Abdominal visceral fat measurement using dual-energy X-ray: Association with cardiometabolic risk factors


  • Funding agencies: The study was supported in part by a grant from the Italian Society for the Study of Atherosclerosis Lombardy Chapter (SISA). GE Healthcare funded a portion of this project related to data acquisition and data analysis.

  • Disclosure: MPR, JX, WKW, YX, and DLE are employees of GE Healthcare or its affiliates. No other potential conflicts of interest related to this article were reported.

Correspondence: David L. Ergun(ergundavid@gmail.com)



To examine the association between cardiometabolic risk factors and visceral adipose tissue (VAT) measurements using a dual-energy X-ray absorptiometry (DXA) based approach.

Design and Methods

An analysis of cross-sectional relationships between DXA VAT measured using CoreScan (GE Healthcare) and cardiometabolic indicators was conducted on a sample of 939 subjects (541 females and 398 males; average age, 56 years; average BMI, 26 kg/m2) who had previously undergone a total body DXA scan as well as measurements of key cardiometabolic risk factors.


Sex-specific, age-adjusted multivariable regression analysis showed that for both men and women, DXA VAT was significantly associated with increased odds of hypertension, impaired fasting glucose, metabolic syndrome, and type 2 diabetes (P < 0.001). After additional model adjustment for BMI and waist circumference, the odds ratio (per SD change in VAT) for type 2 diabetes was 2.07 for women and 2.25 for men. Similarly, the odds ratio for metabolic syndrome for women was 3.46 and for men was 1.75.


VAT measured using DXA showed a significant association with cardiometabolic risk factors and disease. These relationships persist after statistical adjustment for age, BMI, and waist circumference. DXA VAT may provide a new accessible option for quantifying VAT-related cardiometabolic risk.


With the increase in obesity prevalence in the developed world, there has been a concurrent increase in cardiometabolic conditions such as hypertension, dyslipidemia, metabolic syndrome, and type 2 diabetes. However, large epidemiological studies have shown that not all individuals who are obese using BMI criteria have adverse metabolic profiles [1]. There is also a subset of individuals with normal or modestly overweight BMI who are in poor cardiometabolic health [1, 2, 5, 6]. These findings suggest that BMI alone may be inadequate for identification of patients likely to have cardiometabolic disease [1].

Visceral adipose tissue (VAT) has long been considered to be a metabolically active pathogenic fat depot [9]. Because of both its cytokine profile and physical proximity to the liver, VAT has been implicated in insulin resistance as well as number of related cardiovascular and metabolic conditions, including type 2 diabetes [2, 4, 13]. Studies have shown both cross sectional and longitudinal associations with between VAT accumulation and hypertension [18, 19], and impaired glucose tolerance [17, 20, 21]. Studies have also indicated that a measurement of VAT may discriminate between subjects with impaired and normal glucose tolerance [17, 20, 21]. Strong associations have been observed between high levels of VAT and metabolic syndrome and type 2 diabetes [4, 15, 17, 20]. These studies span populations that are diverse with respect to gender, age, and ethnic background. In many studies, the association between VAT and disease remains significant after statistical adjustments for other measures of obesity and regional adiposity, such as BMI and waist circumference, suggesting that VAT may have an independent effect in determining overall cardiometabolic disease risk [3, 18].

Historically, VAT measurements have been made using computer tomography (CT) and magnetic resonance imaging (MRI). Many groups have developed their own proprietary data acquisition and processing techniques to generate visceral fat area and volume measures [22, 23]. Recently, a new algorithm to measure VAT using dual-energy X-ray absorptiometry (DXA), CoreScan (GE Healthcare, Madison, WI), has been introduced [24]. The technical performance of DXA VAT has been demonstrated using volumetric CT as the reference standard. This study showed a high correlation (r2 = 0.95) and small average difference (56 cm3) between DXA and CT [24].

While the technical performance of DXA VAT has been previously demonstrated against CT, the technical validation study did not include measurements of relevant clinical parameters. The purpose of this study was to demonstrate the association between DXA VAT measured using CoreScan and cardiometabolic disease indicators including hypertension, impaired fasting glucose, metabolic syndrome, and type 2 diabetes. To our knowledge, this will be the first demonstration of the association between VAT measured using an automated DXA-based approach and cardiometabolic disease indicators.

