Estimates of percentage body fat in young adolescents: a comparison

European Journal of Clinical Nutrition (2003) 57, 1402–1410

& 2003 Nature Publishing Group All rights reserved 0954-3007/03 $25.00 www.nature.com/ejcn

ORIGINAL COMMUNICATION Estimates of percentage body fat in young adolescents: a comparison of dual-energy X-ray absorptiometry and air displacement plethysmography D Radley1*, PJ Gately1, CB Cooke1, S Carroll1, B Oldroyd2 and JG Truscott2,3 1 School of Leisure and Sport, Leeds Metropolitan University, Leeds, UK; 2Centre for Bone and Body Composition Research, University of Leeds, Leeds, UK; and 3Imaging Science Group, School of Healthcare Studies, University of Leeds, Leeds, UK

Objective: To evaluate the accuracy of percentage body fat (%fat) estimates from air displacement plethysmography (ADP) against an increasingly recognised criterion method, dual-energy X-ray absorptiometry (DXA), in young adolescents. Design: Cross-sectional evaluation. Setting: Leeds General Infirmary, Centre for Bone and Body Composition Research, Leeds, UK. Subjects: In all, 28 adolescents (12 males and 16 females), age (mean7s.d.) 14.970.5 y, body mass index 21.272.9 kg/m2 and body fat (DXA) 24.2710.2% were assessed. Results: ADP estimates of %fat were highly correlated with those of DXA in both male and female subjects (r ¼ 0.84–0.95, all Po0.001; s.e.e. ¼ 3.42–3.89%). Mean %fat estimated by ADP using the Siri (1961) equation (ADPSiri) produced a nonsignificant overestimation in males (0.67%), and a nonsignificant underestimation in females (1.26%). Mean %fat estimated by ADP using the Lohman (1986) equations (ADPLoh) produced a nonsignificant underestimation in males (0.90%) and a significant underestimation in females (3.29%; Po0.01). Agreement between ADP and DXA methods was examined using the total error (TE) and methods of Bland and Altman (1986). Males produced a smaller TE (ADPSiri 3.28%; ADPLoh 3.49%) than females (ADPSiri 3.81%; ADPLoh 4.98%). The 95% limits of agreement were relatively similar for all %fat estimates, ranging from 76.57 to 77.58%. Residual plot analyses, of the individual differences between ADP and DXA, revealed a significant bias associated with increased %fat (DXA), only in girls (Po0.01). Conclusions: We conclude that ADP, at present, has unacceptably high limits of agreement compared to a criterion DXA measure. The ease of use, suitability for various populations and cost of ADP warrant further investigation of this method to establish biological variables that may influence the validity of body fat estimates. European Journal of Clinical Nutrition (2003) 57, 1402–1410. doi:10.1038/sj.ejcn.1601702 Keywords: air displacement plethysmography; dual-energy X-ray absorptiometry; body composition; children

*Correspondence: D Radley, Leeds Metropolitan University, School of Leisure and Sports Studies, Beckett Park Campus, Fairfax Hall, Leeds LS6 3QS, UK. E-mail: [email protected] Guarantor: D Radley. Contributors: DR performed data collection, statistical analysis and contributed to writing the paper. PJG was the chief project coordinator and contributed to writing the paper. CBC assisted in the design of the study, critically reviewed all parts of the paper and assisted in interpretation of the results. SC assisted in statistical analysis, interpretation of the results and critically reviewed all parts of the paper. BO performed technical measurements and contributed to writing the paper. JGT assisted in the design of the study and critically reviewed all parts of the paper. Received 6 June 2002; revised 7 November 2002; accepted 26 November 2002

Introduction Body composition analysis serves as an important measure for evaluating and monitoring nutritional status and health. As levels of childhood obesity rise (Rudolf et al, 2001), identification of appropriate and accurate methods of assessment will provide information that can be compared with other risk profile data, such as metabolic, fitness and fat distribution variables. Traditional methods for characterising the body utilise a two-component (2-C) molecular level model (Wang et al, 1995). Based on the Archimedes principle, it has long been recognised that the fat and fat-free components of the body can be derived from whole body density (Db), assuming

