A comparison of bioelectrical impedance and near infra-red interactance with dual energy x-ray absorptiometry for the determination of body fat. (Original Research).

Abstract

Objective: To compare measurements of percentage body fat obtained using bioelectrical impedance and near infra-red interactance assessments to those obtained using dual energy x-ray absorptiometry, a reference technique.

Design: Cross-sectional, cross-validation study. Percentage body fat was assessed using bioelectrical impedance analysis (Bodystat 1500, Tanita TBF-622 foot-to-foot analysers with proprietary equations and the equation of Kotler and colleagues) and near infra-red interactance (Futrex 1000). Results were compared with a Norland XR-26 dual energy x-ray absorptiometry scanner.

Subjects: Thirteen males (mean age 35 [+ or -] 14 years) and 17 females (mean age 33 [+ or -] 11 years).

Setting: University of Wollongong, New South Wales, Australia.

Main outcome measures: Assessment of percentage body fat using the different techniques.

Statistical analysis: Bias and limits of agreement using paired tests and correlation coefficients. Independent samples t test to assess differences between sexes.

Results: Bioelectrical impedance performed well. The foot-to-foot scales (mean 24.44 [+ or -] 8.53%), the Bodystat 1500 (mean 26.27 [+ or -] 8.57%) and the equation of Kotler et al. (mean 26.33 [+ or -] 7.79%) all gave mean values for percentage body fat that were similar to the mean value for dual energy x-ray absorptiometry (mean 24.86 [+ or -] 8.01%, P > 0.05). Near infra-red interactance gave an estimate of percentage body fat (30.37 [+ or -] 5.38%) significantly different from the dual energy x-ray absorptiometry value (P < 0.001).

Conclusion: Bioelectrical impedance analysis is an accurate, inexpensive method of assessing body composition. Further research is required before near infra-red interactance could be recommended for use in this subject population. (Nutr Diet 2002;59:120-126)

Key words: Body fat, adult, bioelectrical impedance, near infrared interactance, dual energy x-ray absorptiometry, nutritional assessment 

Introduction

Body composition assessment is a useful clinical tool for dietitians in clinical or research settings to provide feedback to patients or assess the effectiveness of nutritional interventions. Most body composition assessment methods divide the body into two compartments: fat mass and the fat free mass. Relative changes in these components of body composition are of interest in most nutritional interventions and can provide valuable information for outcomes-based research.

Direct analysis of body composition can only be made from sampling of cadavers and few studies have been published (1). Models for indirect measurement of body composition have been developed, however, no research reporting on a comparison between direct and indirect techniques has been published (2). Current methods of indirect body composition analysis can be divided into two types, criterion or reference methods and field methods. A review of currently available techniques can be found elsewhere (3).

The ideal body composition technique would be inexpensive, non-invasive, easily taught, universally applicable and capable of being performed with highly reproducible and accurate results (4). As a technique meeting all these criteria is not yet available, compromises must be made to choose a technique suitable for each situation (5). Criterion techniques are expensive either in the initial outlay for the equipment or the individual assessment and with a couple of exceptions are limited to a setting housing the equipment (usually a major teaching hospital or research centre). Field techniques of body composition have, therefore, been developed to allow cost-effective assessment of body composition in a variety of environments. Combinations of values obtained from field measures of body composition are used in prediction equations developed using the criterion methods for validation. From these analyses, equations can be derived to estimate body composition (6). However, prediction equations developed in one p opulation should not be used in another without cross-validation (5).

Bioelectrical impedance analysis (BIA) and near infrared interactance (NIR) are two relatively inexpensive field methods of assessing body composition. The aim of this study was to compare assessments of percentage body fat obtained using these two field methods to the reference method, dual energy x-ray absorptiometry (DXA).

Methods

Subjects were healthy adults recruited from the university academic and student community by an email advertisement. Exclusion criteria included in the advertisement included the use of a pace maker (exclusion criteria for BIA) or pregnancy (exclusion criteria for the DXA scan). The study was approved by the University of Wollongong human research ethics committee and all participants signed written informed consent. Height was measured using a two-metre stature measure which was fixed to the wall (Handy height scale, Mentone educational, Carnegie, Vic) and weight was measured using the Tanita TBF-622 body fat monitor (Tanita Corporation of Japan, Tokyo). Body fat as a percentage of weight was measured using a Norland XR-26 dual energy x-ray absorptiometry scanner (Norland Company, Fort Atkinson, WI), a Futrex fat analyser (Futrex 1000, Futrex Inc., Gaithersburg, MD), a Bodystat 1500 body composition monitoring unit (Bodystat Ltd, Douglas, Isle of Man) and the Tanita scales. All measures were performed on the same day.

