Several methods are currently available to assess body composition in humans, but densitometry has been the most widely used (12). This method is based on Archimedes' principle and the assumption that there is a constant density value for the fat and the fat-free components of the body (i.e., two-compartment model). This two-compartment model, however, does not take into account individual differences in the water, mineral, and protein content of the fat-free component, which may invalidate the assumed constant density of the fat-free component. To address this limitation, several multicompartment models have been developed that should provide an increased level of accuracy versus the two-compartment model, which has been reviewed elsewhere (10). In densitometric methods, body mass (BM) and body volume (BV) are measured, and body density (BD) is calculated (BD = BM/BV). The resultant BD is then used in a population-specific equation (e.g., Siri or Brozek) to estimate body composition. Historically, BV has been measured routinely using hydrodensitometry. Although this method yields good results, it is both time consuming and physically demanding on the subjects. Also, many subject populations exist for which hydrodensitometry is not a viable option for practical reasons (e.g., subjects who are unable or unwilling to be submerged). Alternatively, BV can be measured using air displacement plethysmography (ADP). Although this method has been used for nearly a century, only recently has a viable method become commercially available (8). This system is known by the trade name Bod Pod® (Life Measurement, Inc, Concord, CA). Compared with hydrodensitometry, the Bod Pod® offers a much quicker assessment that is far less demanding on the subjects and can be safely used in virtually any adult subject population. Since its introduction in the mid-1990s, several studies have tested the validity of ADP using established methods of body composition assessment, with mixed results. Whereas some studies have found agreement between the Bod Pod® and various other established methods (11,13,25), others have suggested that the Bod Pod® yields biased results (6,7,15). Although it is imperative that the validity of the Bod Pod® be clearly established, there are many situations when the reliability of the test is equally important (e.g., when monitoring the effects of a treatment program on body composition changes over time). To date, only a few studies have examined the reliability of the Bod Pod®, and most of these studies have used small, homogeneous samples. Therefore, this investigation was conducted to examine the reliability of the Bod Pod® in a large, heterogeneous sample. Additionally, an attempt was made to determine whether subject characteristics (e.g., age, gender, and percent body fat (% BF)) influence the reliability of the Bod Pod®. For example, wearing clothing will warm the layer of air immediately above the skin and make it more compressible, which, in turn, will cause BV to be underestimated (8). In addition to affecting the validity of the measure, warming the layer of air around the skin with clothing can also decrease reliability of the Bod Pod® (20). It is conceivable that differences in lean mass and the resulting difference in resting metabolic rate could affect the temperature of the air layer immediately above the skin and influence the reliability of the BV measure.
It has been recommended that measures of reliability include both a relative and an absolute measure (22). The intraclass correlation coefficient (ICC) is a unitless relative measure of reliability; the closer the value is to 1.00, the more reliable the test (18). The standard error of measurement (SEM) is an estimate of the absolute measure of reliability that shares the same units as the test variable; the smaller the value, the better the reliability (3). To date, the use of the ICC and SEM in the assessment of the reliability of the Bod Pod® has been underutilized; this is the first report on the reliability of the Bod Pod® using these measures in a large heterogeneous sample.
Before all testing, approval for the study was obtained from the human research ethics review board at the University of Western Ontario. Written informed consent was obtained from the subjects before any testing. In total, 980 adults were studied (Table 1).
Air displacement plethysmography.
Body volume was measured via ADP, using the Bod Pod® and software version 1.69, as outlined by the manufacturer. These procedures have been described in detail previously (8). Briefly, before testing, the scale was calibrated using two 10-kg weights, and the Bod Pod® was calibrated using a cylinder of known volume. The subject's height was measured using a stadiometer, and each subject was weighed wearing only a tight-fitting swimsuit or undergarments and an acrylic swim cap. Subjects sat in the chamber and BV measurements were taken. This measurement was done in duplicate, with each test lasting approximately 40 s. If both measures were within 150 mL of each other, the mean was taken and used in subsequent calculations. If the two measurements differed by > 150 mL, a third measurement was performed. If two of the three measurements were within 150 mL of each other, the mean of those two were taken and used, but if the three measurements were not within 150 mL of each other, the entire process, including the calibration steps, was repeated. The measured BV was adjusted for lung volume and body surface area artifact using prediction equations integral to the system software. This corrected BV was used in combination with the BM to determine BD (BD = BM/BV). The resultant BD was used in the Siri equation (19) to estimate body composition. All calculations were done using the system software. This entire procedure was then repeated immediately, for a total of two (N = 980) repeated measurements. The total time to collect all measurements was 30-40 min per subject.
