Validation of air displacementplethysmography for assessingbody composition

DALE R. WAGNER, VIVIAN H. HEYWARD, and ANN L. GIBSON

Exercise and Sports Science Department, Vanguard University of Southern California, Costa Mesa, CA 92626; andCenter for Exercise and Applied Human Physiology, University of New Mexico, Albuquerque, NM 87131

ABSTRACT

WAGNER, D. R., V. H. HEYWARD, and A. L. GIBSON. Validation of air displacement plethysmography for assessing bodycomposition.Med. Sci. Sports Exerc.,Vol. 32, No. 7, pp. 1339–1344, 2000.Purpose:The purpose of this study was to verify the validityof an air displacement plethysmography device (Bod Pod®) for estimating body density (Db).Methods: The Db from the Bod Pod(DbBP) was compared with the Db from hydrostatic weighing (DbHW) at residual lung volume in a heterogeneous sample of 30 blackmen who varied in age (32.06 7.7 yr), height (180.36 7.5 cm), body mass (84.26 15.0 kg), body fatness (16.16 7.5%), andself-reported physical activity level and socioeconomic status. The Db for each method was converted to relative body fat (%BF) usingrace-specific conversion formulas and subsequently compared with %BF obtained from dual-energy x-ray absorptiometry (%BFDXA).Results: Linear regression, using DbHW as the dependent variable and DbBP as the predictor, produced an R2 5 0.84 and SEE50.00721 gzcc21. However, the mean difference between the two methods (0.004506 0.00718 gzcc21) was significant (P , 0.01). TheBod Pod underestimated the Db of 73% of the sample. The %BF estimates from the Bod Pod, HW, and DXA differed significantly(P , 0.01). The average %BFBP (17.76 7.4%) was significantly greater than %BFHW (15.86 7.5%) and %BFDXA (16.16 7.5%);however, there was no significant difference between %BFHW and %BFDXA. Conclusion:The Bod Pod significantly and systematicallyunderestimated Db, resulting in an overestimation of %BF. More cross-validation research is needed before recommending the BodPod as a reference method.Key Words: HYDRODENSITOMETRY, DUAL-ENERGY X-RAY ABSORPTIOMETRY, BODYDENSITY, BODY FAT, BLACK, AFRICAN-AMERICAN

Multicomponent models, combining the body den-sity (Db), total body water, and total body mineraldata obtained from several different laboratory

procedures, are frequently used to derive criterion measuresof body composition. However, this approach is costly, timeconsuming, and often not practical. Thus, many researchersand clinicians continue to rely on a two-component modelthat simply separates the body into fat and fat-free compo-nents to assess body composition. Commonly, the two-component model involves measuring Db and then using aconversion formula to estimate relative body fat (%BF).Regardless of which body composition model is used, mul-ticomponent or two-component, an accurate measurementof Db is critical to obtain valid and reliable estimates of%BF.

Db is the ratio of body mass to body volume (Vb).Traditionally, hydrostatic weighing (HW) has been used to

determine Vb. This technique, which was pioneered byBehnke et al (2)., is based on Archimedes principle—a bodyimmersed in a fluid is buoyed by a force equal to the weightof the displaced fluid.

The HW procedure requires the subject to perform themaneuver of maximally exhaling while completely sub-merged under water. Additionally, the subject needs toremain relatively motionless underwater in order for atrained technician to get an accurate reading of the subject’sunderwater weight. This procedure, often considered un-pleasant by some subjects, must be repeated multiple timesfor a valid and reliable reading. Furthermore, to ensure anaccurate measure of Vb from HW, residual lung volume(VR) must also be measured. Thus, HW is time-consuming,requires a skilled technician, and is difficult, and sometimesimpossible, for some subjects to perform.

An alternative to HW is the Bod Pod® (Life MeasurementInstruments, Concord, CA), a large, egg-shaped, fiberglasschamber. Based on air displacement plethysmography, theBod Pod uses a pressure-volume relationship to derive Vbfor a subject seated inside the chamber. The Vb is equal tothe volume of air in an empty chamber minus the volume ofair remaining in the chamber after the subject enters thechamber. The physical design and the operating principlesof the Bod Pod have been described in detail elsewhere (5).

0195-9131/00/3207-1339/0MEDICINE & SCIENCE IN SPORTS & EXERCISE®Copyright © 2000 by the American College of Sports Medicine

Submitted for publication May 1999.Accepted for publication September 1999.

