Traditional indirect methods of body composition assessment (densitometry, hydrometry, K40 spectroscopy) are based on a two-component model in which the body mass (BM) is assumed to be comprised of fat and fat-free components. Approximately 40 yr ago, in an attempt to reduce the error associated with percent fat (%fat) estimates based on the two-component model, Siri (33) proposed a three-component model in which %fat was estimated from measures of body water and body density. This reduced the number of assumptions underlying the two-component model and led to more accurate assessment of body composition.
The recent development of dual-energy x-ray absorptiometry (DXA) has enabled researchers to assess another component of the fat-free mass (FFM), bone mineral. The ability to assess this additional chemical constituent has led to the development of a four-component model (31). The four-component model has been used as a criterion model to test the accuracy of body composition estimates based on the simpler two-component model (12,21,24,26). However, in order for the four-component model to be considered a criterion method of body composition assessment, the assumed densities of fat and each of the primary FFM constituents, water, mineral, and protein, must be accurate. Furthermore, accurate measurement of body density, body water, and bone mineral content (BMC) is necessary. Although the densities of fat (0.9007 g·cm−3), water (0.9937 g·cm−3), mineral (3.038 g·cm−3), and protein (1.34 g·cm−3) at 36°C are accepted as being correct (4) and the validity of body density and body water measurements have been established, the accuracy of BMC measurements by DXA can be questioned. Studies have found substantial differences between BMC estimates from DXA instruments manufactured by different companies (22,36). Although this discrepancy in BMC theoretically alters estimates of %fat based on the four-component model, the magnitude of its impact is unknown. According to some investigators, a deviation in mineral by 1% FFM would alter %fat by 5% BM (38), whereas others suggest the deviation is 2% BM (17) or less (21).
In addition to estimating BMC, DXA can separate soft tissue into fat and fat-free components. Because estimates of %fat from DXA have compared favorably with densitometry, have high precision, and are easy to obtain, some researchers have used it as a criterion body composition method (35). Although some studies using a four-component model as a criterion method have shown DXA to be a valid measure of body composition (21,25), other studies (2,5,26) and reviews (13,27) have questioned its accuracy. Furthermore, estimates can vary depending upon the instrument (1,22,36) and software (22,36,37) used. Currently, there are no studies that have assessed body composition estimates from two different DXA instruments simultaneously, using a four-component model as a criterion.
The primary purpose of the present study was to determine the influence of BMC measurements obtained from DXA instruments manufactured by two different companies on %fat estimates from a four-component model. Based on theoretical calculations (21), we hypothesized that using bone mineral estimates from different DXA instruments would have a minimal influence on estimates of %fat from a four-component model (<1% BM). A second purpose was to test the accuracy of %fat estimates from two different DXA instruments using estimates from a four-component model as the criterion. We hypothesized that estimates of %fat from both DXA instruments would agree closely with estimates from a four-component model in a group of young men and women.
Twenty-four healthy young men (13 Caucasian and 1 Asian) and women (9 Caucasian and 1 Asian) participated in the study. The study was approved by the University Institutional Review Boards for Human Subjects. Written consent was obtained before testing. Physical characteristics of the subjects (mean ± SD) were: age = 26.7 ± 6.0 yr, height = 171.6 ± 9.6 cm, BM = 72.6 ± 15.1 kg, body mass index = 24.4 ± 3.5 m·kg−2, and body density = 1.055 ± 0.021 g·cm−3.
Subjects reported for testing following an overnight fast (∼12 h) of food and beverages except water. Physical characteristics, body density, body water, bone mineral, and soft tissue body composition were measured. Bone mineral and soft tissue body composition were measured using the QDR 1000/W DXA (DXAQDR) and the DPX-L DXA (DXADPX-L). All measurements were collected during a single test session (∼5 h) and DXA test order was counterbalanced.
Body density was measured using underwater weighing and Archimedes’ principle to determine body volume. Body mass was measured in air to the nearest 0.01 kg using an electronic scale. Body mass was measured under water to the nearest 0.025 kg using a Chatillon autopsy scale, and residual lung volume was measured simultaneously using a closed-circuit oxygen-rebreathing nitrogen-dilution technique modified from Goldman and Buskirk (9). The volume of gas in the gastrointestinal tract was assumed to be 100 mL. The within-subjects SD of duplicate measurements of body density ∼1 wk apart in five subjects was 0.0016 g·cm−3 (21).
