Dual-energy X-ray absorptiometry based body volume measurement for 4-compartment body composition 1 3

Transcription

1 Dual-energy X-ray absorptiometry based body volume measurement for 4-compartment body composition 1 3 Joseph P Wilson, Kathleen Mulligan, Bo Fan, Jennifer L Sherman, Elizabeth J Murphy, Viva W Tai, Cassidy L Powers, Lorena Marquez, Viviana Ruiz-Barros, and John A Shepherd ABSTRACT Background: Total body volume (TBV), with the exclusion of internal air voids, is necessary to quantify body composition in Lohman s 4-compartment (4C) model. Objective: This investigation sought to derive a novel, TBV measure with the use of only dual-energy X-ray absorptiometry (DXA) attenuation values for use in Lohman s 4C body composition model. Design: Pixel-specific masses and volumes were calculated from low- and high-energy attenuation values with the use of first principle conversions of mass attenuation coefficients. Pixel masses and volumes were summed to derive body mass and total body volume. As proof of concept, 11 participants were recruited to have 4C measures taken: DXA, air-displacement plethysmography (ADP), and total body water (TBW). TBV measures with the use of only DXA (DXA-volume) and ADP-volume measures were compared for each participant. To see how body composition estimates were affected by these 2 methods, we used Lohman s 4C model to quantify percentage fat measures for each participant and compared them with conventional DXA measures. Results: DXA-volume and ADP-volume measures were highly correlated (R 2 ¼ 0.99) and showed no statistically significant bias. Percentage fat by DXA volume was highly correlated with ADPvolume percentage fat measures and DXA software-reported percentage fat measures (R 2 ¼ 0.96 and R 2 ¼ 0.98, respectively) but were slightly biased. Conclusions: A novel method to calculate TBV with the use of a clinical DXA system was developed, compared against ADP as proof of principle, and used in Lohman s 4C body composition model. The DXA-volume approach eliminates many of the inherent inaccuracies associated with displacement measures for volume and, if validated in larger groups of participants, would simplify the acquisition of 4C body composition to a single DXA scan and TBW measure. Am J Clin Nutr 2012;95: INTRODUCTION Over the past 50 y, body volume measurement has changed very little, despite its necessity in some multicompartment body composition models (1). TBV 4 is typically defined as the solid body volume that excludes the air voids of the lungs, intestines, and nasal cavities (2). Currently, TBV measurement devices make many assumptions about these voids, which lead to inaccuracies in body composition measurements. ADP systems are used clinically to measure body volume and to estimate fat mass and fat-free mass. In ADP, an individual who is wearing minimal clothing is enclosed in a capsule while body volume is measured by air displacement, with corrections for residual lung volume and surface area artifacts. Unfortunately, these corrections present challenges in the generation of accurate volume measures (8). Four-compartment models are considered the gold standard in body composition. Lohman s 4C model uses mass, TBV, TBW, and BMC to calculate percentage fat (3). This model is as expressed in the following relation: % Fat ¼ ð2:747=d b 0:714 3 TBW=Mass 1 1:146 3 BMC=Mass 2:050Þ ð1þ where D b is mass divided by TBV. However, because of cost, equipment, and inconvenience, 4C models are rarely used clinically (4). This 4C model currently requires 3 separate measurement devices: TBV by ADP or UWW, TBW by deuterium dilution or bioimpedance analysis, and BMC values by DXA. To the best of our knowledge, this is the first report in which DXA has been used to measure TBV. The objective of this preliminary investigation was to derive a TBV measure from 1 From the University of California, Berkeley University of California, San Francisco Graduate Program in Bioengineering (JPW and JAS), San Francisco, CA; and the Department of Radiology and Biomedical Imaging (JPW, BF, JLS, CLP, LM, and JAS), and the Department of Medicine, Division of Endocrinology and Metabolism, San Francisco General Hospital (KM, EJM, VWT, and VR-B), University of California, San Francisco, San Francisco, CA. 2 Supported by the National Science Foundation Graduate Research Fellowship, which provided the stipend for the lead author. Studies were performed in the University of California, San Francisco, Clinical and Translational Science Institute Clinical Research Center at San Francisco General Hospital, which is supported by NIH/NCRR grant UL1 RR Address correspondence to JA Shepherd, MRSC AC122, 1 Irving Street (Box 0628), San Francisco, CA john.shepherd@ucsf. edu. 4 Abbreviations used: ADP, air-displacement plethysmography; ALS, amyotrophic lateral sclerosis; BMC, bone mineral content; CT, computed tomography; DXA, dual-energy X-ray absorptiometry; 4C, 4-compartment; HE, high-energy attenuation; TBV, total body volume; TBW, total body water; UWW, underwater weighing. Received May 9, Accepted for publication September 23, First published online November 30, 2011; doi: /ajcn Am J Clin Nutr 2012;95: Printed in USA. Ó 2012 American Society for Nutrition 25

