) that was checked daily with a standard 25-kg weight. Height was measured to ±0.5 cm with a wall-mounted stadiometer (Holtain; Crymych, Dyfed, UK), and BMI was calculated as weight divided by height squared.

Tetrapolar BIA

Total body water and FFM were measured by BIA using a Xitron variable frequency impedance machine (Xitron, San Diego, CA). Electrodes were placed on the right wrist and ankle. On the wrist, one electrode was placed to bisect the ulnar head, and the other electrode was placed just behind the middle finger. On the ankle, one electrode was placed to bisect the medial malleolus, and the other electrode was placed just behind the middle toe. Our population was considered obese; therefore, we used equations from Segal et al. (34) to calculate FFM for obese individuals.


Tanita Body Fat Analyzer

BIA from foot to foot was measured using the Tanita 305 body fat analyzer (Tanita Corp., Tokyo, Japan). The subjects stood on the metal plates of the machine. Predictions equations for men and women were provided by the manufacturer.

Equations for body density (BD) for men are as follows:



Plethysmography measures body volume using Boyle's law of the pressure/volume relationship. Thus, body volume is equal to the reduction of volume in the chamber with the introduction of the subject under isothermal conditions, while maintaining a constant temperature. However, it is impossible to maintain a constant temperature throughout the test. Therefore, Poisson's law shows the relationship between pressure and volume under changing temperature (adiabatic conditions). Isothermal air is 40% more compressible than adiabatic air; therefore, it is important to consider both of these relationships to accurately measure body volume. Air close to skin, hair, and clothing will not maintain adiabatic conditions. Therefore, it is important to eliminate or minimize these effects. The subjects are instructed to wear a formfitting bathing suit and a swim cap for the duration of the test to minimize clothing and compress air, thereby eliminating volume variations. Subjects were weighed on an electronic scale. The subject then sat quietly while their body volume was measured. This measurement is also affected by any air located inside the body such as air located in the lungs or thorax. Thoracic gas volume was measured by connecting the subject to a breathing circuit. The process was repeated until a consistent measurement was obtained. BD was calculated as mass divided by volume and corrected for lung volume. The Siri and Brozek formulas were used to calculate percentage body fat (35, 36).


We chose to calculate percentage body fat using both formulas to determine which is more accurate under the present conditions.


DXA uses an X-ray source to achieve a congruent beam of stable, dual-energy radiation with effective energies of 40 and 70 keV. The differential attenuation of the two-photon energies is computer analyzed, and the soft tissue attenuation and R value are calculated. Percent lean mass can be calculated from the known R values of fat, lean tissue, and bone, and the calculated soft tissue. Whole body scans were performed using a Hologic 2000 (Waltham, MA) in the fan beam mode. The scans were analyzed using the Hologic enhanced whole body version 6.8 software. The coefficients of variation for the various body compartments in our laboratory were 0.25% for weight, 1.0% for lean mass, 1.6% for fat mass, 1.5% for percentage fat, and 1.0% for bone mineral content.

Weight Loss Intervention

Subjects were randomized to either a self-help group or commercial weight loss group. Description of the groups are found elsewhere (37).

Statistical Analysis

Statistical analysis was performed using SAS software version 8 (SAS, Cary, NC). Linear regression was used to assess the accuracy of the methods compared with DXA both before and after treatment. Linear regression was also used to assess differences between the methods when measuring changes in percentage fat, FFM, and FM over a period of weight loss. The Bland-Altman approach was used to assess the accuracy of each of the methods against DXA. The difference between each method and DXA was compared with the average of the two methods by simple linear regression. The hypothesis that the slope was equal to zero was also tested in each case. Linear regression was also run on the change in percentage fat, FM, and FFM as detected by each of the methods.


Description of the Study Population

A description of the population is shown in Table 1. The average weight decreased by 6.5 ± 0.4 kg. Table 2 shows the changes in body composition parameters by each method before and after weight loss. We included estimated measurements of total body water by tetrapolar BIA to assess whether changes in body water may have affected the results. This is shown in Table 2 immediately under the tetrapolar BIA data. Absolute percentage fat as measured by DXA decreased by 2.1%. DXA FM decreased by 4.8 kg, whereas FFM decreased by 1.9 kg. There were significant differences in all measures with all methods before and after weight loss. Both Tanita (difference = −2.8) and tetrapolar BIA (difference = −3.7) underestimated percentage fat before weight loss, whereas the BodPod overestimated percentage fat using either the equation of Siri (difference = 5.6) or Brozek (difference = 2.4). Tanita and tetrapolar BIA overestimated FFM before weight loss, whereas BodPod underestimated FFM.

