Luís B Sardinha, Timothy G Lohman, Pedro J Teixeira, Dartagnan P Guedes, and Scott 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 1,2 Luís B Sardinha, Timothy G Lohman, Pedro J Teixeira, Dartagnan P Guedes, and Scott B Going ABSTRACT This study was designed to compare air displacement plethysmography with dual-energy X-ray absorptiometry (DXA) and 3 other field methods for estimation of body composition. Subjects were 62 healthy, white men aged 37.6 ± 2.9 y (weight: 81.8 ± 11.3 kg; height: 171.5 ± 4.9 cm). Body composition was also assessed by using body mass index, single-frequency bioelectrical impedance analysis, multifrequency bioelectrical impedance spectroscopy, and the skinfold-thickness equations of Jackson and Pollock and Durnin and Womersley. Percentage body fat (%BF) with the plethysmograph was 23.4 ± 7.0 and with DXA was 26.0 ± 7.4. The 2.6% mean difference was significant (P< 0.05). Total error was 3.7%BF. As assessed by multiple regression analysis, %BF with the plethysmograph, age, weight, and height yielded a DXAadjusted R 2 value of 89.5% fat and an SEE of 2.4% fat. All other models had higher SEEs and lower adjusted R 2 values: 4.3% and 66.5% for body mass index, 3.3% and 79.8% for bioelectrical impedance analysis, 3.6% and 76.2% for bioelectrical impedance spectroscopy, 3.7% and 74.55% for the equations of Jackson and Pollock, and 3.9% and 71.6% for the equations of Durnin and Womersley, respectively. The plethysmograph also predicted fat mass and fat-free mass more accurately than all other models, with a lower SEE and higher adjusted R 2 value. In conclusion, although %BF was systematically underestimated, body composition was closely estimated with air displacement plethysmography in middle-aged men. Am J Clin Nutr 1998;68:786 93. KEY WORDS Bioelectrical impedance analysis, BIA, bioelectrical impedance spectroscopy, dual-energy X-ray absorptiometry, DXA, air displacement plethysmography, percentage body fat, predictive accuracy, reliability, Portugal, men INTRODUCTION During the past decade, several new technologic developments have introduced alternative methods to determine percentage body fat (%BF), allowing for a more accurate and precise assessment of body composition. In line with early attempts to apply the plethysmometric method to measure body volume (1, 2), recent bioengineering technologies were used to develop a new portable air plethysmograph that measures body volume by air displacement (3). The accuracy, reliability, and linearity of the volume measurement of several objects within the range of body volume has been reported, with an overall mean percentage error between measured and actual volume of <0.1% and a mean CV for repeated trials of 0.026% (3). In a sample including both men and women, good agreement was found between fat assessed with the plethysmograph and underwater weighing, with an average difference of 0.3 ± 0.2%; results were more reliable with the plethysmograph than with underwater weighing (4). Using different methodologic approaches, these were the only 2 scientific studies that evaluated the usefulness of the plethysmograph for estimating body volume. This new, convenient technology enables the assessment of body density in several populations, such as children and obese, elderly, and handicapped persons. In these populations, an accurate assessment of body composition should be based on multicompartment models, accounting for differences in the contents of water, minerals, and proteins in fat-free mass (FFM) (5, 6). The quick and easy operation of the plethysmograph, which is particularly important in these populations, is of great interest to researchers of body composition as an alternative to underwater weighing because it overcomes some of the methodologic and technical constraints of traditional body-density assessment methods. Since its introduction 10 y ago, dual-energy X-ray absorptiometry (DXA) has been used widely to assess body composition in a variety of populations, and many scientific reports (7 13) and reviews (14 17) have analyzed its validity under different physiologic conditions. Although there is still some controversy about whether DXA should be considered a reference method (14, 17), its more criticized technologic assumption the influence of variation in hydration on %BF should introduce minimal error. A change of 1 kg in extracellular fluid induces an estimation error of only 0.6% fat, which is less than one-half the error associated with underwater weighing for the same change in hydration (18). When compared with underwater weighing (10, 13), magnetic 1 From the Exercise and Health Laboratory, Faculty of Human Movement, Technical University of Lisbon, and the Department of Physiology, Body Composition Research Laboratory, University of Arizona, Tucson. 2 Address reprint requests to TG Lohman, Department of Physiology, 101 Ina Gittings Building, University of Arizona, Tucson, AZ. Email: Lohman@u.arizona.edu. Received November 18, 1997. Accepted for publication April 1, 1998. 786 Am J Clin Nutr 1998;68:786 93. Printed in USA. 1998 American Society for Clinical Nutrition

