Anthropometry and methods of body composition measurement for research and eld application in the elderly

Transcription

1 (2000) 54, Suppl 3, S26±S32 ß 2000 Macmillan Publishers Ltd All rights reserved 0954±3007/00 $ Anthropometry and methods of body composition measurement for research and eld application in the elderly SB Heyms eld 1 *, C NunÄez 1, C Testolin 1 and D Gallagher 1 1 Obesity Research Center, Department of Medicine, St Luke's-Roosevelt Hospital Center, Columbia University College of Physicians and Surgeons, New York, NY, USA Evaluation of body composition is important in the study of human energy and protein metabolism as methods are available for quantifying energy stores and protein content at a single point in time; energy ± protein balance can be monitored over time; and dynamic measures of energy and protein metabolism can be referenced to body mass and related measurable components for between-individual comparisons. This review emphasizes the need for considering subject age when developing body composition component prediction models that are applied in elderly populations. An overview of body composition research is provided that emphasizes compartment and level de nitions and interrelations. Two broad method categories, mechanistic and descriptive, are then critically examined in relation to their role in energy-protein metabolism and aging research. Our collective review indicates that all major body composition components are now measurable using one or more methods that are based on non age-dependent assumptions. We also found that some methods, particularly descriptive eld methods (eg anthropometry), may be based on age-sensitive assumptions and measurements and suggestions for future development of these methods are provided. Lastly, as body composition differences between races, cultures, and countries are now recognized, it would be useful to create international cooperative groups with the aim of developing widely applicable descriptive eld methods based on simple available techniques such as anthropometry and bioimpedance analysis. Descriptors: anthropometry; body composition; aging (2000) 54, Suppl 3, S26±S32 `Science is fundamentally an exercise in measurement.' Dr Harold Varmus, NIH Director, Overview Evaluation of human body composition relates to the topic of this workshop in two respects: methods are available for quantifying energy stores and protein content at a single point in time and energy ± protein balance can be monitored over time; and dynamic measures of energy and protein metabolism can be referenced to body mass and related measurable components for between-individual comparisons. The speci c aim of this review is to emphasize the need for considering subject age when developing body composition component prediction models that are applied in elderly populations. Human body composition research The study of human body composition dates back several centuries, although major research interest evolved in the early years following World War II (Wang et al, 1999). The modern study of body composition encompasses three areas Ð body composition rules, methodology, and alterations (Figure 1). Each of these areas interacts with the other two, and this is particularly important in the study of aging. Alterations in body composition that occur with aging may, in turn, modify both component relationships or `rules' and body composition methods. *Correspondence: SB Heyms eld, Weight Control Unit, 1090 Amsterdam Avenue, 14th Floor, New York, NY 10025, USA. SBH2@Columbia.edu Human body components and their organization The rst area, body composition rules, involves component and level de nitions and also includes the study of quantitative component relationships. The 35 main body composition components are organized into ve levels of increasing complexity, atomic, molecular, cellular, tissuesystem, and whole body. The rst four levels and the respective main components at each level are depicted in Figure 2 (Heyms eld et al, 1998). Subjects who are weight stable over short time periods (ie several weeks) are in a body composition `steady state' in which component mass balance approaches zero. An important aspect of body composition research is establishing the component relationships that exist under these conditions. These include, for example, the water content of fat-free body mass (ie total body water=fat-free body mass) and the potassium content of fat-free body mass (ie total body potassium=fat-free body mass). These relationships are important and interesting because of their intrinsic scienti c value (Wang et al, 2000) and because they are often the formative basis of body composition model development (Wang et al, 1995). This is a critical area in the study of aging as component relationships may change with senescence, physical activity, and nutritional status. Body composition models developed in young subjects may therefore not be accurate in older individuals. Body composition method organization Body composition methods can be broadly classi ed as in vitro and in vivo (Wang et al, 1995; Heyms eld et al, 1996). This review only examines available in vivo