2 26 WILSON ET AL DXA soft tissue measures. Ultimately, we hope our method makes Lohman s 4C model easier to use clinically because it will require only a DXA scan and TBW measurement. In addition to its use in the 4C model, total body and regional DXA-volume measurements could be related to various health outcomes not yet studied. Our DXA-volume measure should have advantages over other displacement techniques because X-ray attenuations intrinsically exclude hollow body regions such as the lungs and bowels. DXA is useful in a wide variety of populations, from neonates (down to 1 kg) to the morbidly obese (up to 205 kg). As a proof of concept for our approach, we compared our DXA-volume measure with the ADP-volume measure in a convenience sample of adults. With the use of both volume techniques, we constructed Lohman s 4C model and compared percentage fat measures with conventional DXA measurements. SUBJECTS AND METHODS We took standard 4C measures on a convenience sample of adults (n ¼ 11) to compare volume techniques and associated 4C percentage fat values. Each participant fasted for 4 h before coming for a morning visit at the study site (University of California, San Francisco, Clinical Research Center at San Francisco General Hospital, San Francisco, CA). Weight was measured with the use of a calibrated scale while the participant wore a hospital gown or other minimal clothing. TBW was measured by deuterium dilution, TBV by ADP, and BMC and soft tissue attenuations by DXA, as described below. The University of California, San Francisco, Committee on Human Research approved the study. All participants gave informed consent. Participants The 11 participants (5 male, 6 female) had the following characteristics (mean 6 SD): age y(range: 23 67y), weight kg (range: kg), and BMI (in kg/m 2 ) (range: ). This population was a mix of healthy participants (n ¼ 8) and those with ALS (n ¼ 3). Deuterium dilution TBW was measured with the use of deuterium dilution. Participants spit into a conical vial so that a baseline saliva sample could be obtained. Participants were then given 2 H 2 Oto drink; this 99.9% enriched 2 H 2 O (Cambridge Isotope Laboratories) was diluted with deionized water and aliquoted into ;6- to 7-g individual doses (;0.1 g/kg body weight). The bottles were weighed before and after each dose administration to determine the exact amount of 2 H 2 O ingested. Saliva samples were then obtained 3 and 4 h after 2 H 2 O administration. No food was allowed for 4 h before the study or during the 4-h sampling period. Saliva specimens were frozen for later batch analysis. A total of 10 (out of 11) participants had TBW measurements by deuterium dilution. One participant was unable to produce saliva because of the progression of her illness. While still included in the deuterium dilution analysis, another subject drank water between the 3- and 4-h time point, which resulted in a very dilute 4-h time point, which was excluded from the analysis. To measure deuterium content, 50-lL aliquots of saliva underwent distillation in the cap of an inverted evaporation tube in a glass bead-filled heating block at 45 C for 3.5 h Water distillate was then collected and analyzed for isotope enrichment with the use of a wavelength-scanned cavity ring down spectrometer (Picarro L1102-i; Picarro Inc) (5). The 2 H isotope concentration was first calculated in delta (d) units relative to a working standard calibrated against Vienna-Standard Mean Ocean Water and was then converted into atom percentage excess as described previously (6). Samples were processed in duplicate and data from 3- and 4-h time points were averaged to give a single postingestion enrichment. TBW was calculated as TBWðkgÞ ¼ ðd=dape%þ 3 ð18:02=1000þ3 ð1=1:04þ where d is the dose of 2 H 2 O given in moles, DAPE% is the difference between the atom percentage excess in the sample before and after 2 H 2 O dose administration expressed as a percentage, 18.02/1000 is the conversion factor for moles of ordinary water to kilograms, and 1.04 is a correction factor to account for loss of 2 H to nonexchangeable pools (eg, amino acids) in the body, which has been estimated at 4% (7). ADP ADP measures were acquired with the use of the Bod Pod (Life Measurement Instruments Inc) to measure total body volume. Before each measurement, the instrument was calibrated by the placement of a hollow cylinder with known mass and volume into the Bod Pod. Participants wore a Lycra or silicone swim cap and