Data were retrieved from the Bod Pod® program using Data Management Software version 1.2 (Life Measurement, Inc., Concord, CA). This program exports the data in a format that can then be imported into Microsoft Excel for manipulation. The Bod Pod® measures data to four decimal places (BM to 0.0001 kg and BV to 0.0001 L), but during normal operation, only two decimal places are displayed. The data retrieved using the data management software were analyzed using four decimal places, but for clarity and convention, data are presented here using fewer decimal places.
Once the data were imported into Excel, they were divided into subsets for further analysis. Although somewhat arbitrarily chosen, the subsets were assigned in an attempt to create groups of individuals that were clearly distinct in an attempt to determine whether the reliability of the measure is affected by obvious subject characteristics. The subsets chosen were subjects < 25 yr (N = 674), subjects > 50 yr (N = 135), all men (N = 548), men < 10% BF (N = 147), men > 30% BF (N = 60), all women (N = 432), women < 20% BF (N = 37), and women > 40% BF (N = 103). Characteristics of these subsets are summarized in Table 1.
Data were analyzed using SPSS version 11 (SPSS Inc., Chicago, IL). A Pearson's correlation between repeated measures was performed followed by a paired sample t-test to determine significant differences. Although neither test is generally considered a valid measure of reliability, they were included in the analysis so that our data can be readily compared with existing published studies. A two-way random effects model ICC was used as a measure of relative reliability, and the SEM was used as a measure of absolute reliability. SEM was calculated as SEM = SD√(1 − ICC). Additionally, the coefficient of variation for repeated measures (CV) and the technical error for a single measurement (TEM) were determined. CV was calculated as (SDd/X)×100, where SDd is the mean of the standard deviation of the repeated trials, and X is the mean of the repeated trials. TEM was calculated as TEM = √[∑d2/2n], where d is the difference between two trials, and n is the number of subjects measured (16). For all analysis, the alpha level was set at P < 0.05.
Reliability in entire sample.
A significant correlation was found between trial 1 and trial 2 for all variables (N = 980), with no significant differences over time for BD or % BF (Table 2). A significant reduction in BM was found from trial 1 to trial 2 (P = 0.001) as well as a tendency for BV to decrease from trial 1 to trial 2 (P = 0.08). The CV and TEM for % BF were 3.09 and 1.07%, respectively. The ICC for % BF was 0.996 (P = 0.001), and the SEM was 0.05% BF (Table 2).
Reliability of % BF in sample subsets.
A significant correlation was found for % BF between trial 1 and trial 2 for all subsets, with no significant differences over time (Table 3). The CV varied from a high of 12.30% for men < 10% BF to a low of 1.44% for subjects > 50 yr. TEM was much less variable, with a range of 0.80 to 1.54% BF for subjects > 50 yr and men >30% BF, respectively. ICC was significant for all subsets and varied from a high of 0.995 for women to a low of 0.902 for men < 10% BF. SEM varied from 0.24% BF in men < 10% BF to 0.05% BF in subjects > 50 yr (Table 3).
The Bod Pod® represents an attractive tool for measuring body composition in a variety of clinical, research, and commercial settings because of its ease of use and excellent subject compliance. Because the technique is relatively new, data addressing its reliability and validity are lacking. This study addressed the reliability of the Bod Pod® by using a large, heterogeneous sample. Regardless of the measure used, the Bod Pod® displayed good test-retest reliability. A significant correlation was seen between trial 1 and trial 2 and a nonsignificant paired Student's t-test for both BD and % BF. This is in agreement with several other studies that also found a significant test-retest correlation in much smaller and more homogenous sample groups (2,15,20,21). The high ICC (ICC = 0.9961, P = 0.001) and the low SEM (SEM = 0.05% BF) for % BF suggest good test-retest reliability. The lack of other studies that have reported these measures, however, prevents direct comparison of these data with previously published results. The observed CV for % BF of 3.09% for two trials on all subjects is similar to values reported in other studied that have ranged from 1.7 to 3.4% (11,13,14,17,21). However, the observed TEM of 1.07% for % BF in the present study is slightly greater than TEM reported in other studies, which have ranged from 0.4 to 0.99% BF (6,9,13,20,21,24). Although it is possible that differing sample sizes and the use of different, predominantly homogeneous samples, contributed to the range of reliability reported in the literature to date, a study by Collins et al. (5), which transported subjects to two different laboratories for measurements on the same day, suggests that there may be a difference in precision between different Bod Pod® units. However, arecent study that used two different Bod Pod® units in the same laboratory has questioned these results (1). This would suggest that Bod Pod® units themselves are similar in their precision, but the possibility remains that different laboratory environments (e.g., the proximity of the Bod Pod® unit to air vents) could have an impact on the precision of the unit. If this is true, then caution should be used when comparing data from different laboratories. Although the TEM of 1.07% found in the present study is greater than that found in other studies, it is similar to values reported for hydrodensitometry (9,23), which suggests the Bod Pod® is at least as reliable as a previously established method for determining BD.