VALIDITY OF AIR DISPLACEMENT PLETHYSMOGRAPHY 787 resonance imaging (19), chemical analysis (9, 12, 20, 21), neutron-activation analysis (8), and different 4-component models (5, 22, 23), DXA has been shown to be a precise and accurate method for the assessment of soft tissue body composition (12). Considering the different calibrating materials and assumptions regarding bone-edge detection and the nonuniformity of soft tissues overlying bone (24), it is necessary to recognize that absolute and relative estimates of composition will vary between manufacturers and software versions (15). The validity of the new air displacement plethysmograph needs to be established in several populations. The purpose of this study was, therefore, to further assess the precision of the plethysmograph as well as its respective accuracy compared with DXA, a more established method of assessing body composition, and other field methods. SUBJECTS AND METHODS Subjects Sixty-two healthy, white men with a mean (± SD) age of 37.6 ± 2.9 y (weight: 81.8 ± 11.3 kg; height: 171.5 ± 4.9 cm) volunteered to participate in this study. The main descriptive characteristics of the sample are shown in Table 1. Subjects were police officers recruited from 5 different police departments. All subjects were informed about the research design and signed a consent form complying with the regulations of the Ethical Committee of the Faculty of Human Movement, Technical University of Lisbon. This study was part of a research project designed to assess the physical and metabolic fitness of police officers. Subjects came to the laboratory after a 12-h fast, and all measurements and tests were carried out on the same morning. TABLE 1 Physical characteristics of the subjects 1 x ± SD Minimum Maximum Age (y) 37.6 ± 2.9 31.0 46.0 Height (cm) 171.5 ± 4.9 162.6 185.7 Weight (kg) 81.8 ± 11.3 58.9 117.7 BMI (kg/m 2 ) 27.8 ± 3.5 18.6 34.5 %BF DXA 26.0 ± 7.4 5.3 42.3 Plethysmograph 23.4 ± 7.0 5.2 39.9 1 n = 62. DXA, dual-energy X-ray absorptiometry; %BF, percentage body fat. Anthropometric measurements Body weight was recorded on a calibrated electronic floor scale (Tanita Corp, Skokie, IL), to the nearest 5 g, that was connected to the plethysmograph. The scale was placed horizontally and was calibrated before each weighing session by using calibration procedures described by the manufacturer. Height was measured with a wall-mounted stadiometer to the nearest 0.5 cm while subjects were standing without shoes on a horizontal surface with their bodies stretched upward to the fullest extension and their heads in the Frankfurt plane. Body mass index (BMI) was derived from weight (kg) and height squared (m). Triceps-, biceps-, subscapular-, suprailiac-, and chest-skinfold thicknesses were measured according to the procedures described in the Anthropometric Standardization Reference Manual (25): 1) With the elbow extended and relaxed, the triceps-skinfold thickness was measured on the midline of the posterior aspect of the upper arm. 2) The biceps-skinfold thickness was a vertical fold measured on the anterior midline of the upper arm over the belly of the biceps at the level of the triceps skinfold. 3) The subscapular-skinfold thickness was measured as a diagonal fold 1 2 cm from the inferior angle of the scapula. 4) Suprailiac-skinfold thickness was a diagonal fold measured on the midaxillary line immediately superior to the iliac crest. 5) The chest-skinfold thickness was a diagonal fold measured at half the distance between the anterior axillary line and nipple. Two measurements were made with a Lange caliper (Cambridge Scientific Instruments, Cambridge, MD) at each site by a trained researcher. The mean was computed for further analysis. Jackson and Pollock s (J-P; 26) 3-site (chest, triceps, and subscapular) generalized equation and Durnin and Womersley s (D- W; 27) 2-site (triceps, biceps, subscapular, and suprailiac) generalized equation were used to predict body density. To convert body density to %BF, Siri s (28) 2-compartment model was used. FFM was calculated as the difference between each subject s body weight and body fat mass (FM). Dual-energy X-ray absorptiometry To estimate %BF, FFM, and FM, DXA measurements were made with a total body scanner (pencil beam mode, software version 5.67, enhanced whole-body analysis, QDR-1500; Hologic, Waltham, MA) that measured the attenuation of X-rays pulsed between 70 and 140 kv synchronously with the line frequency for each pixel of the scanned image. Comprehensive reviews of the theory and methodology of DXA measurements of body composition were published recently (15, 16). For the analysis of tissue composition, a step phantom with 6 fields of acrylic and aluminum of varying thickness and known absorptive properties was scanned alongside each subject to serve as an external standard. FFM was defined as the sum of the fat-free soft tissue and total bone mineral content from the whole-body scans. The same laboratory technician positioned the subjects, performed the scans, and executed the analysis according to the operator s manual, using the standard analysis protocol. The measure of precision selected in this study was the technical error (TE), d 2 /2n (29), where d is the difference between 2 trials and n is the number of paired observations. As estimated from 10 repeated measures, the TE of DXA-measured %BF was 0.5% fat. Bioelectrical impedance spectroscopy Whole-body impedance measurements were performed with a bioelectrical impedance spectroscopy (BIS) analyzer (model 4000B; Xitron Technologies, San Diego). For this experiment, a nonconducting bed was used. Subjects laid down in a supine position with their arms and legs abducted at an angle of 45. After the skin was cleaned with alcohol, 4 electrodes (model IS4000; Xitron Technologies) were placed on the dorsal surfaces of the right hand and right foot. The source electrodes were placed on the hand in the middle of the dorsal surface proximal to the metacarpal-phalangeal joint, and on the foot in the middle of the dorsal surface proximal to the metatarsal-phalangeal joint. The sensor electrodes were placed on the wrist at the midline between the distal prominences of the radius and ulna and on the