Total body water was determined using deuterium oxide dilution (6,19). After an initial blood draw, subjects consumed a solution of deuterium oxide (∼0.3 g·kg−1 BM) and distilled water (100 mL), which was rinsed with 100 mL of distilled water. After a 3-h equilibration period, another blood sample was drawn. Blood samples were immediately centrifuged for 20 min at 3000 rpm and plasma was stored at −70°C. Urine was collected every hour during the equilibration period. Plasma samples were purified by a diffusion method (6), in which equal volumes of plasma and deionized water (∼1.5 mL) were incubated at 37°C for 48 h in 8.3 cm Conway diffusion dishes (Bel-Air Products, Pequannock, NJ). The purification procedure was validated in our laboratory using known quantities of deuterium mixed with plasma. Recovery of the initial deuterium concentration averaged 98% in 10 samples (range = 95–102%), identical to results reported by DeLuca and Friedl (7). After incubation, purified water samples were analyzed using a single-beam infrared spectrophotometer with a 4-μm fixed filter (Miran 1FF, Foxboro, MA) at 16°C. Deuterium concentrations were determined using a standard curve.
Total body water was corrected for deuterium lost in urine and reduced by 4% to account for hydrogen exchange between deuterium, and protein and carbohydrate during the 3-h equilibration period (30). The within-subjects SD of duplicate measurements of body water ∼1 wk apart in five subjects was 0.75 L (21). The within-subjects SD of replicate (2–4) measurements of body water using the same plasma sample on a given day was 0.71 L (21), which was considered the technical error associated with body water measurement. The body water measurement procedure using the infrared spectrophotometer to measure deuterium concentration was validated in our laboratory against measures in which an isotope-ratio mass spectrometer (Finnigan, Bremen, Germany, and MAT 251) was used to determine deuterium concentration. Body water values determined using the two procedures were highly correlated and not different when measured in 50 women (r = 0.99, mean difference = 0.06 ± 0.89 L).
Dual-energy x-ray absorptiometry.
BMC, bone mineral density (BMD), and soft tissue body composition (fat mass and fat-free soft tissue) were measured using DXAQDR (Enhanced Whole Body Analysis software version 5.71; Hologic, Inc., Waltham, MA) and DXADPX-L (software version 1.3Z; medium scan mode; Lunar, Corp., Madison, WI) instruments. One subject was taller than the scanning region of the DXAQDR instrument. Therefore, BMC, BMD, fat mass, and fat-free soft tissue (FFST) were determined using a two-scan approach, in which head, trunk, and arm measurements from one scan were combined with leg measurements from a second scan and regression equations were employed. This technique has been found to accurately predict BMC, BMD, fat mass, and FFST from one complete scan (23). As recommended by the DXADPX-L manufacturer, subjects with a body thickness greater than 28 cm (N = 2) were scanned using the slow-scan mode. All other subjects were scanned using the medium-scan mode. Fat-free mass from DXA was determined by combining BMC and FFST.
To ensure consistency, one trained technician performed and analyzed all scans at each laboratory. A calibration step wedge, consisting of thermoplastic resin (68% fat) and thermoplastic resin-aluminum (−10% fat) steps calibrated against stearic acid (100% fat) and water (8.6% fat; Hologic, Inc.) and placed adjacent to each subject during scans using DXAQDR, was used to calibrate fat mass and FFST. Quality control for DXAQDR was checked before each test session by scanning a lumbar spine phantom consisting of calcium hydroxyapatite embedded in a cube of thermoplastic resin (model DPA/QDR-1; Hologic x-caliber anthropometric spine phantom). The coefficient of variation of 449 scans of the phantom during a 5-yr period was 0.29%. The DXADPX-L instrument was calibrated daily using a standard calibration block, which consisted of a thermoplastic resin housing with three bone-equivalent chambers filled with hydroxyapatite. BMD and BMC measurements for each calibration scan were required to be within 2% of the known BMD and BMC values of the calibration block before subjects could be tested.