Bioelectrical impedance analysis

Impedance is defined as 'the frequency dependent opposition of a conductor, animate or inanimate, to the flow of an administered alternating electrical current' (7,8). The impedance of the body can be measured by passing a small alternating electric current (< 1 mA) through the body and measuring the resulting potential difference (voltage drop). The current is administered and detected through electrodes on the right hand and foot or in the case of the Tanita model by foot pads. BIA assesses the amount of total body water. As the fat mass is assumed to be anhydrous the current only passes through the fat free mass which is assumed to be 73% water (9). The current is below the threshold of perception of the subject by a factor of more than 1000 (10). The measure of impedance (or one of its component vectors, resistance and reactance) is usually combined with height, weight, age and/or sex in a regression equation based on comparison with a criterion technique (usually hydrometry or underwater weighing) (5).

In this study three BIA equations were used. The first is the default equation used by the Tanita TBF-622 body fat monitor. This equation is proprietary and requires height, weight, sex, and a factor for adult or athelete. The Tanita model does not present any raw impedance data that could be used to predict body composition by an alternative equation. The second is the default equation utilised by the Bodystat 1500 body fat monitor. This equation requires height, weight, sex, age, and activity level (low, low to medium, medium, medium to high, very high) to be entered. The Bodystat 1500 reports the impedance on the display. The impedance value can then be used in other equations to predict body composition. In this study the impedance value obtained was used in the Kotler et al. equation (11), a published, validated and highly employed equation in both healthy and diseased subjects.

For males the equation is:

fat-free mass

= 0.50 x ([height.sup.1.48]/[impedance.sup.0.55] x 1.0/1.21) + 0.42 x (weight) + 0.49.

For females the equation is:

fat-free mass

= 0.88 x ([height.sup.1.93]/[impedance.sup.0.49] x 1.0/22.22) + 0.081 x (weight) + 0.07

where height is in centimetres and weight in kilograms.

Near infra-red interactance (NIR)

This technique is based on the principles of light absorption and reflection using near-infra-red spectroscopy. When electromagnetic radiation strikes a material, the energy is reflected, absorbed or transmitted depending on the specific properties of the material sampled (12). The Futrex 1000 emits electromagnetic radiation in the form of infra-red light at a single wavelength of approximately 940 nm into the biceps brachii muscle (personal communication Futrex Inc., 13 February 2001). A silicon-based detector then measures the intensity of the re-emitted light and compares the measure to that obtained from a calibration standard measured before each assessment. The interactance ratio is inverted, log transformed and the second derivative is calculated. The value obtained is expressed as optical density (5). The prediction equation for percentage body fat used by the Futrex 1000 is proprietary and not available from the manufacturers. The variables, other than the optical density, are weight and height and a sex factor is not included. The precision for the Futrex 1000 reported by the manufacturers is 0.41% body fat (pers. comm., Futrex Inc., 8 Feburary 2002).

Dual energy x-ray absorptiometry (DXA)

DXA directly measures the total body bone mineral content, the fat mass of soft tissue and the lean mass of soft tissue independently of other variables such as weight and height.

The Norland XR-26 scanner used in the present study uses a k-edge filter (samarium) to produce two photon peaks at 44 and 100 keV. This scanner makes a series of transverse scans from head to toe and the total scan takes about 20 minutes to complete. The precision for whole body fat estimates reported by the manufacturer is 2.4%. Bone mass is estimated from the relative attenuation of the two energies through bone-containing pixels. Each of the bone-containing pixels must be corrected for the overlying soft tissue mass. Beam attenuation through soft tissue depends on both the mass and proportion of fat and lean tissue components and constant attenuations have been derived for pure fat and lean tissue that are used in equations to calculate the fat and lean content (13). Studies have shown that the soft tissue attenuation ratio (the ratio of beam attenuation at the lower energy to higher energy) and percentage fat are inversely and linearly related (14). Therefore the percentage fat in soft tissue is estimated directly from the attenuation ratio. Percentage fat in the soft tissue multiplied by the soft tissue mass equals the fat mass. The lean tissue mass can then be calculated by deduction.

