REVIEW Development of imaging methods to assess ... - Nature

International Journal of Obesity (2008) 32, S76–S82 & 2008 Macmillan Publishers Limited All rights reserved 0307-0565/08 $32.00 www.nature.com/ijo

REVIEW Development of imaging methods to assess adiposity and metabolism SB Heymsfield Global Center for Scientific Affairs, Merck & Co., Rahway, NJ, USA Body composition studies were first recorded around the time of the renaissance, and advances by the mid-twentieth century facilitated growth in the study of physiology, metabolism and pathological states. The field developed during this early period around the ‘two-compartment’ molecular level model that partitions body weight into fat and fat-free mass. Limited use was also made of X-rays as a means of estimating fat-layer thickness, but the revolutionary advance was brought about by the introduction of three-dimensional images provided by computed tomography (CT) in the mid 1970s, followed soon thereafter by magnetic resonance imaging (MRI). Complete in vivo reconstruction of all major anatomic body compartments and tissues became possible, thus providing major new research opportunities. This imaging revolution has continued to advance with further methodology refinements including functional MRI, diffusion tensor imaging and combined methods such as positron emission tomography þ CT or MRI. The scientific advances made possible by these new and innovative methods continue to unfold today and hold enormous promise for the future of obesity research. International Journal of Obesity (2008) 32, S76–S82; doi:10.1038/ijo.2008.242 Keywords: body composition; nutritional assessment; magnetic resonance imaging; nuclear magnetic resonance computed tomography

Introduction Early civilized humans viewed body composition as the four humors: black bile, yellow bile, blood and phlegm.1 Sanctorius, working in the early seventeenth century, was among the first to quantify insensible body water loss through his careful longitudinal weight measurements (Figure 1).1 Progress through the first half of the twentieth century provided investigators with advanced body composition methods that partitioned body weight into two body compartments, fat and lean body mass, now referred to as fat-free mass.1,2 Among these methods were whole-body 40K counting and underwater weighing, considered the ‘reference standards’ of their era for body fat measurement.3 These methods, however, were based on assumptions such as the constancy of fat-free mass potassium content and density, respectively.2,3 The two-compartment model assumptions were recognized as of limited accuracy under certain conditions, for example, with disturbances in hydration. Simple and practical methods were also available to early investigators, such as anthropometry techniques based on Correspondence: Dr SB Heymsfield, Clinical Research, Metabolism, Merck Research Laboratories, 126 E. Lincoln Avenue, PO Box 2000, RY34A-A238, Rahway, NJ 07065-0900, USA. E-mail: [email protected]

skinfold and circumference measurements.4 New developments during this period in radiographic methods led Stuart et al.5 to report in 1940 the first use of conventional X-ray films to capture fat and muscle ‘shadows’. With the arrival of the 1960s, ‘radiogrammetry’, as it was referred to, was by then a well-developed method providing investigators with estimates of subcutaneous adipose tissue layer ‘thickness’ and muscle ‘widths’.6 The low contrast and two-dimensional nature of these radiographic images limited the radiogrammetry approach to rough approximations of regional adiposity-related body compartments. Investigators during this formative period had a limited capability to fully quantify body composition as we recognize it today.7 Only the rudimentary anthropometric and radiogrammetric methods were available for quantifying adipose tissue and related components at the ‘tissue-system’ level of body composition (Figure 2).

Imaging revolution Computed tomography Historical overview. Matters were to change radically early in the 1970s, and the impact of these and related discoveries continues unabated today. The United Kingdom engineer

Imaging method development SB Heymsfield

S77 adipose tissue and lean tissues have CT numbers of 1000, approximately 70 and 20, respectively. A CT number is assigned to each of the image picture elements (pixels) or volume elements (voxels) that give the image contrast. The image is usually reconstructed using mathematical techniques based on two-dimensional Fourier analysis or filtered back-projection alone or in combination.

Figure 1 Santorio Sanctorius (1561–1636), the seventeenth century Paduan physician who spent many years eating, sleeping and working in a ‘statical chair’ connected to the arm of a calibrated scale. Sanctorius described insensible fluid losses and he paved the way for the contemporary science of metabolism.

