SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

The aim of this study was to determine the accuracy of dual-energy X-ray absorptiometry (DXA)-derived percentage fat estimates in obese adults by using four-compartment (4C) values as criterion measures. Differences between methods were also investigated in relation to the influence of fat-free mass (FFM) hydration and various anthropometric measurements. Six women and eight men (age 22–54 years, BMI 28.7–39.9 kg/m2, 4C percent body fat (%BF) 31.3–52.6%) had relative body fat (%BF) determined via DXA and a 4C method that incorporated measures of body density (BD), total body water (TBW), and bone mineral mass (BMM) via underwater weighing, deuterium dilution, and DXA, respectively. Anthropometric measurements were also undertaken: height, waist and gluteal girth, and anterior-posterior (A-P) chest depth. Values for both methods were significantly correlated (r2 = 0.894) and no significant difference (P = 0.57) was detected between the means (DXA = 41.1%BF, 4C = 41.5%BF). The slope and intercept for the regression line were not significantly different (P > 0.05) from 1 and 0, respectively. Although both methods were significantly correlated, intraindividual differences between the methods were sizable (4C-DXA, range = −3.04 to 4.01%BF) and significantly correlated with tissue thickness (chest depth) or most surrogates of tissue thickness (body mass, BMI, waist girth) but not FFM hydration and gluteal girth. DXA provided cross-sectional %BF data for obese adults without bias. However, individual data are associated with large prediction errors (±4.2%BF). This error appears to be associated with tissue thickness indicating that the DXA device used may not be able to accurately account for beam hardening in obese cohorts.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

Overweight and obesity are associated with increased morbidities and mortality. Mortality risk is increased among those with BMI >25.0 kg/m2, and is greatly elevated among those with BMI exceeding 30.0 kg/m2 (1). There is robust evidence of increased incidences of coronary heart disease, hypertension, diabetes, and certain forms of cancer among overweight and obese adults (2). The high incidence of obesity in western societies (3,4) has therefore provided impetus to accurately describe body composition and interventional changes in body composition in obese cohorts.

The use of dual-energy X-ray absorptiometry (DXA) to determine body composition, both whole body and regional, has burgeoned over the last decade. This expedient technique involves subjects lying on a scan platform for ∼5 min as an X-ray source and detector move over the body in a rectilinear fashion. Differential attenuation of the low- and high-energy photons in the beam allows for the determination of the relative amounts of bone mineral free lean tissue, fat, and bone mineral in each pixel of the scan area. The resultant radiation exposure from a whole body scan is relatively minor (∼0.3 µSv) and is less than the unavoidable daily background radiation (∼2.0 µSv/24 h) thereby making DXA suitable for most applications including pediatric testing.

Although many researchers have embraced DXA for determining body composition and some have come to consider it as a criterion method, significant limitations have been identified. Two large studies (5,6) reported that DXA significantly underestimated percent body fat (%BF) in lean individuals. van der Ploeg et al. (6) revealed a large underestimation (3.7%BF) of body fat by DXA in lean, physically active individuals compared with values derived via their four-compartment (4C) criterion method. Moreover, a Bland-Altman plot of the entire data set (n = 152; %BF range: 6.5–36.6%BF) demonstrated a definite trend for DXA to progressively underestimate the %BF of leaner individuals. However, it was not possible to determine from these data whether the trend would progress to an overestimation of %BF via DXA in those with very high levels of fat (≥35%BF) because only a few subjects fell in this range. Work with children and adolescents (7) has revealed that DXA not only underestimates %BF in lean individuals but it overestimates fat in the obese when compared with 4C values.

Although no true in vivo criterion method currently exists for body composition analysis, 4C models, which divide the body into fat, water, bone mineral, and residual components, provide the best reference methods available (8). However, the application of multicompartment models is costly and time consuming. In view of the need to expediently characterize the body composition profile of obese adults and changes in %BF as a result of interventions, it is imperative that the robustness of using DXA with this population is tested. Correction equations may be developed if DXA is found to systematically overestimate %BF in obese populations. The aim of this study was therefore to determine the accuracy, precision, and bias of DXA-derived estimations of body composition relative to those obtained using a 4C body composition model in a cohort of obese men and women.

Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

Subjects

Eight men and six women sedentary (2+ years of self-reported minimal habitual physical activity) white (mean ± s.d.: 36.5 ± 12.9 years; 170.4 ± 7.0 cm; 98.20 ± 14.99 kg) aged 22–54 years volunteered for this study, which was approved by the University of South Australia's Human Research Ethics Committee. Written informed consent was obtained in accordance with the established protocol for human subjects.

