Comparison Between the Air Displacement Method and Dual Energy X-Ray Absorptiometry for Estimation of Body Fat

Air displacement plethysmography (ADP) is a method for the determining percent body fat (%BF) using the two-compartment model, in which the body is partitioned into body-fat mass and fat-free mass (FFM). Although this model assumes a constant density of FFM as 1.10 g/ml, its density may depend upon the bone mineral content (BMC) and total body water (TBW) which vary with age, gender, and race/ethnicity. This study compared %BF determined from ADP (ADP%BF) with %BF obtained by dual-energy x-ray absorptiometry (DXA%BF), and also investigated the effects of BMC, TBW, and other factors on its value. The subjects were 721 female and male Japanese aged 40 to 79 years. Body density was measured by ADP and %BF was calculated using Brozek et al's equation. BMC and body-fat volume were measured using DXA, and TBW was measured by multifrequency bioelectrical impedance. A series of anthropometric measurements was taken. Although ADP%BF was highly correlated with DXA%BF (female: r=0.89, male: r=0.90) (p<0.001), ADP%BF differed significantly from DXA%BF (female: -1.30*0.14 % (mean*s.e.m.), male: 1.22*0.13 %) (p<0.001). The difference in %BF (ADP%BF-DXA%BF) was negatively associated with BMC/FFM but not with TBW/FFM in both genders. The difference in %BF was also positively correlated with waist circumference. Considering previous studies, this result may be explained by the underestimation of DXA%BF, rather than by the overestimation of ADP%BF. In conclusion, ADP may be a useful method to measure %BF. However, BMC should be taken into consideration. Furthermore , DXA%BF may be underestimated in people with large waists. J Epidemiol, 2000 ; 10 : S82-S89 .


INTRODUCTION
Hydrostatic weighting (HW) is a traditional method for estimating the percent body fat (%BF).It is a reference for the prediction equations for estimation of %BF in Japanese people on the basis of skinfold thickness 1) and bioelectrical impedance (Tanita co., Tokyo) 2), respectively.However, it is relatively hard to measure elderly and handicapped persons' body volume by HW.Dual-energy x-ray absorptiometry (DXA) is another method that can measure body-fat mass (BF), bone mineral content (BMC), and bone-free lean tissue mass (LTM).Several animal studies in vivo 3,4) and comparison of DXA with other methods5) have indicated that DXA provids valid estimates of %BF (DXA%BF).However, DXA is an invasive method and thus is sometimes difficult to use repeatedly for the aims of our health check.There is a need for noninvasive, easy, and reliable methods for measuring body composition.Air displacement plethysmography (ADP) is a new method for determining the %BF using the traditional twocompartment model in a similar way to HW.The ADP utilizes Boyle's law (the inverse relationship between pressure and volume of a gas at constant temperature 6) to measure body volume directly.McCrory and colleagues 6) reported that the %BF obtained by ADP (ADP%BF) was not significantly different from that determined by HW (HW%BF).
Some previous studies indicated that %BF using a two-compartment model was overestimated in elderly people 7) a and in males 7-9) compared with DXA%BF.Snead et al. a indicated that the HW%BF of elderly U.S. females was 4 to 5 % higher, and that of U.S. males was 5 to 6 % higher than with DXA%BF.Going et al. 8) found that the HW%BF was 1.8 % (mean, p=0.04) higher than the DXA%BF in U.S. males aged 19-31 years.Those differences may be due to the assumption of a two-compartment model.
In the two-compartment model, the body is partitioned into BF and fat-free mass (FFM).The two-compartment model assumes a constant BF density of 0.90 g/ml and an FFM of 1.10 g/ml 10).However, the density of FFM varies with gender 11), age [11][12][13][14][15], and race/ethnicity [16][17][18][19].For instance, loss of bone minerals decreases the density of FFM.A density of FFM lower than 1.10 g/ml would lead to an overestimation of BF in the two-compartment model.The correlation or the difference between ADP%BF and DXA%BF in Japanese was unclear.In order to use ADP to measure %BF in an epidemiological study, basic data on the relationships between ADP%BF and DXA%BF by age and gender is required.Therefore, the purpose of this study was to compare ADP%BF with DXA%BF, and to investigate the effects of BMC, TBW, and other factors on its value.