Estimates of percentage body fat in young adolescents D Radley et al

1403 constant densities of their constituents (Stern, 1901). It was not, however, until 1942 that Behnke et al perfected the measurement technique using hydrostatic weighing (HW). HW has, for many years, been considered the gold standard of body composition analysis (Going, 1996). HW is, however, time consuming and necessitates a high degree of both subject and technician skills, particularly in relation to the measurement of residual lung volume. In addition to the above factors cumbersome equipment makes it unsuitable for certain research facilities and populations (eg children and the obese). A recent alternative to hydrodensitometry for assessment of Db is air displacement plethysmography (ADP). The first commercially available ADP (ie the BOD POD) is quick, non-invasive, accommodating to a variety of populations and requires less technical expertise than HW (McCrory et al, 1995). Additionally, the cost makes it a viable tool for many institutes that are unable to conduct more ‘precise’ methods (eg magnetic resonance imaging, in vivo neutron activation analysis (IVNA), four-component (4-C) analysis and dual-energy X-ray absorptiometry (DXA)). The major disadvantage of techniques based upon the 2-C model is the fundamental assumption regarding the proportions of the water and bone mineral constituents of the fatfree mass (FFM). The most commonly used equation for conversion of Db to percentage body fat (%fat) assumes a constant FFM density of 1.10 g/cm3 (Siri, 1961). Children’s FFM, however, is acutely sensitive to variability in hydration status (Boileau et al, 1984) and bone mineral content (BMC) (Lohman et al, 1984). Lohman (1986) subsequently devised prediction equations utilising age- and gender-specific constants to estimate %fat from Db. To date, there have been relatively few studies considering the accuracy of ADP measurements in children (Nunez et al, 1999; Dewit et al, 2000; Fields & Goran, 2000; Lockner et al, 2000; Nicholson et al, 2001; Demerath et al, 2002) of which

only two (Fields & Goran, 2000; Nicholson et al, 2001) utilise age- and gender-specific constants and none of which were based solely on the adolescent age range under consideration in the present study. Therefore, the present study was designed to compare ADP and DXA measurements of %fat in young adolescents. DXA is a quick, safe and relatively noninvasive technique for measuring body composition. Originally developed for measuring BMC (Mazess et al, 1981), DXA is able to provide information about soft tissue composition and has quickly become established as a criterion measure of body composition (Bray et al, 2001). It has been shown to be reliable (Pintauro et al, 1996) and a valid standard, based on: carcass analysis of animal models (Svendsen et al, 1993; Ellis et al, 1994), and 4-C analysis (Prior et al, 1997) and DXA is a suitable criterion measure because it measures BMC and has been reported to be relatively unaffected by differences in hydration of the FFM (Going et al, 1993; Testolin et al, 2000). In addition, although DXA measurement is not assumption free (Pietrobelli et al, 1996), it is an independent measure of %fat, whereas the 4-C model incorporates Db.

Methods Subjects A total of 28 adolescents (12 males and 16 females) participated in the study. Subjects were recruited from the local community by advertisement in newspapers and doctors surgeries. The main characteristics of the sample are shown in Table 1. All measurements were performed on the same day within a 3 h period. The study protocol was approved by Leeds Teaching Hospital Research Ethics Committee, and both participant written informed assent and parental/guardian written informed consent were obtained prior to testing.

Table 1 Subject characteristics

Age (y) Body mass (kg) Height (m) BMI (kg/m2) DXA Lean tissue mass (kg) Fat mass (kg) Bone mineral content (kg) Body fat (%) BMC/FFM (%) ADP Body volume (l) Thoracic gas volume (l)

Males (n=12)

Females (n=16)

Combined (n=28)

14.8770.41 64.12711.02 1.7470.05 21.1472.91

15.0070.62 57.1578.78 1.6570.07 21.2272.92

14.9470.54 60.14710.23 1.6970.08 21.1972.87

50.0075.16 12.5879.52 2.5870.39 18.05710.87 4.9070.39

38.6375.18 16.9675.62 2.2770.40 28.7776.87 5.5370.48

43.5077.65 15.0877.71 2.4070.42 24.17710.17 5.2670.54

61.20711.58 3.6570.26

55.1778.72 3.1470.26

57.76710.30 3.3670.36

Values are means 7s.d.

European Journal of Clinical Nutrition


1404 Height and weight Body mass was measured to the nearest 0.01 kg using calibrated Tanita electronic scales, with participants wearing swimsuits. Height was measured to the nearest 0.1 cm using a floor-standing Seca stadiometer (model 220), with subjects standing erect without shoes. Body mass index (BMI) was calculated as weight/height2, where weight is expressed in kg and height in m (Quetelet, 1842).

DXA BMC, fat mass (FM) and lean tissue mass (LM) measurements were obtained using the GE/Lunar Prodigy densitometer. The Prodigy utilises a narrow fan beam (4.51) orientated parallel to the longitudinal axis of the body. A dual-energy X-ray source with peak X-ray energy of 80 kVp and a current of 3 mA with a K-edge filter (cerium 300 mg/cm2) gives effective energies of approximately 38 and 70 keV. The detector consists of an array of energy-sensitive cadmium zinc telluride detectors of length 5 cm (16 elements each 3 mm wide) allowing rapid photon counting. Imaging is typically over a 24 mm length with longitudinal steps of 17 mm. Pixel size for total body standard scan mode is 4.8 13.0 mm2 (Mazess et al, 2000). Scan time for a total body measurement is approximately 5 min. Software version 2.26 was used for analysis. %Fat was calculated as FM relative to total body mass (BM) as estimated by the Prodigy sum of parts (BM ¼ BMC þ FM þ LM). For ethical reasons repeat measurements were not performed on the children in this study. In our laboratory, the between-trial coefficient of variation (CV) for measurement of %fat in 10 adult subjects measured twice, with repositioning, on the same day is 2.7% (data not shown).