Protocol

All subjects were tested during one ∼4-h morning session (beginning between 8 and 11 am to minimize diurnal variation) when they were postabsorptive, euhydrated and had not exercised for 36 h. They were requested to void before testing to eliminate any flatus in the gastrointestinal tract. Height was measured to within 1 mm implementing the stretch stature protocol of the International Society for the Advancement of Kinanthropometry (9) using a wall-mounted stadiometer (Model 222; Seca GMBH, Hamburg, Germany) and body mass determined to the nearest 20 g with a calibrated electronic scale (model FW-150K; A&D Mercury, Thebarton, Australia). This was followed by the collection of baseline saliva samples and ingestion of a deuterium dose. In addition to weight and height, further anthropometric measures (detailed below) were taken prior to the commencement of the body composition procedures.

Anthropometric measurements

Repeat measures of the waist and gluteal girth and anterior-posterior (A-P) chest depth were performed in accordance with the protocols of the International Society for the Advancement of Kinanthropometry (9) by an ISAK accredited Level 2 anthropometrist, using a metal tape measure (Lufkin W606PM; Cooper Industries, Houston, TX) and an anthropometer with curved branches (GPM, Siber-Hegner, Zürich, Switzerland). This anthropometrist had previously demonstrated acceptable (0.48–0.61%) technical error of measurement for these measurements.

Body composition

Body density (BD), total body water (TBW), and bone mineral mass (BMM) were measured using hydrodensitometry, isotopic (deuterium) dilution, and DXA, respectively. The procedures for these techniques have been fully described previously (10). Briefly, BD was measured by underwater weighing with the associated gas volume in the respiratory system determined by O2 dilution. This procedure involves subjects' immersed mass being recorded after they have expired to a comfortable level (approximately functional residual capacity) for four trials with the lung volume being measured immediately on surfacing. This technique correlates very highly (intraclass correlation coefficient = 0.999 and standard error of estimate = 0.4%BF) with the traditional method of exhaling to residual volume while also enhancing subject comfort during the procedure (11). TBW was determined from saliva samples by deuterium dilution (dose of 40 mg2 H2O/kg) using a Europa Scientific Geo 20–20 (Europa, Crewe, UK) isotope ratio mass spectrometer calibrated against Vienna Standard Mean Ocean Water and International Atomic Energy Agency enriched standards 302A and 302B. As recommended by Schoeller et al (12), a 4% correction factor for isotopic exchange with nonaqueous hydrogen was incorporated into calculations of the isotope dilution space. Bone mineral content was determined by DXA using a Lunar Prodigy scanner (General Electric, Madison, WI, using enCORE 2003 software version 7.52.002), which was calibrated daily against the phantom supplied by the manufacturer. Bone mineral content was multiplied by 1.0436 to obtain BMM (13,14). %BF was then calculated using the following formulae:

Four-compartment model

  • image

This equation was derived by assuming that the residual mass component of the body, after accounting for TBW, storage fat, and BMM has a density of 1.404 g/cm3 (10). Although other 4C equations have been generated using slightly different densities for the body components, they return very similar %BF values (10). Because the 4C models incorporate the determination of three parameters (TBW, BMM, BD) their %BF estimations are exposed to error propagation. That is the summation of the individual measurement errors for each parameter. However, we (15) have reported high reliability (intraclass correlation = 0.993, technical error of measurement = 0.6%BF) for the previously described model using test-retest reliability with a heterogeneous group (n = 11, %BF range 12.7–35.8).

Lean tissue and fat tissue masses along with %BF were also provided by the DXA output file.

Statistical analysis

A t-test was performed on the 4C/DXA %BF differences comparing men and women. The association between 4C and DXA %BF values was evaluated by linear regression and a Bland-Altman plot (16). A dependent t-test was applied to determine the significance (P ≤ 0.05) of the mean %BF difference between methods. Fourteen subjects provided a power of 0.95 for the aforementioned comparison. This value was based on a between-method difference of 3.7%BF that was reported by van der Ploeg et al. (6) for their lean subjects. The effects of fat-free mass (FFM) hydration and tissue thickness (body mass, BMI, A-P chest depth, waist circumference, gluteal girth) on %BF estimates were evaluated by linearly regressing each variable on the %BF differences between the two methods.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

The descriptive statistics for the sample are presented in Table 1. An initial t-test comparing men and women 4C-DXA %BF differences failed to demonstrate a significant difference by gender (P = 0.5258). Consequently, men and women data are presented and analyzed together. All subjects were overweight (BMI > 25 kg/m2), with %BF > 30%.