Geoffrey N Hounsfield first introduced computed tomography (CT) for brain imaging in 1971, and he reported his seminal findings in 1973.8 The conceptual framework and mathematical algorithms required for cross-sec4tional image reconstruction were developed earlier by Raden in 19179 and subsequently by Cormack in 1963.1 Physical basis. The basic CT system as originally conceived by Hounsfield consists of an X-ray tube and receiver that rotate in a plane perpendicular to the subject. The generated X-rays are attenuated as they pass through tissues, an effect expressed as the linear attenuation coefficient or CT number, a measure of photon attenuation relative to air and water. The air and water CT numbers are empirically defined as 1000 and 0 Hounsfield units (HU), respectively. The X-ray beam is attenuated by coherent scattering, photoelectric absorption and Compton interactions. The main determinant of attenuation and thus CT number is physical density, and a linear correlation is present between CT number and tissue density.10 With water at 0 HU, air,

Early applications. Unlike the crude radiogrammetric approach, even early CT scanners provided high-resolution cross-sectional images through any region of the body. Clear definable anatomic boundaries could be identified between adipose tissue, skeletal muscle, visceral organs, brain and skeleton. Heymsfield and colleagues during this period were exploring the validity of simple anthropometric assessment methods, as they applied to seriously ill hospitalized patients. The installation of a CT scanner in the investigators’ hospital provided an important opportunity to examine anthropometric assessment methods by applying CT measurements as the ‘reference’. The group’s use of CT to estimate skeletal muscle mass was reported in 1979,11 and subsequent reports described methods of measuring visceral organ volumes,12 visceral adipose tissue,13 and liver fat.10–13 Widespread recognition of CT-measured adipose tissue was prompted by the seminal 1982 publication of Borkan et al.14 detailing age-related differences in adipose tissue distribution. Tokunaga et al.15 in 1983 reported an early segmentation approach for deriving whole body and regional adipose tissue volumes from measured cross-sectional CT images ¨ stro ¨ m et al.16 reported a series of studies on (Figure 3). Sjo their ‘whole-body’ multicomponent CT model in the mid1980s. Technological advances. During the early period of CT, as more instruments became available, epidemiological studies began to appear that linked ‘upper body’ adipose tissue distribution and visceral adipose tissue with obesity comor¨ rntorp17). There arose a clear bidities (for example, Bjo rationale for quantifying visceral adipose tissue, and CT rapidly became the method of choice. With more than two decades elapsing since these early efforts, the advances in CT imaging are breathtaking. Scanners have proliferated worldwide, thus providing instrument access to most interested investigators. Some relevant examples of important developments include the following: greatly improved image resolution with lower radiation exposure; introduction of spiral or ‘helical’ methods that allow capture of whole-body regions with rapid organ and tissue volume reconstructions; gated scanning protocols with capture of high-resolution images of structures such as the heart and major blood vessels. Accurate flow estimates, including those derived using radiographic contrast agents; micro-CT methods for small animal and bone evaluation; dual-energy CT for vascular and bone imaging; and development of protocols for quantifying hepatic lipid content. International Journal of Obesity


S78

Head

N & Other Elements

Minerals, CHO and other molecules

Protein

Hydrogen

Extracellular Solids Extracellular Fluid

Lipid Carbon Adipocytes Water Oxygen

Other Tissues Visceral Organs

Trunk

Bone Skeletal Muscle

Appendages

Adipose Tissue

Cells

Whole-Body Level

Tissue-Organ Level Cellular Level Molecular Level Atomic Level Figure 2 The five level model of body composition. N, nitrogen; CHO, carbohydrates. Adapted with permission from Wang et al.7

childbearing women and pregnant women. The introduction of non-X-ray magnetic resonance imaging (MRI) in the 1980s was thus a welcome addition to those interested in studying regional and whole-body adiposity.

Magnetic resonance imaging Historical overview. The historical roots for MRI appeared even earlier than those for CT. Bloch et al.18 and Purcell et al.19 independently described the fundamental phenomena that ultimately led to nuclear magnetic resonance spectroscopy in 1946. Differences between malignant tumors and normal tissue were studied using magnetic resonance methods in 1971 by Damadian,20 who patented his system in 1974 (Figure 4). Lauterber21 first reported the creation of phantom images using the signal from proton nuclear magnetic resonance in 1973, and Mansfield22 showed how the MRI radio signals could be mathematically transformed into a useful image in 1977. By the early 1980s, MRI systems became available on a widespread basis.

Figure 3 The whole-body adipose tissue reconstruction method based on computed tomography proposed by Tokunkuna et al.15 in 1983.