Table 1. . Subject characteristics
VariableBefore weight lossAfter weight loss
Age52 ± 8.6 
Height (cm)166 ± 9 
Weight (kg)93.1 ± 11.586.6 ± 12.0
Table 2. . Subject body composition by method
MethodVariableBefore weight lossAfter weight lossDifferenceSignificance
  • *

    Total body water was calculated using the obese equation of Kushner (38).

DXAPercent fat (%)46.1 ± 7.944.0 ± 8.5−2.1 ± 2.80.000
 FFM (kg)50.5 ± 11.448.6 ± 10.7−1.9 ± 2.20.000
 FM (kg)42.9 ± 7.738.1 ± 8.8−4.8 ± 4.50.000
BIAPercent fat (%)42.4 ± 5.841.9 ± 7.0−0.5 ± 1.70.000
 FFM (kg)53.5 ± 10.550.7 ± 9.6−2.8 ± 2.20.000
 FM (kg)39.2 ± 6.036.6 ± 7.6−2.6 ± 4.40.000
Total body water*TBW (liters)41.7 ± 8.639.9 ± 7.6−1.8 ± 2.20.000
TanitaPercent fat (%)43.2 ± 7.740.5 ± 8.1−2.7 ± 3.80.000
 FFM (kg)53.2 ± 11.451.6 ± 10.3−1.6 ± 3.00.000
 FM (kg)40.3 ± 7.735.1 ± 8.5−5.2 ± 4.70.000
 SiriPercent fat (%)51.7 ± 6.948.8 ± 9.0−2.9 ± 3.80.001
 FFM (kg)45.3 ± 9.644.3 ± 9.8−1.0 ± 2.30.000
 FM (kg)48.2 ± 8.342.3 ± 10.3−5.9 ± 5.40.000
 BrozekPercent fat (%)48.5 ± 6.546.1 ± 8.6−2.4 ± 3.60.004
 FFM (kg)47.8 ± 9.846.2 ± 10.1−1.6 ± 2.30.000
 FM (kg)44.9 ± 7.739.6 ± 9.7−5.3 ± 5.10.000

Regression Coefficients before and after Weight Loss

Table 3 shows the regression coefficients for percentage fat, FFM, and FM for each of the methods in relation to DXA. Regression coefficients were high (r2 > 0.8) for percentage fat, FFM, and FM for tetrapolar BIA and the BodPod. However, regression coefficients using Tanita were lower (r2 <0.7) for percentage fat and FM. After weight loss, the regression coefficient for FM with Tanita increased (before = 0.35; after = 0.83). Regression coefficients for tetrapolar BIA and BodPod increased slightly.

Table 3. . Body composition measured before and after weight loss compared with DXA
  Before weight lossAfter weight loss
  1. Correlation coefficients, slopes, y-intercepts, and p values are given for each of the methods calculated from measured percent body fat and weight compared with DXA.

BIAPercent fat0.881.2−6.10.0000.911.3−11.20.000
TanitaPercent fat0.610.7611.10.0000.760.926.70.000
 SiriPercent fat0.831.0−7.70.0000.850.880.350.000
 Brozek 0.831.1−9.30.0000.840.91−6.90.000
 Brozek 0.931.1−5.40.0000.941.1−1.80.000
 Brozek 0.820.920.590.0000.940.893.30.000

Bland-Altman Analysis

The results for percentage fat from Bland-Altman statistics are shown in Figure 1. Figure 1A shows the comparison between DXA and the BodPod using the equation of Brozak. There was a significant positive slope (p < 0.001), indicating that percentage body fat was overestimated by BodPod using the equation of Brozek at lower body fat (<40% fat), but it seemed to underestimate percentage body fat with increasing body fat. Although there was still an overestimation after weight loss (Figure 1B), the slope disappeared (p = 0.878). A similar pattern was seen with FFM. There were no differences in FM after weight loss. Percent fat was also overestimated using the equation of Siri at all body sizes (Figure 1C). The slope, however, was not significant (p = 0.016). There were similar trends with FFM and FM. Using the Siri equation, there were no changes after weight loss (Figure 1D). Using tetrapolar BIA, there was a significant positive slope (p < 0.000), indicating overestimation at lower body fat (<45%) but underestimation at body fat >45% (Figure 1E). No change was noted after weight loss (p < 0.000; Figure 1F). Similar trends were also seen in FM and FFM. After weight loss, improvements were seen in FM, with no changes in FFM. There was high variability in percentage fat using Tanita (Figure 1G). This improved only slightly with weight loss (Figure 1H). There was also high variability with FFM and FM.

Figure 1.

Bland-Altman plots for percentage fat. (A) BodPod Brozek before weight loss. (B) BodPod Brozek after weight loss. (C) Siri BodPod before weight loss. (D) Siri BodPod after weight loss. (E) Xitron BIA before weight loss. (F) Xitron BIA after weight loss. (G) Tanita before weight loss. (H) Tanita after weight loss.