ADP Body volume (Vb) was measured using the BOD POD whole body ADP (Life Measurement Instruments, Concord, CA) and software version 1.69, in accordance with the manufacturer’s operation instructions. A detailed explanation of the measurement procedures has been described previously (Dempster & Aitkens, 1995). Briefly, prior to any measurements, a system testing calibration, using a cylinder of known volume (50.146 l) and manual calibration of the scales, using two 10 kg weights was completed. After the calibration, subjects were weighed, then entered the Bod Pod chamber wearing only a tight fitting swimsuit and swim cap. To calculate raw body volume (Vraw) two 50 s measurements were performed. If these measurements were within 150 ml they were accepted and the mean volume used for further calculations. If, however, they differed by more than 150 ml, a third test was administered. If two of the three measurements fell within 150 ml their mean was used for calculations, if not, the whole process (including calibration) was repeated. European Journal of Clinical Nutrition

Actual Vb was computed from Vraw correcting for a surface area artefact (SAA) (a term used to correct for the effects of isothermal air near the subject’s body surface area (BSA)) and thoracic gas volume (Vtg) (Dempster and Aitkens, 1995): Vb ðLÞ ¼ Vraw ðlÞ SAAðlÞ þ 0:40Vtg ðlÞ SAA was predicted by estimation of BSA using the Du Bois & Du Bios (1916) formula, multiplied by a negative constant k (personal communication, Life Measurement Instruments, Concord, CA, USA): SAAðLÞ ¼ kðl=cm2 Þ:BSAðcm2 Þ Vtg was predicted by estimation of functional residual capacity (FRC) and tidal volume (Vt) (McCrory et al., 1998): Vtg ðlÞ ¼ FRCðlÞ þ 0:5Vt ðlÞ Vtg was calculated by the supplied software using the FRC equations of Crapo et al (1982) and assumed Vt constants, 1.2 l for men and 0.7 l for women (personal communication, Life Measurement Instruments, Concord, CA). To obtain %fat estimates Vb was first converted to Db (body mass/Vb). %Fat was then calculated using the general equation of Siri (1961) (ADPSiri) and the age and gender specific equations of Lohman (1986) (ADPLoh). All equations are of the form X %fat ¼ Y 100 Db where X and Y are derived from the equation 1 d1 d2 d2 %fat ¼ 100 Db d1 d2 d1 d2 where d2 represents the constant density of fat, 0.9 g/cm3, and d1 represents the assumed density of the FFM (Dffm) compartment as given by Siri (1961) and Lohman (1986). In our laboratory the between-trial CV for measurement of %fat in 10 female adolescents measured twice on the same day is 7.6% (data not shown).

Statistical Analysis The relation between DXA and BOD POD %fat estimates was examined using paired sample t-tests, the correlation coefficient (r) and least-squares linear regression. The accuracy of the prediction of body fatness by the regression analysis was evaluated using the coefficient of determination (r2) and the standard error of estimate (s.e.e.). Agreement between body composition estimates was examined by two qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P ^ methods: (1) total error (TE) ðY YÞ2 =N , as described by Lohman (1981), where Y^ is %fat BOD POD, Y is %fat DXA and N is number of subjects, and (2) by calculation of the 95% limits of agreement as described by Bland and Altman (1986). Additionally, potential bias between %fat estimates by DXA and the BOD POD was examined using residual plots. Firstly, this analysis reveals whether differences in the estimates of %fat vary across the range of fatness, as a function of %fat DXA. Secondly, it determines whether


1405 differences between %fat estimates by DXA and ADP are related to variability in the BMC of FFM. For all analysis the alpha level adopted for statistical significance was Po0.05.