The DXAQDR within-subjects SD of duplicate measurements of BMC and %fat ∼1 wk apart in five subjects was 7.2 g and 0.2% BM, respectively (21). The DXADPX-L within-day SD of replicate measurements of BMC and %fat were reported as 28 g and 0.9% BM, respectively (34).
Body composition calculations.
BMC from DXA is assumed to represent bone ash, which is bone mineral minus volatile components that are lost during the ashing process, such as water of crystallization and carbon dioxide from carbonate (4). Bone ash has been used to validate DXA (11) and to calculate bone mineral and body mineral (10). Bone mineral is calculated by multiplying bone ash by 1.0436 (4). Nonbone mineral is determined from bone ash, assuming the nonbone mineral to bone ash ratio is 0.2305 (4). Body mineral is bone mineral and nonbone mineral combined.
Percent fat based on a four-component model was calculated using body density, body water, body mineral, and the following equation of Lohman (14):MATH 1 with Db representing body density, w representing the water fraction of BM, and m representing the mineral fraction of BM. Percent fat based on the four-component model was calculated using and BMC from DXAQDR (4CQDR) or DXADPX-L (4CDPX-L). The within-subject SD of duplicate measurements of %fat from the four-component model ∼1 wk apart in five subjects was 0.7% BM (21).
The density of the fat-free mass (Dffm) was calculated using the following equation:MATH 2 in which wffm is the water fraction of the FFM, dw is the density of water, mffm is the mineral fraction of the FFM, dm is the density of mineral, pffm is the protein fraction of the FFM, and dp is the density of protein. The within-subject SD of duplicate measurements of Dffm ∼ 1 wk apart in five subjects was 0.002 g·cm−3 (21).
A two-way (gender × method) repeated measures analysis of variance and simple contrasts were used to determine significant differences among body composition measurements from 4CQDR, 4CDPX-L, DXAQDR and DXADPX-L. The Bonferroni adjustment was used to hold alpha level for the six comparisons for a given measurement to 0.05. The alpha level for individual pair-wise comparisons was 0.0083. Correlation and linear regression analyses were performed to determine relationships between differences in body composition measurements. Accuracy of regression equations was assessed using the SEE and total error (TE). Total error was determined using the formula: (Σ (Y′ - Y)2/N), where Y′ is the predicted %fat and Y is the criterion %fat (16). The technique described by Bland and Altman (3) was used to determine agreement between body composition measurements from 4CQDR, 4CDPX-L, DXAQDR, and DXADPX-L. The mean difference (meanDiff) and SD of the difference (SDDiff) indicated systematic and nonsytematic differences between %fat estimates from different methods.
There was no difference in the pattern of results between men and women (no gender × method interaction) for %fat comparisons; therefore, the analysis reported is based on all subjects. BMC, BMD, and fat-free soft tissue measurements from DXAQDR and DXADPX-L are reported in Table 1. As expected, BMC and BMD were significantly lower, and FFST higher, using DXAQDR than DXADPX-L. Measures of %fat, fat mass, and FFM from 4CQDR, 4CDPX-L, DXAQDR, and DXADPX-L are presented in Table 2. When body composition measures from the two different four-component models were compared, %fat and fat mass were significantly lower, whereas FFM was significantly higher using 4CQDR than 4CDPX-L. When body composition measures from the two DXA instruments were compared, %fat and fat mass were significantly lower and FFM significantly higher using DXAQDR than DXADPX-L. When measures of body composition from the four-component models and DXA were compared, DXAQDR yielded significantly lower measures of %fat than 4CDPX-L.
Agreement among 4CQDR, 4CDPX-L, DXAQDR, and DXADPX-L measures of %fat is shown in Figure 1. There was a small significant systematic difference between estimates of %fat from 4CQDR and 4CDPX-L (meandiff = 0.7% BM), with little nonsystematic disagreement between methods (SDDiff = 0.2% BM; SEE = 0.2% BM;Fig. 1A). Individual differences ranged from −1.0 to −0.3% BM. Larger systematic and nonsystematic differences in %fat were observed between DXAQDR and DXADPX-L (meanDiff ± SDDiff = 1.7 ± 1.5% BM; SEE = 1.5% BM;Fig. 1B), but differences were still quite small. Individual differences ranged from −3.5 to 3.1% BM.