1339

The Bod Pod method is much easier and more accom-modating for subjects than HW, thereby potentially reduc-ing subject error. Furthermore, it is a faster, more conve-nient, and easier test to administer, thereby reducingpotential error associated with technician skill. Thus, theBod Pod appears to be a promising tool for the determina-tion of Vb and Db, and a method that could potentiallyreplace HW.

Although this device has been extensively advertised andcommercially available for several years, there remains apaucity of research on the Bod Pod. The published researchis limited to only one study that demonstrated its validityand reliability for measuring the volume of inanimate ob-jects (5), and one study that verified that it was a reliable andvalid tool for estimating %BF in adult humans (15). Thus,the purpose of this study was to cross-validate Db measuredby the Bod Pod to that obtained from HW. Additionally, the%BF values estimated from the Bod Pod and HW werecompared with those obtained from dual-energy x-ray ab-sorptiometry (DXA), a method that estimates %BF inde-pendent of Db.

METHODS

Subjects. Thirty black men, ages 19–45 yr, volunteeredto participate in this study. The sample was limited to blackmen as this was a substudy of a larger research study doneto cross-validate the body composition formulas availablefor black men (23). Subjects were recruited from the Albu-querque area via word of mouth, newspaper advertisements,and flyers distributed throughout the university and tohealth/fitness clubs, major corporations, churches, commu-nity centers, and at public events. The subject pool washeterogeneous with respect to age, height, body mass, bodyfatness, physical activity level, and socioeconomic status.All subjects underwent a physical examination by a physi-cian and were deemed to be in good health. The subjectsstayed overnight in the clinical research ward of the hospitalto control for the influence of physical activity and food anddrug intake on body composition measures. The participantswere informed of the purpose, procedures, risks, and bene-fits of the study before signing an informed consent. TheHuman Research Review Committees of the UniversityHospital and the University of New Mexico approved thisstudy.

Body mass and height. With the men in briefs only,body mass was measured twice to the nearest 0.01 kg on acalibrated electronic scale (Bod Pod, Life MeasurementInstruments, Concord, CA), and the average of the twovalues was used for subsequent calculations. Height wasmeasured twice to the nearest 0.1 cm using a stadiometer(Holtain Ltd., Crymych, Dyfed Wales), with the average ofthe two measurements recorded. This measurement wastaken with subjects at mid-inspiration, standing erect, andarms hanging freely at the sides.

Dual-energy x-ray absorptiometry. Relative bodyfat was determined by scanning the subjects with DXA

(Lunar DPX, Lunar Radiation Corp., Madison, WI). TheDXA was calibrated daily with the manufacturer’s “stan-dard block” bone simulating substance of known composi-tion and attenuating capacity and weekly with an aluminumspine phantom. A licensed radiological technologist per-formed all scans. From a supine position, the men werescanned using the medium speed scan mode (20 min) unlesstheir body thickness exceeded 27 cm. In such cases, the slowspeed scan (40 min) was used. Lunar software (version3.6Z) provided an estimate of %BF based on an extrapola-tion of fatness from the ratio of soft tissue attenuation of twox-ray energies in pixels not containing bone (14).

Bod Pod. Before each test, the Bod Pod was calibratedaccording to the manufacturer’s instructions with the cham-ber empty using a cylinder of known volume (50 L). Thesubject, in briefs and swim cap only, then entered and sat inthe fiberglass chamber. The Bod Pod was sealed, and thesubject breathed normally for 20 s while Vb was measured.After this, the subject was connected to a breathing tubeinternal to the system to measure thoracic gas volume(VTG). The subject resumed tidal breathing through the tube.After two or three breathing cycles, a valve in the circuitmomentarily occluded the airway. At this point, the subjectgently “puffed” by alternately contracting and relaxing thediaphragm. This effort produced small pressure fluctuationsin the airway and chamber that were used to determine VTG.This value was used to correct Vb for VTG. This was thesame method used by previous Bod Pod researchers (15,16).

Hydrodensitometry. The Vb and Db were measuredusing HW at VR. The VR was measured via closed-circuitspirometry with a 9-L Collins Helium Analyzer (Warren E.Collins, Inc., Braintree, MA) using a helium dilution tech-nique (17) with the participant seated out of the water. Aminimum of three trials was done with the two closestreadings within 100 mL being averaged and used to correctVb measured from HW. To minimize the error, the sameexperienced technician conducted all VR tests.