Although the current data suggest that the Bod Pod® is a reliable instrument, the use of multiple trials may be beneficial in detecting and eliminating unexplained outliers that have been reported in several studies (4,14,24). In the present study, 32 of the 980 subjects (3%) had a difference in % BF between the two trials of at least 3% (~2× TEM), with one subject having a difference of 12%. Presently, no definitive reason is available to explain these outliers. One individual variant body volume measure could result from an artifact of the subjects' breathing pattern (e.g., a yawn or a sneeze) or a transient change in pressure within the testing room (e.g., an air conditioner turning on or an outside door being opened). A complete body composition measurement, however, requires two body volume measures that are within 150 mL of each other. Therefore, for these outliers to occur, whatever has changed-the subject, the environment, or the Bod Pod® itself-has to remain consistently different over the 3-5 min of measurement. From the present data, the underlying cause for this phenomenon cannot be determined. Regardless of the cause, these outliers contributed negatively to the measures of reliability in this study. When these subjects were removed from the database, the CV for % BF was reduced from 3.53 to 2.76%, and the TEM was reduced from 1.07 to 0.96% BF. Unless it can be determined how to eliminate these outliers, it is strongly advised that at least two repeated measures be performed to identify any outliers.
Analysis of the sample subsets failed to reveal a clear picture of differences in test-retest reliability between the different groups. For example, the CV for % BF for men < 10% BF was 12.3% compared with a CV of 1.91% for men > 30% BF. However, this apparent difference is largely a function of the smaller mean % BF value for the leaner individuals because the SDd was similar for both the lean (0.78) and the fatter men (0.70). Similarly, the ICC for men < 10% BF (0.902) was lower than the ICC for men > 30% BF (0.959), and the SEM (0.24% BF) was higher than the SEM (0.142% BF) for the men > 30% BF, suggesting that the Bod Pod® is not as reliable for lean populations. However, this might also be misleading because the ICC, and subsequently the SEM, are influenced by the amount of between-subject variability (22), and the men < 10% BF had substantially less between-subject variability (data not shown). The TEM shows that the Bod Pod® is actually more precise for men < 10% BF compared with men > 30% BF, which were 1.00 and 1.54% BF, respectively. In general, the ICC, SEM, and specifically the TEM show that the Bod Pod® is a reliable measure for all of the population subsets measured in this study, which is consistent with other studies that have found similar reliability in men and women (5,24) or in female athletes and sedentary women (2). It should be pointed out, however, that all the subjects in this study were adults, and thus these results should not be extrapolated to children who have a much smaller BV because it has been suggested that the precision of the Bod Pod® is compromised with volumes below 40 L (4).
To the best of our knowledge, this is the first study to show a significant reduction in BM with repeated trials. It is unknown whether other studies have failed to examine BM changes with serial measures or whether the large sample size used in the present study allowed us the statistical power to find a small significant difference. Although a significant reduction was seen in BM between repeated trials, no significant difference was noted in BD between trials because BV also decreased absolutely over time (P = 0.08).
In conclusion, using a large, heterogeneous sample, ADP using the Bod Pod® appears to be a reliable measure for determining body composition in adults. Furthermore, the Bod Pod® offers similar precision when comparing differing subsets of the population. Although the Bod Pod® is a reliable measure, there appears to be large discrepancy, for reasons largely undetermined, between repeated measures in about 3% of the tests. For this reason, it is strongly advised that multiple measures be performed to detect outliers. The small but significant BM loss seen with repeated measures, although interesting, does not influence the reliability of the body composition measure because it is accompanied by a decrease in BV.
Support for this study was provided by the Joe Weider Foundation.
1. Ball, S. D. Interdevice variability in percent fat estimates using the BOD POD. Eur. J. Clin. Nutr.
2. Ballard, T. P., L. Fafara, and M. D. Vukovich. Comparison of Bod Pod and DXA in female collegiate athletes. Med. Sci. Sports Exerc.
3. Bruton, A., J. H. Conway, and S. T. Holgate. Reliability: What is it, and how is it measured? Physiotherapy
4. Collins, A. L., and H. D. McCarthy. Evaluation of factors determining the precision of body composition
measurements by air displacement plethysmography. Eur. J. Clin. Nutr.