Estimates of %fat from DXAQDR were systematically lower than estimates from 4CQDR and 4CDPX-L (Fig. 1), but the difference was significant only between DXAQDR and 4CDPX-L (meanDiff = −2.8% BM). Large nonsystematic differences in estimates of %fat were found between DXAQDR and 4CQDR (SDDiff = 4.6% BM), DXADPX-L and 4CDPX-L (SDDiff = 4.3% BM), [DXADPX-L and 4CQDR (SDDiff = 4.3% BM)], and [DXAQDR and 4CDPX-L (SDDiff = 4.7% BM)], as shown in Figure 1C–F. Furthermore, SEE and TE were large for all comparisons, ranging from 4.0 to 4.3% BM and 4.2 to 5.4% BM, respectively. Individual differences between DXA and four-component model estimates of %fat ranged from −12.8 to 10.2% BM. Correlations between differences in DXA and four-component model estimates of %fat and mean %fat were not significant (correlation range = 0.04 to 0.29, P > 0.05), except between 4CQDR and 4CDPX-L (r = 0.59, P < 0.05).
The absolute and fractional composition of the FFM and the Dffm are presented in Tables 3 and 4, respectively. Mineral mass was lower (meanDiff ± SDDiff = 0.43 ± 0.14 kg) and protein mass higher (meanDiff ± SDDiff = 0.91 ± 0.30 kg) using 4CQDR compared to 4CDPX-L, resulting in significantly lower water and mineral fractions and a higher protein fraction of the FFM. The difference in DXAQDR and DXADPX-L estimates of BMC and mineral are perfectly correlated (r = 1.00) with the difference in 4CQDR and 4CDPX-L estimates of Dffm. Because mineral is more dense than protein (3.038 g·cm−3 vs 1.34 g·cm−3), and the ratio of mineral mass to protein mass estimated by 4CQDR (0.273 ± 0.037) was significantly less than the ratio estimated by 4CDPX-L (0.328 ± 0.045), Dffm from 4CQDR was lower than from 4CDPX-L (meanDiff ± SDDiff = 0.0020 ± 0.0005 g·cm−3). The difference between estimates of %fat from 4CQDR and 4CDPX-L was significantly (P < 0.001) correlated with the difference in estimates of Dffm (r = 0.82), FFM (r = −0.82) and BMC (r = 0.82) between the two four-component models.
Previous reports have demonstrated significantly lower estimates of BMC using DXAQDR than DXADPX-L (22,36); however, the impact of using the different measures of BMC on body composition estimates from a four-component model had not been determined. The major finding of the present study was that %fat and fat mass were only slightly lower (<1% BM and 0.4 kg) and FFM slightly higher (0.5 kg) when estimated from a four-component model using the ∼11% lower BMC from DXAQDR compared with DXADPX-L. An additional finding was that body composition estimates from DXAQDR and DXADPX-L were not different from those from a four-component model in which BMC was measured using the same instrument. However, relatively large variability was found between %fat estimated from DXA and %fat estimated from the four-component models. Furthermore, estimates of %fat from DXAQDR were significantly lower than estimates from 4CDPX-L.
During the past 10 yr, there has been an increased use of four-component models to assess the validity of body composition methods based on lesser-order models. Because mineral has the highest density of the major constituents of the FFM, some investigators believed that mineral had the greatest influence on Dffm (20,38). Wang et al. (38) suggested a deviance in the mineral to lean BM ratio of 1% would alter Db by 0.01 g·cm−3 and %fat by 5%. Martin and Drinkwater (20) proposed a similar theory. Based on cadaver work, they suggested normal variation in bone, which is primarily mineral, can cause the assumed Dffm of 1.1 g·cm−3 to vary by 0.01 g·cm−3. According to their calculations, such a variation in Dffm would alter %fat estimates by 4% BM. Conversely, others suggested the impact of mineral on %fat estimates is less dramatic. According to the calculations of Lohman and Going (17), mineral would need to be altered by ∼2% FFM, from 6.87 to 4.8% FFM, for %fat to be altered by 4% BM. Modlesky et al. (21) calculated that a deviation in mineral by ∼1% FFM would alter %fat estimates by ∼ 0.8% BM (21). Data from the present study support our hypothesis that the substantial difference in BMC resulting from the use of different DXA instruments has a minimal influence on the assessment of %fat. The reason for the small impact on %fat estimates is the slight contribution of BMC and total body mineral to the FFM. Although the density of mineral (3.038 g·cm−3) is greater than the other primary constituents of the FFM (water = 0.9937 g·cm−3 and protein = 1.34 g·cm−3), it comprises only ∼6.8% of the FFM. Thus, the overall contribution of mineral to Dffm (m/Dffm) is only ∼17.2% as determined in the following equation derived from :MATH 3
In contrast, an 11% error in body water measurement would result in a marked change in %fat estimation (∼4.5% BM), because water is the largest contributor to the FFM (∼61.1%).
The importance of measuring water rather than mineral in body composition assessment was also demonstrated in studies that compared %fat estimates based on a four-component model with estimates based on three-component models that use either body water or mineral in addition to body density in their calculations. Roemmich et al. (26) found that the bias associated with %fat estimates from a two-component model using body density alone was significantly reduced when body density measures were combined with body water or bone mineral (5.15% vs 0.40% and 0.75% BM, respectively) in children. However, the TE in %fat was 2.4-fold lower when body density was combined with water rather than bone mineral. Modlesky et al. (21) reported that measuring mineral in addition to body density reduced the systematic difference in %fat between 4CQDR and a two-component model using body density in weight trainers (0.3% vs 4.1% BM, respectively), but considerable disagreement between methods remained (SDDiff = 1.9% BM). In controls, the systematic difference was larger (2.7% vs 0.5% BM), and the large disagreement remained (SDDiff = 2.7% BM). When the three-component model employed water rather than mineral, %fat was only slightly lower than estimates from 4CQDR in weight trainers (1.0% BM) and controls (0.9% BM), with very close agreement within each group (SDDiff = 0.6% and 0.4% BM, respectively). These data further support the greater importance of measuring water than mineral in body composition assessment.
Whether DXAQDR or DXADPX-L provide a more accurate measure of BMC is not known. Tothill et al. (36) found BMC and BMD of a moderately anthropomorphic phantom, consisting of aluminum, acrylic, and thin vinyl plastic, to be underestimated by DXAQDR (7% and 7%, respectively) and overestimated by DXADPX-L (8% and 2%, respectively). Ho et al. (11) found DXAQDR to yield in vitro BMC estimates ∼8.9% lower than actual bone ash measures from human lumbar vertebrae (N = 11), similar to the discrepancy between DXAQDR and DXADPX-L found in the present study. The underestimation of BMC in vitro using DXAQDR was further supported by Louis et al. (18), who found lumbar vertebrae BMC (N = 34) to be lower than estimates from neutron activation analysis (BMCNAA = 1.016·BMCDXA + 0.99). These reports suggest DXAQDR underestimates BMC and DXADPX-L may overestimate BMC.
To account for the underestimation of BMC by DXAQDR, some investigators (26,32) have divided the BMC measurement by 0.88. This correction is based on the work of Sabin et al. (29), who reported volumetric BMD, measured using anteroposterior and lateral spine scans, to be 12% lower than actual volumetric BMD in lumbar vertebrae. However, because Sabin et al. (28) did not report BMC or bone mineral area, it was unclear whether the underestimation of BMD was due to an underestimation of BMC, an overestimation of bone area, or both. In a more extensive report of their data (28), anterior posterior lumbar spine BMC from DXAQDR was found to be 14% lower than bone ash, similar to the correction suggested by Siconolfi et al. (32). These data suggest applying a correction factor to BMC from DXAQDR may result in more accurate measures of bone ash; however, based on the present study, the impact of a BMC correction on %fat estimates from a four-component model is minimal.
Previous studies have tested the accuracy of body composition estimates from DXAQDR and DXADPX-L using a four-component model as the criterion; however, their accuracy had never been tested simultaneously. Penn et al. (24) found DXADPX-L to yield estimates of %fat no different than estimates from 4CDPX-L in 10 male long-distance runners (meanDiff = 0% BM) and 10 controls (meanDiff = 0.6% BM). Modlesky et al. (21) found similar results with DXAQDR in male weight trainers and controls (meanDiff ± SDDiff = 0.4 ± 2.5% BM). Conversely, Bergsma-Kadjik et al. (2) found DXADPX-L to significantly underestimate estimates of %fat based on a four-component model in young (19–27 yr; 3.1 ± 1.8% BM) and elderly (65–78 yr; 5.3 ± 3.8% BM) women.
Lohman (15) suggested the accuracy of %fat estimation from a new method be determined based on the SEE of predicting %fat determined from a criterion method, with a SEE lower than 3.0% BM indicating good accuracy. Another measure of the accuracy of predicting a criterion estimate of %fat is the TE, which combines the SEE and the mean difference between the predicted and actual %fat from a criterion method (16). Using SEE and TE, some reports suggest DXAQDR and DXADPX-L provide valid measures of %fat (8,25). Prior et al. (25) found that DXAQDR, validated against 4CQDR, accurately assessed %fat in young adults who varied in gender, race, athletic status, body size, musculoskeletal development, and body fatness (meanDiff = −0.4 ± 2.9% BM; SEE = 2.8% BM; TE = 2.9% BM). To support their findings, Prior et al. (25) examined data (N = 10) previously reported by Friedl et al. (8). Percent fat from DXADPX-L was found to agree well with %fat from 4CDPX-L (meanDiff = 0.4 ± 1.9% BM; SEE = 1.9% BM). Conversely, the accuracy of DXA has been questioned using SEE and TE (26). Roemmich et al. (26) found that %fat from DXAQDR overestimated %fat from a four component model (23.6 vs 21.7%) in 24 boys and 23 girls and that the prediction had a large TE (4.4% BM). Clasey et al. (5) found DXAQDR to have a large SEE (5.0% BM) when predicting %fat from 4CQDR in 41 older adults (67.0 ± 5.1 yr). In the present study, when %fat estimates from DXA were compared with estimates from a four-component model, only differences between DXAQDR and 4CDPX-L were significant. However, SEE and TE were large for all comparisons, ranging from 4.0 to 4.3% BM and 4.2 to 5.4% BM, respectively.
Roemmich et al. (26) suggested the higher water fraction of the FFM in children (∼75%) lead to the erroneous estimates of %fat from DXAQDR. Our finding of a lower water fraction of the FFM (4CQDR = 70.9%; 4CDPX-L = 71.5%) than is assumed in lower-order body composition models (72–73.8%) and the underestimation of %fat from DXAQDR compared to 4CDPX-L (2.8 ± 4.7% BM) supports their contention. However, no relationship was found between the difference in DXA and four-component model estimates of %fat, and the water fraction of the FFM. Moreover, because DXA interprets pure water as ∼ 8% fat, a ± 5% FFM deviation in the water fraction would alter fat mass or FFM minimally (<0.5 kg; 13). Further studies are needed to assess the accuracy of DXAQDR and DXADPX-L.
It is possible that measurement error or an overcorrection for hydrogen exchange in the body may have deflated body water values in the present study, causing an overestimation of %fat using the four-component model. However, the body water procedure in our laboratory has been validated using mass spectrophotometry and the correction factor for deuterium exchange has been recommended following a careful review of the literature (30). Furthermore, %fat estimates from the two DXA instruments were not statistically different than estimates from those from the respective four-component model.
In conclusion, the results of the present study suggest use of BMC from different DXA instruments has minimal impact on %fat, fat mass, and FFM estimates from a four-component model. Although estimates of %fat, fat mass, and FFM from DXAQDR and DXADPX-L were not different than estimates from those from the respective four-component model, there was considerable variability between methods. Furthermore, %fat from DXAQDR was lower than %fat from 4CDPX-L. Further studies that test larger sample sizes (100–400 subjects; 16) and specific population groups are needed to further assess the accuracy of body composition measurements from DXA.
We thank Linda B. Rosskopf and Teresa K. Snow for their technical assistance and all participants for their time and effort.
1. Abrahamsen, B., J. Gram, T. B. Hansen, and H. Beck-Nielsen. Cross calibration of QDR-2000 and QDR-1000 dual-energy x-ray densitometers for bone mineral and soft-tissue measurements. Bone 16:385–390, 1995.