788 SARDINHA ET AL ankle joint at the line between the malleoli. Measurements were taken after subjects had been lying down for 5 min. Data were sampled at 50 programmed logarithmically spaced frequencies from 5 to 500 khz. This impedance spectra was modeled with the Cole-Cole cell suspension model (30) to derive a theoretical impedance at zero and infinity frequency, based on a nonlinear curve fitting from the measured resistance and reactance. Intracellular water (ICW) and extracellular water (ECW) were predicted from the Hanai mixture theory (31), and total body water (TBW) was estimated as the sum of ICW and ECW. A recent technologic review addressed the principles of BIS and the validity of the Hanai theory to estimate ECW and ICW volumes (32). Predicted %BF by BIS used the manufacturer s equation, which is based on a 2-compartment model with a fixed percentage of water in the FFM. For further analysis, the resistance at 50 khz was registered. This single frequency was used to derive FFM by using the fatness-specific equations of Segal et al (33). Air displacement plethysmography Body density was assessed with a new air displacement plethysmograph (BOD POD; Life Measurement Instruments, Concord, CA). The plethysmograph is a dual-chamber, single unit with an electronic-controlled volume perturbation diaphragm between the 2 chambers. The diaphragm induces volume perturbations and pressure fluctuations in the front test chamber and in the rear reference chamber, where the logic boards are housed. These pressure fluctuations are used to assess the chamber volume based on Poisson s Law. Before each testing session, a calibration procedure was performed and a brief description of the procedures was presented to the subjects. Each subject wore a swimming suit that had been weighed to the nearest 5 g with an electronic scale connected to the plethysmograph computer. For each subject, a 2-point chamber calibration, with the chamber empty and with a 50-L cylinder, was performed. The testing involves a 4-step process. First, the volume of the empty chamber was estimated. Then, a 50-L calibration cylinder was introduced so that a regression equation identifying the relation between any given volume and the ratio of the pressure amplitudes in both chambers could be developed internally. After this calibration, the subjects entered the chamber wearing a swim cap, and 2 body volume assessments were made. The door remained open between measurements. Whenever the difference between the 2 measurements was >150 ml, a third body volume measurement was performed. Finally, and after attaching a nose clip, the subject was connected to the breathing circuit housed in the rear chamber via a disposable filtered tube and was instructed to breath normally until the moment the system induced an airway occlusion. After airway occlusion, subjects were instructed to gently contract and relax the diaphragm. The computer then yields a figure of merit, ie, a mathematic figure that assesses subjects compliance with the procedures based on the agreement between the changes in airway pressure and chamber pressure. Whenever the value was >1, subjects repeated the 3 5-min procedure. It was rarely necessary for the procedure to be repeated. Tidal breathing was determined on the basis of the change in volume detected by the pressure transducers during normal breathing, and thoracic gas volume was calculated on the basis of changes in pressure in the lungs detected by a transducer in the breathing circuit during air occlusion. Final body volume was computed based on the initial body volume corrected for thoracic gas volume and a surface area artifact computed automatically. This accounted for a negative volume due to a more compressible air induced by the isothermal conditions associated with the skin surface area (3). Based on 19 repetitions, the TE for body density was 0.0021 kg/l. Statistical analysis Correlation analysis and a paired t test were used to analyze the plethysmograph s trial-to-trial precision. Pure or total error between DXA and the plethysmograph was also calculated, (Y 1 Y 2 ) 2 /n (34), where Y 1 is %BF assessed by DXA and Y 2 is %BF estimated by the plethysmograph. Multiple regression analysis was used to assess the significance of selected independent variables on %BF, FFM, and FM assessed by DXA. For each model, age, weight, height, BMI, and %BF were used as predictor variables, together with %BF, FFM, or FM measured by the plethysmograph, and body composition determined by BIS with frequencies from 5 to 500 khz, with BIA with a 50- khz single frequency, with J-P, and with D-W. The bias and 95% limits of agreement between DXA and the plethysmograph were calculated according to the method of Bland and Altman (35). The level of significance was set at P < 0.05. RESULTS Values for %BF and body density from repeated trials on the plethysmograph and related descriptive statistics are presented in Table 2. Correlation coefficients between the 2 trials were high (r = 0.98, P< 0.05) and no differences between means were observed as assessed by the paired t test. The TE was found to be <1.0%BF, corresponding to a TE of 0.0021 kg/l for body density. The correlations among selected methods for assessing %BF are shown in Table 3. The correlations were all significant (P < 0.01), ranging from 0.47 for the correlations between BMI and BIS-measured %BF, to 0.95 for the correlations between %BF TABLE 2 Precision of repeated measurements of percentage body fat (%BF) and body density with the plethysmograph 1 Values Intrasubject difference r (95% CI) t P Technical error %BF First trial 25.35 ± 7.4 2 Second trial 24.96 ± 7.7 0.396 ± 1.348 0.98 (0.96, 0.99) 1.22 > 0.05 0.966 Body density (g/ml) First trial 1.0414 ± 0.0163 Second trial 1.0423 ± 0.0171 0.0009 ± 0.0031 0.98 (0.96, 0.99) 1.18 > 0.05 0.0021 1 n = 19. x ± SD for the 2 trials on the same day.