Subjects
The subjects were a gender and age-stratified random sampling from the NILS-LSA (National Institute for Longevity Sciences -Longitudinal Study of Aging) from the area of our institute from November, 1997 until November, 1998.A total of 721 Japanese (366 female and 355 male) aged 40 to 79 years participated (Table 1).The design of the NILS-LSA has been described elsewhere 20).The purpose of this study was explained to each subject before obtaining written consent and was approved by the Ethical Committee of Chubu National Hospital.
Air displacement method Body volume and body weight were measured with an ADP (BOD POD; Life Measurement Instruments, Concord, CA), and body density was calculated as body volume divided by body weight.The mechanism and methods are already described in detail elsewhere 6,21,r2).Percent body fat was calculated from body density using Broiek et al's equation 10).ADP%BF = (4.57/Bodydensity -4.142) X 100 BF volume was calculated as body weight multiplied by ADP%BF.Software version 1.5 was used for BOD POD.
Dual-energy x-ray absorptiometry BMC, LTM, BF, and DXA%BF were measured using a DXA (Hologic QDR-4500, Waltham, MA) with whole-body analysis.FFM was calculated as LTM plus BMC, and the body weight estimated by DXA (DXA-wt) as FFM plus BF.Total body water Total body water (TBW) was measured with a multifrequency bioelectrical impedance spectroscopy analyzer (model 4000C; XITRON Technologies, SanDiego).Subjects laid down in a supine position and electrodes were placed on the dorsal surfaces of the right hand and the right foot.Measurements were taken after subjects had been lying down for 5 min.The mechanism and methods are already described in detail elsewhere 23).

Anthropometric measurements
Anthropometric measurements were taken.Waist circumference at the level of the umbilicus was measured using a fiberglass tape.The sagittal abdominal diameter from the abdomen to the back at the level of the umbilicus was measured horizontally with a spreading caliper (Martin Type).Subjects were in an upright position for all measurements.
DXA%BF were performed after lunch, and other measurements were performed in the morning after at least a 12-h fast.

Data Analysis
The data was analyzed using SAS® statistical software packages (release 6.12, SAS Institute, Cary, NC, USA) .The paired t test was applied to measurements of DXA and ADP.The student t test was utilized to compare measurements between genders.Pearson's correlation coefficients were calculated between ADP%BF and DXA%BF, and among the differences of %BF, FFM, BMC fraction of FFM (BMC/FFM), sagittal abdominal diameter, and waist circumference.Probability values below 0.05 were regarded as significant.