Methods and Procedures

Patient population and protocol

Participants included in this analysis were drawn from the 8-year follow-up time point of the Progression of Carotid Intimal Lesions (PLIC) study, a prospective, population-based study conducted at The Atherosclerosis Center in Bassini Hospital, Department of Pharmacological Sciences (University of Milan, Italy). The overarching goal of the study is to examine how carotid intimal media thickness is related to other biomarkers of cardiovascular disease in a large population. Participants, who were referred to the study by their general practitioners, were free of coronary heart disease, hypertension, type 2 diabetes, liver and kidney disease, and thyroid dysfunction at their baseline visit. Subjects were asked to complete periodic follow up visits, and some subjects developed disease after the initial visit. Details of the cohort and study design have been previously described [25, 26]. The study was approved by the ethical committee of the center. All subjects provided written informed consent prior to enrolling in the study.

Because the purpose of this study was to determine the relationship between DXA measurements of VAT and signs of cardiometabolic disease, a subsample of the total PLIC cohort (n = 939) who consented to total body DXA measurement of body composition was used for analysis. During a single study visit, each participant underwent a complete medical history questionnaire, measures of height, weight, waist circumference, blood pressure, and provided a blood sample, which was used to derive lipid and glucose measurements. Subjects also completed a total body DXA scan.

DXA VAT assessment

Total body DXA scans were acquired using the Lunar iDXA (GE Healthcare, Madison, WI) instrument. Daily quality control scans were acquired during the study period. All subjects were scanned wearing a hospital gowns with all metal artifacts removed from their body. Trained operators performed all scans. Patient positioning and data acquisition was conducted in accordance with the operator's manual. Lunar iDXA scans were analyzed with the enCORE software (version 14.0) (GE Healthcare, Madison, WI).

DXA VAT was computed automatically over the DXA android region, a region-of-interest automatically defined by the enCORE software, whose caudal limit is automatically placed at the top of the iliac crest and whose height is set to 20% of the distance from the top of the iliac crest to the base of the skull to define its cephalad limit.

Cardiometabolic disease determination

All biochemical measures were derived from a single fasting blood sample. Type 2 diabetes was defined based on fasting plasma glucose >125 mg/dl. Impaired fasting glucose was defined as fasting plasma glucose between 100 and 125 mg/dl. Hypertension was defined as either having a systolic blood pressure ≥140 mmHg or a diastolic blood pressure ≥90 mmHg. Metabolic syndrome was defined using the modified Adult Treatment Panel (ATP) criteria. Blood pressure, lipid-lowering, and diabetes medications were used to establish the clinical classification of each subject.

Statistical analysis

Subject demographics are reported as mean +/− standard deviation (minimum, maximum). The Pearson partial correlation controlling for the effect of age between DXA VAT and all continuous variables was computed. Multivariable linear regression was performed comparing DXA VAT with both continuous measures and dichotomous variables (e.g., presence or absence of disease). Results are reported as the effect size (continuous variables) or odds ratio (dichotomous variables) per SD of DXA VAT. Regression results adjusted for age are reported with and without adjustments for BMI and waist circumference. All statistical analysis was performed in R Version 2.13 (www.R-project.org).


Table 1 shows the descriptive statistics of the subjects separated by gender. A total of 541 female subjects and 398 male subjects were included in this analysis. The average age of subjects was 56 years, and the average BMI was 26 kg/m2, with men having a slightly higher average BMI than women. While the prevalence of metabolic syndrome was similar between the genders (22% for women and 23% for men), men showed a substantially higher prevalence of impaired fasting glucose (46% relative to 27%), type 2 diabetes (8% relative to 4%), and hypertension (43% relative to 33%) relative to women. Men also had a higher DXA VAT volume than women (1614 cm3 relative to 839 cm3), although the DXA VAT value is not currently normalized for body size.