tight-fitting, Spandex-like shorts and (sport bra for women) to reduce the effects of isothermal air. The participant was first sealed in the Bod Pod capsule. Body volume was measured by the body s air displacement with corrections for residual lung volume and surface area artifacts (8). Thoracic gas volume was measured in each individual while the participant breathed through a tube connected to a filter and reference chamber, following the manufacturer s instructions. After a few normal breaths, the participant was signaled by the operator to exhale with 3 small, consecutive huffs. The surface area artifact was automatically calculated by the Bod Pod software. These 2 adjustments (thoracic gas volume and surface area artifact) were factored into the overall calculation of body volume. Volumetric density was that actual measurement displayed by the Bod Pod; TBV was calculated as the scale weight divided by this density value. DXA imaging Participants were scanned on a Discovery Wi DXA system (Hologic Inc) while dressed in a hospital gown or the attire used for the ADP measures. Participants lay flat on the DXA table and the whole-body scan took ;7 min. The Hologic software version for analysis was Discovery 12.6; measures reported included the following used in this study: total body mass, percentage fat, and BMC. Hologic s NHANES population option was used in the analysis for the percentage fat measurement. This option is intended to improve the agreement of the DXA soft tissue measures with those of the 4C models (9). The NHANES option multiplies the lean mass by and the ð2þ

3 TOTAL BODY VOLUME VIA DXA 27 fat mass by In addition, all DXA images were analyzed with the use of custom software that we developed to quantify TBV and mass, as further described below. The pixel size of the DXA images was 0.80 cm 2. Areal Densityðg=cm 2 Þ¼ 6: : HE 1 2: HE 2 1 6:990 3 RðR 2 ¼ 0:99; RMSE ¼ 0:250 g cm 2 ; CV ¼ 1:594%Þ ð3þ Derivation of DXA volume In commercial DXA systems, X-ray attenuation values are used to solve directly for the mass of fat and lean soft tissue. Calibration phantoms of biologically equivalent materials with known composition are scanned on each DXA machine to create the relation between DXA attenuation values and the masses of fat and lean tissue. With a known pixel size, fat mass and lean mass can be represented as thickness or volume. Instead of solving for fat mass and lean mass, our approach was to use a similar calibration process to solve for volume and mass with the use of the following general steps: 1) Create a soft tissue phantom from biologically equivalent materials. 2) Generate calibration equations to relate phantom X-ray attenuations to volume and mass. 3) Isolate bone pixels from soft tissue pixels. 4) Interpolate the soft tissue attenuation values of bone pixels. 5) Estimate each pixel s soft tissue volume and mass from the calibration equations. A soft tissue calibration phantom was created to relate pixel mass and volume to DXA attenuation values of various thicknesses and compositions. Materials that mimic the X-ray properties of biological materials were used such that regions could be defined with specific compositions of lipid, protein, and water. Lipid was represented as machinable wax (McMaster Carr Inc), water as Plastic Water LR (CIRS Inc) (10), and protein as the thermoplastic material Delrin (11). Delrin has been shown previously to represent protein s X-ray properties well (11). The distributions of protein, lipid, and water values were chosen to fall within typical biological concentration ranges (12). Delrin sheets (which represented protein) were placed beneath blocks of machinable wax (lipid) and solid water (water) to create unique 2 3 4cm 2 -sized regions of interests. The low- and high-energy X-ray attenuations were defined separately as the difference between the pixel values and the background ( air ) values. The low-energy attenuation and HE values were used in our analysis as an R value, defined as the ratio of low-energy attenuation to HE. DXA scan analysis was performed with the use of custom software that we developed in MATLAB R2010a (MathWorks Inc). For calibration purposes, the R value and the HE were measured within regions of interest of known areal density, thickness, and composition. Twentyseven calibration regions with unique thicknesses from to cm and areal densities from 9.62 to g/cm 2 were defined. Regression techniques were used to determine the bestfit calibration equations of R and HE value to thickness and areal density. SAS software, version 9.2 (SAS Institute Inc) was used for all statistical analysis. Calibration equations were fit to the phantom data with coefficients up through cubic powers and cross terms. Coefficients with P values 0.05 remained in the model. Only the Hologic Discovery W was available for scanning of the calibration