2. Bergsma-Kadijk, J. A., B. Baumeister, and P. Deurenberg. Measurement of body fat in young and elderly women: comparison between a four-component model and widely used reference methods. Br. J. Nutr. 75:649–657, 1996.
3. Bland, J. M., and D. G. Altman. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1:307–310, 1986.
4. Brozek, J., F. Grande, J. T. Anderson, and A. Keys. Densitometric analysis of body composition
: revision of some quantitative assumptions. Ann. N. Y. Acad. Sci. 110:113–140, 1963.
5. Clasey, J. L., M. L. Hartman, J. Kanaley, et al. Body composition
by DEXA in older adults: accuracy and influence of scan mode. Med. Sci. Sports Exerc. 29:560–567, 1997.
6. Davis, J. M., D. R. Lamb, W. A. Burgess, and W. P. Bartoli. Accumulation of deuterium oxide in body fluids after ingestion of D2
O-labeled beverages. J. Appl. Physiol. 63:2060–2066, 1987.
7. Deluca, J. P., and K. E. Friedl. Validation of a simplified method of total body water measurement using deuterium oxide and diffusion dish purification. Med. Sci. Sports Exerc. 24:S8, 1992.
8. Friedl, K. E., J. P. Deluca, L. J. Marchitelli, and J. A. Vogel. Reliability of body-fat estimations from a four-compartment model by using density, body water, and bone mineral measurements. Am. J. Clin. Nutr. 55:764–770, 1992.
9. Goldman, R. F., and E. R. Buskirk. Body volume measurement by underwater weighing: description of a method. In: Techniques for Measuring Body Composition
, J. Brozek (Ed.). Washington, DC: National Academy of Sciences, 1961, pp. 77–89.
10. Heymsfield, S. B., S. Lichtman, R. N. Baumgartner, et al. Body composition
of humans: comparison of two improved four-component models that differ in expense, technical complexity, and radiation exposure. Am. J. Clin. Nutr. 52:52–58, 1990.
11. Ho, C. P., R. W. Kim, M. B. Schaffler, and D. J. Sartoris. Accuracy of dual-energy radiographic absorptiometry of the lumbar spine: cadaver study. Radiology 176:171–173, 1990.
12. Hopkinson, J. M., N. F. Butte, K. J. Ellis, W. W. Wong, M. R. Puyau, and E. O’Brian Smith. Body fat estimation in late pregnancy and early postpartum: comparison of two-, three, and four-component models. Am. J. Clin. Nutr. 65:432–438, 1997.
13. Kohrt, W. M. Body composition
by DXA: tried and true? Med. Sci. Sports Exerc. 27:1349–1353, 1995.
14. Lohman, T. G. Applicability of body composition
techniques and constants for children and youths. In: Exercise and Sports Science Reviews, K. Pandolf (Ed.). Baltimore: Williams & Wilkins, 1986, pp. 325–357.
15. Lohman, T. G. Dual energy x-ray absorptiometry. In: Human Body Composition
, A. F. Roche, S. B. Heymsfield, and T. G. Lohman (Eds.). Champaign, IL: Human Kinetics, 1996, pp. 63–78.
16. Lohman, T. G. Skinfolds and body density and their relation to body fatness: a review. Hum. Biol. 53:181–225, 1981.
17. Lohman, T. G., and S. B. Going. Multi-component models in body composition
research: opportunities and pitfalls. In: Human Body Composition
: In Vivo Methods, Models, and Assessment. K. J. Ellis and J. D. Eastman (Eds.). New York: Plenum Press, 1993, pp. 53–58.
18. Louis, O., P. Van Den Winkel, P. Covens, A. Schoutens, and M. Osteaux. Dual-energy x-ray absorptiometry
of lumbar vertebrae: relative contribution of body and posterior elements and accuracy in relation with neutron activation analysis. Bone 13:317–320, 1992.
19. Lukaski, H. C., and P. E. Johnson. A simple, inexpensive method of determining total body water using a tracer dose of D2
O and infrared absorption of biological fluids. Am. J. Clin. Nutr. 41:363–370, 1985.
20. Martin, A. D., and D. T. Drinkwater. Variability in the measures of body fat: assumptions of technique? Sports. Med. 11:277–288, 1991.