Data analysis

Data were analysed using the SPSS statistical package (SPSS Inc., Chicago, IL, version 10.0.5, 1999). Pearson product-moment correlation coefficients were calculated to determine the relationship between DXA and the three field techniques. Paired t tests were used to assess differences between the field techniques and DXA and the technique of Bland and Altman (15) was used to determine the mean bias and limits of agreement (mean [+ or -] 2SD). The differences between the two estimates (bias) were plotted against the mean percentage body fat of the two estimates (DXA and the field method) to examine the effect of increasing percentage body fat on the agreement between the two techniques (16). With complete agreement, the mean bias is zero, and the regression slope will be zero (i.e. there is no change in the bias with increases in percentage body fat). Independent sample t tests were used to determine if there were differences in bias between the males and females.

Results

Thirty subjects participated in the study. The subject characteristics are shown in Table 1. All participants were Caucasian and 22 participated in regular exercise. Most subjects participated in social team sports or walked for exercise. Two subjects were competitive long distance runners, one participated in competitive kayaking, several subjects participated in regular weight training or aerobic classes and eight participants were sedentary. The sample size is consistent with that reported in the literature for comparison studies of this nature, that is between ten and 77 subjects (17-24).

As shown in Table 2 all the prediction equations for the field methods correlated significantly with the reference method (DXA). The three equations for BIA used in this sample produced an estimate of percentage body fat that was not significantly different from the DXA measure. The NIR estimate significantly overestimated percentage body fat when compared with the DXA measure (P <0.001).

Figures 1 to 4 show the bias plots and limits of agreement (15) for the percentage body fat assessment using the field methods compared with the DXA measures. Figure 1 comparing the results using the Kotler equation with those from DXA results shows that the bias for this equation is positive (1.47%, see Table 2) and that most of the values fall randomly around this line (the slope is non-significantly negative), only one value falls outside the limits of agreement. Figure 2 shows that the bias is very small (-0.42%, see Table 2), the slope of the bias is non-significantly positive and only one value falls outside the limits of agreement. Figure 3 shows that the bias for the equation for Futrex 1000 is large (5.51%, see Table 2) and the slope is significantly negative, and again a single value falls outside the limits of agreement. Figure 4 shows the comparison of the estimate of body fat from Bodystat 1500 with the values from DXA, the bias is positive but small (1.41%, see Table 2), with a positive non-sign ificant slope, and a single value outside the limits of agreement. The bias using each of the field methods was not significantly different between the sexes (Kotler equation for BIA +1.48 kg female, +1.46 kg male, P = 0.987; Futrex 1000 NIR +4.82 kg female, +6.42 kg male, P = 0.497; Tanita BIA -0.79 kg female, +0.06 kg male, P = 0.694; Bodystat BIA +2.32 kg female, +0.21 kg male, P = 0.173).

Discussion

This study showed that BIA was accurate in assessing fat mass as a percentage of body weight in this sample population as the bias was small and the mean value obtained was not significantly different from the DXA value. Near infra-red interactance significantly overestimated percentage body fat on average, however, the bias decreased as the percentage body fat increased. These results suggest that BIA is suitable for measuring percentage body fat in our sample population, however further research is required before NIR could be recommended.

All the field methods showed a highly significant correlation with the DXA estimate of percentage body fat. However, it is the bias (average difference of the field methods estimate from the reference measure) and the limits of agreement that allow comparison of different techniques (15). In this study the Tanita scales gave an estimate of percentage body fat with the smallest bias, however, the limits of agreement were large, the Bodystat 1500 and use of the Kotler equation with BIA assessments had a larger bias with smaller limits of agreement. The Futrex 1000 NIR gave an estimate with both a large bias and limits of agreement. While the assessment provided using the Bodystat 1500, Kotler and Tanita BIA equations are acceptable, the results obtained using the Futrex 1000 NIR have discouraged us from further use of this equipment in our population setting.

To examine the existence of a relationship between the bias and the mean of the two techniques the correlation coefficient is used. If the bias changes across a range of different percentages of body fat this not only implies a change in the bias across body fatness in a cross-sectional analysis such as this one, it also presents doubt over the ability of the technique to measure with a constant bias in longitudinal studies where a change in body composition is anticipated, for example weight loss studies. The correlation coefficient for the bias versus the mean measurement was not significant for the three BIA techniques used in this study. The Futrex 1000 MR estimate of bias had a negative slope indicating that the bias decreased as percentage body fat increased. This slope was influenced by an outlier (the subject with the greatest body fat), nevertheless the relationship is consistent with a previous study reporting a similar relationship when NIR was compared with hydrodensitometry (25,26) and suggests t his technique would be inappropriate for use in a longitudinal study.

Previously a study in both healthy and hospitalised patients (27) with various conditions, using the same Bodystat 1500 model and equation as the present study, also showed good results compared with DXA with a bias of 0.07% body fat and 2SD limit of agreement of 7.76%. The equation that was developed by Kotler et al. used a large sample of healthy and chronically ill subjects of different ethnicities (11,28) and has been cross-validated in different ethnic groups and in a disease state (HIV) (11,23,28). Based on the previous application of these equations to a variety of subjects, their accuracy is not unexpected.

Most of the BIA equations in common use have been developed in white sedentary adults (29,30). Successful use of BIA in other populations, for example other ethnic groups or body composition types such as elite athletes, is likely to require the development of specialised prediction equations. Studies of cross-validation of equations in various age groups were reviewed by Houtkooper (29), and in both different age and racial groups were reviewed by Guo et al. (30). BIA has been trialled in many different clinical situations such as acromegaly (19), HIV/AIDS (23), Crohn's disease (20), haemodialysis (21), obesity (31,32), athletes (22,24,33-35), the elderly (18,36), and chronic respiratory insufficiency (37,38). Different BIA machines and many different equations have been used in these studies making the interpretation of results and comparison between studies difficult. This variation in prediction equations and equipment used reinforces the need to cross-validate the technique in each study population. BIA measures total body water, therefore, the use of this technique in conditions where abnormalities in body water compartments are common should be regarded with caution. The limitations of the use of BIA to assess body composition of malnourished patients during refeeding (39), critically ill patients (40), patients receiving dialysis (41) and patients with liver disease (42) have been reviewed.

Previous studies using the Futrex 1000 are limited. Most published research using NIR uses the dual wavelength model (Futrex 5000) as this is the model recommended by the company for research purposes. Smith et al. (26) compared the Futrex 1000 and Futrex 5000 (dual wavelength models) estimates for percentage body fat against those obtained using hydrodensitometry in female high school gymnasts. In this study the F1000 significantly overestimated percentage body fat by 4.5% (n = 89) compared to the estimates obtained using hydrodensitometry, whereas the F5000 model gave a mean estimate not significantly different from the hydrodensitometry result (-0.3%, n = 52). In general however, studies show that MR using the Futrex 5000 model overestimates percentage body fat in children and underestimates it in adults (17,31,43-47). Further development of the prediction equations for percentage body fat is necessary before this relatively new technique is recommended for routine body composition assessment.

The foot-to-foot method of body composition assessment is also a new technique and validation studies are limited. Jebb et al. (48) found that the Tanita foot-to-foot scales estimated fat mass with a bias of 0.8 kg when compared with a four compartment model (assessed using densitometry, hydrometry and DXA). Bell et al. (49) compared estimates of total body water (which may be extrapolated to obtain fat free mass and hence fat mass) obtained using the Tanita foot-to-foot scales and found a non-significant bias of 0.71 when compared with hydrometry in their group of 57 healthy subjects.

The Tanita analyser has some advantages compared with the standard hand-to-foot Bodystat monitor: assessments can be made with the subject standing rather than fully reclined; it is readily available and approximately one-tenth the cost; and electrocardiographic electrodes are unnecessary.

DXA was used as the reference method in the present study because it was readily available. The equipment is found in most hospital and private radiology practices in Australia. The use of DXA as a gold standard technique is controversial because of software differences between different types of DXA. However, DXA has high subject acceptability, a very low radiation dose and scanning time is only ten to 30 minutes.

Conclusions

Bioelectrical impedance analysis is a simple easy-to-use and relatively inexpensive technique that proved to be accurate for assessing percentage body fat in our adult population when compared with DXA. Further research is necessary before near infra-red intereactance could be recommended for accurate body composition assessment in our population.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

Table 1

The subject characteristics

                                   Male (n = 13)
                                 Mean [+ or -] SD

Age (years)                        35 [+ or -] 14
Weight (kg)                     79.05 [+ or -] 6.80
Body mass index (kg/[m.sup.2])  24.65 [+ or -] 1.95
Body fat DXA (%)                20.44 [+ or -] 5.40

                                  Female (n = 17)
                                 Mean [+ or -] SD

Age (years)                        33 [+ or -] 11
Weight (kg)                     63.24 [+ or -] 11.92
Body mass index (kg/[m.sup.2])  23.28 [+ or -] 4.80
Body fat DXA (%)                28.24 [+ or -] 8.15

                                All subjects (n = 30)
                                  Mean [+ or -] SD

Age (years)                         34 [+ or -] 12
Weight (kg)                      70.09 [+ or -] 12.69
Body mass index (kg/[m.sup.2])   23.87 [+ or -] 3.84
Body fat DXA (%)                 24.86 [+ or -] 8.01

DXA dual energy x-ray absorptiometry
Table 2

Correlation, bias and limits of agreement of the field methods compared
with dual energy X-ray absorptiometry (DXA) for assessing percentage
body fat mass

Method of
assessing                                                    Correlation
percentage                                         Paired t  coefficient
body fat     Mean      SD      Bias    Limits (a)  test (b)   with DXA

            (% body  (% body  (% body   (% body       (P
             fat)     fat)     fat)       fat)      value)

DXA          24.86    8.01
BIA          26.27    8.57     1.41       8.34      0.075       0.876
Bodystat
1500
BIA Kotler   26.33    7.79     1.47       8.52      0.068       0.855
BIA Tanita   24.44    8.53    -0.42      11.44      0.692       0.763
NIR Futrex   30.37    5.38     5.51      12.44     <0.001       0.631

Method of
assessing                Bias vs  Slope of
percentage                mean    the bias
body fat    Correlation  (slope)  vs mean

                                     (P
             (P value)             value)

DXA
BIA            0.000      0.138    0.468
Bodystat
1500
BIA Kotler     0.000     -0.054    0.778
BIA Tanita     0.000      0.097    0.612
NIR Futrex     0.000     -0.466    0.009

n = 30

BIA, Bioelectrical impedance analysis, Kotler, Kotler et al. equation
(11), Tanita, Tanita TBF-622 body fat scales, NIR Futrex, Futrex 1000
near infra-red interactance analyser.

(a) Limits refers to limits of agreement or 2SD from the mean.

(b) Significance of the difference of the mean percentage body fat from
the field method with the mean percentage body fat from the DXA scan.
This is the P value of the bias.

References

(1.) Clarys JP, Martin AD, Marfell-Jones MJ, Janssens V, Caboor D, Drinkwater DT. Human body composition: A review of adult dissection data. Am J Hum Biol 1999;11:167-74.

(2.) Brodie DA. Techniques of measurement of body composition. Part I. Sports Medicine 1988;5:11-40.

(3.) Ellis KJ. Human body composition: In vivo methods. Physiological Reviews 2000;80:649-80.

(4.) Garrow JS. New approaches to body composition. Am J Clin Nutr 1982;35:1152-8.

(5.) Lukaski HC. Methods for the assessment of human body composition: traditional and new. Am J Clin Nutr 1987;46:537-56.

(6.) Roche AF. Anthropometry and ultrasound. In: Roche AF, Heymsfeld SB, Lohman TG, editors. Human Body Composition. Champaign, IL: Human Kinetics; 1996.p.167-89.

(7.) Chumlea WC, Guo SS. Bioelectrical impedance and body composition: present status and future direction. Nutr Rev 1994;52:123-31.

(8.) Lukaski HC. Biological indexes considered in the derivation of the bioelectrical impedance analysis. Am J Clin Nutr 1996;64:397S-404S.

(9.) Wang ZM, Deurenberg P, Wang W, Pietrobelli A, Baumgartner RN, Heymsfeld SB. Hydration of fat-free body mass: review and critique of a classic body-composition constant. Am J Clin Nutr 1999;69:833-41.

(10.) Foster KR, Lukaski HC. Whole-body impedance--what does it measure? Am J Clin Nutr 1996;64:388S-96S.

(11.) Kotler DP, Burastero S, Wang J, Pierson Jr RN. Prediction of body cell mass, fat-free mass, and total body water with bioelectrical impedance analysis: effects of race, sex, and disease. Am J Clin Nutr 1996;64:489S-97S.

(12.) Conway JM, Norris CH, Bodwell CE. A new approach for the estimation of body composition: infrared interactance. Am J Clin Nutr 1984;40:1123-30.

(13.) Lohman TG. Dual energy X-ray absorptiometry. In: Roche AF, Heymsfield SB, Lohman TG editors. Human body composition. Champaign, IL: Human Kinetics; 1996. p.63-78.

(14.) Mazzess RB, Barden HS, Bisek JP, Hanson J. Dual-energy x-ray absorptiometry for total-body and regional bone-mineral and soft-tissue composition. Am J Clin Nutr 1996;51:1106-12.

(15.) Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307-10.

(16.) Wong WW, Cochran WJ, Klish WJ, Smith EO, Lee LS, Fiorotto ML, et al. Body fat in normal adults estimated by oxygen-18- and deuterium-dilution and by anthropometry: a comparison. Eur J Clin Nutr 1988;42:233-42.

(17.) Vehrs P, Morrow JR, Butte N. Reliability and concurrent validity of Futrex and bioelectrical impedance. Int J Sports Med 1998;19:560-6.

(18.) Bussolotto M, Ceccon A, Sergi G, Giantin V, Beninca P, Enzi G. Assessment of body composition in elderly: accuracy of bioelectrical impedance analysis. Gerontology 1999;45:39-43.

(19.) Brummer RJ, Lonn L, Bengtsson BA, Kvist H, Bosaeus I, Sjostrom L. Comparison of different body composition models in acromega]y. Growth Regulation 1996;6:191-200.

(20.) Royall D, Greenberg GR, Allard JP, Baker JP, Harrison JE, Jeejeebhoy KN. Critical assessment of body-composition measurements in malnourished subjects with Crohn's disease: the role of bioelectric impedance analysis. Am J Clin Nutr 1994;59:325-30.

(21.) Formica C, Atkinson MG, Nyulasi I, McKay J, Heale W, Seeman E. Body composition following hemodialysis: studies using dual-energy X-ray absorptiometry and bioelectrical impedance analysis. Osteoporos Int 1993;3:192-7.

(22.) Webster BL, Barr SI. Body composition analysis of female adolescent athletes: comparing six regression equations. Med Sci Sports Exerc 1993;25:648-53.

(23.) Batterham MJ, Garsia R, Greenop P. Measurement of body composition in people with HIV/AIDS: a comparison of bioelectrical impedance and skinfold anthropometry with dual-energy x-ray absorptiometry. J Am Diet Assoc 1999;99:1109-1l.

(24.) Clark RR, Kuta JM, Sullivan JC. Cross-validation of methods to predict body fat in African-American and caucasian collegiate football players. Res Q Exerc Sport 1994;65:21-30.

(25.) Fogelholm M, van Marken Lichtenbelt W. Review. Comparison of body composition methods: a literature analysis. Eur J Clin Nutr 1997;51:495-503.

(26.) Smith DB, Johnson GO, Stout JR, Housh TJ, Evetovich TK. Validity of near-infrared interactance for estimating relative body fat in female high school gymnasts. Int J Sports Med 1997;18:531-7.

(27.) Ghosh S, Meister D, Cowen S, Hannan WJ, Ferguson A. Body composition at the bedside. Eur J Gastroenterol Hepatol 1997;9:783-8.

(28.) Kotler DP, Thea DM, Heo M, Allison DB, Engelson ES, Wang J, et al. Relative influences of sex, race, environment, and HIV infection on body composition in adults. Am J Clin Nutr 1999;69:432-9.

(29.) Houtkooper LB, Lohman TG, Going SB, Howell WH. Why bioelectrical impedance analysis should be used for estimating adiposity. Am J Clin Nutr 1996;64:436S-48S.

(30.) Guo S, Chumlea WC, Cockram DB. Use of statistical methods to estimate body composition. Am J Clin Nutr 1996;64:428S-35S.

(31.) Heyward VH, Cook KL, Hicks VL, Jenkins KA, Quatrochi JA, Wilson WL. Predictive accuracy of three field methods for estimating relative body fatness of nonobese and obese women. Int J Sport Nutr 1992;2:75-86.

(32.) Fuller NJ, Sawyer MB, Elia M. Comparitive evaluation of body composition methods and predictions, and calculation of density and hydration fraction of fat-free mass, in obese women. Int J Obes 1994;18:503-12.