An obstacle to the use of CT is radiation exposure, limiting ‘whole-body’ longitudinal human studies, particularly those that include closely spaced cross-sectional images and largely excluding non-clinical CT use in young children, potential International Journal of Obesity

Physical basis. Proton MRI, the most common form applied in research and clinical settings, is based on the interaction among nuclei of hydrogen atoms with the system’s generated magnetic fields. The nuclei of hydrogen, protons, have a non-zero magnetic moment that leads them to behave similar to small magnets. These magnetic moments are randomly oriented in the earth’s weak magnetic field, and the net result is that they tend to cancel. Once inside the MRI scanner’s magnet, however, the field strength is about 10 000 times greater than that of the earth, and the proton magnetic moments align with the magnet’s field. Aligned


S79

Figure 4 Raymond Damaidian’s apparatus and method for detecting cancer in tissue, an early precursor to modern magnetic resonance imaging (MRI) scanners. The image is from the US Patent and Trademark Office, patent no. 3789832, filed 17 March 1972, issued 5 February 1974.

protons are then ‘pulsed’ with a radio-frequency signal and some protons absorb this energy and flip. Turning off the radio-frequency field allows the protons to gradually return to their original positions and to release the absorbed energy in the process. The ‘relaxation times’ describe the rate at which the absorbed energy is released. Measured relaxation times are used to generate MRIs. Early applications. Foster et al.23 in 1984 were the first to report that compared with cadaver samples, MRI could be used to distinguish adipose tissue from the adjacent skeletal muscle. MRI was first used several years later in 1988 to characterize the distribution of human subcutaneous adipose tissue.24 Many studies have since used MRI methods to describe adipose tissue and/or skeletal muscle tissue distribution among in utero, newborns, adolescents, normal adult males and females, obese males and females, metabolic syndrome and diabetic patients and elderly patients.25 Often the observations in these studies, similar to CT, are made with a single cross-sectional image. As multiple images can be obtained without any recognized health risks to the patient, MRI is ideally suited for assessment of whole-body adipose tissue or skeletal muscle tissue distribution. An important benefit of multislice image acquisitions is that the volumes of various tissues can be derived, thus eliminating the restrictions imposed by observations based on a single image. Fowler et al.26 in 1991 acquired 28 MRIs over the entire body and Ross et al.27 in 1992 and 199328 reported using a 41-image model to measure adipose tissue distribution in healthy adults. Technological advances. As with CT, spectacular advances have also been made with magnetic resonance technology over the past two decades. Image contrast has improved,

greater field strength magnets have been installed, contrast agents became available, and gated imaging was introduced. Additionally, important advances to those working in the obesity field include the following: functional MRI for tracking the metabolic activity of organs and tissues;29 diffusion tensor imaging for tracing muscle fiber and neuronal tissue tracts;30 proton and phosphorous spectroscopy;31 water-fat imaging for discerning tissue lipid and water content;32 sodium MRI; and whole-body nuclear magnetic resonance systems for quantifying total body lipid and water in animals and humans.33 Moreover, the concept of ‘visceral adipose tissue’ as a single compartment is advancing to the recognition of the existence of many anatomically localized depots with unique metabolic activities such as those located in bone marrow, epicardial area, around the kidneys (that is, peri and para-renal), and in the retro-orbital space.34 Although some of these compartments can be directly visualized and therefore segmented into their component areas, in some cases contiguous adipose tissue compartments cannot be easily identified as to their origin. For example, within the visceral compartment, peritoneal and mesenteric adipose tissue depots are not readily separated from the larger intra-abdominal adipose tissue compartment. Accordingly, methods are still needed to improve the separation of adipose tissue compartments into their appropriate anatomic and functional groups.

Combined anatomyFmetabolism evaluations Both CT and MRI provide excellent anatomic detail when combined with other methods appropriate for evaluating metabolic processes in vivo. The most notable of these is the combined use of positron emission tomography with CT to International Journal of Obesity


S80 localize specific metabolic processes.35 Many positron emission tomography tracers are available and provide investigators to go beyond ‘whole-body’ physiological studies to specific organs and tissues. These new combined methods are breakthrough technologies for evaluating an array of obesity-related topics including visualization of brown fat,36 weight loss drug receptor binding occupancy37 and energy expenditure of specific organs and tissues. Although these technological advances are providing novel scientific insights, a critical issue is their limited availability to most potential investigators. These approaches tend to be complex, costly and require multidisciplinary teams for their implementation. The path forward requires consideration of these issues at a national and international level.