5. Collins, A. L., S. Saunders, H. D. McCarthy, J. E. Williams, and N. J. Fuller. Within- and between-laboratory precision in the measurement of body volume
using air displacement plethysmography and its effect on body composition
assessment. Int. J. Obes. Relat. Metab. Disord.
6. Collins, M. A., M. L. Millard-Stafford, P. B. Sparling, et al. Evaluation of the BOD POD for assessing body fat in collegiate football players. Med. Sci. Sports Exerc.
7. Demerath, E. W., S. S. Guo, W. C. Chumlea, B. Towne, A. F. Roche, and R. M. Siervogel. Comparison of percent body fat estimates using air displacement plethysmography and hydrodensitometry in adults and children. Int. J. Obes. Relat. Metab. Disord.
8. Dempster, P., and S. Aitkens. A new air displacement method for the determination of human body composition
. Med. Sci. Sports Exerc.
9. Dewit, O., N. J. Fuller, M. S. Fewtrell, M. Elia, and J. C. Wells. Whole body air displacement plethysmography compared with hydrodensitometry for body composition
analysis. Arch. Dis. Child.
10. Heyward, V. H. Evaluation of body composition
. Current issues. Sports Med.
11. Levenhagen, D. K., M. J. Borel, D. C. Welch, et al. A comparison of air displacement plethysmography with three other techniques to determine body fat in healthy adults. JPEN. J. Parenter. Enteral. Nutr.
12. McArdle, W. D., F. I. Katch, and V. L. Katch. Exercise Physiology
,: Williams & Wilkins, 1996, pp. 541-575.
13. McCrory, M. A., T. D. Gomez, E. M. Bernauer, and P. A. Mole. Evaluation of a new air displacement plethysmograph for measuring human body composition
. Med. Sci. Sports Exerc.
14. Miyatake, N., K. Nonaka, and M. Fujii. A new air displacement plethysmograph for the determination of Japanese body composition
. Diabetes Obes. Metab.
15. Radley, D., P. J. Gately, C. B. Cooke, S. Carroll, B. Oldroyd, and J. G. Truscott. Estimates of percentage body fat in young adolescents: a comparison of dual-energy x-ray absorptiometry and air displacement plethysmography. Eur. J. Clin. Nutr.
16. Sale, D. G. Testing Strength and Power. In: Physiological Testing of the High-Performance Athlete
, 2nd ed., J. D. MacDougall, H. A. Wenger, and H. J. Green. Champaign, IL: Human Kinetics Books, 1991, pp. 21-106.
17. Sardinha, L. B., T. G. Lohman, P. J. Teixeira, D. P. Guedes, and S. B. Going. Comparison of air displacement plethysmography with dual-energy x-ray absorptiometry and 3 field methods for estimating body composition
in middle-aged men. Am. J. Clin. Nutr.
18. Shrout, P. E., and J. L. Fleiss. Intraclass correlations: uses in assessing rater reliability. Psychol. Bull.
19. Siri, W. E. Body composition
from fluid spaces and density: analysis of methods. In: Techniques for Measuring Body Composition
, J. Brozek and A. Henschel (Eds.). National Research Council, Washington, DC: National Academy of Sciences, pp. 223-244, 1961.
20. Vescovi, J. D., S. L. Zimmerman, W. C. Miller, and B. Fernhall. Effects of clothing on accuracy and reliability of air displacement plethysmography. Med. Sci. Sports Exerc.
21. Vescovi, J. D., S. L. Zimmerman, W. C. Miller, L. Hildebrandt, R. L. Hammer, and B. Fernhall. Evaluation of the BOD POD for estimating percentage body fat in a heterogeneous group of adult humans. Eur. J. Appl. Physiol.
22. Weir, J. P. Quantifying test-retest reliability using the intraclass correlation coefficient and the SEM. J. Strength Cond. Res.
23. Wells, J. C., I. Douros, N. J. Fuller, M. Elia, and L. Dekker. Assessment of body volume
using three-dimensional photonic scanning. Ann. N. Y. Acad. Sci.
24. Wells, J. C., and N. J. Fuller. Precision of measurement and body size in whole-body air-displacement plethysmography. Int. J. Obes. Relat. Metab. Disord.
25. Yee, A. J., T. Fuerst, L. Salamone, et al. Calibration and validation of an air-displacement plethysmography method for estimating percentage body fat in an elderly population: a comparison among compartmental models. Am. J. Clin. Nutr.