VALIDITY OF AIR DISPLACEMENT PLETHYSMOGRAPHY 789 TABLE 3 Correlation coefficients among selected methods to estimate percentage body fat (%BF) 1 50 khz BIS Plethysmograph D-W DXA J-P 50 khz BIS 0.64 Plethysmograph 0.84 0.63 D-W 0.83 0.48 0.75 DXA 0.88 0.68 0.93 0.85 J-P 0.81 0.51 0.79 0.95 0.87 BMI 0.86 0.47 0.77 0.84 0.81 0.83 1 All correlations are significant, P = 0.01 level (two-tailed t test). 50-kHz, single-frequency assessment with the equations of Segal et al (33); BIS, bioelectrical impedance spectroscopy; D-W, equations of Durnin and Womersley (27); DXA, dual-energy X-ray absorptiometry; J-P equations of Jackson and Pollock (26). assessed by the J-P and D-W anthropometric equations. The second highest correlation was 0.93, between DXA-measured %BF and plethysmograph-measured %BF. This relation is depicted in Figure 1, with respective 95% CIs between 0.89 and 0.96. The predictive abilities of 7 different models for assessing DXA-measured %BF are given in Table 4. For each model, the regression coefficients of the selected independent variables, intercepts, adjusted R 2 values, and SEEs are presented. As assessed by the highest adjusted R 2 value (89.5) and by the lowest SEE (2.4%BF), the best model was the one that included the plethysmograph plus age, weight, and height. The model that included age, weight, and height independently (model 1) and the model that used a combination of weight and height to compute BMI (model 2) gave similar adjusted R 2 values, 66.4 and 66.5, respectively, and the same SEE, 4.3%BF. Similarly, the 2 anthropometric models that used skinfold thicknesses (models 6 and 7) gave similar adjusted R 2 and SEE values. Compared with BIS, the 50-kHz, single-frequency assessment of %BF with the equations of Segal et al (33) gave higher adjusted R 2 (79.8 compared with 76.2) and lower SEE (3.3 kg compared with 3.6 kg) values. Predicted fat and FFM from the same models are presented in Tables 5 and 6. The highest adjusted R 2 and lowest SEE values were seen with the plethysmograph model. SEEs as low as 1.9 kg were found for both FM and FFM, with high adjusted R 2 values of 94.9 and 90.0, respectively. This model performed much better than all other models. Models 4 and 5, which included singlefrequency BIA and BIS, predicted FM and FFM similarly. As shown in Table 6, when regressed against DXA, FFM, BMI, and age yielded less predictive results, accounting for only 22% of the variance, with an SEE of 5.2 kg FFM. In Figure 2, the difference between DXA-measured %BF and plethysmograph-measured %BF is plotted against the average of the 2 methods, as described by Bland and Altman (35). Horizontal lines are drawn at the mean difference and at the mean difference plus and minus 1.96 times the SD of the differences. A systematic difference between the methods, with a mean difference of 2.6%BF (95% CI: 1.91, 3.29; t [60] = 7.58, P < 0.05) was found. Compared with DXA, the plethysmograph underestimated %BF. The total error was 3.7%BF. intrasubject difference between trials of 0.40 ± 1.35% was not different from zero (P > 0.05), yielding a precision error of 1.0%BF, which was higher than the 0.4%BF found by McCrory et al (4). The between-trial CV reported by these authors was almost half the value found in our study, 1.7% compared with 3.3%. Inanimate volume measurements are extremely reliable with the plethysmograph (CV = 0.025%) (3), whereas reliability for human body volume assessment is somewhat poorer, probably because of the variability of thoracic gas volume measurements. A volume error of 0.02% was found with inanimate objects, which corresponds to an error of 0.1%BF and an SEE of 0.004 L, when a series of volumes were assessed and regressed against the actual values (3). McCrory et al (4) reported an SEE of 1.8% between body composition measured with the plethysmograph and that by underwater weighing. In the present study, which had a more homogeneous sample, a smaller SEE of 2.4% was found (Table 4). When age, weight, and height were included in the plethysmograph model, the adjusted R 2 value of 89.5 was lower than the value of 0.93 found in the study by McCrory et al. When values for both sexes were combined, McCrory et al found a non- DISCUSSION The precision of 0.0021 kg/l for body density on repeated measurements with the plethysmograph was close to the 0.0020 kg/l error accepted as characteristic of trial-to-trial variation during the same day for underwater weighing (6). The mean FIGURE 1. Regression analysis of percentage body fat determined by dual-energy X-ray absorptiometry (%BF DXA) compared with percentage body fat determined with the plethysmograph (%BF plethysmograph) with 95% CIs (dashed lines) and prediction intervals (solid lines).