Correlation Coefficients for Change in Body Composition

Regression analyses for changes in body composition are displayed in Figures 2, 3, 4, 5. Using Brozek (Figure 2 A-C), coefficients for BodPod were higher for Δ percentage fat (r2 = 0.24) and Δ FM (r2 = 0.64). Similar results were also seen with Siri (Figure 3 A-C). There was no correlation between DXA and the BodPod for FFM using the equation of either Brozek (r2 = 0.02) or Siri (r2 = 0.00). Correlation coefficients for percentage fat (Figures 4 A-C and 5, A-C) were higher using tetrapolar BIA than Tanita (r2 = 0.57 vs. r2 = 0.27). Similar results were seen with FFM (r2 = 0.44 vs. r2= 0.09). However, correlation coefficients for FM were higher using Tanita (r2 = 0.61) than tetrapolar BIA (r2 = 0.51).

Figure 2.

Change in body composition measured by BodPod (Brozek) compared with DXA. (A) Percent fat: r2 = 0.24, slope = 0.7, y-intercept = 0.5. (B) FFM: r2 = 0.02, slope = 0.1, y-intercept = 1.4. (C) FM: r2 = 0.63, slope = 1.0, y-intercept = 0.2.

Figure 3.

Change in body composition measured by BodPod (Siri) compared with DXA. (A) Percent fat: r2 = 0.28, slope = 0.7, y-intercept = −0.6. (B) FFM: r2 = 0.00, slope = 0.0, y-intercept = 1.3. (C) FM: r2 = 0.65, slope = 1.0, y-intercept = −0.3.

Figure 4.

Change in body composition measured by BIA compared with DXA. (A) Percent fat: r2 = 0.57, slope = 0.5, y-intercept = −0.8. (B) FFM: r2 = 0.44, slope = 0.6, y-intercept = −1.0. (C) FM: r2 = 0.51, slope = 0.4, y-intercept = −0.5.

Figure 5.

Change in body composition measured by Tanita compared with DXA. (A) Percent fat: r2 = 0.27, slope = 0.7, y-intercept = −1.3. (B) FFM: r2 = 0.09, slope = 0.7, y-intercept = −1.5. (C) FM: r2 = 0.61, slope = 0.9, y-intercept = −0.9.


The widespread obesity epidemic has led to the need for accurate and feasible measures of body composition. In addition, body weight and BMI may not be the best assessments of health or success of a weight loss program. The purpose of this study was to compare body composition by two bioelectrical impedance methods and ADP during weight loss, using DXA as the criterion method. A number of studies have been conducted to assess the accuracy of various body composition methods compared with reference methods. However, DXA has consistently been proven to be an accurate assessment of body composition (39). The novel aspect of this study was the assessment of body composition after a small amount of weight loss. Body composition measurements are based on a number of different assumptions depending on the technique being used. For example BIA and ADP assess BD using different principles. Therefore, it was not the goal of this study to determine whether these measures are interchangeable. Rather, it was the goal of this study to determine whether the measures are sensitive enough to pick up differences relatively similar to what DXA would pick up from the amount of weight loss seen in this study.

Studies measuring body fat in athletes and leaner individuals have found that BIA correlates very well with DXA (19, 23). The studies comparing BIA to DXA in other populations have found that BIA overestimates percentage fat (31). More specifically, electrode placement on the wrist and ankles correlates more closely with DXA than the foot-to-foot method (40). The results of this study are in agreement with others indicating that, before weight loss, BIA may overestimate FFM, leading to an underestimation of FM and percentage fat compared with DXA (41). ADP has also been shown to correlate well with DXA in athletes and leaner individuals (19, 25, 40). Wells and Fuller (42) concluded that plethysmography is accurate over a wide range of body size. However, other studies have found mixed results. A number of studies have found that ADP overestimates FFM, leading to underestimation of FM. For example, Levenhagen et al. (23) compared ADP to DXA and found that ADP underreported FM compared with DXA. Similar results were also found in a number of other studies (21, 24). There were differences in body size between this and previous studies (BMI, <25 vs. >33.3 kg/m2). The results of this study are in agreement with those that suggest that ADP underestimates FFM in overweight individuals and, therefore, overestimates FM (22, 26, 27, 28). These studies are similar in that they have included subjects over a wide range of body sizes. The regression line comparing ADP to DXA crosses the line of identity at a point somewhere around the range of normal weight. Therefore, at higher body weights, there is an overestimation, whereas at lower body weights, there is an underestimation. This is in agreement with our findings—there is an improvement in this overestimation once weight is lost, which would explain why it has been found to be more accurate in leaner individuals. ADP may become less reliable with increasing body fat.

