Location of body fat and body size impacts DXA soft tissue measures: a simulation study

(2008) 62, 553 559 & 2008 Nature Publishing Group All rights reserved 0954-3007/08 $30.00 www.nature.com/ejcn ORIGINAL ARTICLE Location of body fat and body size impacts DXA soft tissue measures: a simulation study RJ Valentine, MM Misic, RB Kessinger, MC Mojtahedi and EM Evans Department of Kinesiology and Community Health, University of Illinois at Urbana-Champaign, Urbana, IL, USA Objective: The aim of this study was to determine the ability of dual X-ray absorptiometry (DXA) to detect exogenous fat in men and women simulating typical sex-specific weight changes. Subjects: A diverse sample including 29 elderly (52 83 years) and 61 young (18 40 years) individuals (45 women, 45 men) of varying body mass index (BMI; M7s.d.: 26.174.9 kg/m 2, range ¼ 16.4 39 kg/m 2 ). Methods: Whole body (WB) DXA scans (Hologic QDR 4500A) were completed with Scan 1 performed as a normal baseline scan, Scan 2 with 1 kg packet of lard placed on each thigh and Scan 3 with two 1 kg lard packets placed on the abdomen (men) or chest and abdomen (women). Results: Measurement error of fat mass (FM) was more pronounced in the trunk as lard packets were detected with 59% accuracy (error ¼ 0.8270.42 kg, Po0.001), whereas 94% of thigh lard (error ¼ 0.1170.45 kg, Po0.001) was determined to be FM, while the remainder in both conditions was determined to be mineral free lean mass. Initial FM (r ¼ 0.37, Po0.001) for thigh loading and trunk bone mineral content (r ¼ 0.30, Po0.01) for trunk-loaded conditions had the most impact on measurement error of WB FM. Conclusions: Regional differences impact measurement error of simulated changes in FM with greater measurement error in the trunk compared to the thigh region and initial FM and higher levels of bone mineral content in the trunk region impacting error. (2008) 62, 553 559; doi:10.1038/sj.ejcn.1602770; published online 25 April 2007 Keywords: body composition; regional adiposity; technique Introduction Dual X-ray absorptiometry (DXA) is the most commonly used method to assess body composition in obesity research (Schoeller et al., 2005). The rapidly rising availability of DXA in clinical and research settings has generated increased interest in the use of this instrument to assess and track changes in body composition. Interventions to reduce fat mass (FM) are of eminent public health importance given the expanding prevalence of obesity (NIH National Heart Correspondence: Dr EM Evans, Department of Kinesiology and Community Health, University of Illinois, 215 Freer Hall, MC-052, 906 S Goodwin Avenue, Urbana, IL 61801, USA. E-mail: elevans@uiuc.edu Contributors: RJV contributed to statistical analysis and is the primary author; MMM contributed to this manuscript; RBK was involved in the subject recruitment, and data collection; MCM was involved in the data quality control and contributed to drafting of this manuscript; EME is the senior author of laboratory and contributed to all aspects of the manuscript including research design, obtainment of funding, statistical analysis oversight and final editing. Received 29 November 2006; revised 5 March 2007; accepted 8 March 2007; published online 25 April 2007 Lung and Blood Institute, 1998). The reliability of DXA to determine body composition and appropriately measure body composition change has become of increasing importance. Indeed, the Hologic QDR 4500A DXA instrument has been used in the most recent National Health and Nutrition Examination Survey (NHANES) (Centers for Disease Control and Prevention, 2006). However, the ability of fan-beam DXA to accurately assess changes in fatness in regions of the body that differ in size, depth and bone mineral content (BMC) in a sample representing diverse body composition and body morphology is unknown. Several confounding factors may impact DXA accuracy. Abdominal adiposity and body thickness have been correlated with imprecision in larger individuals (Tothill et al., 2001). The aging process initiates changes in body morphology, as adipose tissue in both men (Villareal et al., 2005) and post-menopausal women (Espeland et al., 1997) is predominantly stored in central regions. This type of fat-patterning may introduce measurement error with repeated measures sampling due to changes in body morphology, even in the absence of appreciable changes in body weight. It has been demonstrated that fan-beam DXA significantly underesti-