phantom. The transformation equations that resulted are below: ThicknessðcmÞ ¼ 23: : HE 20:077 3 R ðr 2 ¼ 0:99; RMSE ¼ 0:353 cm; CV ¼ 2:193%Þ ð4þ Pixel mass was determined by multiplication of the areal density by the physical pixel size in squared centimeters. Pixel volume was determined by multiplication of the thickness by the physical pixel size in squared centimeters. These calibration equations were implemented with the use of a custom MATLAB routine to report DXA volume and mass for either whole-body or arbitrary regions of interest. Bone pixels were empirically isolated from soft tissue pixels with the use of a threshold function. For a given material composition, the R value decreases as thickness (or HE) increases (13). Based on first principles of X-ray attenuation for our DXA machine, we generated theoretic curves of R compared with HE values for a variety of compositions. A function was selected that best isolated bone from soft tissue in our data set. The threshold function generated was the following: Bone Threshold R Value ¼ 0: HE 3 1 0:004 3 HE 2 0: HE 1 1:26 ð5þ As a visual example, the R value compared with HE for a sample image at each step of the bone isolation process is displayed in Figure 1. Sample HE and R value images can be seen in Figure 1, A and B, respectively. The smooth curve represents the threshold function. Any pixels with R values greater than the threshold function for the corresponding HE were considered bone pixels, as in Figure 1C. Pixels below this threshold were considered soft tissue pixels, as shown in Figure 1D. To estimate soft tissue values above and below bone, we used a nearest-neighbor interpolation of soft tissue values, as shown in Figure 1E. A pixel window was used to interpolate soft tissue values for bone pixels. For those bone pixels that had no soft tissue pixels within this pixel neighborhood (such as some in the skull), a larger ( pixel) interpolation window was used. The output of this interpolation process was a soft tissue image (with bone pixels replaced by soft tissue estimates), as shown in Figure 1F. This soft tissue image was fed into our calibration equations described earlier, in which individual pixel mass and volume were calculated. Total mass and TBV were calculated as the sum of all individual pixel values. Body composition equations To see how these different volume measurement methods affected body composition, we used Lohman s 4C body composition model to quantify percentage fat as expressed in Equation 1 (3), in which BMC is the BMC as reported by the DXA analysis software. When ADP volume was used in this model, scale

4 28 WILSON ET AL FIGURE 1. Process of identification of bone and estimation of soft tissue. A: HE image. B: R value image. C: Bone pixels are selected as pixels above the bone threshold function (solid line). D: Image of pixels identified as soft tissue, which are below the bone threshold function. E: Soft tissue values for bone pixels are estimated by nearest neighbor and next-nearest neighbor interpolation of pixels from D. F: Complete soft tissue image with no bone, a combination of D and E. DXA, dual-energy X-ray absorptiometry; HE, high-energy attenuation. weight was used for mass; when DXA volume was used in this model, the mass (as calibrated from Equation 3) was used for mass. These percentage fat measures from ADP volume and DXA volume were calculated and compared with standard DXA measurements. DXA-volume test tool We have made a demonstration release of our DXA-volume test tool available for download on the laboratory website ( In this first release, a Hologic scan file is selected as input. After analysis, the test tool outputs the following items: 1) Excel (Microsoft Inc) file that contains the filename, total mass in kilograms, and TBV in liters; 2) Excel-readable mass image; and 3) Excel-readable volume image. Statistics All statistical analysis was done with the use of SAS software, version 9.2. We used Bland-Altman analysis (14) to compare DXA volumes and ADP volumes and to compare percentage fat values from different techniques, including the clinical measure of percentage fat reported from the DXA analysis software. Bland-Altman relationships were compared in linear regression models with the use of the proc glm function within SAS. Mass, volume, and percentage fat measures were also compared in linear regression models with the use of the proc glm function within SAS. The proc glm function calculated correlation coefficients, root mean square error, slope and intercept values, and SE of slope and intercept values. Statistical significance from zero was defined at the P, 0.05 threshold. RESULTS A representative DXA-volume image of one of our 4C study participants is shown in Figure 2. A pseudo-color version of this image can be found under "Supplemental data" in the online issue. DXA-volume and