Dual-photon absorptiometry Both CT and MRI procedures tend to be costly, and their availability is typically limited in busy hospital settings. Dual-photon absorptiometry methods that overcome this limitation and also provide novel body composition measurements were introduced for clinical and research applications within the same decade as CT and MRI.

Historical overview The renowned Harvard surgeon, Francis Moore, tried with little success more than five decades ago to estimate skeletal weight in vivo.38 In vivo neutron activation analysis, introduced in 1964,39 subsequently led to the first analysis of skeletal weight in living subjects. The relatively high radiation exposure and complexity of neutron activation analysis provided the rationale for a new strategy of estimating bone mineral, first reported in 1963 by Cameron and Sorenson.40 Cameron and Sorenson placed a radionuclide source on one side of the wrist, a body region with minimal soft tissue, and a photon detector was positioned on the opposite side of the wrist. Elements within the bone mineral compartment attenuated the photon beam, and the reduced flux was measured by the opposed photon detector. Single-photon absorptiometry, as it was referred to, was an important development in the assessment of wrist bone mineral content. Hip and spine fractures are common manifestations of osteoporosis, and there was a need to study bone mineral in these regions in addition to that of the wrist. When photons pass through soft tissues overlying hip and vertebral bones, they are attenuated by bone minerals and additionally by the traversed fat and lean soft tissues. Establishing bone mineralization required removal of the soft tissue contribution to total photon attenuation. This problem was subsequently solved by considering the photon pathway as a mixture of two components consisting of bone minerals and soft tissue. International Journal of Obesity

Two photon energies, usually provided by radionuclide sources, were used to evaluate the bone mineral and soft tissue content of each pixel. Referred to originally as ‘dualphoton absorptiometry’, the development of X-ray photon sources led to the methods now referred to as dual energy X-ray absorptiometry (that is, ‘DXA’ or ‘DEXA’).41 Today DXA systems are available worldwide for regional and total body bone mineral evaluation.

Early applications Dual-photon systems required an estimate of fat and lean soft tissue in pixels adjacent to bone to provide a soft tissue attenuation estimate for pixels overlying bone. Whole-body DXA systems thus apply the attenuation characteristics to solve for three components, fat, lean soft tissue and bone mineral.42 The currently available whole-body DXA systems combined with appropriate software can estimate these three components for the whole-body and selected regions. As appendicular fat-free soft tissue is primarily muscle, DXA also can provide regional and whole-body skeletal muscle mass estimates.43

Technological evolution Early DPA systems required long scan times and frequent replacement of the radionuclide source. The introduction of DXA systems combined with advanced software greatly increased the usefulness of this technology for evaluation of total body and regional body composition. The original DPA and DXA systems were based on a ‘pencil beam’ source and detector. Newer systems frequently collect data using a fan beam approach that reduces the scan time. Scanning platforms are increasing in width, and wider overweight and obese patients can therefore usually be fully accommodated. Alternatively, half- or partial-body scans are also appropriate for obese patients whose size exceeds that of the scanner marker guides.44 Many studies now report the accuracy and precision of DXA body composition estimates. Accordingly, DXA has emerged as the adiposity evaluation method of choice for clinical trials, epidemiology studies and individual patients. The introduction of whole-body DXA systems provided the opportunity to realize a long-held vision of developing an agreed-upon reference method for quantifying body fat. Two-compartment methods, such as the one based on underwater weighing, rely on models that are known to have limitations under certain conditions. Siri45 and others in the 1950s introduced a three-compartment model based on combined measurements of body density and total body water. The availability of DPA and later DXA provided the first practical measure of bone mineral mass and related total body mineral. This major advance then provided the opportunity to partition body mass into four components (fat, total body water, minerals, and residual mass) using measured body density, total body water, and bone mineral mass.46 This model is based on very robust assumptions and


S81 is often considered the reference against which other methods are compared.

Conclusions This 20th anniversary of the Pennington Center is a benchmark reflecting evolution of the global obesity epidemic and concordant efforts to understand and treat human obesity. Timed with this major public health and research event are the introduction and evolution of imaging methods that provide a unique window into the adipose tissue compartment and related metabolic processes. The scientific advances brought about by these methods continue to unfold today and hold enormous promise for the future of obesity research.

13

14

15

16

17 18 19 20

Acknowledgements I acknowledge the outstanding creative and supportive collaborations that have provided the basis for my research program over the past three decades.