(33.) Hortobagyi T, Israel RG, Houmard JA, O'Brien KF, Johns RA, Wells JM. Comparison of four methods to assess body composition in black and white athletes. Int J Sport Nutr 1992;2:60-74.

(34.) Pacy PJ, Quevedo M, Gibson NR, Cox M, Koutedakis Y, Millward J. Body composition measurement in elite heavyweight oars-women: a comparison of five methods. J Sports Med Phys Fitness 1995;35:67-74.

(35.) Fornetti WC, Pivarnik JM, Foley JM, Fiechtner JJ. Reliability and validity of body composition measures in female athletes. J Appl Physiol 1999;87:1114-22.

(36.) Reilly JJ, Murray LA, Wilson J, Durnin JVGA. Measuring the body composition of elderly subjects: a comparison of methods. Br J Nutr 1994;72:33-44.

(37.) Pichard C, Kyle UG, Janssens J-P, Burdet L, Rochat T, Slosman DO, et al. Body composition by x-ray absorptiometry and bioelectrical impedance in chronic respiratory insufficiency patients. Nutrition 1997;13:952-8.

(38.) Kyle UG, Pichard C, Rochat T, Slosman DO, Fitting JW, Thiebaud D. New bioelectrical impedance formula for patients with respiratory insufficiency: comparison to dual-energy X-ray absorptiometry. Eur Respir J 1998;12:960-6.

(39.) Pencharz PB, Azcue M. Use of bioelectrical impedance analysis measurements in the clinical management of malnutrition. Am J Clin Nutr 1996;64:485S-8S.

(40.) Jacobs DO. Use of bioelectrical impedance analysis measurements in the clinical management of critical illness. Am J Clin Nutr 1996;64:498S-502S.

(41.) Kushner RE, De Vries PM, Gudivaka R. Use of bioelectrical impedance analysis measurements in the clinical management of patients undergoing dialysis. Am J Clin Nutr 1996;64:503S-9S.

(42.) Schloerb PR, Forster J, Delcore R, Kindscher JD. Bioelectrical impedance analysis in the clinical evaluation of liver disease. Am J Clin Nutr 1996;64:510S-4S.

(43.) Israel RG, Houmard JA, O'Brien KF, McCammon MR, Zamora BS, Eaton AW. Validity of near-infrared spectrophotometry device for estimating human body composition. Res Q Exerc Sport 1989;60:379-83.

(44.) Heyward VH, Jenkins KA, Cook KL, Hicks VL, Quatrochi JA, Wilson W, et al. Validity of single-site and multi-site models of estimating body composition of women using near-infrared interactance. Am J Hum Biol 1992;4:579-93.

(45.) McLean KP, Skinner JS. Validity of Futrex-5000 for body composition determination. Med Sci Sports Exerc 1992;24:253-8.

(46.) Cassady SL, Nielsen DH, Janz KF, Wu YT, Cook JS, Hansen JR. Validity of near infrared body composition analysis in children and adolescents. Med Sci Sports Exerc 1993;25:1185-91.

(47.) Wilmore KM, McBride PJ, Wilmore JH. Comparison of bioelectric impedance and near-infrared interactance for human body composition assessment in a population of self-perceived overweight adults. Int J Obes 1994;18:375-81.

(48.) Jebb SA, Cole TJ, Doman D, Murgatroyd PR, Prentice AM. Evaluation of the novel Tanita body-fat analyser to measure body composition by comparison with a four-compartment model. Br J Nutr 2000;83:115-22.

(49.) Bell NA, McClure PD, Hill RJ, Davies PS. Assessment of foot-to-foot bioelectrical impedance analysis for the prediction of total body water. Eur J Clin Nutr 1998;52:856-9.

(50.) Tothill P, Avenell A, Love J, Reid DM. Comparisons between Hologic, Lunar and Norland dual-energy X-ray absorptiometers and other techniques used for whole-body soft tissue measurements. Eur J Clin Nutr 1994;48:781-94.

Smart Foods Centre and Department of Biomedical Science, University of Wollongong, New South Wales

M. Batterham, PhD, APD, Research Fellow, Smart Foods Centre

L. Tapsell, BSc, DipNutrDiet, MHPEd, PhD, APD, Associate Professor

A. Jenkins, PhD, Associate Professor, Department of Biomedical Sciences

Correspondence: M. Batterham

Smart Foods Centre

University of Wollongong, Northfields Avenue

Wollongong, NSW 2522.

Email: [e-mail]