ADP-volume measurements were highly correlated (R 2 ¼ 0.99, P, ) as seen in Table 1 and Figure 3. Bland-Altman analysis of DXA volume and ADP volume (found in the supplementary materials) showed no significant bias between the measures but did display an apparent curvature. The differences between DXA volume and ADP volume ranged from L to 0.97 L. One ALS and one healthy participant were unable to give a valid residual lung volume measurement with ADP, so the device calculated a gross estimate of the total body volume. For the ALS participant, the average volume measurement was L, and the DXA volume was 1.52 L smaller than the ADP volume. For the healthy participant, the average volume was L, and the DXA volume was 0.47 L smaller than the ADP volume. The mass, based on Equation 3, was highly correlated to scale weight (R 2 ¼ 0.99, P, ) and had no statistically significant bias (see Table 1). We constructed Lohman s 4C model for each participant to compare percentage fat measures with both volume methods, as in Equation 1. A very high correlation between percentage fat measures from DXA volume and ADP volume (R 2 ¼ 0.96; root mean square error ¼ 1.82%; P, ) is shown in Table 1. Both 4C percentage fat measures (from DXA volume and ADP volume) were highly correlated to DXA software-reported percentage fat (see Table 1). With the use of Bland-Altman analysis, we observed a significant bias between 4C DXA-volume percentage fat and 4C ADP-volume percentage fat (slope ¼ 20.25, intercept ¼ 8.35%)

5 TOTAL BODY VOLUME VIA DXA 29 FIGURE 2. Left: Volume contour map for a 28-y-old study participant with a dual-energy X-ray absorptiometry volume of L; notice the depression in the lung area where air is filled. Top right: Top view of volume image. Bottom right: Side view of volume image. (Figure 4). We also observed a significant bias between 4C DXA-volume percentage fat and DXA software-reported percentage fat (slope ¼ 20.16, intercept ¼ 5.56%). The differences between DXA-volume percentage fat and ADP-volume percentage fat ranged from 22.59% to 7.61%. The participant with the 7.61% fat difference had a 0.60-L difference between DXA volume and ADP volume. The participant with the 22.59% fat difference had a L difference between DXA volume and ADP volume. The difference between DXA-volume percentage fat and DXA software-reported percentage ranged from 21.62% to 4.92%. DISCUSSION This article showcases the derivation of a DXA-based method to measure total body volume. With DXA volume, Lohman s 4C model can be fully constructed with the use of only a single DXA scan and TBW measure; our method eliminates the need for a separate body volume device. We showed DXA volume and ADP volume to be highly correlated with no significant bias, despite calibration with the use of independent standards and methods. We used these 2 volume measures to calculate percentage fat for each individual with a TBW measure. We showed that all percentage fat measures were highly correlated. Importantly, our DXA-volume method uses a single setup for a large range of ages (neonates to the elderly) and body types (underweight to obese) (15, 16). Currently, there are separate ADP devices marketed for infants (Pea Pod), toddlers (Tod Pod), and adults (Bod Pod). DXA volume requires only that a subject lie flat for 5 min, without immersion in water or encapsulation. With the DXA-volume method, accurate total and regional body volumes can be measured with fewer assumptions than with ADP and UWW. For both ADP and UWW, residual lung volume must be estimated to calculate the TBV accurately. For UWW, the participant exhales deeply while a measure of buoyancy is taken. For the Bod Pod, a specific breathing protocol is used in which participants do not fully exhaust the lungs. The accuracy of these 2 techniques has long been debated (17, 18) and neither is appropriate for all participants. In our study, for instance, one ALS subject and one healthy participant were not able to provide a valid ADP-volume measure because of breath weakness and breathing protocol issues, respectively. In these 2 cases, the Bod Pod used estimates of lung volume based on body size to calculate TBV (19). Despite no statistically significant bias between DXA volume and ADP volume in Bland-Altman analysis, the data display an apparent curvature, which is likely because of the small sample size. With UWW, infants cannot be submerged under water, nor can they follow breathing instructions. Our DXA-volume method does not require a lung volume measure because the X-ray attenuations are normalized to air. Thus, DXA volume includes only solid (non-air) volume in its pixel volume measures. UWW has been considered the gold standard for body volume measures, and for the past decade ADP has been considered a direct substitute for UWW (20) and has largely supplanted UWW in studies because of its