21

22 23

24

Conflict of interest Steven B Heymsfield has declared no financial interests.

25

26

References 1 Wang ZM, Wang ZM, Heymsfield SB. History of the study of human body composition: a brief review. Am J Hum Biol 1999; 11: 157–165. 2 Heymsfield SB, Lohman TG, Wang ZM, Going SB. Human Body Composition, 2nd edn. Human Kinetics: Champaign, IL, 2005. 3 Forbes GB. Human Body Composition. Growth, Aging, Nutrition and Activity. Springer-Verlag: New York, 1987. 4 Lohman TG, Roche AF, Martorell R (eds). Anthropometric Standardization Reference Manual. Human Kinetics: Champaign, III, 1988. 5 Stuart HC, Hill P, Shaw C. The growth of bone, muscle and overlying tissue as revealed by studies of roentgenograms of the leg area. Monogr Soc Res Child Dev 1940; 5: 1–190. 6 Garn SM. Roentgenogrammetric determinations of body composition. Hum Biol 1957; 29: 337–353. 7 Wang ZM, Pierson Jr RN, Heymsfield SB. The five-level model: a new approach to organizing body-composition research. Am J Clin Nutr 1992; 56: 19–28. 8 Hounsfield GN. Computerized transverse axial scanning (tomography). Br J Radiol 1973; 46: 1016–1022. 9 Raden J. On the determination of functions from their integrals along certain mainfolds. Math Phys Klasse 1917; 69: 262–277. 10 Heymsfield SB, Noel R, Lynn M, Kutner M. Accuracy of soft tissue density predicted by CT. J Comput Assist Tomogr 1979; 3: 859–860. 11 Heymsfield SB, Olafson RP, Kutner MH, Nixon DW. A radiographic method of quantifying protein-calorie undernutrition. Am J Clin Nutr 1979; 32: 693–702. 12 Heymsfield SB, Fulenwider BT, Nordlinger R, Balow R, Sones P, Kutner M. Accurate measurement of liver, kidney, and spleen

27

28

29

30 31

32

33

34

volume and mass by computerized axial tomography. Ann Intern Med 1979; 90: 185–187. Heymsfield SB, Noel R. Radiographic analysis of body composition by computerized axial tomography. In: Ellison NM, Newell GR (eds). Nutrition and Cancer, vol. 17. Raven Press: New York, 1981, pp 161–172. Borkan GA, Gerzof SG, Robbins AH, Hults DE, Silbert CK, Silbert JE. Assessment of abdominal fat content by computerized tomography. Am J Clin Nutr 1982; 36: 172–177. Tokunaga K, Matsuzawa Y, Ishikawa K, Tarui S. A novel technique for the determination of body fat by computed tomography. Int J Obes 1983; 7: 437–445. ¨ stro ¨ m L, Kvist H, Cederblad A, Tylen U. Determination of total Sjo adipose tissue and body fat in women by computed tomography, 40 K, and tritium. Am J Physiol 1986; 250: E736–E745. ¨ rntorp P. Adipose tissue distribution and function. Int J Obes Bjo 1991; 5 (Suppl 2): 67–81. Bloch F, Hansen WW, Packard ME. Nuclear introduction. Phys Rev 1946; 70: 460–474. Purcell EM, Torrey HC, Pound RV. Resonance absorption by nuclear magnetic moments in a solid. Phys Rev 1946; 69: 37–38. Damadian R. Tumor detection by nuclear magnetic resonance. Science 1971; 171: 1151–1153. Lauterber PC. Image formation by induced local interactions: examples employing nuclear magnetic resonance. Nature 1973; 242: 190–191. Mansfield P. Multi-planar image formation using NMR spinechoes. J Phys Chem 1977; 10: L55. Foster MA, Hutchison JMS, Mallard JR, Fuller M. Nuclear magnetic resonance pulse sequence and discrimination of highand low fat tissues. Magn Reson Imaging 1984; 2: 187–192. Hayes PA, Sowood PJ, Belyavin A, Cohen JB, Smith FW. Subcutaneous fat thickness measured by magnetic resonance imaging, ultrasound, and calipers. Med Sci Sports Exerc 1988; 20: 303–309. Ross R. Advances in the application of imaging methods in applied and clinical physiology. Acta Diabetol 2003; 40 (Suppl 1): S45–S50. Fowler PA, Fuller MF, Glasby CA, Foster MA, Cameron GG, McNeill G et al. Total and subcutaneous adipose tissue distribution in women: the measurement of distribution and accurate prediction of quantity by using magnetic resonance imaging. Am J Clin Nutr 1991; 54: 18–25. Ross R, Le´ger L, Morris D, de Guise J, Guardo R. Quantification of adipose tissue by MRI: relationship with anthropometric variables. J Appl Physiol 1992; 72: 787–795. Ross R, Shaw KD, Martel J, de Guise Y, Avruch L. Adipose tissue distribution measured by magnetic resonance imaging in obese women. Am J Clin Nutr 1993; 57: 470–475. Farooqi IS, Bullmore E, Keogh J, Gillard J, O’Rahilly S, Fletcher PC. Leptin regulates striatal regions and human eating behavior. Science 2007; 317: 1355. Izhikevich EM, Edelman GM. Large-scale model of mammalian thalamocortical systems. Proc Natl Acad Sci 2008; 105: 3593–3598. Petersen KF, Dufour S, Shulman GI. Decreased insulin-stimulated ATP synthesis and phosphate transport in muscle of insulinresistant offspring of type 2 diabetic parents. PLoS Med 2005; 2: e233. Kim H, Taksali SE, Dufour S, Befroy D, Goodman TR, Petersen KF et al. Comparative MR study of hepatic fat quantification using single-voxel proton spectroscopy, two-point dixon and threepoint IDEAL. Magn Reson Med 2008; 59: 521–527. Napolitano A, Miller SR, Murgatroyd PR, Coward WA, Wright A, Finer N et al. Validation of a quantitative magnetic resonance method for measuring human body composition. Obesity 2008; 16: 191–198. Shen W, Wang Z, Punyanita M, Lei J, Sinav A, Kral JG et al. Adipose tissue quantification by imaging methods: a proposed classification. Obes Res 2003; 11: 5–16.