790 SARDINHA ET AL TABLE 4 Multiple regression coefficients for the prediction of percentage body fat with dual-energy X-ray absorptiometry (%BF DXA) from selected body-composition variables 1 Dependent variable %BF DXA Independent variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Age (y) 0.443 2 0.399 2 0.318 2 0.422 2 0.216 2 0.122 0.104 Weight (kg) 0.587 2 0.173 2 0.144 0.451 2 Height (cm) 0.693 2 0.207 2 0.134 0.587 2 BMI (kg/m 2 ) 1.755 2 %BF Plethysmograph 0.785 2 50 khz 1.199 2 BIS 0.498 2 J-P (26) 0.986 2 D-W (27) 1.432 2 Intercept 113.5 7.8 40.9 23.2 87.4 8.5 6.4 R 2 100, adjusted 66.4 66.5 89.5 79.8 76.2 74.5 71.6 SEE (%BF) 4.3 4.3 2.4 3.3 3.6 3.7 3.9 1 50 khz, single-frequency assessment with the equations of Segal et al (33); BIS, bioelectrical impedance spectroscopy; D-W, equations of Durnin and Womersley (27); J-P, equations of Jackson and Pollock (26). 2 P< 0.05. significant 0.3 ± 0.2 (95% CI: 0.6, 0.0) difference between %BF measured with the plethysmograph and by underwater weighing, with limits of agreement for the individual values between 4.0%BF and 3.4%BF. In our study, a difference of 2.6 ± 2.7%BF (95% CI: 1.9, 3.3) was observed between the 2 methods, with DXA giving higher estimates of %BF than the plethysmograph. Overall, the plethysmograph model performed better than the other models with DXA-measured %BF, FM, and FFM as dependent variables. BMI had the poorest relation with these bodycomposition variables. The error of 4.3%BF was equal to the value reported by Jackson et al (36) and lower than values found in the validation samples of Jackson and Pollock (5.8%; 26) and Durnin and Womersley (5.9%; 27). It was higher than the 3.3% reported by Hansen et al (10) and the 3.9% reported by Womersley and Durnin (37) for females. The J-P equation had a somewhat higher %BF error than the equation reported originally (26) (3.7% compared with 3.5%), whereas the D-W equation (27) showed a better prediction error than that found in the validation sample (3.9% compared with 4.6%). Fat and FFM predicted by BIS and by the fatness-specific equations of Segal et al (33) yielded similar SEEs. The SEE of 2.8 kg FFM found was within the range of 2.5 3.0 kg reported by Segal et al (33) for nonobese and obese men lower (38) and similar to values from other cross-validation studies (39, 40) and somewhat higher than the value of 2.6 kg found with BIS in a sample of 14 women and 10 men (41). Contrary to these results, Chumlea et al (42) found better results for FFM prediction with BIS than with BIA in a sample of 31 men, with DXA as the reference method. When compared with BIS, the 50-kHz, single-frequency fatness-specific equations of Segal et al (33) had a slightly higher adjusted R 2 value (79.8 compared with 76.2) and a lower SEE (3.3%BF compared with 3.6%BF). To compute FFM and %BF, BIS uses a 2-compartment approach with a fixed percentage of water in the TABLE 5 Multiple regression coefficients for the prediction of fat mass (FM) by dual-energy X-ray absorptiometry (DXA) from selected body-composition variables 1 Dependent variable FM DXA Independent variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Age (y) 0.361 2 0.403 2 0.265 2 0.362 2 0.146 0.124 0.119 Weight (kg) 0.721 2 0.199 2 0.081 0.479 2 Height (cm) 0.519 2 0.151 2 0.111 0.431 2 BMI (kg/m 2 ) 2.148 2 FM (kg) Plethysmograph 0.796 2 50 khz 1.184 2 BIS 0.540 2 J-P 1.049 2 D-W 1.286 2 Intercept 65.5 22.7 25.8 23.3 52.6 6.7 0.9 R 2 100, adjusted 81.9 81.0 94.9 88.0 88.5 85.2 83.5 SEE (kg) 3.5 3.6 1.9 2.8 2.8 3.2 3.3 1 50 khz, single-frequency assessment with the equations of Segal et al (33); BIS, bioelectrical impedance spectroscopy; D-W, equations of Durnin and Womersley (27); J-P, equations of Jackson and Pollock (26). 2 P< 0.05.