ease of use and reported TABLE 1 Linear regression analysis for measurements taken in this study 1 y Measure x Measure N R 2 RMSE Slope (SE) Intercept (SE) DXA reported mass Scale weight kg (0.02) 3.40 (1.64) Equation 3 mass Scale weight kg (0.02) (1.91) DXA volume ADP volume L (0.03) 2.04 (2.08) 4C DXA-volume % fat 4C ADP-volume % fat % (0.06) 7.84* (1.73) 4C DXA-volume % fat DXA % fat % (0.04) 5.39* (1.30) DXA % fat 4C ADP-volume % fat % (0.05) 2.95 (1.63) 1 DXA reported mass is the mass value reported by the DXA analysis software. Equation 3 mass is the total mass calculated by multiplication of each pixel s areal density (from Equation 3) by the pixel area. DXA volume is highly correlated to ADP volume. Both 4C percentage fat measures were highly correlated with DXA measures. ADP, air-displacement plethysmography; DXA, dual-energy X-ray absorptiometry; 4C, 4-compartment; RMSE, root mean square error. 2 Statistically different from zero, P, 0.05 (F-test).

6 30 WILSON ET AL FIGURE 3. Scatter plot with linear regression of DXA-volume compared with ADP-volume measures for the 4-compartment population (n ¼ 11). The linear regression had an R 2 ¼ 0.99, significant slope of 0.97 (P, ), and no statistically significant intercept. Statistical significance from zero was defined at the P, 0.05 threshold. ADP, air-displacement plethysmography; ALS, amyotrophic lateral sclerosis; DXA, dual-energy X-ray absorptiometry. repeatability. Although DXA volume may have advantages over UWW and ADP because of lung volume errors, DXA does involve radiation. For this reason, ADP is more appropriate in some special circumstances, such as pregnancy. In this first preliminary description of the DXA-volume method, we calibrated directly to convenient biologically equivalent phantom materials and clinically compared our method with ADP. Two techniques that may be considered gold standards for TBV are CT and MRI. CT is commonly used for measurement of organ volumes and densities (21), muscle volumes, and visceral and subcutaneous fat cross-sectional volumes (22). Studies of wholebody CT volume are rare because of the high doses involved. However, MRI has been used extensively for whole-body composition in both whole-body and regional modes (23). With.30,000 systems in the United States and 50,000 worldwide, DXA can measure total body mass in adults with test retest reliability better than 100 g. DXA is the de facto gold standard body composition technique because of its high reliability, extensive availability, and good safety profile (24 26). In future studies, when combined with a TBW measurement, our DXA-volume method could calculate protein mass indirectly by subtraction of the fat, water, and bone mineral components from the whole body mass. As a small component measured indirectly in the 4C model, protein mass would be very sensitive to measurement errors in the other compartments. Direct validation of protein measures with neutron activation analysis measurements would be ideal, but these potassium counters are expensive, hard to find, and not clinically feasible. Our study has limitations. The soft tissue calibration phantom used to convert R values and HEs to masses and volumes was small and represented a limited number of pixel thicknesses ( cm) and masses ( g). Because of these limited sizes, the calibration phantom was scanned with the use of the PA Spine Scan mode for better spatial resolution. The wholebody scan mode pixels were too large to sample the phantom calibration regions adequately. A larger phantom with a wider range of masses, thicknesses, and compositions that can be scanned in whole-body mode would further improve the mass and volume calibrations. Our calibration phantom was made from stable polymer materials that mimicked biological materials; previous studies have shown these materials to be similar to their biological (protein, lipid, and water) counterparts (11). However, further calibration may be needed against cadaveric samples to fully verify their relation to biological materials. Additionally, our study population was small (11 total participants, 10 with TBW measures). Future work with a larger number of participants would allow for a better description of technique results. The 4C DXA-volume percentage fat had FIGURE 4. Left: Bland-Altman analysis of 4C DXA-volume percentage fat and 4C ADP-volume percentage fat (n ¼ 10). This relation had an R 2 ¼ 0.59, significant slope of 20.25, and significant intercept of Right: Bland-Altman analysis of 4C DXA-volume percentage fat and DXA percentage fat (n ¼ 10). This relation had an R 2 ¼ 0.54, significant slope of 20.16, and significant intercept of Statistical significance from zero was defined at the P, 0.05 threshold. ADP, air-displacement plethysmography; ALS, amyotrophic lateral sclerosis; DXA, dual-energy X-ray absorptiometry; 4C, 4-compartment.