554 mates abdominal fat compared to computer tomography (CT) (Glickman et al., 2004); however, the impact of changes in body composition and fat distribution on DXA measurement error remains inadequately characterized. Although data from DXA instruments that use pencilbeam technology to assess known quantities of exogenous fat exist (Milliken et al., 1996; Kohrt, 1998), minimal research (Glickman et al., 2004) to date has been published investigating similar questions using the new generation fan-beam instrument. Moreover, the impact of differing body size, thickness and fat distribution on the ability of DXA to measure small changes in fat mass is unknown. Because total and trunk FM are often utilized in the research setting to assess efficacy of various weight loss interventions, insights regarding the ability of fan-beam DXA to measure changes in FM are of high interest. In this context, the primary aim of this study was to assess the ability of fan-beam DXA to accurately assess modest changes in exogenous fat (i.e. lard packets) in magnitudes that typically accompany weight loss interventions. Exogenous fat was positioned to simulate typical sex-specific weight changes in men and women. A secondary aim of this study was to assess the impact of initial body size and thickness on the ability of fan-beam DXA to accurately measure exogenous fat in different locations. Methods Subjects A total of 90 individuals (79 Caucasian, 3 Asian, 3 black, 5 Hispanic) were tested including 34 young (16 male, 18 female; aged 18 40 years) normal weight (defined as a body mass index (BMI) o25 kg/m 2 ), 27 young overweight (15 male, 12 female; BMI 425 kg/m 2 ) and 29 older (14 male and 15 female; 450 years) individuals (BMI ¼ 16.4 39.0 kg/m 2 ). All young individuals were free from known metabolic or bone disease. The elderly individuals were communitydwelling and representative of typical older adults in that they had various body morphologies, health status and medication usage. Notably, the targeted recruitment aimed to provide a total sample diverse in body fatness and thickness (range of BMI values), bone size (males and females) and body morphology (young and old). Morbidly obese individuals were excluded due to weight limitations on the DXA machine (136 kg) and/or the inability to fit within the dimensions of the scanning table. All subjects signed a consent form approved by the university s Institutional Review Board upon enrollment in the study. Anthropometric measures Standing height and weight measurements were completed with subjects wearing light-weight clothing and no shoes. Height was obtained using a stadiometer with measures obtained to the nearest 0.1 cm. Weight was measured on a calibrated digital scale (Tanita, Model BWB-627A). Body morphology was further characterized by waist and hip measures performed with the subject in a standing position using a Gulick II retractable measuring tape (Country Technologies Inc., Gay Mills, WI, USA). Waist was designated as the smallest circumference between the lowest rib and the top of the iliac crest. Measures were also taken while in the supine position using a sliding caliper (Holtain Ltd) to characterize body thickness at the thigh, chest and abdomen. DXA Subjects wore light-weight cotton clothing free from metal during the scanning (i.e. medical scrubs). Before testing, the DXA instrument was calibrated as per the manufacturer s guidelines. Precision for DXA measures of interest are 1 1.5% in our laboratory with CV% calculated from duplicate scans of young adults and postmenopausal women. Subjects were positioned on a DXA (Hologic QDR 4500A, software version 11.1:3) as per the manufacture s recommendations and whole body (WB) scans were performed. WB scans were repeated with lard packets (2.54 cm thick; 25.4 17.8 cm; 1 kg each) placed over the thighs (all subjects) or trunk area (chest and abdomen for women or abdomen for men, see Figure 1) to simulate the typical sex-specific change in adiposity in response to weight change. Lard packets were constructed using commercially available (i.e. food store) lard, assumed to be 100% fat, that was packaged in thin plastic, sealed and vacuum packed to remove air. Packets were of uniform thickness and identical in size, shape and weight. To optimize precision, subjects were not repositioned between the baseline scans and the lard condition scans. All scans for a given participant were performed on the same day and analyzed by the same research technician. Lard-loaded conditions Placement for the lard packets was as follows: Thigh: One lard packet was placed on each thigh (Figure 1c). The packets were positioned so that the top edge of each bag was even with the pubic bone Abdomen (men): Two lard packets were placed on the trunk with the edge of the first bag facing the head of the participant and the bottom edge of the first packet touching the top edge of the second packet. The second packet was placed with the bottom edge at the anterior superior iliac spine (Figure 1a): Chest/abdomen (women): Two lard packets were placed on the trunk with one packet placed towards the head of the participant, with the bottom edge even with the xiphoid process. The second packet was positioned to meet the bottom edge of the first packet at the xiphoid process, and the bottom edge was placed at the anterior superior iliac spine (Figure 1b). Statistical analysis Data were analyzed with SPSS for Windows version 12.0 (SPSS Inc., Chicago, IL, USA). Descriptive statistics were

555 Figure 1 Simulation of lard packet placement on the trunk of a male (a) and female (b) and thighs of a female (c). performed on participant characteristics and DXA outcomes of interest. Primary DXA outcomes were also assessed for normality using skewness and kurtosis statistics. Paired t-tests were performed to determine if lard packet mass and composition was accurately detected by DXA. The sum of total body mass from the baseline scan and the scale mass of the packets (reference condition) was compared to DXAdetermined total body mass under lard-loaded conditions. Lard packets were assumed to be 100% fat; thus reference values represent baseline values plus lard packet mass for fat composition only (other reference compositions do not differ from baseline values). To further characterize the degree of individual variability and impact of bone size and body thickness on measurement error (directional bias), regression analysis was performed on Bland Altman plots (Altman and Bland, 1983; Bland and Altman, 1986) that were generated using the reference value and the difference or error scores (reference scan value lard loaded condition scan value). An a-level of 0.05 was considered significant. Results Descriptive statistics are presented in Table 1. The recruitment strategy was successful and provided a sample diverse in body size, shape and fatness. All primary variables used in subsequent analyses were normally distributed, having skewness and kurtosis values less than 2.0. The composition of packets placed on the thighs was more accurately detected than the same packets placed over the trunk; however, DXA erroneously detected fat as mineral free lean mass (MFLM) in both conditions (Table 2). Exogenous fat was underestimated by DXA, corresponding to 94 and 59% accuracy when fat was placed on the thighs and trunk, respectively (Figure 2). Although small, erroneous detection of the lard packets led to small but significant differences between reference total body mass (TBM; baseline þ lard) and TBM when packets were placed on the thighs and trunk resulting in a similar error of 0.3% (Po0.001). Both lard-loaded conditions were compared to the reference scan (baseline þ packet) to examine the influence of inaccurate determinations of packet composition on absolute and relative whole body soft tissue composition (Table 2). The largest discrepancy between lard-loaded and reference conditions occurred with packet placement on the trunk (Po0.001; 1.1% underestimation of WB% FM and 1.4% overestimation of WB% MFLM). Thigh-loaded and reference conditions produced similar estimates of body composition; however, significant differences were found (Po0.001), corresponding to an underestimation of 0.2% FM and an overestimation of 0.4% MFLM. A