DISCUSSION
Although some previous studies suggested that DXA was an accurate method for measurement of BF 3-50), DXA is an invasive method.ADP is a non-invasive method.Moreover, it is an easier and more comfortable method to measure %BF than HW.The results of this study showed the relationship between ADP%BF and DXA%BF was strong.Dempster and Aitkens 21) demonstrated that reliability of ADP over repeated trials was good.Therefore, ADP may be a highly reliable and valid method for determining body density.
However, ADP%BF differed significantly from DXA%BF (p<.001).In 30 % of females ADP%BF was 2.5% or less than DXA%BF, and conversely, ADP%BF was 2.5% or more than DXA%BF in 30% of males.The difference may be due to the  assumption of a two-compartment model that the density of FFM is 1.10 g/ml.A density of FFM lower than 1.10 g/ml would lead to an overestimation of BF with the two-compartment model.In contrast, a density of FFM higher than 1.10 g/ml would lead to an underestimation of BF.The difference in %BF was negatively correlated with BMC/FFM in both genders.The results of this study agreed with some previous studies 11`2.A decline in bone minerals could lead to a decrease in the density of FFM and an overestimation of body fat with the two-compartment model.Although ADP%BF was correlated with BMC/FFM, DXA%BF was not.Consequently, DXA should be used to measure BF, because it is free from BMC bias.
Snead and colleagues » showed that there were small differences between HW%BF and DXA%BF in young and middleaged females, and DXA%BF was 4 to 5 % lower than HW%BF in elderly females.These results differed from ours slightly.In the present study, on the average ADP%BF was 1.9 to 3.3 % lower in females aged 40-49 and 50-59 years, and was not significantly different from DXA%BF in females of 60-69 or 70-79 years.This suggested that the density of FFM may be higher than 1.10 g/ml in Japanese females aged 40-59 years, and that translated into an underestimated %BF when calculated from a two-compartment model.In support of this, Miyamoto et al. 19) found that young Japanese females had an FFM density higher than 1.10 g/ml.In females aged 60-79 years, the density of FFM may be close to 1.10 g/ml.Some European or U.S. studies considered that the high density of FFM was due to high BMC.For example, black females have a higher BMC than white females [16][17][18].Thus, the density of FFM in black females was higher than 1.10 g/ml.However, premenopausal Japanese females had a 5 % lower bone miner-al density than that of U.S. and European whites 26).Therefore, lower LTM, for example of skeletal muscles, in Japanese females may account for the high density of FFM.If skeletal muscles decline more than their bones, BMC as a percent of FFM increases.According to previous studies, BMC/FFM was significantly higher in females than in males X-9).In this study, the BMC/FFM of females aged 40-49 and 50-59 years was also higher than that of males at the same age.However, the BMC/FFM was lower in females of 60-69 and 70-79 years than that of males.In males, the mean ADP%BF was 0.8 to 2.0 % higher than DXA%BF in age groups of 50-59, 60-69, and 70-79 years in this study.These results were similar to previous studies ~~~.They suggested that the density of FFM may be lower than 1.10 g/inl in Japanese males of 50-79 years due to lower BMC/FFM and/or higher TBW/FFM than that of females.
The TBW may also have an effect on the difference in %BF obtained by the two-compartment model5,11,30) and on the density of FFM.Increases in TBW contributed to decreases in the density of FFM, because the density of water is 0.99 g/ml '01 Pace and Rathbun 3U found that TBW as a percentage of FFM was 73.2 %, and some studies showed that TBW/FFM did not change or the changes of TBW/FFM were small 11,32.33)However, other studies suggested that TBW/FFM increased with age 14,34)_ Results were not consistent.In the results of this study, although TBW/FFM was at least different between genders, TBW/FFM was not correlated with a difference in %BF in either gender.
The difference in %BF was positively correlated with waist circumference and sagittal abdominal diameter.These results may be explained by the underestimation of DXA%BF rather than by the overestimation of ADP%BF.Some previous stud-ies7,88)found that fat thickness influenced measurement of BF using DXA.Snead et al. 7) examined DXA%BF to evaluate subjects with packets of lard (2 to 3 kg) overlying either the thigh or the trunk region.Only 55 % of the added mass was identified as fat when the lard was placed over the trunk, compared with 96 % when it was positioned over the legs.
One limitation of our study was to measure DXA%BF after lunch.Although DXA-wt was correlated with scale body weight (r=0.997p<.001), the mean of difference between them was 2.1 kg for females and 2.6 kg for males.Prior et al. 0 made a comparison between scale body weight and DXA-wt in 172 young males and females.The mean DXA-wt was significantly less than body weight by 0.6*0.5 kg.According to another study 8), body weight and DXA-wt were highly correlated (r>=0.999)and the average values were within 0.22 to 0.45 kg.Further, we compared the BF obtained by DXA and the BF from ADP.It is possible that BF obtained by DXA is greater than BF by ADP due to food intake.Alternatively, BF obtained by DXA and BF by ADP are similar, because the food component was regarded as water or others except fat.However, in males aged 50-79 years BF by DXA was lower than BF by ADP.We could not distinguish whether the discrepancies in weight and BF were due to food intake or DXA.Therefore, the comparison between %BF is more desirable than that with BF.However, these results were consistent with the results of %BF.
Broiek et al's equation 10) was established by a study in which subjects were Caucasian.However, the BMC and/or the FFM of Caucasians are different from those of Japanese zn.Future studies should construct an equation estimating %BF for Japanese using a four-compartment model, in which the body is partitioned into fat, water, bone mineral, and others considering age and gender.
In conclusion, ADP is a useful method for measuring middle-aged and elderly people's %BF.However, %BF using a two-compartment model may be underestimated in Japanese middle-aged females and overestimated in Japanese middleaged and elderly males, and DXA%BF may be underestimated in people with large sagittal abdominal diameters or waists, because the BMC/FFM ratio and the fat thickness may influence the %BF estimate.

TnhIP 1 .
Snhii t characteristics a: Body weight measured by DXA (BMC+LTM+FM) b: Body mass index.Body weight (kg) divided by height squared (m)z c: Percent body fat measured by DXA d: Percent body fat measured by ADP e: Fat mass measured by DXA f: Fat mass measured by ADP g: Bone mineral content measured by DXA h: Bone-free lean tissue mass measured by DXA i Total body water measured by impedance j: Bone mineral content divided by fat free mass (BMC+LTM) k: Total body water divided by fat free mass (BMC+LTM) l: Standard deviation

Figure 1 .
Figure 1.Relationship between percent body fat measured by DXA and air displacement plethysmography.The diagonal is an identical line.

Figure 2 .
Figure 2. The frequency of the difference in percent body fat measured by DXA and air displacement plethysmography.

Figure 3 .
Figure 3. Relationship between bone mineral content divided by fat-free mass and the difference in percent body fat (ADP%BF-DXA%BF) .

Table 2 .
Pearson's correlation coefficients among differences in percent body fat and other factors.Female (n=366) Male (n=355) *p< .05,**p<.Ol, ***p<.001a: Percent body fat measured with air displacement plethysmography minus percent body fat measured by DXA b: Bone mineral content divided by fat free mass (BMC+LTM) c: Bone-free lean tissue mass d: Total body water divided by fat free mass (BMC+LTM) ).Furthermore, ADP%BF, but not DXA%BF, was negatively associated with BMC/FFM However, BF by DXA was greater than BF by ADP in females of 40-49 years (2.0±0.2 kg (mean±s.e.m.), p<.001) and 50-