7 TOTAL BODY VOLUME VIA DXA 31 statistically significant biases compared with DXA softwarereported percentage fat and 4C ADP-volume percentage fat. This bias must be further explored before DXAvolume is clinically used for estimation of percentage fat. A larger number of participants will help determine whether these biases are a function of limited sample size. Whereas the calibration phantom had a thickness range that was similar to human body thickness, our DXA-volume method neglects any fan beam magnification effects, which introduce small pixel volume error. We hope to quantify these parallax effects and create corrections in the future. The spatial resolution in whole-body mode also limited this study; with larger pixel sizes in this mode, partial volume artifacts affected pixels on the air/body interface. Our attenuation data set had pixel areas of 0.80 cm 2, whereas the native resolution of the whole-body scan is higher. Other DXA systems, such as the idxa (GE Health care) may provide higher spatial resolution. When we constructed the 4C model, we used BMC values reported by the manufacturer s software because our calibration phantom included only soft tissue materials. Future work will include construction of a phantom with bone material, and validation of our bone segmentation algorithm on a larger scale. We estimated soft tissue values for pixels that contained bone with the use of a small pixel neighborhood. This small neighborhood worked well for most of the body, but the head needed a larger ( ) interpolation window to estimate soft tissue values. Additional study of the head is necessary to understand how this soft tissue estimate affects the region s volume measurements. Although ADP is an innovative technology that is used extensively to measure body volume because of its cost and availability, the technique has many assumptions that make us question its gold standard claim. We would like to validate DXA volume against measures from more accurate, volumetric imaging modalities such as CT or MRI, which do not have residual lung volume issues. A novel method to calculate TBV with the use of a clinical DXA system was developed, compared against ADP as proof of principle, and used in Lohman s 4C body composition model. The DXAvolume approach eliminates many of the inherent inaccuracies associated with displacement measures for volume and, if validated in larger groups of participants, would simplify the acquisition of 4C body composition to a single DXA scan and TBW measure. The authors responsibilities were as follows JPW and JAS: designed the research; JPW, KM, JLS, VWT, LM, and CP: conducted the research; JPW, BF, EJM, VWT, VRB, and JAS: analyzed the data; JPW, KM, EJM, VWT, and JAS: wrote the article; JPW: had primary responsibility for final content; JPW, KM, BF, JLS, EJM, VWT, CLP, LM, VR-B, and JAS: read and approved the final manuscript. None of the authors had a conflict of interest to disclose. REFERENCES 1. Withers RT, LaForgia J, Pillans RK, Shipp NJ, Chatterton BE, Schultz CG, Leaney F. Comparisons of two-, three-, and four-compartment models of body composition analysis in men and women. J Appl Physiol 1998;85: Goldman RF, Buskirk ER. Body volume measurement by underwater weighing: description of a method. In: Brozek J, Henschel A, eds. Techniques for measuring body composition. Natick, MA: National Academy of Sciences, Lohman TG, Going SB. Multicomponent models in body composition research: opportunities and pitfalls. Basic Life Sci 1993;60: Fields DA, Goran MI. Body composition techniques and the fourcompartment model in children. J Appl Physiol 2000;89: Brand WA, Geilmann H, Crosson ER, Rella CW. Cavity ring-down spectroscopy versus high-temperature conversion isotope ratio mass spectrometry; a case