regional body composition analysis on summed thighs and trunk was also performed (Table 2). When lard packets were placed over the thighs, regional analysis of both legs combined revealed an underestimation in total mass (TM), FM and BMC of 0.9, 6.3 and 2.0%, respectively and an overestimation in MFLM of 1.9% (Po0.001). When packets were loaded on the trunk, FM and BMC were underestimated by 8.7 and 2.9%, respectively, while MFLM was overestimated by 3.4% (Po0.001) and TM did not differ compared to reference values (P ¼ 0.76). The influence of anthropometric (waist circumference, chest, abdomen and thigh depths, BMI) and DXA measures (WB and trunk FM and bone measures) on measurement error was evaluated using linear regression analysis of Bland Altman plots. WB FM measurement error due to lard placed on the thigh was most strongly impacted by initial WB FM (r ¼ 0.37, Po0.001; see Figure 3a), with a strong trend for WB bone area to also influence error in an opposite manner (r ¼ 0.20, P ¼ 0.057; see Figure 3b). Measurement error in WB

556 Table 1 Subject characteristics Women (n ¼ 45) Men (n ¼ 45) Total (n ¼ 90) Mean7s.d. Range Mean7s.d. Range Mean7s.d. Range Age (years) 38723 18 83 36722 18 78 37722 18 83 Height (cm) 163.475.9 148.1 177.8 176.777.0 163.8 188.9 170.179.3 148.1 188.9 Weight (kg) 68.7714.7 41.0 107.6 82.5714.6 55.0 117.3 75.6716.1 41.0 117.3 BMI (kg/m 2 ) 25.775.3 16.4 38.4 26.474.6 20.2 39.0 26.174.9 16.4 39.0 Percent fat 31.477.1 17.7 47.6 19.276.1 9.1 33.2 25.379.0 9.1 47.6 DXA TBM (kg) 69.7714.9 41.4 109.0 83.7714.9 55.9 119.3 76.70716.38 41.4 119.3 DXA FM (kg) 22.679.5 7.6 48.6 16.576.7 5.8 32.6 19.5478.74 5.8 48.6 DXA MFLM (kg) 44.977.1 29.1 62.1 64.4710.6 47.0 94.7 54.62713.31 29.1 94.7 DXA BMC (kg) 2.270.4 1.1 3.1 2.970.5 2.1 4.2 2.5570.54 1.1 4.2 Waist circumference (cm) 85.7712.6 66.1 123.3 91.1711.7 69.9 114.3 88.4712.4 66.1 123.3 Thigh thickness (cm) 11.572.1 7.9 17.7 12.271.9 7.7 16.7 11.8572.05 7.7 17.7 Chest thickness (cm) 15.572.6 10.2 21.2 17.072.6 11.4 23.2 16.2772.68 10.2 23.2 Abdomen thickness (cm) 14.573.2 9.9 23.9 15.172.7 11.0 22.7 14.7772.95 9.9 23.9 Abbreviations: BMC, bone mineral content; BMI, body mass index; DXA, dual X-ray absorptiometry; FM, fat mass; MFLM, mineral-free lean mass; TBM, total body mass. Table 2 Absolute and relative body composition measured by DXA in reference and lard-loaded conditions Lard condition Reference Thigh-loaded Trunk-loaded Whole body TM (kg) 78.70716.38 78.91716.33* 78.95716.39* FM (kg) 21.5478.74 21.4278.92w 20.7278.78* FM (%) 27.3278.77 27.1079.06* 26.1878.89* MFLM (kg) 54.62713.31 54.94713.43* 55.71713.50* MFLM (%) 69.4078.44 69.8478.80* 70.8278.64* BMC (kg) 2.5570.54 2.5570.52 2.5370.53* BMC (%) 3.2770.52 3.2870.51 3.2570.51* Figure 2 Lard packet composition as detected by DXA. Regional thigh TM (kg) 29.5875.86 29.3275.87* FM (kg) 9.4473.51 8.8473.62* MFLM (kg) 19.1674.90 19.5274.96* BMC (kg) 0.9870.24 0.9670.22* Regional trunk TM (kg) 38.3878.32 38.3878.29 FM (kg) 10.9274.68 9.9774.73* MFLM (kg) 27.4676.15 28.4076.24* BMC (kg) 0.6870.18 0.6670.17* Abbreviations: BMC, bone mineral content; DXA, dual X-ray absorptiometry; FM, fat mass; MFLM, mineral-free lean mass; TM, total mass. Values are means7s.d. Reference ¼ baseline scan þ composition of the 2 kg lard packets. w Different from reference, Po0.05, *Po0.001. FM due to lard placed on the trunk was similarly influenced by trunk BMC (r ¼ 0.30, P ¼ 0.004; see Figure 4a), WB bone area (r ¼ 0.29, P ¼ 0.006; see Figure 4b) and WB BMC (r ¼ 0.28, P ¼ 0.008; see Figure 4c). Directionally, as these values increased, WB FM was increasingly underestimated. No significant relationships existed between circumferences, body thickness at any site, or BMI and measurement error due to lard placed on the trunk or thighs (data not shown, all P40.05). Discussion The primary aim of this study was to assess the ability of fan-beam DXA to accurately detect changes in soft tissue composition with exogenous fat placed on the thighs or trunk in men and women varying in body and skeletal size, adiposity and morphology. Our primary findings are that lard placement simulating fat gain is (1) underestimated by DXA in a site-specific manner as exogenous fat placed on the trunk is poorly detected compared to placement on the thighs, (2) measurement error is impacted by initial fat mass and bone size with greater adiposity resulting in overestimation and larger bone mass resulting in underestimation of exogenous fat placed on the thighs and (3) measurement error is impacted by initial fat mass and bone size with greater adiposity and bone mass resulting in an underestimation of exogenous fat placed on the trunk. Based upon the present results, care should be taken when using

557 Figure 3 Relation between (a) initial WB fat mass (FM; r ¼ 0.37, Po0.001) and (b) WB bone area (r ¼ 0.20, P ¼ 0.057) and measurement error (reference scan lard loaded condition scan) in WB FM after lard placement on thighs. Solid lines indicate mean error, dashed lines indicate 72 s.d. and dotted line represents regression line. DXA in soft tissue measurement in larger individuals and for repeated measures. The current findings are consistent with similar studies conducted using pencil-beam DXA, which demonstrated comparable measurement error influenced by exogenous fat location, quantifying thigh and trunk-loaded exogenous fat with B96 and 55% accuracy, respectively (Snead et al., 1993; Milliken et al., 1996). However, a study employing lard packets located similarly to the present study by Kohrt found DXA to accurately detect exogenous fat as B96%, irrespective of lard positioning (Kohrt, 1998). It is important to note that previous studies used various pencil-beam instruments, and sample sizes examined were somewhat limited with subjects representing relatively young lean individuals. To date, only one previous study has assessed the ability of fan-beam DXA to detect changes in exogenous fat. Glickman et al. (2004), using lard-loading over a small abdominal Figure 4 Relation between (a) trunk bone mineral content (BMC; r ¼ 0.30, P ¼ 0.004), (b) WB bone area (r ¼ 0.29, P ¼ 0.006) and (c) WB BMC (r ¼ 0.28, P ¼ 0.008) and measurement error (reference scan lard loaded condition scan) in WB FM after lard placement on trunk. Solid lines indicate mean error, dashed lines indicate 72 s.d. and dotted line represents regression line. region (L1-L4), found DXA (Lunar DPX-IQ) to account for 78% of fat loaded on the abdomen. The present study reinforces the growing body of literature supporting the inadequacy of DXA to detect changes in centrally located fat depots compared to other regions of the body. The

558 inaccurate detection of trunk fat by DXA may be generalizable, as the present sample was diverse in age, gender and anthropometrics. However, translation of these results is limited to Hologic QDR 4500A measures, as manufacturers, scan mode and DXA scanner technology (pencil, narrow or traditional fan-beam) all impact measurement artifact (Tothill et al., 2001). A major assumption inherent in DXA technology is that body areas not analyzed for soft tissue composition are similar to body areas that are analyzed. It is estimated that 40 45% of the 21 000 pixels in a typical whole-body scan contain bone in addition to soft tissue, and these pixels are excluded from the calculated soft tissue measurements. Moreover, the influence of the thorax on the total body composition estimates may be under-represented because of the relatively large areas of bone in these regions (Lohman, 1996). Our results are theoretically logical in that trunk loading elucidated greater measurement error than thigh loading. Importantly, little data are available to guide the researcher or clinician in estimating the impact of changes in mass on measures of adiposity, particularly as it relates to regional body composition. Insights into the ability of fan-beam DXA to accurately assess soft tissue outcomes after weight change are essential to aid studies involving weight loss interventions and to adequately characterize body fatness in the clinical realm. It is important to note that the mass of the exogenous fat utilized in this study is relatively moderate in comparison to weight reductions typically seen in interventions, often eliciting weight loss of 5 15 kg. In this regard, the measurement error witnessed in the current study would likely escalate when applied to many clinical settings if it is assumed that the fat gain and fat loss impacts on