Table 1.  Descriptive statistics for sample (n = 14)
inline image

As shown in Figure 1, DXA-derived %BF and 4C-derived %BF were highly correlated (interclass correlation = 0.946, standard error of estimate = 2.1%BF). Mean values for %BF from the two body composition methods were not significantly different (DXA = 41.13%, 4C = 41.48%, P = 0.57). Among individual subjects, differences (4C − DXA) between DXA %BF and 4C %BF ranged from −3.04%BF to 4.01%BF. Prediction bias was not apparent because the slope and intercept of the regression line were not significantly different (P ≥ 0.05) from 1 and 0, respectively. Moreover, Bland-Altman analysis indicated no significant trend toward greater differences in %BF between the two methods with increasing 4C %BF (r2 = 0.18, P = 0.190).

image

Figure 1. Percent body fat (%BF) comparison between the two body composition methods (closed triangle, men; open triangle, women).

Download figure to PowerPoint

Figures 24 display the relationships between the differences in %BF estimates from the two methods and FFM hydration; A-P chest depth; body mass; BMI; waist girth; and gluteal girth. FFM hydration (Figure 2) was unrelated to %BF difference (r2 = 0.157, P = 0.16), although tissue thickness and most surrogates of tissue thickness were significantly and positively correlated with %BF difference: A-P chest depth, Figure 3, r2 = 0.431, P = 0.011; body mass, Figure 4a, r2 = 0.333, P = 0.031; BMI, Figure 4b, r2 = 0.452, P = 0.008; and waist girth, Figure 4c, r2 = 0.456, P = 0.008. Gluteal girth (r2 = 0.0.023, P = 0.606) was unrelated to %BF difference between the two methods (Figure 4d).

image

Figure 2. Relationship of percent body fat difference between methods compared to fat-free mass (FFM) hydration (closed triangle, men; open triangle, women).

Download figure to PowerPoint

image

Figure 4. Relationship of percent body fat difference between (a) methods to mass (kg, closed triangle, men; open triangle, women), (b) methods to BMI (kg/m2, closed triangle, men; open triangle, women), (c) methods to waist girth (cm, closed triangle, men; open triangle, women), (d) methods to gluteal girth (cm, closed triangle, men; open triangle, women).

Download figure to PowerPoint

image

Figure 3. Relationship of percent body fat difference between methods to anterior-posterior (A-P) chest depth (closed triangle, men; open triangle, women).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

The accuracy and precision of DXA-derived body composition values in a cohort of obese male and female subjects were determined in this study by using a 4C model as a criterion method. Recruitment of obese subjects for the collection of 4C data is generally difficult because of the requirement for underwater weighing. Reduced mobility among the obese along with body image concerns contributes to a reluctance to undergo underwater weighing. Consequently, our pilot data comprise an important addition to the very limited literature involving obese cohorts. Our subjects ranged from 31.3%BF to 52.6%BF with only three falling below 35%BF. However, we are cognizant of the limitations imposed by our sample size, which limits our ability to identify factors, which may contribute to between-method differences such as anthropometric parameters and FFM hydration status.

Although the correlation between DXA and 4C %BF values was high (r2 = 0.894, Figure 1) and the mean values (DXA = 41.13%BF; 4C = 41.48%BF) were not significantly different (P = 0.57) for our sample, considerable %BF differences between the methods existed. DXA values ranged between −3.04 and 4.01%BF from the respective 4C measures with a mean difference of 0.4%BF. Linear regression identified large 95% confidence intervals (±4.2%BF) for the prediction of 4C %BF from DXA measures. However, the slight departure of the trend line from the line of identity in Figure 1 indicates no systematic bias in our obese cohort. Bland-Altman analysis therefore indicated that the differences in %BF between the 4C and DXA values were not significantly correlated (r2 = 0.189, P = 0.190) with %BF over the range tested. In relation to the aims of this study, our results suggest questionable precision regarding the estimation of %BF from DXA (given the large 95% confidence intervals), but no evidence of systematic bias in this cohort.

The range of differences in %BF between the two methods reported in this study are smaller than those presented by van der Ploeg et al. (6) (−2.6–7.3%BF: mean 1.8%BF). However, the larger differences reported by these investigators appeared to be associated with greater bias associated with the lean subjects in their sample. These investigators (6) found that DXA progressively underestimated %BF compared with the 4C model as %BF decreased. In their subcohort of lean subjects (%BF ≤ 10: n = 32) the mean underestimation was 3.7%BF. It was unclear whether the DXA bias uncovered in their study projected to an overestimation of %BF in obese individuals because only two of their 152 subjects were in the vicinity of 35%BF. The data of Sopher et al. (7), however, did display a trend showing DXA overestimates %BF in obese children and adolescents ranging in age from 6 to 18 years. As mentioned above, the obese group measured in this study, which generally fell within a similar body mass range (55.8–115.21 kg) to that reported by van der Ploeg et al. (6), did not display a bias in the DXA %BF measurements. It could therefore be argued that DXA may well be used to provide reasonable %BF estimates for obese adult cohorts; however, we found that the body size of our heavier subjects was at the limit of the scan area for the DXA machine used.