International Journal of Obesity


S82 35 Erondu N, Gantz I, Musser B, Suryawanshi S, Mallick M, Addy C et al. Neuropeptide Y5 receptor antagonism does not induce clinically meaningful weight loss in overweight and obese adults. Cell Metab 2006; 4: 275–282. 36 Nedergaard J, Bengtsson T, Cannon B. Unexpected evidence for active brown adipose tissue in adult humans. Am J Physiol Endocrinol Metab 2007; 293: E444–E452. 37 Addy C, Wright H, Van Laere K, Gantz I, Erondu N, Musser BJ et al. The acyclic CB1R inverse agonist taranabant mediates weight loss by increasing energy expenditure and decreasing caloric intake. Cell Metab 2008; 7: 68–78. 38 Moore FD, Oleson KH, McMurray JD, Parker HV, Ball MR, Boyden CM. The Body Cell Mass and its Supporting Environment. WB Saunders: Philadelphia, 1963. 39 Anderson J, Osborn SB, Tomlinson RWS, Newton D, Rundo J, Salmon L et al. Neutron-activation analysis in man in vivo: a new technique in medical investigation. Lancet 1964; 2: 1201–1205. 40 Cameron JR, Sorenson J. Measurement of bone mineral in vivo. Science 1963; 42: 230–232.

International Journal of Obesity

41 Mazess 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 1990; 51: 1106–11012. 42 Pietrobelli A, Formica C, Wang ZM, Heymsfield SB. Dual-energy X-ray absorptiometry body composition model: review of physical concepts. Am J Physiol 1996; 271: E941–E951. 43 Heymsfield SB, Smith R, Aulet M, Bensen B, Lichman S, Wang J et al. Appendicular skeletal muscle mass: measurement by dualphoton absorptiometry. Am J Clin Nutr 1990; 52: 214–218. 44 Tataranni PA, Ravussin E. Use of dual-energy X-ray absorptiometry in obese individuals. Am J Clin Nutr 1995; 62: 730–734. 45 Siri WE. Body composition from fluid spaces and density: analysis of method. In: Brozek J, Henschel A (eds). Techniques for Measuring Body Composition. National Academy of Sciences, National Research Council: Washington, DC, 1961, pp 223–244. 46 Heymsfield SB, Lichtman S, Baumgartner RN, Wang J, Kamen Y, Aliprantis A et al. Body composition of humans: comparison of two improved four-compartment models that differ in expense, technical complexity, and radiation exposure. Am J Clin Nutr 1990; 52: 52–58.