Results A summary of the results relating to the accuracy of and bias in measurement of %fat as assessed by ADP relative to DXA are presented in Table 2. When comparing all subjects, mean (7s.d.) %fat determined by ADPSiri (23.7479.12%) showed a nonsignificant difference compared to DXA (24.17710.17%). In contrast, %fat determined by ADPLoh (21.9179.15%) was significantly less (Po0.01) than that obtained from DXA. Combining males and females masks an important gender disparity. In males, %fat ADPSiri (18.73711.16%) produced a nonsignificant overestimation, and %fat ADPLoh (17.15711.42%) a nonsignificant underestimation, compared to %fat DXA (18.05710.87%). In females, %fat ADPSiri (27.5174.81%) produced a nonsignificant underestimation, and %fat ADPLoh (25.4874.84%) a significant underestimation (Po0.01), compared to %fat DXA (28.7776.87%). All %fat estimates by ADP were significantly correlated with %fat DXA (rZ0.84, Po0.001). In all cases the intercept and slope were not significantly different from zero and one, respectively. However, for all subjects, regression analysis highlighted the differences between the Siri (y ¼ 1.043x0.592) and Lohman (y ¼ 1.030x þ 1.612) equations. Comparison combining all subjects masked gender differences in the ADP estimated %fat in relation to those derived by DXA. Figure 1 illustrates the disparity in the regression for males and females. The variation around the regression line (SEE) was slightly greater for females (ADPSiri 3.68%; ADPLoh 3.89%) than males (ADPSiri 3.42%; ADPLoh 3.51%). TE, the actual mean difference between DXA %fat and ADP estimates, confirmed the regression findings. In all cases, the TE was greater for the Lohman compared with the

Siri conversion. Additionally, males produced a smaller TE (ADPSiri 3.28%; ADPLoh 3.49%) in comparison with females (ADPSiri 3.81%; ADPLoh 4.98%). As summarised in Table 3, the 95% limits of agreement were relatively similar for all %fat estimates, ranging from 76.57% to 77.58%. For all subjects ADPSiri produced slightly lower individual differences (77.12%) than ADPLoh (77.54%) (Figure 2). Analysing males and females separately highlighted that individual differences for males (ADPSiri76.57; ADPLoh76.89) were slightly less than for females (ADPSiri77.28; ADPLoh77.58). Residual plot comparisons considering whether differences in DXA and ADP estimates of %fat vary across the range of fatness as a function of %fat DXA for all subjects, revealed a significant ADP bias across the range of fatness (r ¼ ADPSiri 0.45; ADPLoh 0.44, Po0.05). However, considering all subjects masks the true bias in our sample. For males no significant ADP bias was evident, whereas among females there was a significant ADP bias (r ¼ ADPSiri 0.74; ADPLoh 0.73, Po0.01) with increasing body fat content (ADPSiri Figure 3). Additionally, residual plot comparisons considering whether differences in DXA and ADP estimates of %fat were related to variability in the BMC of FFM for all subjects, revealed a significant inverse relation (r ¼ ADPSiri 0.58; ADPLoh 0.58, Po0.01). However, considering all subjects masks the true relation in our sample. For males no significant bias was evident, whereas among females there was a significant bias (r ¼ ADPSiri 0.67; ADPLoh 0.63, Po0.01) with variability in the BMC/FFM ratio (ADPSiri Figure 4).

Discussion The present study was designed to investigate the ability of the BOD POD ADP to accurately assess %fat in young adolescents, compared to a DXA criterion measure. As expected, all ADP %fat estimates were significantly correlated

Table 2 Summary of the association between ADP and DXA %fat estimates Model

Mean (7s.d.)

r

r2

Intercept

Slope

s.e.e.

TE

All subjects (n=28) DXA ADPSiri ADPLoh

24.17710.17 23.7479.12 21.9179.15**

0.94*** 0.93***

0.87 0.86

0.592 1.612

1.043 1.030

3.68 3.91

3.59 4.40

Males (n=12) DXA ADPSiri ADPLoh

18.05710.87 18.73711.16 17.15711.42

0.95*** 0.95***

0.91 0.90

0.655 2.519

0.929 0.906

3.42 3.51

3.28 3.49

Females (n=16) DXA ADPSiri ADPLoh

28.7776.87 27.5174.81 25.4874.84**

0.86*** 0.84***

0.73 0.70

4.886 1.512

1.223 1.188

3.68 3.89

3.81 4.98

**Po0.01, ***Po0.001.



1406 Table 3 Mean differences 795% limits of agreement between ADP and DXA %fat estimates

50

All subjects (n=28)

Males (n=12)

Females (n=16)

0.4377.12 2.2677.54

0.6776.57 0.9076.89

1.2677.28 3.2977.58

40

DXA (%fat)

ADPSiri ADPLoh 30

20

10

0 0

10

20

30

40

50

ADPSiri (%fat) 50

DXA (%fat)

40

30

20

10

0 0

10

20

30

40

50

ADPLoh (%fat)

Figure 1 Regression between %fat determined by DXA and %fat estimated by ADP using the Siri and Lohman equations (dashed line: ’, male; solid bold line: n, female).