2 Figure 1 The study of human body composition: three research areas. From Wang et al (1995), with permission. Mechanistic methods are often based on models that have an underlying physical or biological basis. The typical form of mechanistic models is C ˆ b Q where b is usually a model relating Q to C. Some typical models are the ratios of total body water and potassium to fat-free body mass (ie TBW=FFM ˆ 0.73 and TBK= FFM ˆ 60 mmol=kg) (Wang et al, 1995). As with descriptive methods, age should be considered when developing mechanistic methods. S27 Available methods Mechanistic All of the methods in this category are presently formulated on mechanistic models. Figure 2 Some of the main components at the rst four body composition levels. From Heyms eld et al (1998), with permission. methods. In vivo methods can be further classi ed according to any one of several descriptors: measurement location (laboratory= eld); measurement frequency (static=dynamic); mathematical function type (model=descriptive); and portion of the body evaluated (regional=whole-body). The primary organization paradigm applied in this review is based upon mathematical function type. All in vivo methods are by necessity indirect (Wang et al, 1995). That is, some measurable somatic or physical property is exploited in quantifying the component of interest. A useful and informative approach for organizing body composition methods is based on mathematical function type. The basic formula for estimating components (C) in vivo is C ˆ f Q where f is mathematical function and Q is measurable quantity. Measurable quantities include various properties (eg tissue conductivity) and other components. Two broad categories of mathematical function are recognized, descriptive and mechanistic. Descriptive mathematical functions have the general form C ˆ a b Q where a and b are regression line intercepts and slopes, respectively. A reference method is selected for measuring the component of interest in a well-characterized subject group. The property or predictor component (ie Q) is also measured in the subjects and regression analysis is then used to develop the component estimation model. A key point is that descriptive models are population speci c. Thus, if methods are to be applied in the elderly, they must be developed and cross validated in an appropriate older population. In vivo neutron activation. Nitrogen, carbon, hydrogen, phosphorus, sodium, chlorine, calcium and oxygen are all measurable in vivo by a group of methods referred to as neutron activation analysis. A source emits a neutron stream that interacts with subject tissues. The resulting decay products of activated elements can be counted by detectors and elemental mass established. By linking elements with molecular level components using simultaneous equations, mechanistic models can be prepared for all major molecular level components (Heyms eld et al, 1996). For example, C, N and Ca can be used to solve for total body fat, protein and bone mineral mass (Kehayias et al, 1991). Although limited in availability, neutron activation analysis is extremely valuable in body composition research as there are no presently known age or gender-dependencies of currently applied models. Moreover, components such as total body protein are best evaluated using neutron activation methods (Chettle & Fremlin, 1984). Neutron activation methods thus afford the opportunity to quantify molecular level components without potential age-related bias in derived estimates. Whole-body 40 K counting. A small and constant percentage of total body potassium (TBK) is radioactive ( 40 K) and emits a characteristic g-ray. With appropriate shielding from background, this g-ray can be counted using scintillation detectors. As the ratio of 40 Kto 39 K is known and constant, 39 K and `total body potassium' can be estimated accordingly. All of the body's potassium is within the fat-free body mass component and the TBK=FFM ratio is reasonably stable in the same subject over time and between different subjects of the same gender. This allows development of mechanistic models based on assumed TBK ratios for sex and potentially age (Forbes, 1987). An example of the relationship between TBK and fatfree body mass is shown in Figure 3. The subjects are healthy weight stable adults and TBK and fat-free body mass were measured using 40 K whole-body counting and dual-energy X-ray absorptiometry (DXA), respectively. There is a strong correlation between TBK and fat-free body mass (r ˆ 0.97), although the regression line intercept is signi cantly different from zero. Multiple regression analysis with TBK as dependent variable reveals that, in

3 S28 for estimating the important visceral adipose tissue compartment (Abate et al, 1994). Figure 3 Total body potassium (TBK) vs fat-free body mass (FFM) in 190 healthy adults. The univariate regression model is shown in the gure and the multiple regression model is presented below the gure. addition to fat-free body mass, sex and age are signi cant predictor variables. Two important points arise from this example. First, this example highlights the critical importance of considering age in model development. That is, the multiple regression formula suggests that the ratio of TBK to fat-free body mass is lower in older subjects compared to their younger counterparts. Second, even though mechanistic TBK models (ie assumed stable TBK=FFM ratio) for estimating fat-free body mass have been used for half a century (Forbes, 1987), the highly signi cant (r ˆ 0.98) descriptive multiple regression model shown in the gure suggests that more than likely a descriptive model would give equally good or even more accurate fat-free body mass estimates. A likely explanation for these observations is a change with age in the proportion of organs and tissues that constitute fat-free body mass. Magnetic resonance imaging=computed tomography. Although available for over 25 y, computed tomography (CT) was only minimally applied in body composition research because of expense and radiation exposure (Sjostrom, 1991). The introduction of magnetic resonance imaging (MRI) and the technical re nements that followed led to the widespread use of this important method in body composition research (Fowler et al, 1991). Although still costly, MRI studies are safe and instruments are available in most medical centers. The importance of both CT and MRI is that both methods produce cross-sectional images of tissue-system level components at pre-de ned anatomic locations. Image analysis software then allows estimation of the adipose tissue, skeletal muscle, and other tissue-system level component pixels. Pixels, or `picture elements', translate to respective tissue areas. Acquiring images at prede ned intervals and then integrating tissue component areas permits reconstruction of whole-body components such as skeletal muscle mass (Ross et al, 1992; Kvist et al, 1986). CT and MRI can now quantify the mass of all major organs and tissues. These two imaging methods are based on mechanistic models that have no known age dependence. CT and MRI also afford the important and novel opportunity to quantify tissue-system level components. Lastly, both imaging methods provide the only presently fully accepted approach Hydrodensitometry=air plethysmography. An important early advance was the introduction by Behnke and his colleagues of a re ned accurate method of measuring body volume by underwater weighing (Going, 1996). Today there are a number of additional methods for measuring body volume, including air displacement plethysmography (McCrory et al, 1995). Behnke rst suggested and others later re ned the nowclassic `two-compartment' model for estimating total body fat and fat-free body mass (Behnke et al, 1942). Body volume is used in this model with the assumption of known and assumed constant densities of fat and fat-free body mass. While fat, or triglyceride, has a stable density in all mammals so far studied, the density of fat-free body mass can vary depending on the proportions of water, protein, and minerals (Lohman, 1986). Hydration and bone mineral mass change from birth onward and the density of the fat-free body mass compartment is thus likely not `constant' across the lifespan. Substantial changes in the density of fat-free body mass in adults were not observed by Visser et al (1997). Baumgartner and colleagues, however, suggest that subclinical edema and osteoporosis of variable degree in the elderly may produce large individual differences in the density of fat-free body mass (Baumgartner et al, 1995). The classic two-compartment mechanistic `body volume' method is thus susceptible to model errors when applied in elderly populations. Total body uids. Dilution methods are available for quantifying body uid and water spaces (Forbes, 1987). The most important of these methods is dilution of stable or radio-labeled water (Schoeller, 1996). There is a long-standing observation that the hydration of fat-free body mass is stable in mammals and humans in particular with TBW=FFM ˆ0.73 (Figure 4; Pace & Rathburn, 1945). A two-compartment model for estimating fat-free body mass from measured total body water can be developed from this assumed stable hydration of about 0.73 (Schoeller, 1996). Fat mass can then be calculated as body weight minus fat-free body mass. There are a number of studies that have examined the constancy of fat-free body mass hydration as a function of age (Schoeller, 1996; Wang et al, 1999). Hydration is high very early in life and declines to a mature level in early adulthood (Figure 5). There is then no or only a small measured increase in hydration from maturity onward. As with the related hydrodensitometry two compartment model, the two-compartment total body water model for estimating fat-free body mass may thus be inaccurate in individual or groups of elderly subjects. Dual-energy X-ray absorptiometry (DXA). The DXA method evolved from earlier single and dual photon absorptiometry methods for evaluating bone mineral (Mazess et al, 1990). DXA systems share in common an X-ray source that, after appropriate ltration, emits two effective photon energy peaks (Figure 6). The attenuation of the two energy peaks relative to each other depends on the elemental content of tissues through which the photons

4 S29 Figure 4 Fat-free body mass (FFM) hydration vs log body mass in nine mammals (left). Total body water vs FFM observed in nine human cadavers (right). From Wang et al (1999c), with permission. Figure 5 Fat-free body mass hydration versus age. The data points represent mean values from actual previous published studies and the smooth curve is a derived function based on the theoretical model of Wang et al. From Wang et al (1999c) with permission. pass. Bone, fat, and lean soft tissues are relatively rich in calcium=phosphorus, carbon, and oxygen, respectively (Figure 7; Pietrobelli et al, 1996). DXA systems are designed to separate pixels based on appropriate models and relative attenuation into bone mineral, fat, and lean soft tissue (Figure 6). There are no known factors, including hydration effects that signi cantly in uence the validity of Figure 7 Mass fraction of four elements (H, C, N, O) in three components (lean soft tissue, fat, and bone mineral). Residual includes remaining elements including calcium. DXA fat and bone mineral estimates. Excessive or de cient uid volume would be appropriately registered as changes in lean soft tissue (Pietrobelli et al, 1998). Accordingly there is no reason to speci cally challenge the validity of DXA body composition measurements in healthy elderly populations. In addition to availability, low per-subject cost, and safety, DXA provides the opportunity to reproducibly measure the mass of total body and regional components such as bone mineral that could not easily be measured in the past. Figure 6 Simpli ed schematic of DXA main features (left) and threecomponent model (right). Multi-component models. The limitations of two-compartment methods led Siri (1961), Anderson (1963), and others to propose more complex multicomponent methods. There are now several well accepted multicomponent methods (Table 1) that are designed to quantify three or more components in addition to those mentioned earlier for DXA, MRI and CT. Several generalizations regarding multicomponent methods are important when considering application of these methods to elderly populations. First, adding more measurements and components to the developed model usually reduces model error (Guo et al, 1996). For example, purported changes with aging in fat-free body mass hydration would cause model errors for total body water and hydrodensitometry two-compartment methods. These

5 S30 Table 1 Mechanistic models for estimating total body fat mass (kg) based on measured body weight and volume Model Measurable properties Known component (s) Two-compartment fat ˆ 4.956BV BW BV, BW none Three-compartment fat ˆ BV TBW BW BV, BW TBW Four-compartment fat ˆ BV mineral BW BV, BW mineral fat ˆ 2.756BV TBW mineral BW BV, BW TBW, mineral fat ˆ 2.756BV TBW Mo BW BV, BW TBW, Mo fat ˆ BV TBW Mo BW BV, BW TBW, Mo Abbreviations: BV, body volume (L); BW, body weight (kg); Mo, bone mineral; TBW, total body water (kg). Modi ed from Heyms eld et al (1996) with permission. model errors are minimized by measuring both body volume and total body water as proposed in the threecompartment model of Siri (1961). Combinations of body volume, total body water, and bone mineral mass now permit development of four or more component models (Heyms eld et al, 1996). The second generalization is that with increasing numbers of model components, measurement error increases (Guo et al, 1996). That is, each measurement of a component or property is accompanied by measurement error and this error is propagated to the nal component estimate. Multicomponent methods are important because they afford the opportunity to quantify more components of biological interest with less potential age-related model error. Descriptive An important observation arising from review of mechanistic methods is that some, although possibly not all, of the available methods provide accurate component estimates regardless of subject age. Hence, there are appropriate available reference methods for preparing descriptive body composition equations in elderly populations. Although there are several available descriptive methods, two will be reviewed here as examples, anthropometry and bioimpedance analysis (BIA). Anthropometry. Anthropometric methods apply skinfold thickness, circumference, and other somatic measurements to the assessment of component mass. Estimates of total body fat (Durnin & Womersley, 1974), skeletal muscle (Doupe et al, 1997), and other components are based on prediction models usually developed using simple or multiple regression analysis. A simple example reveals some of the important factors that should be considered when developing methods for use in elderly populations. Subjects were healthy adult men and women varying widely in age and body weight. Each subject had a head-totoe MRI scan consisting of about 40 ± 50 slices at 4 cm increments. Scans were analyzed for adipose tissue in each slice and subcutaneous, visceral and total body adipose tissue were calculated from the collective subject images. Subjects also had four skinfolds measured: biceps, triceps, subscapular, and iliac crest (Durnin & Womersley, 1974). The four skinfolds were summed as a measure of collective skinfold thickness. These four skinfold sites and the sum of the four measured skinfolds are used in the classic Womersley ± Durnin anthropometric method of estimating total body fat (Durnin & Womersley, 1974). Figure 8 Sum of four skinfolds (SF) vs subcutaneous (S) adipose tissue (AT) mass derived by MRI in 190 healthy adults. The multiple regression model is presented below the gure. The relationship between summed skinfold thickness and MRI-derived subcutaneous adipose tissue is presented in Figure 8. The correlation is quite strong, suggesting that the measured skinfolds are collectively a good measure of subcutaneous adiposity. However, multiple regression analysis, as shown in the gure, indicates that age is a signi cant independent variable in prediction of subcutaneous adipose tissue after controlling rst for skinfold sum. Older subjects have more subcutaneous adipose tissue for the same skinfold sum than do young subjects. Gender did not enter as a signi cant independent variable in the model. One possible explanation for this observation is that subcutaneous adipose tissue distribution changes with aging and therefore the selected skinfold sum translates to a different total subcutaneous adipose tissue mass in old subjects than it does in young subjects. Whatever the Figure 9 Sum of four skinfolds (SF) vs total adipose tissue (AT) mass derived by MRI in 190 healthy adults. The multiple regression model is presented below the gure.

6 explanation, age enters the model as a signi cant predictor variable. Extending this analysis to the next step, the association between skinfold sum and total body adipose tissue is presented in Figure 9. Again, the correlation is quite strong and supports the link between measured skinfold thickness and total body adiposity. However, multiple regression analysis also reveals a signi cant age term in the total body adipose tissue regression model presented in the gure. A portion of the age term presence must by necessity carry over from the age-related skinfold sum=subcutaneous adipose tissue relation presented for the subjects in Figure 8. However, there is another agerelated effect that is easily observed in the data. It has been known for some time that older subjects have a larger visceral adipose tissue mass than do young subjects (Borkan et al, 1982). Skinfolds, which provide a measure of subcutaneous adipose tissue, cannot account for variation in visceral adipose tissue when estimating total body adiposity. This possibility was explored by plotting the ratio of subcutaneous adipose tissue to total adipose tissue versus age in these subjects (Figure 10). It is clear that in both men and women, but particularly in men, the fraction of total adipose tissue sequestered in the subcutaneous compartment becomes lower with greater age. These observations provide only an example of what has been recognized for some time: adipose distribution, both within the subcutaneous compartment and between the subcutaneous and visceral compartments, varies with age (Borkan et al, 1982). Descriptive anthropometric methods must therefore be carefully developed and cross-validated for speci c use in older subjects. An important consideration with respect to all body composition methods, particularly the ones that are descriptive, is the source and nature of subjects on whom prediction formulas are developed. In particular, a prediction formula developed on one ethnic or age group may not be accurate when applied to another ethnic or age group. For example, in a recent study we observed signi cant differences in body fat, after controlling rst for age, gender, and body mass index, between Caucasian subjects in the USA and UK, and Asian subjects in Japan (Gallagher et al, 2000). Others have made similar observations when carrying out cross-ethnic and=or cross-country studies (Roche, 1995). Again, this highlights the potential population speci city of developed descriptive body composition methods. Bioimpedance analysis. The BIA method is widely applied in the eld as a means of estimating fat-free body mass (Baumgartner et al, 1990), total body water, and total body fat (ie body weight 7 fat-free body mass). BIA systems generate a single frequency alternating current and a pair of receiver electrodes (Foster & Lukaski, 1996) detects the voltage drop across a measured tissue bed. Multiple frequency BIA systems are also available (Foster & Lukaski, 1996) as are systems that measure impedance, resistance, reactance, and phase angle in speci c body segments and regions (Tan et al, 1997). Most simple BIA systems include body composition software based upon descriptive models. Virtually all comprehensive published BIA prediction formulas include an age term (Lukaski et al, 1985; Houtkooper et al, 1996). If lean tissues and their associated electrolyte containing uids are the primary electrical conductors, what explains the presence of age terms in various component prediction models? The answer to this question is not entirely clear, although several explanations are that: weakly conducting adipose tissue increases with age and may contribute to measured impedance in the elderly (Baumgartner et al, 1998); lean, particularly skeletal muscle, distribution may change with age, thus altering the nature of electrical pathways (Nunez et al, 1999); and muscle, the main conductor tissue in arms and legs, may change in composition and speci c resistivity with age. While these are speculations, the importance of age-related effects is that any developed BIA descriptive component prediction models should be developed