21. Modlesky, C. M., K. J. Cureton, R. D. Lewis, B. M. Prior, M. A. Sloniger, and D. A. Rowe. Density of the fat-free mass and estimates of body composition
in male weight trainers. J. Appl. Physiol. 80:2085–2096, 1996.
22. Modlesky, C. M., R. D. Lewis, K. A. Yetman, et al. Comparison of body composition
and bone mineral measurements from two DXA instruments in young men. Am. J. Clin. Nutr. 64:669–676, 1996.
23. Modlesky, C. M., B. M. Prior, R. D. Lewis, and K. J. Cureton. Measurement of body composition
in tall individuals using dual-energy x-ray absorptiometry
. Med. Sci. Sports Exerc. 29:S37, 1997.
24. Penn, I., Z. Wang, K. M. Buhl, D. B. Allison, S. E. Burastero, and S. B. Heymsfield. Body composition
and two-compartment model assumptions in male long distance runners. Med. Sci. Sports Exerc. 26:392–397, 1994.
25. Prior, B. M., K. J. Cureton, C. M. Modlesky, et al. In vivo validation of whole-body composition
estimates from dual-energy x-ray absorptiometry
. J. Appl. Physiol. 83:623–630, 1997.
26. Roemmich, J. N., P. A. Clark, A. Weltman, and A. D. Rogol. Alterations in growth and body composition
during puberty. I. Comparing multicomponent body composition
models. J. Appl. Physiol. 83:927–935, 1997.
27. Roubenoff, R., J. J. Kehayias, B. Dawson-Hughes, and S. B. Heymsfield. Use of dual-energy x-ray absorptiometry
in body composition
studies: not yet a “gold standard.” Am. J. Clin. Nutr. 58:589–591, 1993.
28. Sabin, M. A., G. M. Blake, S. M. Maclaughlin-Black, and I. Fogelman. The accuracy of volumetric bone density measurements in dual x-ray absorptiometry. Calcif. Tissue Int. 56:210–214, 1995.
29. Sabin, M. A., G. M. Maclaughlin, G. M. Blake, and I. Fogelman. A study of the accuracy of volumetric spinal bone density measurements with the Hologic QDR-2000. In:Ninth International Bone Densitometry Workshop Traverse City Michigan 1992
, Waltham, MA: Hologic, 1992, p. 6.
30. Schoeller, D. A. Hydrometry. In: Human Body Composition
. A. F. Roche, S. B. Heymsfield, and T. G. Lohman (Eds.). Champaign, IL: Human Kinetics, 1996, pp. 25–43.
31. Selinger, A. The body as a three-component system. Doctoral Dissertation. University of Illinois, Urbana, 1977.
32. Siconolfi, S. F., R. J. Gretebeck, and W. W. Wong. Assessing total body protein, mineral, and BMC from total body water and body density. J. Appl. Physiol. 79:1837–1843, 1995.
33. Siri, W. E. The gross composition of the body. Adv. Biol. Med. Phys. 4:239–280, 1956.
34. Svendsen, O. L., J. Haarbo, C. Hassager, and C. Christiansen. Accuracy of measurements of body composition
by dual-energy x-ray absorptiometry
in vivo. Am. J. Clin. Nutr. 57:605–608, 1993.
35. Svendsen, O. L., J. Haarbo, B. L. Heitmann, A. Gotfredsen, and C. Christiansen. Measurement of body fat in elderly subjects by dual-energy x-ray absorptiometry
, bioelectrical impedance, and anthropometry. Am. J. Clin. Nutr. 53:1117–1123, 1991.
36. Tothill, P., A. Avenell, and D. M. Reid. Precision and accuracy of measurements of whole-body bone mineral: comparisons between Hologic, Lunar and Norland dual-energy x-ray absorptiometers. Br. J. Radiol. 67:1210–1217, 1994.
37. Van Loan, M. D., N. L. Keim, K. Berg, and P. L. Mayclin. Evaluation of body composition
by dual energy x-ray absorptiometry and two different software packages. Med. Sci. Sports Exerc. 27:587–591, 1995.
38. Wang, J., S. B. Heymsfield, M. Aulet, J. C. Thorton, and R. N. Pierson. Body fat from body density: underwater weighing vs. dual-photon absorptiometry. Am. J. Physiol. 256:E829–E834, 1989.