to those derived by DXA (r ¼ 0.84 to 0.95, all Po0.01). For all subjects, ADPSiri (r ¼ 0.94) is comparable to those previously reported in children and adolescents by Nunez et al (1999) (r ¼ 0.90) and Lockner et al (2000) (r ¼ 0.94). It has been suggested that the statistical principles established by Lohman (1981) be utilised to validate new methods of body composition analysis (Sinning et al, 1985; Clark et al, 1993). Lohman (1981) suggested analysis should include evaluation of mean7s.d., mean difference (MD), s.e.e. and TE. In addition, individual agreement in the European Journal of Clinical Nutrition

present study was analysed using the methods of Bland and Altman (1986), and bias assessed by residual plots. In the present study conversion of Db using the Siri (1961) equation led to mean %fat estimates for all subjects that agreed better with that determined by DXA, compared with the age- and gender-specific equations of Lohman (1986) (Table 2). However, considering all subjects masks an important gender division. In males, there was no significant difference between mean %fat estimates by ADP and that determined by DXA using the Siri (1961) or Lohman (1986) equations. The Siri (1961) equation overestimated %fat (0.7%) while the Lohman (1986) equations underestimated %fat (0.9%). In females, both the Siri (1961) and Lohman (1986) equations revealed an underestimation of %fat (1.3 and 3.3%, respectively), of which the ADPLoh was significantly (Po0.001) different from the DXA mean value. In children, HW comparisons with DXA also provide inconsistent results. Bray et al (2001) reported an underestimation of %fat in both males (2.0 and 7.4%) and females (2.0 and 7.5%), by the Siri (1961) and Lohman (1986) equations, respectively. In contrast, compared to DXA, the data reported by Roemmich et al (1997) suggest an overestimation of %fat by HW using the Siri (1961) equation and a male overestimation and female underestimation using the Lohman (1986) equations, the magnitude of error depending on gender and maturational stage. Similarly, Wells et al (1999) reported slight underestimation’s of %fat (0.1% males, 1.9% females) utilising age- and gender-specific equations. To date, only one study in children has directly compared ADP and DXA %fat estimated using the Lohman (1986) equations (Nicholson et al, 2001), and studies using the Siri equation are equivocal (Table 4). In males, the present study supports the results of Roemmich et al (1997), finding an ADPSiri %fat overestimation (0.7%). These findings are expected because the Siri (1961) equation assumes a greater Dffm (1.1 g/cm3) than would be found in children (Lohman, 1986), thus producing an overestimation of %fat. The smaller magnitude of the overestimation in the present study (0.7% compared to 5.2 and 3.4%), may be explained by the greater age in our sample (14.8770.41 y compared to 10.970.3 and 13.470.5 y), and thus a resultant Dffm closer to 1.1 g/cm3. The findings in females using the Siri (1961) equation are surprising. Similar to boys, it is expected that female adolescents have a Dffm lower than 1.1 g/cm3 (Lohman, 1986), therefore resulting in an overprediction of %fat. These findings are supported by Nunez et al (1999) and Nicholson


1407 15

ADPSiri - DXA (%fat)

10

5

0

-5

-10

-15 0

10

20

30

40

50

Mean DXA + ADPSiri (%fat)

15

ADPLoh - DXA (%fat)

10

5

0

-5

-10

-15 0

10

20

30

40

50

Mean DXA + ADPLoh (%fat)

Figure 2 Bland and Altman plots comparing body fat content determined by DXA and ADP for all subjects using the Siri and Lohman equations (’, male; n, female).

10 8

r = -0.07 (NS) r = -0.74 (p < 0.01)


6 4 2 0 -2 -4 -6 -8 0

10

20

30

40

50

DXA (%fat)

Figure 3 Residual plot displaying relationship between differences in %fat estimates by DXA and ADPSiri, as a function of %fat DXA (dashed line: ’, male; solid bold line: n, female). NS ¼ not significant.



1408 10

r = -0.34 (NS) r = -0.67 (p < 0.01)

8 6


4 2 0 -2 -4 -6 -8 -10 4.0

4.5

5.0

5.5

6.0

6.5

7.0

BMC/FFM (%)

Figure 4 Residual plot displaying relationship between differences in %fat estimates by DXA and ADPSiri as a function of the BMC/FFM ratio (dashed line: ’, male; solid bold line: n, female). NS ¼ not significant.

Table 4 Summary of studies comparing ADP and DXA %fat estimates in children and adolescents Mean %fat (ADP – DXA) Reference

n

Sex

Age range

Nunez et al (1999)

26 22 14 11 12 42 56 63 12 16

M F M F M F M F M F

6–19

Fields and Goran (2000)* Lockner et al (2000) Nicholson et al (2001) Present study

Siri (1961)

9–14

0.0 0.1 3.9

10–18

2.1

6.1–14 6.1–14 14.4–15.5 13.9–15.9

0.6 3.0 0.7 1.3

Lohman (1986)

5.1 8.5 0.9 3.3

*Data from the original study was re-evaluated and reported by Fields et al (2002).