A load cell system (Precision Biomedical, State College,PA), integrated to an analog signal acquisition system(Biopac Systems, Inc., Goleta, CA) and a personal com-puter, was used to measure underwater weight. The analogsignal was acquired at 200 Hz and processed using com-mercially available software (AcqKnowledge; Biopac Sys-tems, Inc., Goleta, CA). For this assessment, the subjectassumed a hands and knees position on the load cell plat-form. A forced expiratory reserve volume maneuver wasperformed as the subject lowered his head below the surfaceof the water to become completely submerged. The under-water weight was obtained by averaging the flattest regionof the waveform over approximately 1 s after completeexpiration. Three trials within 100 g were averaged and usedas the underwater weight for calculation of Vb and Db (4).

Statistical analysis. The Db from the Bod Pod wascompared to the Db from HW at VR using cross-validationcriteria developed by Lohman (13). The criteria for accept-ing the predictive accuracy of the Bod Pod included: (a) nosignificant difference between average DbBP and DbHW, (b)a substantial correlation (ry,y9 $ 0.80) and shared variance

1340 Official Journal of the American College of Sports Medicine http://www.msse.org

(r2) between DbBP and DbHW, (c) a standard error of esti-mate (SEE) and total error (E) of# 0.0080 gzcc21, (d) anonsignificant correlation (ry,res) between the DbHW and theresidual scores (e.g., DbHW – DbBP), and (e) the slope andintercept of the lines of best fit not differing significantlyfrom the line of identity. Additionally, residual scores (e.g.,DbHW –DbBP) were analyzed using the Bland and Altman(3) method to determine the percentage of subjects whoseBod Pod-derived Db was estimated within6 0.0080 gzcc21

of DbHW. For this analysis, the Db from both methods wasaveraged and plotted against each subject’s residual score.

After the analysis of Db, the Db data were converted to%BF using the race-specific conversion formulas of Schutteet al. (20) and Wagner (23) so that the densitometric meth-ods could be compared with DXA. The Schutte et al. for-mula has customarily been used to convert Db to %BF forblack men. However, Wagner recently showed that thefat-free body density of black men was 1.10570 gzcc21

rather than the 1.113 gzcc21 estimated by Schutte et al.Thus, %BF was estimated from both conversion formulasand compared with %BFDXA using a one-way repeatedmeasures ANOVA. The alpha level for significance was setat P # 0.05. After a significant main effect, Tukey’sposthoc tests were used to determine which methods differedsignificantly.

The statistical analyses employed in this study assumethat all the variables are normally distributed. Extreme caseswill have a deleterious effect on regression solutions, andcross-validation procedures and should be excluded fromthe analysis (22). Therefore, before data analysis, the nor-mality of the distributions was examined by inspecting theskewness and kurtosis of the variables. Statistical outlierswere defined as individual scores exceeding6 3.29 SDfrom the mean for that variable (22). All variables werenormally distributed with no potential outliers (z-scores,3.29); therefore, all subjects were included in subsequentanalyses. Data were analyzed using the Statistical Packagefor the Social Sciences (SPSS version 8.0 for Windows).

RESULTS

Descriptive characteristics. The physical character-istics of the sample (N 5 30) are presented in Table 1. Asevidenced by the data in the table, the sample was diverse.

Db comparison. The Db from the Bod Pod and HWwere highly correlated (r5 0.91,P , 0.01). This relation-ship is depicted in Figure 1. Linear regression produced anR2 5 0.84, SEE5 0.00721 gzcc21, and E5 0.00837 gzcc21.The correlation between the reference measure (DbHW) andresidual scores (DbHW –DbBP) was not significant (ry,res 50.25,P . 0.05). The slope did not differ significantly fromone (0.93,P . 0.05), and the intercept was not significantlydifferent from zero (0.08,P . 0.05).

Although the mean difference in Db between HW and theBod Pod was only 0.00450 gzcc21, the pairedt-test revealedthat the DbBP (1.062946 0.0171 gzcc21) was significantlyless (P , 0.01) than DbHW (1.067446 0.0174 gzcc21).Furthermore, the Bland and Altman plot of residual scoresshowed that the Bod Pod accurately estimated the Db of67% of the subjects within6 0.0080 gzcc21 of HW; how-ever, 73% of the subjects had a DbBP that was less thanDbHW (Fig. 2).

%BF comparison. The Schutte et al. (20) and Wagner(23) conversion formulas produced significantly different(P , 0.01) estimates of %BF. However, regardless of whichformula (Schutte et al. (20) or Wagner (23)) was used toconvert Db to %BF, the ANOVA indicated a significantoverall main effect for testing method (P , 0.01). Althoughaverage %BFHW (Wagner5 15.86 7.5%BF; Schutte et al.5 17.16 6.8%BF) did not differ significantly (P . 0.05)from %BFDXA (16.16 7.5%BF), the mean %BFBP (Wag-ner 5 17.7 6 7.4%BF; Schutte et al.5 18.8 6 6.7%BF)was significantly greater than %BF estimates from HW andDXA (P , 0.01). The observed statistical power for theANOVA was 95.6% with an eta-squared value of 22.5%.The relationship between the %BFDXA and the %BF fromHW and the Bod Pod is represented in Figure 3.

Figure 1—Relationship between the body density (Db) obtained fromhydrostatic weighing (HW) and that obtained from the Bod Pod.Solid lineis line of identity, and dashed lineis regression line (r 5 0.91, SEE50.00721 gzcc21

).

TABLE 1. Physical characteristics of the sample (N 5 30).

Variable Mean 6 SD Minimum Maximum

Age (yr) 31.97 6 7.71 19 45HT (cm) 180.32 6 7.47 157.0 194.6BM (kg) 84.15 6 14.98 56.28 129.45BMI (kgzm22) 25.85 6 4.24 18.91 40.45DbHW (gzcc21) 1.0674 6 0.0174 1.0141 1.0940DbBP (gzcc21) 1.0629 6 0.0171 1.0159 1.0925%BFDXA 16.07 6 7.55 5.00 35.50VR (L) 1.76 6 0.41 0.98 2.56VTG (L) 3.99 6 0.87 1.86 6.17

HT, height; BM, body mass; BMI, body mass index; DbHW, body density as measuredby hydrostatic weighing; DbBP, body density as measured by the Bod Pod; %BFDXA,relative body fat as measured by dual-energy x-ray absorptiometry; VR, residual lungvolume as measured by helium dilution; and VTG, thoracic gas volume as measured bythe Bod Pod.

CROSS-VALIDATION OF BOD POD Medicine & Science in Sports & ExerciseT 1341

DISCUSSION

Although the correlation between DbHW and DbBP washigh and the regression data (SEE, slope, and intercept) metthe cross-validation criteria, the Bod Pod underestimated theaverage Db of this sample by a small, but significant,amount (0.004506 0.00718 gzcc21). To date, there has onlybeen one published study comparing human subjects in theBod Pod with HW (15). In their study, Db data were notreported, rather it was converted to %BF using the Siriformula (21); therefore, data were expressed in terms of%BF rather than Db. Using the Siri formula (21) and themean %BF values reported for their entire sample, weestimated the mean difference between DbHW and DbBP inthe McCrory et al. (15) study to be only 0.00044 gzcc21.

We converted the Db obtained in our study to %BF usingthe race-specific conversion formulas of Schutte et al. (20)and Wagner (23). Regardless of which conversion formulawas used, the Bod Pod significantly overestimated the mean%BFHW of this sample (1.73% BF using Schutte et al. and1.92% BF using Wagner). This result differs greatly fromthe small, nonsignificant mean underestimation of 0.3% BFfor the Bod Pod compared with HW reported in the Mc-Crory et al. (15) study.

McCrory et al. (15) reported that the Bod Pod estimatedthe %BF of 75% of their subjects within6 2.0% BF. Theplot of the residual scores in our study (Fig. 2) also foundgood individual agreement between methods with the Db oftwo-thirds of our sample estimated within6 0.0080 gzcc21.However, unlike the results of McCrory et al., close inspec-tion of our Bland and Altman (3) plot revealed that 73% ofthe residual scores were positive; thus, there was a system-atic tendency for the Bod Pod to consistently underestimateDb in the present study.

HW is not an error-free reference measure, and the pro-cedure is certainly more difficult to perform than what isrequired to obtain Bod Pod measurements. Therefore, itcould be argued that the difference between methods inmeasuring Db might be due more to measurement errorsassociated with the HW procedure than errors from the BodPod. In fact, McCrory et al. (15) reported a smaller, althoughnot statistically significant, between-trial coefficient of vari-ation for the Bod Pod (1.76 1.1%) compared with HW(2.3 6 1.9%).

To test which method (HW or Bod Pod) had greatervalidity, the Dbs from both methods were converted to %BFand compared with %BFDXA. DXA was selected as thecomparison method because it yields an estimate of %BFthat is completely independent of Db. DXA has been shownto be an accurate method to estimate %BF (8,9,12,18). Onaverage, the difference between %BFDXA and %BFHW wasonly 0.25% BF (P . 0.05) compared with21.67% BF forthe Bod Pod (P , 0.01). Additionally, the shared variance(r2 5 0.90) was greater and the standard error of estimatewas lower (SEE5 2.47% BF) for %BFHW than for %BFBP

(r2 5 0.86, SEE5 2.84% BF) when the two methods wereregressed on %BFDXA. These data suggest that HW pro-vides a more accurate measure of Db and subsequent esti-mation of %BF than the Bod Pod.

There are several potential reasons why the Bod Podunderestimated average Db in this sample. Close scrutiny ofthe Bod Pod formula for estimating Db is one item toconsider when searching for potential sources of error:

Db 5 BM/(Vbraw 1 0.40 VTG 2 SAA)

where BM is body mass in air, Vbraw is the raw, uncorrectedbody volume, and SAA is the surface area artifact (5). Withproper calibration, errors in BM and Vbraw should be neg-ligible. Thus, if there is an error in the Bod Pod measure-

Figure 2—Analysis of the individual residual body density (Db) scoresfor the Bod Pod. An SEE within 6 0.0080 gzcc21

is considered “good”when evaluating the predictive accuracy of Db measurements (13).

Figure 3—Relationship between the relative body fat (%BF) esti-mated from dual-energy x-ray absorptiometry (DXA) and that esti-mated from densitometry. %BF for each densitometry method (HWand Bod Pod) was estimated using a race-specific conversion formula(20).

1342 Official Journal of the American College of Sports Medicine http://www.msse.org

ment, it most likely resides in the estimates and assumptionsassociated with the SAA and VTG.

The SAA is the product of body surface area, as estimatedby the DuBois formula (6), and a constant derived from thetesting of plastic films and aluminum foil (5). Surfacesproduce apparent negative volumes when measured in theBod Pod due to the fact that air under isothermal conditionsis more compressible than air under adiabatic conditions;thus, SAA must be considered. The constant in the SAAformula was derived from a regression equation createdfrom measuring the surface area of aluminum or plasticfilms and then recording the apparent volume measured bythe Bod Pod (personal communication, Life MeasurementInstruments).

The VTG, which includes the volume of air in the lungsand any air trapped in the thorax, is measured when thesubject puffs against a closed airway. This maneuver pro-duces pressure changes that, when correlated to tidal breath-ing, yields VTG. It is assumed that Vb needs to be increasedby 40% of VTG to account for the difference betweenisothermal air in the thorax and adiabatic air in the Bod Podchamber (5).

When measured at the end of an exhalation, VTG shouldbe equal to functional residual capacity (FRC) (19). FRC isthe sum of expiratory reserve volume (ERV) and VR. Thus,the VTG from the Bod Pod should equal the sum of the ERVand VR from gas dilution. One potential explanation for theoverestimation of %BFBP seen in this study is that the BodPod overestimated VTG. Unfortunately, the Bod Pod is pro-grammed to occlude the airway and signal the subject tocommence the puffing maneuver at the midpoint of exha-lation rather than at the end of an exhalation, making itdifficult to compare VTG from the Bod Pod to FRC from gasdilution.

Given that VTG values are more than double VR values,one might assume that VTG errors could produce large errorsin the determination of Db. However, as McCrory et al. (16)pointed out, only 40% of VTG enters the Bod Pod formulaused to calculate Db, whereas 100% of VR enters into thecalculation of Db by HW. The largest source of error inmeasuring Db via HW is the measurement of VR (1). Thus,an error in the VR measurement during the original valida-tion of the Bod Pod must also be considered as a potentialreason for the differences in DbHW and DbBP observed in thepresent study.

An underestimation of VTG will result in an overestima-tion of Db and an underestimation of %BF. Conversely, anunderestimation of VR will yield an underestimation of Dband an overestimation of %BF. The subjects in this studywere of similar age (32.06 7.7 yr) and stature (180.36 7.5cm) to the men in the McCrory et al. (15) Bod Pod valida-tion study (38.66 9.1 y and 179.76 5.3 cm). The averageVTG was also similar (8.7% difference; 3.996 0.87 L to3.676 0.86 L), but there was a greater difference (15%) inthe average VR (1.766 0.41 L to 1.536 0.35 L). Given theage and stature of the subjects, the mean VR value for themen in the McCrory et al. study seems low and is consid-erably less than what would be estimated from a VR pre-

diction equation (11). Thus, if the VR was underestimated inthe McCrory et al. validation study, then the %BF from HWwould have been overestimated in that study. Increasing theVR in the McCrory et al. study would result in lowering their%BF from HW and produce an outcome consistent with ourresults: a small but systematic overestimation of %BF fromthe Bod Pod.

Variations in the measurement of VR could potentiallyaffect the results. For example, Girandola et al. (10) reportedthat there was an increase of 6.7% in VR, resulting in a 0.6%reduction in %BF when VR was measured in the waterinstead of outside the tank. However, both the present studyand that of McCrory et al. (15) measured VR outside of thetank, making this a nonissue for this comparison. Also,McCrory et al. (15) used oxygen dilution to measure VR,whereas we used a helium dilution procedure, but researchshows that both techniques produce comparable estimates ofVR (7,17).

Because the Bod Pod is easy to use, requires little efforton the part of the subject, and is a quick and convenientmethod of estimating Db and %BF, it is tempting to con-sider this method as a replacement for the cumbersome HWmethod. Previous research (15) found the Bod Pod to behighly reliable and valid compared with HW; thus, it hasbeen widely marketed as an alternative to HW. However,our results showed a small, but significant, underestimationof Db, resulting in an overestimation of %BF using thisdevice. Furthermore, the underestimation of Db was sys-tematic. Using DbHW as the criterion, we recommend thefollowing correction to DbBP: 0.932 (DbBP) 1 0.077. Thus,although the Bod Pod may be a promising replacement forHW in the future, the results from this study raise someconcern about its validity and use as a reference method.Certainly, more research needs to be conducted before theBod Pod can be accepted as a valid reference method.

We speculate that an overestimation of VTG by the BodPod and/or an underestimation of VR during the validationstudy of McCrory et al. (15) might have contributed to theoverestimation of %BFBP seen in this study. If the Bod Podcould be reprogrammed to signal the puffing maneuver andocclude the airway at the end of an exhalation instead ofduring mid-exhalation and if researchers record ERV aswell as VR, comparisons between FRC from gas dilution toVTG from the Bod Pod can be made. This may help toquantify and clarify small differences between the Bod Podand HW. We also recommend taking the VR measurementwith the subject seated out of the water to minimize differ-ences to the original validation study of McCrory et al. (15).Finally, because DXA estimates %BF independent of Db,we recommend including it as an “external check” whencomparing the Bod Pod and HW as was done in this study.

This research was supported by the General Clinical ResearchCenter at the University of New Mexico (NIH NCRR Grant 5 M01RR00997), the Office of Graduate Studies (RPT Grant), and theGraduate and Professional Students Association (SRAC Grant). Wethank Dr. David James for conducting prescreening physicals, andthe GCRC and Clinical Nutrition Laboratory personnel for their as-sistance in data collection.

CROSS-VALIDATION OF BOD POD Medicine & Science in Sports & ExerciseT 1343

Address for correspondence: Dale R. Wagner, Ph.D., Exer-cise and Sports Science Department, Vanguard University

of Southern California, 55 Fair Dr., Costa Mesa, CA 92626;E-mail: DWagner@ vanguard.edu.

REFERENCES

1. AKERS, R., and E. R. BUSKIRK. An underwater weighing systemutilizing “force cube” transducers. J. Appl. Physiol. 26:649–652,1969.

2. BEHNKE, A. R., B. G. FEEN, and W. C. WELHAM. The specificgravity of healthy men. JAMA118:495–498, 1942.

3. BLAND, J. M., and D. G. ALTMAN . Statistical methods for assessingagreement between two methods of clinical measurement. Lancet307–310:1986.

4. BONGE, D., and J. E. DONNELLY. Trials to criteria for hydrostaticweighing at residual volume. Res. Q. Exerc. Sport60:176–179,1989.

5. DEMPSTER, P., and S. AITKENS. A new air displacement method forthe determination of human body composition. Med. Sci. SportsExerc. 27:1692–1697, 1995.

6. DUBOIS, D., and E. F. DUBOIS. A formula to estimate the approx-imate surface area if height and weight be known. Arch. Intern.Med. 17:863–871, 1916.

7. FORSYTH, R., M. J. PLYLEY , and R. J. SHEPHARD. Residual volumeas a tool in body fat prediction. Ann. Nutr. Metab. 32:62–67, 1988.

8. FRIEDL, K. E., J. P. DELUCA, L. J. MARCHITELLI, and J. A. VOGEL.Reliability of body-fat estimations from a four-component modelby using density, body water, and bone mineral measurements.Am. J. Clin. Nutr. 55:764–770, 1992.

9. FULLER, N. J., S. A. JEBB, M. A. LASKEY, W. A. COWARD, and M.ELIA . Four-component model for the assessment of body compo-sition in humans: comparison with alternative methods, and eval-uation of the density and hydration of fat-free mass. Clin. Sci.82:687–693, 1992.

10. GIRANDOLA, R. N., R. A. WISWELL, J. G. MOHLER, G. T. ROMERO,and W. S. BARNES. Effects of water immersion on lung volumes:implications for body composition analysis. J. Appl. Physiol. 43:276–279, 1977.

11. GOLDMAN, H. I., and M. R. BECKLAKE. Respiratory function tests:normal values at medium altitudes and the prediction of normalresults. Am. Rev. Tuberculosis Resp. Dis. 79:457–467, 1959.

12. KOHRT, W. M. Preliminary evidence that DEXA provides anaccurate assessment of body composition. J. Appl. Physiol. 84:372–377, 1998.

13. LOHMAN, T. G. Advances in Body Composition Assessment: Cur-rent Issues in Exercise Science Series. Champaign, IL: HumanKinetics, 1992, p. 44.

14. MAZESS, R. B., H. S. BARDEN, J. P. BISEK, and J. HANSON. Dual-energy x-ray absorptiometry for total-body and regional bone-mineral and soft-tissue composition. Am. J. Clin. Nutr. 51:1106–1112, 1990.

15. MCCRORY, M. A., T. D. GOMEZ, E. M. BERNAUER, and P. A. MOLE.Evaluation of a new air displacement plethysmograph for mea-suring human body composition. Med. Sci. Sports Exerc. 27:1686–1691, 1995.

16. MCCRORY, M. A., P. A. MOLE, T. D. GOMEZ, K. G. DEWEY, andE. M. BERNAUER. Body composition by air-displacement plethys-mography by using predicted and measured thoracic gas volumes.J. Appl. Physiol. 84:1475–1479, 1998.

17. MOTLEY, H. L. Comparison of a simple helium closed method withoxygen circuit method for measuring residual air. Am. Rev. Tu-berculosis Pulm. Dis. 76:601–615, 1957.

18. PRIOR, B. M., K. J. CURETON, C. M. MODLESKY, et al. In vivovalidation of whole body composition estimates from dual-energyx-ray absorptiometry. J. Appl. Physiol. 83:623–630, 1997.

19. RUPPELL, G. Manual of Pulmonary Function Testing.St. Louis,MO: Mosby, 1994, pp. 11–15.

20. SCHUTTE, J. E., E. J. TOWNSEND, J. HUGG, R. F. SHOUP, R. M.MALINA , and C. G. BLOMQVIST. Density of lean body mass isgreater in Blacks than in Whites. J. Appl. Physiol. 56:1647–1649,1984.

21. SIRI, W. E. Body composition from fluid spaces and density:analysis of methods. In:Techniques for Measuring Body Compo-sition, J. Brozek and A. Henschel (Eds.). Washington, DC: Na-tional Academy of Sciences, 1961, pp. 223–244.

22. TABACHNICK, B. G., and L. S. FIDELL. Using Multivariate Statistics.New York: Harper Collins, 1996, pp. 66–70.

23. WAGNER, D. R. Cross-validation of body composition equationsfor black men using a multicomponent model. Doctoral Disserta-tion. University of New Mexico, Dept. of Phys. Performance andDevelopment. Albuquerque, NM, December 1997.

1344 Official Journal of the American College of Sports Medicine http://www.msse.org