Although the variation in FFM hydration has been proposed by some investigators (17,18,19) to be responsible for the errors in DXA determined body composition values, this was not apparent in the data from van der Ploeg et al. (6) or the current study. FFM hydration was not significantly correlated (R2 = 0.157: P = 0.16) with 4C/DXA %BF differences in our obese subjects. van der Ploeg et al. regarded the lack of association between variation in FFM hydration and %BF error as understandable given the theoretical calculations of Pietrobelli et al. (20). The latter determined that hydration changes of 1–5% would only lead to small DXA errors (<1%). Interestingly, the subjects in the current study displayed a higher mean FFM hydration (73.7%) than that reported (72.4%) for nonobese individuals (10). This trend would be expected because the water content of the fat-free adipose tissue at 88% is higher than the hydration of other fat-free components (21).

Early body composition work using DXA found that large errors in %BF estimates were associated with increasing tissue thickness. The thicker the tissue under analysis the greater the degree of beam hardening which involves the preferential attenuation of the lower energy X-rays. Goodsitt (22) reported that the largest errors would be associated with tissue thicknesses >20 cm when overestimations of tissue fat would occur. Work by Sopher et al. (7) support the aforementioned in that their DXA %BF values for children and adolescents underestimated and overestimated 4C measures in lean and obese subjects, respectively. Our data included one standard anthropometric measurement of tissue thickness (A-P chest depth), which was significantly correlated with 4C/DXA %BF differences (R2 = 0.406: P = 0.011). Interestingly these data have indicated that DXA generally underestimated body fat at higher tissue thicknesses; however, subject numbers were small. Three of seven subjects who had A-P chest depth exceeding 20 cm displayed large DXA underestimations of %BF (3.8, 4.0, 3.7%BF), although one subject recorded a 3.0%BF overestimation. The remaining three subjects displayed smaller deviations (−1.4 to −1.0%BF). The departure of our findings from those of Sopher et al. (7) could be related to the use of DPX/DPX-L systems by the latter and a Prodigy system in the current study. Although Prodigy devices are calibrated to allow results to be obtained over a wide tissue thickness range (23), errors remain apparent at larger tissue thicknesses.

Most of our surrogate measures for body thickness were also significantly correlated with 4C/DXA %BF differences (body mass: R2 = 0.333; P = 0.031, BMI: R2 = 0.452; P = 0.008, minimum waist circumference: R2 = 0.455; P = 0.008). Gluteal girth, however, did not display an association with %BF differences (R2 = 0.023; P = 0.606). These findings, except for the latter, support the observations of van der Ploeg et al. (6) in relation to their large heterogeneous cohort with an extensive range of body fat (6.5–37.5%BF). Our small sample along with a narrower gluteal girth range (101.45–123.65 cm) compared with the data of van der Ploeg et al. (6) (85.6–119.1 cm) may have mitigated against this parameter explaining differences between 4C and DXA %BF values.

Our pilot data on obese adult subjects are consistent with previous findings indicating that although the correlation between 4C and DXA %BF values is high, large individual variability in %BF estimates were evident. However, our work did not reveal any systematic DXA %BF bias in the obese range tested. Body fat differences between DXA and the criterion method could be associated with the inability of DXA to accurately accommodate the phenomenon of beam hardening at larger tissue thicknesses. Although the obese have a higher FFM hydration than the nonobese, variation in this parameter does not account for the inaccuracy of DXA %BF estimates. In summary, DXA may provide reasonable descriptive cross-sectional body composition data for obese cohorts but is less robust in the provision of accurate individual values.

Acknowledgment

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

Emeritus Professor Robert T. Withers (Bob) passed away on the 23 September 2007, after a long battle with illness. As an internationally renowned scientist he has left us the legacy of an enormous body of work in body composition and metabolic research. We are very grateful for his mentorship over the years, and specifically with regard to this study.