VALIDITY OF AIR DISPLACEMENT PLETHYSMOGRAPHY 791 TABLE 6 Multiple regression coefficients for the prediction of fat-free mass (FFM) by dual-energy X-ray absorptiometry (DXA) from selected body-composition variables 1 Dependent variable FFM DXA Independent variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Age (y) 0.361 2 0.134 0.265 2 0.362 2 0.145 0.138 0.137 Weight (kg) 0.279 2 0.005 0.266 2 0.019 Height (cm) 0.519 2 0.151 2 0.111 0.431 2 BMI (kg/m) 0.811 2 FFM (kg) Plethysmograph 0.873 2 50 khz 1.184 2 BIS 0.540 2 J-P 0.886 2 D-W 0.780 2 Intercept 65.5 32.3 25.8 23.3 52.6 0.9 7.5 R 2 100, adjusted 64.8 22.0 90.0 77.1 77.8 72.0 64.8 SEE (kg) 3.5 5.2 1.9 2.8 2.8 3.1 3.5 1 50 khz, single-frequency assessment with the equations of Segal et al (33); BIS, bioelectrical impedance spectroscopy; D-W, equations of Durnin and Womersley (27); J-P, equations of Jackson and Pollock (26). 2 P< 0.05. FFM, after predicting ICW and ECW to calculate total body water with a spectrum of frequencies (32). The single-frequency FFM estimation also assumes a fixed percentage of water in the FFM and relies on total body water that is assessed by the regression equation. Any major change in hydration due to fatness or leanness will cause an overestimation and underestimation of FFM, respectively. It seems that the fatness-specific equations of Segal et al (33) control for this potential source of error, yielding results similar to those with BIS. On the basis of the results from the present study and on the basis of Lohman s (6) subjective ratings of accuracy to estimate %BF and FFM, respectively, the models were rated as follows: the plethysmograph model as excellent and ideal, the single-frequency model as good and very good, the BIS model as fairly good and very good, the J-P and D-W models as fairly good and good, and the BMI model as fair and not recommended. Clearly, as assessed by the highest adjusted R 2 value and the lowest SEE value for this sample of middle-aged men, the plethysmograph model was much more accurate than any other model analyzed. Note that this better accuracy was based on a convenient methodology used to assess body density, making use of expensive technologic developments, and not comparable with the relatively low operating costs of the other field methods used. The present data are good examples of the need to use alternative approaches to assess estimation procedures based on comparisons of estimated and measured scores and their means (35). The mean difference of 2.6%BF between the plethysmograph and DXA measurements was significantly different from zero, and 75% of the subjects had higher %BF values estimated by DXA (95% CI: 2.6, 7.8). The difference between both means suggests that these 2 methods gave somewhat different mean estimates, even though the prediction accuracies of %BF and FFM estimates are rated as excellent and ideal, respectively. In a recent study conducted to evaluate the accuracy of DXA in measuring body composition, DXA estimates of %BF with a similar densitometer and the same whole-body software version used in the present study were compared with measures from traditional hydrodensitometry (43). The correlation was high (%BF: r = 0.95, P < 0.01) but DXA yielded higher FM and FFM values for women than did hydrodensitometry, lower values for men (P < 0.05), but similar sex-related differences. Interestingly, when assumed density of FFM (44) was replaced with a value accounting for variance in water, protein, and mineral, DXA overestimated %BF in 0.7% of the women and in 0.5% of the men. A literature analysis comparing several studies to assess %BF with densitometry by underwater weighing reported a 95% CI for the bias of 2.1 to 1.4 for the densitometer used compared with underwater weighing (45). Taken together, these results indicate that whereas in some studies DXA overestimated %BF, in others DXA underestimated %BF when compared with underwater weighing densitometry. In addition, sex-related differences may be important, suggesting that variances in the density of the FFM (which is likely to affect densitometry measures more than it affects DXA) may explain some of the discrepancies. Nevertheless, although body density estimated by the plethysmograph is based on different principles than is underwa- FIGURE 2. Scatter diagram of intermethod differences between dual-energy X-ray absorptiometry (DXA) determined and plethysmograph-determined percentage body fat (%BF) plotted against the averages of the 2 measurements.