and cross-validated in elderly subjects if they are to be applied in this population. Conclusion Evaluating the amount of and changes in energy stores and protein content is fundamental to the study of aging in relation to energy and protein metabolism. Quantifying the mass of `metabolically active' compartments is also an important aspect of aging research. There are now a number of available body composition reference methods that do not rely on age-dependent models and that can serve as standards for developing descriptive body composition methods. In summary: 1. Body composition changes with age and associated processes such as immobility and malnutrition. 2. Some currently assumed stable component relationships may actually change with age. 3. All major body composition components are measurable using one or more methods that are based on models that are not dependent on age-related assumptions. 4. Some methods, particularly descriptive eld methods, may be based on age-sensitive assumptions and measurements. These and other descriptive methods should be developed using component reference methods chosen for their insensitivity to subject age. 5. As body composition differences may exist between races, cultures, and countries, it would be useful to create international cooperative groups that develop widely applicable descriptive eld methods based on methods such as anthropometry and BIA. S31 Figure 10 Fraction of total adipose tissue (AT) as subcutaneous adipose tissue (SAT) vs age in 190 healthy adults. Acknowledgements ÐThis work was supported by National Institutes of Health grant PO1-DK

7 S32 References Abate N, Burns D, Peshock RM, Garg A & Grundy SM (1994): Estimation of adipose tissue mass by magnetic resonance imaging: validation against dissection in human cadavers. J. Lipid Res. 35, 1490 ± Anderson EC (1963): Three-component body composition analysis based on potassium and water determination. Ann. NY Acad. Sci. 110, 189 ± 212. Baumgartner RN, Chumlea WC & Roche AF (1990): Bioelectric impedance for body composition. Exercise Sport Science Review 18, 193 ± 224. Baumgartner RN, Heyms eld SB & Roch AF (1995): Human body composition and chronic disease. Obes. Res. 3, 73 ± 95. Baumgartner RN, Ross R, & Heyms eld SB (1998): Does adipose tissue in uence bioelectrical impedance in obese men and women? J. Appl. Physiol. 84, 257 ± 262. Behnke AR, Feen BG & Welham WC (1942): The speci c gravity of healthy men. J. Am. Med. Assoc. 118, 495 ± 498. Borkan GA, Gerzof SG & Robbins AH (1982): Assessment of abdominal fat content by computerized tomography. Am. J. Clin. Nutr. 36, 172 ± 177. Chettle DR & Fremlin JH (1984): Techniques of in vivo neutron activation analysis. Phys. Med. Biol. 29, 1011 ± Doupe MB, Martin AD, Searle MS, Kriellaars DJ & Giebrecht GG (1997): A new formula for population-based estimation of whole body muscle mass in males. Can. J. Appl. Physiol. 22, 598 ± 608. Dumin JV & Womersley GA (1974): Body fat assessed from total body density and its estimation from skinfold thickness measurements on 481 men and women aged 16 to 72 y. Br. J. Nutr. 32, 77 ± 97. Forbes GB (1987): Human Body Composition. New York: Springer. Foster KR & Lukaski HC (1996): Whole body impedance Ð what does it measure? Am. J. Clin. Nutr. 64, 388S ± 396S. Fowler PA, Fuller MF, Glasby CA, Foster MA, Cameron GG, McNeil G & Maughan RJ (1991): Total and subcutaneous AT distribution in women: the measurement of distribution and accurate predication of quantity by using magnetic resonance imaging. Am. J. Clin. Nutr. 54, 18 ± 25. Gallagher D, Heyms eld SB, Heo M, Jebb S, Murgatroyd P & Sakamoto Y (2000): Body mass index guidelines: corresponding % fat standards based on three country study (in press). Going SB (1996): Densitometry. In: Human Body Composition, ed. AF Roche, SB Heyms eld, TG Lohman, pp 3 ± 24. Champaign IL: Human Kinetics. Guo SS, Chumlea WC & Cockram DB (1996): Use of statistical methods to estimate body composition. Am. J. Clin. Nutr. 64, 428S ± 435S. Heyms eld SB, Wang ZM & Withers R (1996): Multicomponent molecular-level models of body composition analysis. In: Human Body Composition, ed. AF Roche, SB Heyms eld, TG Lohman, pp 129 ± 148. Champaign, IL: Human Kinetics. Heyms eld SB, Baumgartner RN, Ross R, Allison DB & Wang ZM (1998): Evaluatiion of total and regional body composition. In: Handbook of Obesity, ed. GA Bray, C Bouchard, WPT James, chap 3, pp 41 ± 77. New York: Marcel Dekker. Houtkooper LB, Lohman TG, Going SB & Howell WH (1996): Why bioelectrical impedance analysis should be used for estimating adiposity. Am. J. Clin. Nutr. 64, 436S ± 448S. Kehayias JJ, Heyms eld SB, LoMonte AF, Wang J & Pierson RN (1991): In vivo determination of body fat by measuring total body carbon. Am. J. Clin. Nutr. 53, 1339 ± Kvist H, Sjostrom L & Tylen U (1986): Adipose tissue volume determinations in women by computer tomography: technical considerations. Int. J. Obes. 10, 53 ± 67. Lohman TG (1986): Applicability of body composition techniques and constants for children and youths. Exerc. Sport Sci. Rev. 14, 325 ± 357. Lukaski HC, Johnson PE, Bolonchuk WW & Lykken GI (1985): Assessment of fat-free mass using bioelectrical impedance measurement of the human body. Am. J. Clin. Nutr. 41, 810 ± 817. Mazess RB, Barden HS, Bisek JP & Hanson J (1990): Dual-energy x-ray absorptiometry for total-body and regional bone-mineral and soft-tissue composition. Am. J. Clin. Nutr. 51, 1106 ± McCrory MA, Gomez TD, Bernauer EM & Mole PA (1995): Evaluation of a new air displacement plethysmograph for measuring human body composition. Med. Sci. Sports Exerc. 27, 1686 ± Nunez C, Gallagher D, Grammes J, Baumgartner RN, Ross R, Wang ZM, Thorton J & Heyms eld SB (1999): Bioimpedance analysis: potential for measuring lower limb skeletal muscle mass. JPEN 23, 96 ± 103. Pace N & Rathbun EN (1945): Studies on body composition (III): the body water and chemically combined nitrogen content in relation to fat content. J. Biol. Chem. 158, 685 ± 691. Pietrobelli A, Formica C, Wang ZM & Heyms eld SB (1996): Dualenergy X-ray absorptiometry body composition model: review of physical concepts. Am. J. Physiol. 271, 941 ± 951. Pietrobelli A, Wang ZM, Formica C & Heyms eld SB (1998): Dualenergy X-ray absorptiometry: fat estimation errors due to variation in soft tissue hydration. Am. J. Physiol. 274, E808 ± E816. Roche AF (1995): Body composition of ethnic groups in the US. Asia Paci c J. Clin. Nutr. 4, 63 ± 67. Ross R, Leger L, Morris D, de Guise J & Guardo R (1992): Quanti cation of adipose tissue by MRI: relationship with anthropometric variables. J. Appl. Physiol. 72, 787 ± 795. Schoeller DA (1996): Hydrometry. In: Human Body Composition, ed. AF Roche, SB Heyms eld, TG Lohman, pp 25 ± 44. Champaign IL: Human Kinetics. Siri WE (1961): Body composition from uid spaces and density: analysis of methods. In: Techniques for Measuring Body Composition, ed. H Brozek, A Henschel, pp 223 ± 244. Washington DC: NAS Ð National Research Council. Sjostrom L (1991): A computer-tomography based multicomponent body composition technique and anthropometric predictions of lean body mass, total and subcutaneous adipose tissue. Int. J. Obes. 15, 19 ± 30. Tan YX, Nunez C, Sun Y, Zhang K, Wang ZM & Heyms eld SB (1997): New electrode system for rapid whole-body and segmental bioimpedance assessment. Med. Sci. Sports Exerc. 29, 1269 ± Visser M, Gallagher D, Deurenberg P, Wang J, Pierson RN & Heyms eld SB (1997): Density of fat-free body mass: relationship with race, age, and level of body fatness. Am. J. Physiol. 272, 781 ± 787. Wang Z, Wang ZM & Heyms eld SB (1999a): History of the study of human body composition: a brief review. Am. J. Hum. Biol. 11, 157 ± 165. Wang ZM, Deurenberg P, Wang W, Pietrobelli A, Baumgartner R, Heyms eld SB (1999b): Hydration of fat-free body mass: review and critique of a classic body composition constant. Am. J. Clin. Nutr. 69, 833 ± 841. Wang ZM, Deurenberg P, Wang W, Pietrobelli A, Baumgartner RN & Heyms eld SB (2000): Hydration of fat-free body mass: new physiological modeling approach. Am. J. Physiol. (in press). Wang ZM, Heshka S, Pierson RN, Heyms eld SB (1995): Systematic organization of body composition methodology: overview with emphasis on component-based methods. Am. J. Clin. Nutr. 61, 457 ± 465.