REFERENCES

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES
  • 1
    Bender R, Traunter C, Spraul N, Berger M. Assessment of excess mortality in obesity. Am J Epidemiol 1998; 147: 4248.
  • 2
    Seidell JC, Verschuren WMM, van Leer EM, Kronhout D. Overweight, underweight, and mortality. A prospective study of 48287 men and women. Arch Int Med 1996; 156: 958963.
  • 3
    Thornburn AW. National prevalence of obesity prevalence of obesity in Australia. Obes Rev 2005; 6: 187.
  • 4
    Ford ES, Giles WH, Dietz WH. Prevalence of the metabolic syndrome among US adults. JAMA 2002; 287: 356359.
  • 5
    Gallagher D, Heymsfield SB, Heo M et al. Healthy percentage body fat ranges: an approach for developing guidelines based on body mass index. Am J Clin Nutr 2000; 72: 692701.
  • 6
    van der Ploeg GE, Withers RT, LaForgia J. Body composition via DEXA: comparisons with a four compartment model. J App Physiol 2003; 94: 499506.
  • 7
    Sopher AB, Thornton JC, Wang J et al. Measurement of percentage body fat in 411 children and adolescents: A comparison of dual-energy X-ray absorptiometry with a four compartment model. Pediatrics 2004; 113: 12851290.
  • 8
    Wang Z, Shen W, Withers RT, Heymsfield, SB. Multicomponent molecular-level models of body composition analysis. In: Heymsfield SB, Lohman TG, Wang Z, Going SB (eds). Human Body Composition, 2nd edn. Human Kinetics: Champaign, IL, 2005, pp 161175.
  • 9
    Marfell-Jones M, Olds T, Stewart A, Carter L. International Standards for Anthropometric Assessment. Potchefstroom, South Africa: International Society for the Advancement of Kinanthropometry, 2006.
  • 10
    Withers RT, LaForgia J, Pillans RK et al. Comparisons of two-, three-, and four compartment models of body composition analysis in men and women. J Appl Physiol 1998; 85: 238245.
  • 11
    van der Ploeg GE, Gunn SM, Withers RT, Modra AC, Crockett AJ. Comparison of two hydrodensitometric methods for estimating percent body fat. J Appl Physiol 2000; 88: 11751180.
  • 12
    Schoeller DA, Ravussin E, Schutz Y et al. Energy expenditure by doubly labelled water: validation in humans and proposed calculation. Am J Physiol Regul Integr Comp Physiol 1986; 250: R823R830.
  • 13
    Heymsfield SB, Wang J, Kehayias J et al. Chemical determination of human body density in vivo: relevance to hydrodensitometry. Am J Clin Nutr 1989; 50: 12821289.
  • 14
    Heymsfield SB, Lichtman S, Baumgartner RN et al. Body composition of humans: comparison of two improved four-compartment models that differ in expense, technical complexity, and radiation exposure. Am J Clin Nutr 1990; 52: 5258.
  • 15
    LaForgia J, Gunn SM, Withers RT, Leaney F. Reliability of percent body estimates determined via multicompartment models. Int J Body Comp Res 2007; 5: 5156.
  • 16
    Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307310.
  • 17
    Deurenberg-Yap M, Schmidt G, van Staveren WA, Hautvast JG, Deurenberg P. Body fat measurements among Singaporean Chinese, Malays and Indians: a comparative study using a four compartment model and different two compartment models. Br J Nutr 2001; 85: 491498.
  • 18
    Friedl KE, Deluca JP, Marchitelli LJ, Vogel JA. Reliability of body fat estimations from a four-compartment model by using density, body water, and bone mineral measurements. Am J Clin Nutr 1992; 55: 764770.
  • 19
    Fuller NJ, Jebb SA, Laskey MA, Coward WA, Elia M. Four-component model for the assessment of body composition in humans: comparison with alternative methods, and evaluation of the density and hydration of fat-free mass. Clin Sci (Lond) 1992; 82: 687693.
  • 20
    Pietrobelli A, Wang Z, Formica C, Heymsfield, SM. Dual-energy X-ray absorptiometry: fat estimation errors due to variation in soft tissue hydration. Am J Physiol Endocrinal Metab 1998: 274: E808E816.
  • 21
    Withers RT, Laforgia J, Heymsfield SB, Wang Z-M, Pillans RK. Two, three and four-compartment chemical models of body composition analysis. In: Norton KI, Olds TS (eds). Anthropometrica. University of New South Wales Press: Sydney, Australia, 1996, pp 199231.
  • 22
    Goodsitt MM. Evaluation of a new set of calibration standards for the measurement of fat content via DPA and DXA. Med Phys 1992; 19: 3544.
  • 23
    Mazess RB, Hanson JA, Payne R, Nord R, Wilson M. Axial and total-body bone densitometry using narrow angle fan beam. Osteoporos Int 2000; 11: 158166.