measurement error are linear. Measurement errors induced by gains in trunk fat are noteworthy and have implications for clinical practice. Weight gain, especially in men or after menopause in women, increases abdominal obesity and is associated with several co-morbidities including insulin resistance, chronic inflammation and atherosclerosis (Deprés, 2006). On the basis of the current findings, weight gain inherently introduces some degree of measurement error and insufficient quantification of fat mass, having obvious implications on clinical research regarding FM change and metabolic disease risk. Although FM is the primary focus of body composition investigations, the inaccuracy of soft tissue measurement also influences determinations of MFLM. Indeed, when lard was positioned on the trunk region, B41% of the added fat was detected as MFLM. It is therefore likely that individuals with android body morphology have significantly overestimated values of MFLM. In this context, if an individual gains 10 kg of fat, stored centrally, MFLM values would concurrently increase by 4.1 kg on the WB level according to the results from this study. Similarly, when fat mass increased on the thighs, thigh MFLM increased. Collectively, these overestimations of MFLM may have important implications for the diagnoses of sarcopenia and especially sarcopenic obesity, as we found measurement error to increase with escalating levels of WB fat mass, potentially leaving an important health concern undetected. Theoretically, weight loss would be expected to elicit an opposite effect, as loss of thigh or abdominal fat would be partially detected as a loss of MFLM, creating an overestimation of muscle wasting. A primary aim of most weight loss interventions is to reduce fat mass while maintaining lean mass; therefore, it is critical that both components be accurately and reliably measured. The present study is not without limitations. A primary criticism is that the exogenous fat loading protocol utilized in this study does not truly mimic physiological gains of in vivo adipose tissue. However, using lard, which has similar X-ray attenuation characteristics as fat, to emulate weight change has been investigated by several investigators using various designs (Snead et al., 1993; Milliken et al., 1996; Kohrt, 1998; Glickman et al., 2004). Currently, no existing criterion method exists to segregate measurement artifact during weight change from true physiological change in humans with regard to DXA technology. A second potential limitation is that the lard packet mass uniformly covered the subject s body while lying supine on the scanning table (i.e. lard packet was placed on top of the body and did not extend beyond the body outline) simulating sex-specific changes in tissue distribution observed in weight loss studies of adults in our laboratory. Although uniform thickness of the packets placed entirely on top of the patient does not precisely mimic accumulation of adipose tissue, placement of the lard packets between the patient and the scanning table was less optimal as this would alter the distance between the body and X-ray beam. Additionally, such placement would emulate posterior fat gain, which is not physiologically accurate and would prevent uniform lard thickness due to compression by the body. Although DXA is the most often used technology to assess body composition in clinical research, it is not considered by many experts to be a criterion method. Recent studies have demonstrated an underestimation of FM of B5% assessed by fan-beam DXA as compared to four-component criterion measures (Visser et al., 1999; Deurenberg-Yap et al., 2001; Tylavsky et al., 2003). However, the biological error does not explain the technological error of the magnitude seen with the large 41% inaccuracy in fat detection under the trunkloaded condition. Finally, by increasing mass measurement, precision by DXA is reduced and potentially contributes to measurement error (Cordero-MacIntyre et al., 2002). And as stated previously, due to differences in DXA manufacturers, the results of this study are only directly applicable to Hologic QDR 4500A (fan-beam) DXA instruments using similar software. In summary, this simulation study suggests that measurement error of FM and MFLM is larger when changes in FM occur on the trunk versus thigh and absolute FM and bone