In the Bland-Altman analysis, the difference between two methods for measuring a feature was compared with the average value of the two measurements. Using DXA as the criterion, ADP systematically underestimated body fat as body fat increased. This was seen using the equations of Siri and Brozek; however, it was more significant using the equation of Brozek. Even a small amount of weight loss improved this discrepancy in the equation of Brozek but not of Siri. However, this may be because differences were more significant before weight loss with the equation of Brozek. Similar trends were seen with tetrapolar BIA. However, there was no improvement in this measurement once weight was lost. Although these trends were not seen with the Tanita, there was very high variability with the Tanita both before and after weight loss.

Few studies have evaluated the sensitivity of these methods to changes in body composition over a period of weight loss (32, 43). The amount of weight loss in the study of Hendel et al. (43) was 11 kg—slightly higher than in this study. Weyer et al. (32) used a weight loss of 4.5 kg, a value closer to this study. BIA (Xitron) correlated relatively well with DXA for percentage fat and FFM. Coefficients for FM were not as high. Furthermore, Tanita did not correlate as well with DXA as Xitron for detecting changes in weight loss. As stated previously, previous research indicates that, before weight loss, BIA may overestimate FFM, leading to an underestimation of FM and percentage fat. This discrepancy may be because of differences in FFM hydration status in obese individuals compared with leaner ones. Measurement of body composition by DXA is not affected by small changes in hydration (8). Earlier studies have indicated that there is large variation in the water content of FFM (44). Furthermore, research indicates that fatter children have a higher hydration status of FFM than leaner children (45). These differences can lead to an underestimation of FFM with BIA in the dehydrated state and an overestimation in the overhydrated state (46). In addition, Leone et al. (47) indicated that previously obese individuals are overly hydrated compared with never obese individuals. Total body water calculated for this population was 41.7 liters before weight loss and 39.9 liters after weight loss. Hydration status was calculated from FFM measured by DXA, and total body water was calculated from Kushner. Hydration status was 82.6% before weight loss and 82.1% after weight loss. This is higher than the assumed hydration status (69.0% to 72.2%) calculated for leaner individuals (44). These results indicate that our individuals may have been overly hydrated, which would result in an overestimation of FFM both before and after weight loss. These results are also in agreement with Carella et al. (48), who also found that BIA overestimated FFM compared with hydrodensity weighing.

Correlation coefficients for the BodPod were highest for FM, with a slight increase in R2 using the equation of Siri. Otherwise, coefficients were highest with the equation of Brozek. One possibility is that an overestimation of thoracic volume may have led to the overestimation of percentage fat (28). Further research needs to be done to fully explain this discrepancy.

One final point to consider is that changes in FFM were small, and these changes were comparable with measurement errors of the various methods. We did have a wide range of weight loss over the group. Therefore, we used the median to divide the group into two subgroups: those who lost >8 kg and. those who lost <8 kg. Δ Percentage fat measured by ADP, using the equations of either Brozek or Siri, correlated better with DXA in the group who lost more weight than in the group who lost less (Brozek, r2 = 0.36 vs. r2 = 0.24). There was no difference in Δ FFM. Furthermore, although the correlation for FM did not improve in the group who lost more weight, the correlation was much lower in the group who lost less weight (r2 = 0.000). There was an improvement in the correlation with Δ percentage fat (r2 = 0.70 vs. r2 = 0.57) and FFM (r2 = 0.58 vs. r 2 = 0.44) in the group who lost more weight using tetrapolar BIA. These correlations were nonsignificant in the group who lost less weight. Finally, there was no improvement with any of the measures between DXA and Tanita when the groups were split. This supports the idea that these measures may be more accurate with larger changes in weight loss. Similar results were also found when comparing DXA to potassium counting and deuterium dilution (39).

In conclusion, both BIA and ADP are accurate compared with DXA when assessing body composition. However, these results, along with previous studies including athletes and leaner individuals, indicate that these methods may be less reliable with increasing body fat, which may have to do with hydration status. Overall, the correlation coefficients between each of the methods for Δ body composition compared with DXA were high. In addition, each of the methods was sensitive enough to detect changes with weight loss. Brozek was slightly more accurate in predicting changes with weight loss compared with Siri. BIA using the hand/foot method was also more accurate than the Tanita scale method. However, it would not be beneficial to use these methods interchangeably to detect body compositional changes over weight loss.


We thank the Pennington clinical staff for data collection and Mary Beth Burnett for help in preparing the manuscript.

Funding for the study was provided by Pennington Biomedical Research Center.


  • 1

    Nonstandard abbreviations: FM, fat mass; FFM, fat-free mass; UWW, underwater weighing; ADP, air displacement plethysmography; BIA, bioelectrical impedance analysis; BD, body density.