et al (2001), who reported an underestimation ADPSiri in female subjects, by 0.1 and 3.0%, respectively. While these findings might suggest a systematic error in Vb measurements, in female subjects, it should be noted that at present it is not possible to determine if the underprediction is a function of sex per se or of body fat content (Fields et al, 2002). When converting Db to %fat using the Lohman (1986) equations, the present study found an underestimation in both males (0.9%; nonsignificant) and females (3.3%; Po0.01). It would be expected that the Lohman equations give closer %fat estimates to DXA than the Siri conversion because the equations are based on age- and gender-specific Dffm. As explained above, the underestimation in females subjects may, in part, be because of a systematic error in Vb measurement. Additionally, the large mean difference (3.3%) found in the present study, consistent with the HW findings of Roemmich et al (1997), suggests an underprediction of %fat by the Lohman (1986) equations. Fields et al (2002) have European Journal of Clinical Nutrition

recently outlined the necessity to consider gender independently in adult studies. The above findings and those of Nicholson et al (2001) indicate that independent analysis by gender is also essential in child and adolescent studies. To date few studies have assessed the accuracy of ADP Vb measurement in children. In the six papers that have compared ADP and HW in children, findings are equivocal (Demerath et al, 2002; Fields et al, 2002). Owing to the limited literature available and inconsistent findings, it is not at present possible to make any conclusive judgements concerning the accuracy of Vb by ADP in children. A recent review of ADP (Fields et al, 2002) highlighted that differences in findings are attributable in part to biological errors which have not been explained. Further investigation is required to establish biological and technical errors during ADP measurements, a comprehensive list of which have been suggested by Fields et al (2002). To further assess the accuracy of ADP %fat estimates, the s.e.e. for prediction of %fat by regression analysis was


1409 considered. The s.e.e.’s in the present study ranged from 3.42 to 3.91%, which are classified as very good and good, respectively, by Lohman (1992) and are similar to those reported previously, 3.41–4.10% (Nunez et al, 1999; Fields & Goran, 2000; Lockner et al, 2000). More recently, however, Lohman (1996) has suggested the s.e.e. must be o3% for a new method to be accepted as accurate. First described by Lohman (1981), Clark et al (1993) later suggested the TE as the best single measure for evaluating differences between a new and criterion measure. TE represents the true mean sample error between two measures. TE in the present study ranged from 3.28 to 4.98%, indicating a fair to poor level of agreement. The methods described by Bland and Altman (1986) were used to further examine the level of individual agreement between DXA and ADP %fat estimates. Where the TE reveals the true mean error between measures, the 95% limits of agreement reveal true variability among individuals. The relatively large limits of agreement found in the present study, ranging from 76.57 to 77.58%, denote a wide range of individual variability between ADP and DXA %fat estimates. Fields and Goran (2000) and Nicholson et al (2001) reported similar limits of agreement, ranging from 76.8% to 78.1% between ADP and DXA. It is important to note that these limits are because of inaccuracies in DXA, ADP Vb measurement and errors arising from the conversion of Db to %fat. Residual plot analysis firstly examined the effect of varying degrees of fatness on the accuracy of ADP estimates of %fat. It is evident that consideration of all subjects masks distinct gender diversity. Figure 3 illustrates that in males there was no significant ADP bias while in females there was a significant (Po0.01) ADP bias as body fat content increased. The negative association indicates an overestimation of %fat by ADP in leaner subjects and an underestimation of %fat in fatter subjects. As previously highlighted, because males were leaner than females it is not possible to determine whether the gender diversity is a result of gender per se or differences in body fat content. Residual plot analysis secondly examined the relation between variability in the BMC of FFM and differences in DXA and ADP %fat estimates. Again, consideration of all subjects masks distinct gender diversity. Figure 4 illustrates that in males there was no significant relationship while in females there was a significant (Po0.01) ADP bias as the BMC/FFM ratio increased. As would be expected, ADP tends to overestimate %fat in female subjects with a low BMC/FFM and underestimates %fat in female subjects with a high BMC/FFM, when compared to DXA. The reason a similar significant relation was not found in males is unclear. However, it should be noted that differences between DXA and ADP %fat estimates that are attributable to variability in the Dffm are a function of the cumulative effects of variation in mineral, water and protein content. Furthermore, in our opinion, the residual plot analyses must be interpreted with caution given the small sample size and difference in body fat content between males and females.

One variable that may be construed as a limiting factor in the present study is the use of software predicted estimates of thoracic gas volume (Vtg). Consistent with Nunez et al (1999), Wells et al (1999) and Lockner et al (2000), we found the procedure necessary to obtain Vtg was not well accepted by children, a large proportion being unable to perform the necessary manoeuvre after several attempts. McCrory et al (1995) and Demerath et al (2002) have previously reported that in adults, there is no significant difference between predicted and measured Vtg. In contrast, Lockner et al (2000) and Demerath et al (2002) found that in children the ADP software produced a slight overestimation, when compared to measured Vtg. In addition, Dewit et al (2000) reported a mean overestimation by the ADP software of 0.28 l when compared to child-specific prediction equations. Support for the use of predicted Vtg has been highlighted by Dewit et al (2000) and Fields et al (2002). Dewit et al (2000) noted that a potential disadvantage in using measured Vtg arises because the assessment procedure takes place separately from the Vb measurement. In their review, Fields et al (2002) stated the need for the validation of measured Vtg. Further, it should also be noted that because only 40% of Vtg enters the BOD POD formula used to calculate Vb, the impact on %fat values is small. An overprediction of 0.5 l Vtg, for a 60 kg adolescent would result in a 1.7% overestimation in %fat.

Conclusion Collectively the coefficients of determination, mean differences and s.e.e.’s found in the present study suggest the potential for accurate %fat estimates using ADP with young adolescents. In addition, its ease of use, suitability for various populations, low cost and lack of radiation make it a most promising tool for future body composition analysis. However, at present, ADP has unacceptably high limits of agreement compared to a criterion DXA measure. Further investigation is needed, not only to establish biological variables that may affect the accuracy of Vb measurement, but also to enable the derivation of more accurate conversion formulas from Db to %fat based on not only age and gender but also pubertal status, body fatness and race.

Acknowledgements This study was supported by the National Heart Research Fund, Suite 12D, Joseph’s Well, Leeds, LS3 1AB. References Behnke AR, Feen BG & Welhman WC (1942): The specific gravity of healthy man. J. Am. Med. Assoc. 118, 495–498. Bland JM & Altman DG (1986): Statistical methods for assessing agreement between two methods of clinical measurement. Lancet i, 307–310. Boileau RA, Lohman TG, Slaughter MH, Ball TE, Going SB & Hendrix MK (1984): Hydration of the fat-free body in children during maturation. Hum. Biol. 56, 651–666.



1410 Bray GA, DeLany JP, Harsha DW, Volaufova J & Champagne C (2001): Evaluation of body fat in fatter and leaner 10-y-old African American and white children: the Baton Rouge Children’s Study. Am. J. Clin. Nutr. 73, 687–702. Clark RR, Kuta JM & Sullivan JC (1993): Prediction of percent body fat in adult males using dual energy X-ray absorptiometry, skinfolds, and hydrostatic weighing. Med. Sci. Sports. Exerc. 25, 528–535. Crapo R, Morris AH, Clayton PD & Nixon CR (1982): Lung volumes in healthy non smoking adults. Bull. Eur. Physiopamol. Respir. 18, 419–425. Demerath EW, Guo SS, Chumlea WC, Towne B, Roche AF & Siervogal RM (2002): Comparison of percent body fat estimates using air displacement plethysmography and hydrodensitometry in adults and children. Int. J. Obes. Relat. Metab. Disord. 26, 389–397. Dempster P & Aitkens S (1995): A new air displacement method for the determination of human body composition. Med. Sci. Sports. Exerc. 27, 1692–1697. Dewit O, Fuller NJ, Fewtrell MS, Elia M & Wells JCK (2000): Whole body air displacement plethysmography compared with hydrodensitometry for body composition analysis. Arch. Dis. Child. 82, 159–164. Du Bois D & Du Bois EF (1916): A formula to estimate the approximate surface area if height and weight be known. Arch. Intern. Med. Clin. North. Am. 17, 863–871. Ellis K, Shypailo R, Pratt JA & Pomd WG (1994): Accuracy of dualenergy X-ray absorptiometry for body-composition measurements in children. Am. J. Clin. Nutr. 60, 660–665. Fields DA & Goran MI (2000): Body composition techniques and the four-compartment model in children. J. Appl. Physiol. 89, 613–620. Fields DA, Goran MI & McCrory A (2002): Body-composition assessment via air-displacement plethysmography in adults and children: a review. Am. J. Clin. Nutr. 75, 453–467. Going S, Massett M, Hall MC, Bare LA, Root PA, Williams DP & Lohman TG (1993): Detection of small changes in body composition by dual-energy X-ray absorptiometry. Am. J. Clin. Nutr. 57, 845–850. Going SB (1996): Densitometry. In Human Body Composition, eds A Roche, S Heymsfield & T Lohman, pp 3–23. Champaign, IL: Human Kinetics. Lockner DW, Heyward VH, Baumgartner RN & Jenkins KA (2000): Comparison of air-displacement plethysmography, hydrodensitometry, and dual X-ray absorptiometry for assessing body composition of children 10 to 18 years of age. Ann. N. Y. Acad. Sci. 904, 72–78. Lohman TG (1981): Skinfolds and body density and their relation to body fatness: a review. Hum. Biol. 53, 181–225. Lohman TG, Slaughter MH, Boileau RA, Bunt J & Lussier L (1984): Bone mineral measurements and their relation to body density in children, youths and adults. Hum. Biol. 56, 667–679. Lohman TG (1986): Applicability of body composition techniques and constants for children and youths. In Exercise and Sport Sciences Reviews, ed. KB Pandolf, pp 325–357. New York: Macmillan. Lohman TG (1992): Advances in Body Composition Assessment. Champaign, IL: Human Kinetics. Lohman TG (1996): Dual energy X-ray absorptiometry. In Human Body Composition, eds A Roche, S Heymsfield & TG Lohman, pp 63–78. Champaign, IL: Human Kinetics. Mazess RB, Peppler WW, Chesnut CH, Nelp WB, Cohn SH & Zanzi I (1981): Total body bone mineral and lean body mass by dualphoton absorptiometry. II. Comparison with total body calcium by neutron activation analysis. Calcif. Tissue Int. 33, 361–363. Mazess RB, Hanson JA, Payne R, Nord R & Wilson M (2000): Axial and total-body bone densitometry using a narrow-angle fan-beam. Osteoporos. Int. 11, 158–166.


McCrory MA, Gomez TD, Bernauer EM & Mole PA (1995): Evaluation of a new air displacement plethysmograph for measuring human body composition. J. Med. Sci. Sports Exerc. 27, 1686–1691. McCrory MA, Mole PA, Gomez TD, Dewey KG & Bernauer EM (1998): Body composition by air-displacement plethysmography by using predicted and measured thoracic gas volumes. J. Appl. Physiol. 84, 1475–1479. Nicholson JC, McDuffie JR, Bonat SH, Russell DL, Boyce KA, McCann S, Michael M, Sebring NG, Reynolds JC & Yanovski JA (2001): Estimation of body fatness by air displacement plethysmography in African American and white children. Pediatr. Res. 50, 467–473. Nunez C, Kovera AJ, Pietrobelli A, Heshka S, Horlick M, Kehayias JJ, Wang Z & Heymsfield SB (1999): Body composition in children and adults by air displacement plethysmography. Eur. J. Clin. Nutr. 53, 382–387. Pietrobelli A, Formica C, Wang ZM & Heymsfield SB (1996): Dualenergy X-ray absorptiometry body composition model: review of physical concepts. Am. J. Physiol. Endocrinol. Metab. 271, E941–E951. Pintauro SJ, Nagy TR, Duthie CM & Goran MI (1996): Crosscalibration of fat and lean measurements by dual-energy X-ray absorptiometry to pig carcass analysis in the pediatric body weight range. Am. J. Clin. Nutr. 63, 293–298. Prior BM, Cureton KJ, Modlesky CM, Evans EM, Sloniger MA, Saunders M & Lewis RD (1997): In vivo validation of whole body composition estimates from dual-energy X-ray absorptiometry. J. Appl. Physiol. 83, 623–630. Quetelet LAJ (1842): A treatise on man and the development of his faculties. In Comparative Statistics in the 19th Century, ed. T Sinbert, Edinburgh, Scotland: William and Robert Chambers. Roemmich JN, Clark PA, Weltman A & Rogol AD (1997): Alterations in growth and body composition during puberty. I. Comparing multicompartment body composition models. J. Appl. Physiol. 83, 927–935. Rudolf MC, Sahota PP, Barth J & Walker J (2001): Increasing prevalence of obesity in primary school children: cohort study. BMJ 322, 1094–1095. Sinning WE, Dolny DG, Little KD, Cunningham LN, Racaniello A, Siconolfi SF & Sholes JL (1985): Validity of ‘‘generalized’’ equations for body composition analysis in male athletes. Med. Sci. Sports. Exerc. 17, 124–130. Siri WE (1961): Body composition from fluid spaces and density: analysis of methods. In Techniques for measuring body composition, eds. J Brozek & A Henschel, pp 223–243. Washington, DC: National Academy of Sciences, National Research Council. Stern H (1901): Investigations on corporeal specific gravity and on the value of this factor in physical diagnosis. Med. Record. 59, 204–207. Svendsen O, Haarbo J, Hassager C & Christiansen C (1993): Accuracy of measurements of body composition by dual-energy X-ray absorptiometry in vivo. Am. J. Clin. Nutr. 57, 605–608. Testolin CG, Gore R, Rivkin T, Horlick M, Arbo J, Wang Z, Chiumello G & Heymsfield SB (2000): Dual-energy X-ray absorptiometry: analysis of pediatric fat estimate errors due to tissue hydration effects. J. Appl. Physiol. 89, 2365–2372. Wang ZM, Heshka S, Pierson Jr RN & Heymsfield SB (1995): Systematic organization of body-composition methodology: an overview with emphasis on component-based methods. Am. J. Clin. Nutr. 61, 457–465. Wells JC, Fuller NJ, Dewit O, Fewtrell MS, Elia M & Cole TJ (1999): Four-component model of body composition in children: density and hydration of fat-free mass and comparison with simpler models. Am. J. Clin. Nutr. 69, 904–912.