mass impact measurement error. The relation between weight change and body composition detection is clinically significant and has implications for public health. Precise measurement of soft tissues during weight change in vivo is currently problematic to investigators and clinicians invested in body composition-related pathologies, and measures should be viewed judiciously. Due to current trends in obesity, continued focus on technological advancements in DXA instruments and software programs may reduce this measurement error and facilitate accurate physiological measures of body composition. Additional imaging technologies, including magnetic resonance imaging and CT, will undoubtedly be increasingly used and should be utilized to facilitate DXA technological improvements. Acknowledgements This work was sponsored by, Pilot Grant from International Society for Clinical Densitometry (ISCD) to EM Evans. References Altman DG, Bland JM (1983). Measurement in medicine: the analysis of method comparison studies. Statistician 32, 307 317. Bland JM, Altman DG (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1, 307 310. Centers for Disease Control and Prevention (2006). NHANES 1999 2000 public data release file documentation. Internet http:// www.cdc.gov/nchs/data/nhanes/gendoc.pdf.accessed July 26, 2006. Cordero-MacIntyre ZR, Peters W, Libanati CR, Espana RC, Abila SO, Howell WH et al. (2002). Reproducibility of DXA in obese women. J Clin Densitom 5, 35 44. Deprés JP (2006). Is visceral obesity the cause of the metabolic syndrome? Ann Med 38, 52 63. Deurenberg-Yap M, Schmidt G, van Staveren WA, Hautvast JG, Deurenberg P (2001). Body fat measurement among Singaporean Chinese, Malays and Indians: a comparative study using a fourcompartment model and different two-compartment models. Br J Nutr 85, 491 498. Espeland MA, Stefanick ML, Kritz-Silverstein D, Fineberg SE, Waclawiw MA, James MK et al. (1997). Effect of postmenopausal hormone therapy on body weight and waist and hip girths. Postmenopausal Estrogen Progestin Intervention study investigators. J Clin Endocrinol Metab 82, 1549 1556. Glickman SG, Marn CS, Supiano MA, Dengel DR (2004). Validity and reliability of dual-energy X-ray absorptiometry for the assessment of abdominal adiposity. J Appl Physiol 97, 509 514. Kohrt WM (1998). Preliminary evidence that DEXA provides an accurate assessment of body composition. J Appl Physiol 84, 372 377. Lohman TG (1996). Dual energy X-ray absorptiometry. In: Roche AF, Heymsfield SB, Lohman TG (eds) Human Body Composition, pp 63 78. Human Kinetics: Champaign. Milliken LA, Going SB, Lohman TG (1996). Effects of variations in regional composition on soft tissue measurements by dual-energy X-ray absorptiometry. Int J Obes 20, 677 682. NIH National Heart Lung and Blood Institute (1998). Clinical guidelines on the identification, evaluation and treatment of overweight and obesity in adults: the Evidence Report. Obes Res 6, 51S 209S. Schoeller DA, Tylavsky FA, Baer DJ, Chumlea WC, Earthman CP, Fuerst T et al. (2005). QDR 4500A dual-energy X-ray absorptiometer underestimates fat mass in comparison with criterion methods in adults. Am J Clin Nutr 81, 1018 1025. Snead DB, Birge SJ, Kohrt WM (1993). Age-related differences in body composition by hydrodensitometry and dual-energy x-ray absorptiometry. J Appl Physiol 74, 770 775. Tothill P, Hannan WJ, Wilkinson S (2001). Comparisons between a pencil beam and two fan beam dual energy X-ray absorptiometers used for measuring total body bone and soft tissue. Br J Radiol 74, 166 176. Tylavsky F, Lohman T, Blunt BA, Schoeller DA, Fuerst T, Cauley JA et al. (2003). QDR 4500A DXA overestimates fat-free mass compared with criterion methods. J Appl Physiol 94, 959 965. Villareal DT, Apovian CM, Kushner RF, Klein S (2005). Obesity in older adults: technical review and position statement of the American College for Nutrition and NAASO, The Obesity Society. Am J Clin Nutr 82, 923 934. Visser M, Fuerst T, Lang T, Salamone LM, Harris TB (1999). Validity of fan-beam dual-energy X-ray absorptiometry for measuring fat-free mass and leg muscle mass. J Appl Physiol 87, 1513 1520. 559