study on delta H-2 and delta O-18 of pure water samples and alcohol/water mixtures. Rapid Commun Mass Spectrom 2009;23: Schoeller DA, Ravussin E, Schutz Y, Acheson KJ, Baertschi P, Jequier E. Energy expenditure by doubly labeled water: validation in humans and proposed calculation. Am J Physiol 1986;250:R Wolfe RR, Chinkes DL. Isotope tracers in metabolic research: principles and practice of kinetic analysis. New York, NY: Wiley-Liss, Fields DA, Goran MI, McCrory MA. Body-composition assessment via air-displacement plethysmography in adults and children: a review. Am J Clin Nutr 2002;75: Schoeller DA, Tylavsky FA, Baer DJ, Chumlea WC, Earthman CP, Fuerst T, Harris TB, Heymsfield SB, Horlick M, Lohman TG, Lukaski HC, Shepherd JA, Siervogel RM, Borrud LG. QDR 4500A dual-energy X-ray absorptiometer underestimates fat mass in comparison with criterion methods in adults. Am J Clin Nutr 2005;81: Henderson NK, Price RI, Cole JH, Gutteridge DH, Bhagat CI. Bone density in young women is associated with body weight and muscle strength but not dietary intakes. J Bone Miner Res 1995;10: Laidevant AD, Malkov S, Flowers CI, Kerlikowske K, Shepherd JA. Compositional breast imaging using a dual-energy mammography protocol. Med Phys 2010;37: Wang ZM, Pierson RN Jr, Heymsfield SB. The five-level model: a new approach to organizing body-composition research. Am J Clin Nutr 1992;56: Blake GM, Wahner HW, Fogelman I. The evaluation of osteoporosis: dual energy X-ray absorptiometry and ultrasound in clinical practice. 2nd ed. London, England: Martin Dunitz, Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1: Koo WW, Massom LR, Walters J. Validation of accuracy and precision of dual energy X-ray absorptiometry for infants. J Bone Miner Res 1995;10: Rigo J, Nyamugabo K, Picaud JC, Gerard P, Pieltain C, De Curtis M. Reference values of body composition obtained by dual energy X-ray absorptiometry in preterm and term neonates. J Pediatr Gastroenterol Nutr 1998;27: Vescovi JD, Zimmerman SL, Miller WC, Hildebrandt L, Hammer RL, Fernhall B. Evaluation of the BOD POD for estimating percentage body fat in a heterogeneous group of adult humans. Eur J Appl Physiol 2001; 85: Wells JC, Douros I, Fuller NJ, Elia M, Dekker L. Assessment of body volume using three-dimensional photonic scanning. Ann N Y Acad Sci 2000;904: McCrory MA, Mole PA, Gomez TD, Dewey KG, Bernauer EM. Body composition by air-displacement plethysmography by using predicted and measured thoracic gas volumes. J Appl Physiol 1998;84: Fields DA, Hunter GR, Goran MI. Validation of the BOD POD with hydrostatic weighing: Influence of body clothing. Int J Obes Relat Metab Disord 2000;24: Heymsfield SB, Fulenwider T, Nordlinger B, Barlow R, Sones P, Kutner M. Accurate measurement of liver, kidney, and spleen volume and mass by computerized axial-tomography. Ann Intern Med 1979;90: Yoshizumi T, Nakamura T, Yamane M, Islam AHMW, Menju M, Yamasaki K, Arai T, Kotani K, Funahashi T, Yamashita S, Matsuzawa Y. Abdominal fat: standardized technique for measurement at CT. Radiology 1999;211: Shen W, Punyanitya M, Wang Z, Gallagher D, St.-Onge MP, Albu J, Heymsfield SB, Heshka S. Total body skeletal muscle and adipose tissue volumes: estimation from a single abdominal cross-sectional image. J Appl Physiol 2004;97: Ellis KJ, Shypailo RS, Steinberg FM, Lewis RD, Young RL, Wong WW. Reproducibility of fan-beam DXA measurements in adults and phantoms. J Clin Densitom 2004;7: Hagiwara S, Lane N, Engelke K, Sebastian A, Kimmel DB, Genant HK. Precision and accuracy for rat whole body and femur bone mineral determination with dual X-ray absorptiometry. Bone Miner 1993;22: Leonard CM, Roza MA, Barr RD, Webber CE. Reproducibility of DXA measurements of bone mineral density and body composition in children. Pediatr Radiol 2009;39: