1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References


To evaluate the new DXA VAT method on an Asian Chinese population by comparing to a reference method, computed tomography (CT).

Design and Methods

In total, 145 adult men and women volunteers, representing a wide range of ages (19-83 years) and BMI values (18.5-39.3 kg/m2) were studied with both DXA and CT.


The coefficient of determination (r2) for regression of CT on DXA values was 0.947 for females, 0.891 for males and 0.915 combined. The 95% confidence interval for r was 0.940-0.969 for the combined data. The Bland–Altman test showed a VAT bias (CT as standard method) of 143 cm3 for females and 379 cm3 for males. Combined, the bias was 262 cm3 with 95% limits of agreement of −232 to 755 cm3. While the current DXA method moderately overestimates the VAT volume for the study subjects, a further analysis suggested that the overestimation could be largely contributed to VAT movement due to breath-holding status.


For Asian Chinese, VAT measured with DXA is highly correlated to VAT measured with CT. Validation of the DXA VAT tool using a reference method (e.g., CT) needs to carefully control the breath-holding protocol.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

It is well established that abdominal adiposity as measured by waist circumference and waist-to-hip ratio is related to numerous metabolic disturbances, including glucose intolerance, hyperinsulinemia, diabetes mellitus (DM), hypertension, and hypertriglyceridemia, and is associated with coronary heart disease, stroke, and sudden death [1]. The association between abdominal adiposity and metabolic risk may be explained, in large part, by its strong association with visceral adiposity [5, 6]. Visceral adipose tissue (VAT) is a strong, independent predictor of metabolic syndrome (MetS) [7] and future abnormalities in glucose metabolism. The accumulation of VAT is associated with an increased risk of hypertension [10, 11] and cardiovascular diseases [12], and precedes the development of DM [15, 16].

The effects of VAT on cardiometabolic disease risk are of considerable interest in Asian populations where there is a demonstrable increase in metabolic diseases at BMI's historically considered to be in the normal range [17]. It is postulated that these apparently lean subjects have a tendency toward deposition of VAT rather than subcutaneous fat [22]. Currently, analysis of VAT is restricted to the research setting because the reference standards, computed tomography (CT) [25, 26], and magnetic resonance imaging (MRI) [27, 28] require a capital investment in equipment, manual image analysis, and a relatively high radiation dose in the case of CT. Consequently, there has been a desire to develop alternative technologies for rapid measurement of VAT.

A method to quantify VAT using DXA technology has recently been cleared by the United States Food and Drug Administration (FDA). To date, performance of the DXA VAT algorithm has been only demonstrated in one report [29], which showed strong agreement between DXA and volumetric CT in a population of American adults. Due to the predisposition of Asian populations to developing metabolic disease at relatively low BMI, validation of the DXA VAT tool is especially useful in this population. The purpose of this study is to demonstrate the performance of the Lunar VAT algorithm relative to CT in a population of Chinese adults.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References


An independent Ethics Committee located at Shanghai Zhongshan Hospital, Fudan University, approved this study. The study population included 145 adult men and women, ages 18 and up. All subjects were recruited in 2010 from the general population in Shanghai, China, with a wide range of ages (19-83 years) and BMI values (18.5-39.3 kg/m2). Informed written consent was obtained from all the subjects. Study recruiting criteria were used to ensure that subjects were divided roughly equally among men and women, and were distributed among BMI (normal, overweight, and obese) and age categories (18-30, 31-50, 51-60, and >60 years old). Since the study was primarily designed for technical validation, the BMI is slightly higher than the Chinese population average [30, 31] . The exclusion criteria for this study were restricted to pregnancy and eating disorder. For each subject, height, weight, and hip and waist circumferences were measured. DXA and CT measurement for each subject was completed in the fasting state during a single study visit.

Sample size estimation

Sample size was calculated to reliably determine if the correlation coefficient between DXA and CT exceeded a lower limit of r=0.90. Published data from the Framingham VAT subsample were used to estimate the mean and standard error of the estimate (SEE). Power and Precision (Release 2.1, BioStat, 14 N. Dean St., Englewood, NJ) was used to calculate the power of a one-tailed test of the correlation against the value 0.90, α=0.05, applying Fisher's Z-transformation. Using a power of 90% as a basis for calculating sample size, it was determined that approximately 120 subjects (60 men, 60 women) would be sufficient to evaluate performance of the VAT algorithm.

DXA measurement

Total body scans were acquired on a GE Healthcare Lunar iDXA system (software version: enCORE version 13.10) with scan mode automatically determined by the device. For the DXA measurement, all subjects were wearing a hospital gown and had all metal artifacts removed. The iDXA unit was calibrated daily using the GE Healthcare Lunar calibration phantom. A trained operator performed all scans following the operator's manual for patient positioning and data acquisition.

Scans were analyzed with the enCORE software (version 13.60). The Android region of interest (ROI) was automatically determined, with the caudal limit at the top of the iliac crest and the cephalad limit at 20% of the distance from the top of the iliac crest to the base of the skull. VAT mass and volume in the android region were computed using a semi-empirical algorithm that uses measured subcutaneous thickness and total abdominal thickness to estimate the subcutaneous adipose tissue (SAT) geometry. VAT is the difference between total android fat and estimated SAT. Details of the DXA VAT algorithm have been described previously [29].

Computed tomography measurements

CT scans were acquired using a Siemens Healthcare SOMATOM Sensation 16-slice CT scanner (software version: Syngo CT Workplace VB30B). The site performed a standard abdominal scan without contrast, using 120 kVp, and 5 mm slice thickness. During the CT scan, subjects were asked to maintain a supine position with arms above their head. Contiguous cross-sectional abdominal images were captured over 150 mm of the abdomen, beginning at the top of S1 and moving toward the head.

CT images were segmented using a semi-automated method to remove the subcutaneous adipose tissue, and the remaining tissue volumes were reconstructed from DICOM images using the GE Healthcare Advantage Workstation (software version 4.1). To account for differences in patient positioning and variation in attenuation across the subjects, a subject-specific threshold in Hounsfield units was developed based on manual selection of a region of high confidence VAT in a single slice. The threshold was set for each patient at 2 SD from the mean VAT Hounsfield value in the region of interest. This threshold was applied to all slices and the volume of VAT was reported.

CT data were analyzed by a single operator, and ∼20% of the data were randomly selected to be reanalyzed by a second operator. The average difference in CT VAT volume between the two operators was on the order of 3%. This is in agreement with previous reports in the literature showing differences between operators using a similar method [32, 33] .

Statistical analysis

CT data were transferred to GE Global Research Center (Niskayuna, NY), and DXA data were transferred to GE Healthcare Lunar (Madison, WI). CT and DXA images were not shared between the two centers in order to minimize any potential bias in the performance validation process.

The CT VAT volumes were regressed on the resulting iDXA estimated values. 95% confidence limits for the correlation coefficients were calculated using the Fisher Z transformation. Using Minitab, paired t-tests of the mean difference between the CT and iDXA VAT volumes were calculated with and without the subjects identified as outliers in the regression analysis. Using Analyse-it for Microsoft Excel version 2.25 (Analyse-it Software), Bland–Altman analysis comparing the CT and iDXA VAT was performed. The CT was designated as the reference method. Subjects identified as outliers in the regression analysis were included in the Bland–Altman analysis.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

Descriptive statistics of subjects grouped by gender are listed in Table 1. Participants included 73 male subjects and 72 female subjects. Males and females were from a wide range of age (19-84 years) and BMI (18.5-39.3 kg/m2).

Table 1. Descriptive statistics of subjects
 Female (n = 72)Male (n = 73)
Age (years)50.613.420.582.652.016.319.283.7
Height (cm)159.16.1146.2169.9170.15.8152.5185.9
Weight (kg)68.813.342.998.676.510.853.4101.4
BMI (kg/m2)27.14.718.539.326.
Waist (cm)93.711.273.4119.089.88.572.3112.0
DXA android fat (cm3)27431195545595925549627804995
DXA subcutaneous fat (cm3)174364544543069944342312722
DXA visceral fat (cm3)100071534352915607262413214
CT visceral fat (cm3)857623109326111815871542698

The agreement between DXA estimated and CT measured VAT volume is shown in Figure 1. The coefficient of determination (r2) for regression of CT on DXA values was 0.947 for females, 0.891 for males and 0.915 combined. The 95% confidence interval for r was 0.940-0.969 for the combined data.


Figure 1. Correlations between DXA estimated and CT measured VAT volume: (a) combined males and females; (b) females; and (c) males.

Download figure to PowerPoint

Bland-Altman analysis, a method commonly used to show the agreement between two methods of measuring the same variable, was conducted to characterize differences between DXA and CT VAT measurements over the range of VAT volumes included in the analysis. Bland-Altman bias (CT as standard method) was 143 cm3 for females and 379 cm3 for males. The 95% limits of agreement were −210 to 495 cm3 for females and −128 to 886 cm3 for males. Combined, the bias was 262 cm3 with 95% limits of agreement of −232 to 755 cm3 (Figure 2). Regression of the differences on the average values for the combined data yielded a statistically significant slope of 0.215 (P < 0.0001) and insignificant intercept of 14.4 cm3 (P = 0.66).


Figure 2. Bland–Altman analysis between DXA and CT VAT measurements.

Download figure to PowerPoint

We also considered relationships between the DXA VAT measurement and two common proxies for adiposity: BMI and waist circumference (Figure 3). While both indices are seen to be significantly correlated with DXA VAT, the correlation is modest and the scatter is large.


Figure 3. Correlation plots for dual-energy x-ray absorptiometry (DXA) visceral adipose tissue (VAT) and BMI (a) and waist circumference (b).

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

In this study, we have shown that VAT measured by Lunar iDXA is highly correlated (r2=0.91) with the VAT measured by CT. The current DXA method moderately overestimates the VAT volume, and while a systematic trend between Lunar iDXA and CT VAT volume was observed in this trial, we believe this is an artifact of data acquisition, and that the DXA software provides reliable estimation of VAT in an Asian Chinese population.

To attempt to find a hypothesis to explain the systematic offset between a Lunar iDXA and CT VAT volume, we considered a similar study conducted at Oregon Health and Science University (OHSU) [29], where no systematic differences between CT and DXA were observed. A detailed analysis of the data acquisition at the two sites revealed a difference in the breathing instructions given to the subjects for the CT acquisition. Traditionally, abdominal CT images are acquired during a breath hold to ensure that the internal organs and the fat tissue are in a fixed position for image acquisition. Subjects in the OHSU study were given the instruction to take a normal breath and hold. In our study, subjects were coached to take a shallow breath for CT acquisition. While these instructions are inherently subjective, a difference in respiration prior to imaging could lead to a systematic offset in data analysis.

The depth of respiration is important to the performance of the DXA VAT algorithm because the first step in the validation is identifying a common region of interest between the CT and DXA images. DXA images are acquired with arms at the side of the body and in a free breathing condition. In contrast, CT images are acquired with arms above the head during a breath hold condition. Procedurally, our analysis plan called for the identification of the iliac crest on the CT images. Then, a region of interest extending the height of the DXA android region was identified, and VAT was quantified based on the CT slices in this region. This approach assumes that there is either no, or at least a relatively consistent displacement of the internal organs between CT and DXA exams. If internal organs are shifted between the two modalities, you would expect that the relative amounts of fat and lean tissue would be different in the identified regions of interest.

To determine whether the hypothesis that breathing instructions led to differences in the relative positions of soft tissues between DXA and CT exams at OHSU and in this study, we performed a discriminant analysis to show that the position of the diaphragm relative to the iliac crest was systematically different between OHSU and our study. Samples of 20% of the subjects from each study site were included in the analysis. Subjects were selected using principal components analysis to represent the broader study populations at each site. The linear distance between the iliac crest and the diaphragmatic dome from both left and right sides was measured to characterize diaphragm position in CT (Figure 4). The distance was then compared to the same distance measured by iDXA to calculate the difference in distance observed between iDXA and CT. Differences in distance from our study (right 0.50 ± 1.38 cm, left 0.29 ± 1.38 cm) are much smaller than those observed in OHSU study (right 3.96 ± 1.34cm, left 3.70 ± 1.37cm). This is consistent with a deeper breath in the OHSU subjects. A linear discriminant analysis showed that, using this distance alone, 90% of subjects could be correctly assigned to their respective acquisition sites. This implies that there may be a systematic difference in the tissue that is included when data from each site is mapped to CT data.


Figure 4. Representative CT scout image. Yellow markers highlight the top of the iliac crest and the diaphragmatic arch. [Color figure can be viewed in the online issue, which is available at]

Download figure to PowerPoint

There are several other possible explanations for the observed differences between the DXA VAT algorithm and the CT measured volumes. Technical issues, such as the calibration of the DXA and CT instruments, could be ruled out by examining the calibration records for the instruments, and by measuring a multi-modality imaging phantom, which showed agreement between the CT and DXA volumes of the phantom. Another possible explanation is that Asian and Caucasian subjects have different fat densities. However, to our knowledge, there is no body of literature supporting this hypothesis. Systematic geometric differences in the shape of the subcutaneous fat layer could also explain these differences, since the DXA model makes assumptions about the geometry of the abdomen. Visual and anthropometric assessments of the DXA and CT images did not identify a geometric parameter that would explain the observed difference between the two populations.

This study points out some specific risks and concerns involved with validating the DXA VAT tool. Appropriately aligning CT images and planar DXA images collected under different measurement conditions is inherently challenging. For the development of the VAT tool, a standardized approach was used based on a normal breath and alignment of the images based on the iliac crest and the length of the DXA android region. It is clear that validation results will be different under different CT acquisition and analysis conditions. This should be expected. The VAT algorithm may be best tested in phantoms where soft tissue movement is not a consideration. Overall, these results indicate that DXA VAT measurements can be validated against CT when proper considerations are put in place to map the CT acquisition procedure to those used for the development of the tool.

There are limitations in this study. First, while we have demonstrated that there is a systematic difference in the distance from the iliac crest to diaphragmatic dome between our population and a previous trial, we do not have controlled breathing trials to provide definitive proof of our hypothesis. Future work in this area may be required. Also, all subjects were recruited from one geographic region in China and from a single site. While the fundamental study recruitment targets for age and BMI were met, studies from additional regions and sites could help confirm the generalization of the VAT algorithm. The study also did not include subjects that were morbidly obese (BMI > 40 kg/m2), or subjects under the age of 18 years old. It may be useful in the future to expand DXA VAT measurement to include these groups. Finally, the goal of this study was limited to technical utility and did not address clinical use cases. Studies focused on clinical disease association and outcomes need to be done to confirm the clinical utility of the DXA VAT algorithm.

Deploying a tool for VAT measurement in Asian countries may be of substantial benefit over traditional risk stratification approaches, which have largely been developed on Caucasian populations and tend to underestimate risk in Asians who are lean or overweight using western BMI standards. This study showed that Lunar DXA VAT tools could be used in the Chinese population with results comparable to CT measured volumetric VAT in the abdominal region. The difference introduced by different breath holding pattern in a CT acquisition protocol should not impact the DXA VAT measurement. A normal breath rhythm is recommended during DXA measurement. Compared to area VAT measurement commonly used in risk stratification studies, volumetric VAT measurement can give more consistent results. The DXA VAT tool, validated against volumetric CT measurement, could potentially replace CT as the standard for VAT quantification. VAT measurement, in a way that is rapid, relatively inexpensive and involves minimum radiation dosage, could facilitate large-scale population studies establishing linkage between VAT and metabolic syndrome, hypertension, diabetes, and cardiovascular diseases for the particular population. Additional studies, showing the relationship between DXA VAT and other clinical risk factors, are needed to substantiate the value of the DXA VAT tool in the Chinese population.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References
  • 1
    Després JP, Moorjani S, Ferland M, et al. Adipose tissue distribution and plasma lipoprotein levels in obese women. Importance of intra-abdominal fat. Arteriosclerosis 1989;9:203-210.
  • 2
    Janssen I, Katzmarzyk PT, Ross R. Waist circumference and not body mass index explains obesity-related health risk. Am J Clin Nutr 2004;79:379-384.
  • 3
    Kuk JL, Katzmarzyk PT, Nichaman MZ, Church TS, Blair SN, Ross R. Visceral fat is an independent predictor of all-cause mortality in men. Obesity (Silver Spring) 2006;14:336-341.
  • 4
    Bigaard J, Tj⊘nneland A, Thomsen BL, Overvad K, Heitmann BL, S⊘rensen TI. Waist circumference, BMI, smoking, and mortality in middle-aged men and women. Obes Res 2003;11:895-903.
  • 5
    Ross R, Léger L, Morris D, de Guise J, Guardo R. Quantification of adipose tissue by MRI: relationship with anthropometric variables. J Appl Physiol 1992;72:787-795.
  • 6
    Després JP, Lemieux I. Abdominal obesity and metabolic syndrome. Nature 2006;444:881-887.
  • 7
    Ribeiro-Filho FF, Faria AN, Kohlmann NE, Zanella MT, Ferreira SR. Two-hour insulin determination improves the ability of abdominal fat measurement to identify risk for the metabolic syndrome. Diabetes Care 2003;26:1725-1730.
  • 8
    Hayashi T, Boyko EJ, McNeely MJ, Leonetti DL, Kahn SE, Fujimoto WY. Minimum waist and visceral fat values for identifying Japanese Americans at risk for the metabolic syndrome. Diabetes Care 2007;30:120-127.
  • 9
    Pascot A, Lemieux S, Lemieux I, et al. Age-related increase in visceral adipose tissue and body fat and the metabolic risk profile of premenopausal women. Diabetes Care 1999;22:1471-1478.
  • 10
    Fox CS, Massaro JM, Hoffmann U, et al. Abdominal visceral and subcutaneous adipose tissue compartments: association with metabolic risk factors in the Framingham Heart Study. Circulation. 2007 3; 116:39-48.
  • 11
    Ding J, Visser M, Kritchevsky SB, et al. The association of regional fat depots with hypertension in older persons of white and African American ethnicity. Am J Hypertens 2004;17:971-976.
  • 12
    Després JP, Moorjani S, Lupien PJ, Tremblay A, Nadeau A, Bouchard C. Regional distribution of body fat, plasma lipoproteins, and cardiovascular disease. Arteriosclerosis 1990;10:497-511.
  • 13
    Després JP. Inflammation and cardiovascular disease: Is abdominal obesity the missing link? Int J Obes Relat Metab Disord 2003;27 Suppl 3:S22-S24.
  • 14
    Okauchi Y, Kishida K, Funahashi T, et al. 4-year follow-up of cardiovascular events and changes in visceral fat accumulation after health promotion program in the Amagasaki Visceral Fat Study. Atherosclerosis 2010: 212:698-700.
  • 15
    Boyko EJ, Fujimoto WY, Leonetti DL, Newell-Morris L. Visceral adiposity and risk of type 2 diabetes: a prospective study among Japanese Americans. Diabetes Care 2000;23:465-471.
  • 16
    Goodpaster BH, Krishnaswami S, Resnick H, et al. Association between regional adipose tissue distribution and both type 2 diabetes and impaired glucose tolerance in elderly men and women. Diabetes Care 2003;26:372-379.
  • 17
    Yeh WT, Chang HY, Yeh CJ, Tsai KS, Chen HJ, Pan WH. Do centrally obese Chinese with normal BMI have increased risk of metabolic disorders? Int J Obes (Lond) 2005;29:818-825.
  • 18
    Weng X, Liu Y, Ma J, Wang W, Yang G, Caballero B. Use of body mass index to identify obesity-related metabolic disorders in the Chinese population. Eur J Clin Nutr 2006;60:931-937.
  • 19
    Cheng TO. Lower body mass index cutoff is required for Chinese as a risk factor for coronary artery disease and other obesity-related metabolic disorders. Am J Clin Nutr 2004;80:782; author reply 782-783.
  • 20
    Liu Y, Tong G, Tong W, Lu L, Qin X. Can body mass index, waist circumference, waist-hip ratio and waist-height ratio predict the presence of multiple metabolic risk factors in Chinese subjects? BMC Public Health 2011;11:35.
  • 21
    Palaniappan LP, Wong EC, Shin JJ, Fortmann SP, Lauderdale DS. Asian Americans have greater prevalence of metabolic syndrome despite lower body mass index. Int J Obes (Lond) 2011;35:393-400.
  • 22
    Lim U, Ernst T, Buchthal SD, et al. Asian women have greater abdominal and visceral adiposity than Caucasian women with similar body mass index. Nutrition and Diabetes, 2011. 1 (e6; doi:10.1038/nutd.2011.2)
  • 23
    Lear SA, Humphries KH, Kohli S, Birmingham CL. The use of BMI and waist circumference as surrogates of body fat differs by ethnicity. Obesity (Silver Spring) 2007;15:2817-2824.
  • 24
    Lear SA, Humphries KH, Kohli S, Chockalingam A, Frohlich JJ, Birmingham CL. Visceral adipose tissue accumulation differs according to ethnic background: results of the Multicultural Community Health Assessment Trial (M-CHAT). Am J Clin Nutr 2007;86:353-359.
  • 25
    Borkan GA, Gerzof SG, Robbins AH, Hults DE, Silbert CK, Silbert JE. Assessment of abdominal fat content by computed tomography. Am J Clin Nutr 1982;36:172-177.
  • 26
    Kvist H, Chowdhury B, Grangård U, Tylén U, Sjöström L. Total and visceral adipose-tissue volumes derived from measurements with computed tomography in adult men and women: predictive equations. Am J Clin Nutr 1988;48:1351-1361.
  • 27
    Kullberg J, Ahlström H, Johansson L, Frimmel H. Automated and reproducible segmentation of visceral and subcutaneous adipose tissue from abdominal MRI. Int J Obes (Lond) 2007;31:1806-1817.
  • 28
    Liou TH, Chan WP, Pan LC, Lin PW, Chou P, Chen CH. Fully automated large-scale assessment of visceral and subcutaneous abdominal adipose tissue by magnetic resonance imaging. Int J Obes (Lond) 2006;30:844-852.
  • 29
    Kaul S, Rothney MP, Peters DM, et al. Dual-energy X-ray absorptiometry for quantification of visceral fat. Obesity (Silver Spring) 2012;20:1313-1318.
  • 30
    WHO Expert Consultation. Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet 2004;363:157-163.
  • 31
    Li G, Chen X, Jang Y, et al. Obesity, coronary heart disease risk factors and diabetes in Chinese: an approach to the criteria of obesity in the Chinese population. Obes Rev 2002;3:167-172.
  • 32
    Yoshizumi T, Nakamura T, Yamane M, et al. Abdominal fat: standardized technique for measurement at CT. Radiology 1999;211:283-286.
  • 33
    Potretzke AM, Schmitz KH, Jensen MD. Preventing overestimation of pixels in computed tomography assessment of visceral fat. Obes Res 2004;12:1698-1701.