792 SARDINHA ET AL ter weighing, the mean bias of 2.6%BF found in this study was somewhat higher than most of the values found in the literature. Human body-volume assessment requires correction for thoracic gas volume and isothermal conditions associated with skin surface area. Inaccuracies in these methodologic and theoretic approaches may be related to this systematic error. In conclusion, body composition was accurately estimated in middle-aged men with the new, air displacement plethysmograph studied, although %BF was systematically underestimated. Further research using a multicomponent model or other criterion methods is required in other populations to evaluate this new method. REFERENCES 1. Gundlach BL, Nijkrake HGM, Hautvast JGAJ. A rapid and simplified plethysmometric method for measuring body volume. Hum Biol 1980;52:23 33. 2. Gundlach BL, Visscher GJW. The plethysmometric measurement of total body volume. Hum Biol 1986;58:783 99. 3. Dempster P, Aitkens S. A new air displacement method for the determination of human body composition. Med Sci Sports Exerc 1995;27:1692 7. 4. McCrory MA, Gomez TD, Bernauer EM, Molé PA. Evaluation of a new air displacement plethysmograph for measuring human body composition. Med Sci Sports Exerc 1995;27:1686 91. 5. Fuller NJ, Jebb SA, Laskey MA, Coward WA, Elia M. Four-component model for the assessment of body composition in humans: comparison with alternative methods, and evaluation of the density and hydration of fat-free mass. Clin Sci 1992;82:687 93. 6. Lohman TG. Advances in body composition assessment. Champaign, IL: Human Kinetics Publishers, 1992. 7. Clark RR, Kuta JM, Sullivan JC. Prediction of percent body fat in adult males using dual energy x-ray absorptiometry, skinfolds, and hydrostatic weighing. Med Sci Sports Exerc 1992;25:528 35. 8. Ellis KJ, Shypailo R, Schoknecht P, Pond W. Neutron activation analysis: criterion method for evaluation of dual-energy x-ray absorptiometry measurements in infants. J Radioanal Nucl Chem 1995;195:139 44. 9. Goodsitt MM. Evaluation of a new set of calibration standards for the measurement of fat content via DPA and DXA. Med Physics 1992;19:35 44. 10. Hansen NJ, Lohman TG, Going SB, et al. Prediction of body composition in premenopausal females from dual-energy x-ray absorptiometry. J Appl Physiol 1993;75:1637 41. 11. Jensen MD, Kanaley JA, Roust LR, et al. Assessment of body composition with use of dual-energy x-ray absorptiometry: evaluation and comparison with other methods. Mayo Clin Proc 1993;68:867 73. 12. Svendsen OL, Haarbo J, Hassager C, Christiansen C. Accuracy of measurements of body composition by dual-energy x-ray absorptiometry. Am J Clin Nutr 1993;57:605 8. 13. Wellens R, Chumlea WC, Guo S, Roche AF, Reo NV, Siervogel RM. Body composition in white adults by dual-energy x-ray absorptiometry, densitometry, and total body water. Am J Clin Nutr 1994;59:547 55. 14. Kohrt WM. Body composition by DXA: tried and true? Med Sci Sports Exerc 1995;27:1349 53. 15. Lohman TG. Dual energy x-ray absorptiometry. In: Roche AF, Heymsfield SB, Lohman TG, eds. Human body composition. Champaign, IL: Human Kinetics Publishers, 1996:63 78. 16. Pietrobelli A, Formica C, Wang Z, Heymsfield SB. Dual-energy x- ray absorptiometry body composition model: a review of physical concepts. Am J Physiol 1996;271:E941 51. 17. Roubenoff R, Kehayias JJ, Dawson-Hughes B, Heymsfied SB. Use of dual energy x-ray absorptiometry in body composition studies: not yet a gold standard. Am J Clin Nutr 1993;58:589 91. 18. Lohman TG, Going S. Assessment of body composition and energy balance. In: Lamb DR, Murray R, eds. Perspectives in exercise science and sports medicine. Vol 11. Exercise, nutrition, and control of body weight. Carmel, IN: Cooper Publishing Group, 1998:61 106. 19. Ellis KJ, Shypailo RJ, Pratt JA, Pond WG. Accuracy of dual-energy x-ray absorptiometry for body-composition measurements in children. Am J Clin Nutr 1994;60:660 5. 20. Brunton JA, Bayley HS, Atkinson SA. Validation and application of dual-energy x-ray absorptiometry to measure bone mass and body composition in small infants. Am J Clin Nutr 1993;58:839 45. 21. Haarbo J, Gotfredsen A, Hassager C, Christiansen C. Validation of body composition by dual x-ray absorptiometry (DEXA). Clin Physiol 1991;11:331 41. 22. Heymsfield SB, Litchman S, Baumgartner RN, et al. Body composition of humans: comparisons of two improved four-compartment models that differ in expense, technical complexity, and radiation exposure. Am J Clin Nutr 1990;52:52 8. 23. Prior BM, Cureton KJ, Modlesky CM, et al. In vivo validation of whole body composition estimates from dual-energy x-ray absorptiometry. J Appl Physiol 1997;83:623 30. 24. Prentice A. Application of dual-energy x-ray absorptiometry and related techniques to the assessment of bone and body composition. In: Davies PSW, Cole TJ, eds. Body composition techniques in health and disease. New York: Cambridge University Press, 1995:1 13. 25. Harrison GG, Buskirk ER, Carter JEL, et al. Skinfold thicknesses and measurement technique. In: Lohman TG, Roche AF, Martorell R, eds. Anthropometric standardization reference manual. Champaign, IL: Human Kinetics Publishers, 1988:55 70. 26. Jackson AS, Pollock ML. Practical assessment of body composition. Physician Sports Med 1985;13:76 90. 27. Durnin JV, Womersley J. Body fat assessed from body density and its estimation from skinfold thickness measurements on 481 men and women aged from 16 to 72 years. Br J Nutr 1974;32:77 97. 28. Siri WE. Body composition from fluid spaces and density: analysis of method. In: Brozek J, Henschel A, eds. Techniques for measuring body composition. Washington, DC: National Academy of Sciences, 1961:223 44. 29. Mueller WH, Martorell R. Reliability and accuracy of measurement. In: Lohman TG, Roche AF, Martorell R, eds. Anthropometric standardization reference manual. Champaign, IL: Human Kinetics Publishers, 1988:83 6. 30. Cole KS, Cole RH. Dispersion and absorption in dielectrics. J Chem Phys 1941;9:341 51. 31. Hanai T. Electrical properties of emulsions. In: Sherman P, ed. Emulsion science. New York: Academic Press, 1968:354 477. 32. De Lorenzo A, Andreoli A, Matthie J, Withers P. Predicting body cell mass with bioimpedance by using theoretical methods: a technological review. J Appl Physiol 1997;82:1542 58. 33. Segal KR, Van Loan M, Fitzgerald PI, Hogdson JA, Van Itallie TB. Lean body mass estimation by bioelectrical impedance analysis: a four site cross-validation study. Am J Clin Nutr 1988;47:7 14. 34. Guo SS, Chumlea WC. Statistical methods for the development and testing of predictive equations. In: Roche AF, Heymsfield SB, Lohman TG, eds. Human body composition. Champaign, IL: Human Kinetics Publishers, 1996:191 202. 35. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods for clinical measurement. Lancet 1986;8:307 10. 36. Jackson AS, Pollock ML, Graves JE, Mahar MT. Reliability and validity of bioelectrical impedance in determining body composition. J Appl Physiol 1988;64:529 34. 37. Womersley J, Durnin JVGA. A comparison of the skinfold method with extent of overweight and various weight-height relationships in the assessment of obesity. Br J Nutr 1977;38:271 84. 38. Stolarczyk LM, Heyward VH, Van Loan MD, Hicks VL, Wilson

VALIDITY OF AIR DISPLACEMENT PLETHYSMOGRAPHY 793 WL, Reano LM. The fatness-specific bioelectrical impedance analysis equations of Segal et al: are they generalizable and practical? Am J Clin Nutr 1997;66:8 17. 39. Eckerson JM, Housh TJ, Johnson GO. Validity of bioelectrical impedance equations for estimating fat-free weight in lean males. Med Sci Sports Exerc 1992;24:1298 302. 40. Ross R, Leger L, Martin P, Roy R. Sensitivity of bioelectrical impedance to detect changes in human body composition. J Appl Physiol 1989;67:1643 8. 41. Van Loan MD, Withers P, Matthie J, Mayclin PL. Use of impedance spectroscopy to determine extracellular fluid, intracellular fluid, total body water, and fat-free mass. In: Ellis KJ, Eastman JD, eds. Human body composition. In vivo methods, models, and assessment. New York: Plenum Press, 1992:67 70. 42. Chumlea WC, Guo SS, Baumgartner RN, Siervogel RM. Determination of body fluid compartments with multiple frequency bioelectric impedance. In: Ellis KJ, Eastman JD, eds. Human body composition. In vivo methods, models, and assessment. New York: Plenum Press, 1992:23 6. 43. Kohrt WM. Preliminary evidence that DEXA provides an accurate assessment of body composition. J Appl Physiol 1998;84:372 7. 44. Brozek J, Grande F, Anderson JT, Keys A. Densitometric analysis of body composition: revision of some quantitative assumptions. Ann N Y Acad Sci 1963;110:113 40. 45. Fogelholm M, Lichtenbelt van Marken. Comparison of body composition methods: a literature analysis. Eur J Clin Nutr 1997;