Table 1. Descriptive statistics of the study sample by gender. Results are reported as mean ± SD (range) where applicable
 Women (n = 541)Men (n = 398)
  1. HDL, high density lipoprotein; VAT, visceral adipose tissue.
  2. aMedian (25th and 75th percentiles).
  3. bFasting plasma glucose of 100–125 mg/dL, based on those without type 2 diabetes.
  4. cBelow DXA detection limit.
Age (years)56 ± 12 (19–81)55 ± 13 (21–79)
BMI (kg/m2)25 ± 4 (19–40)27 ± 3 (19–40)
Waist circumference (cm)86 ± 11(64–125)97 ± 9 (73–139)
Triglyceridesa (mg/dL)85 (62–115)101 (72–143)
HDL cholesterol (mg/dL)61 ± 15 (27–115)51 ± 12 (28–111)
Total cholesterol (mg/dL)230 ± 47 (24–510)218 ± 42 (27–376)
Systolic blood pressure (mmHg)124 ± 18 (75–185)130 ± 16 (90–190)
Diastolic blood pressure (mmHg)75 ± 10 (50–130)80 ± 10 (50–200)
Hypertension (%)3343
Fasting plasma glucoseb (mg/dL)96 ± 15 (64–215)102 ± 19 (48–241)
Impaired fasting glucose (%)2746
Type 2 diabetes (%)48
Metabolic syndrome (%)2223
VAT (cm3)839 ± 622 (0c–3665)1614 ± 857 (13–4963)

The gender-specific correlations between DXA VAT and cardiometabolic risk factors are shown in Table 2. DXA VAT was significantly correlated with all measured cardiometabolic risk factors, except total cholesterol. The strongest associations were with BMI (r = 0.76 for women, r = 0.77 for men; P < 0.001) and waist circumference (r = 0.83 for women, r = 0.81 for men; P < 0.001). The correlations between DXA VAT and cholesterol components (log triglycerides and high density lipoprotein) were similar in magnitude to the correlation with age and were higher than the correlations between DXA VAT and fasting plasma glucose or systolic and diastolic blood pressure (r < 0.25).

Table 2. Correlations controlling for the effect of age (except for age) between DXA VAT and cardiometabolic risk factors
 Women (n = 541)Men (n = 398)
  1. HDL, high density lipoprotein.
  2. aP < 0.001.
Waist circumference0.83a0.81a
Log triglycerides0.41a0.39a
HDL cholesterol−0.32a−0.28a
Total cholesterol0.030.07
Systolic blood pressure0.14a0.22a
Diastolic blood pressure0.22a0.21a
Fasting plasma glucose0.22a0.25a

Results of multivariable linear regression for DXA VAT on cardiometabolic risk factors and disease states are shown in Table 3. Results are presented as both age-adjusted only, as well as adjusted for age, BMI, and waist circumference. In the analysis adjusted for age only, significant effect sizes were observed for all variables in men and women (P < 0.05). In general, the effect size and odds ratio were higher for women than for men. The largest odds ratio observed for both genders was for metabolic syndrome with an odds ratio of 10.73 in women and 4.07 in men.

Table 3. Sex-specific multivariable linear regressions for VAT on cardiometabolic risk factors and disease states. Results reported as effect size (average change in risk factor ± standard error) per 1 SD of adipose tissue for continuous variables, and as odds ratio for the condition per 1 SD of adipose tissue with 95% confidence intervals for dichotomous variables
 FemaleMaleP for sex interaction
Age-adjusted effect size or odds ratioPAge, BMI, WC-adjusted effect size, or odds ratioPAge-adjusted effect size or odds ratioPAge, BMI, WC-adjusted effect size or odds ratioP
  1. DPB, diastolic blood pressure; FPG, fasting blood glucose; HDL, high density lipoprotein cholesterol; HTN, hypertension; IFG, impaired fasting glucose; MetS, metabolic syndrome; SBP, systolic blood pressure; TG, triglycerides; T2D, type 2 diabetes; WC, waist circumference.
SBP3.45 ± 1.030.00081.06 ± 1.870.57153.42 ± 0.76<0.00014.57 ± 1.370.00090.0681
DBP2.96 ± 0.58<0.00011.00 ± 1.050.34081.68 ± 0.4<0.00012.11 ± 0.720.00340.0005
FPG3.45 ± 0.68<0.00012.19 ± 1.210.07082.74 ± 0.55<0.00013.60 ± 0.960.00020.4109
Log TG0.27 ± 0.03<0.00010.27 ± 0.05<0.00010.18 ± 0.02<0.00010.15 ± 0.040.0001<0.0001
HDL−7.33 ± 0.93<0.0001−4.85 ± 1.700.0046−3.21 ± 0.56<0.0001−0.78 ± 1.010.4383<0.0001
HTN1.69 (1.25–2.29)<0.00071.09 (0.62–1.91)0.16791.59 (1.27–1.99)<0.00011.69 (1.13–2.53)0.13320.2781
IFG2.19 (1.61–2.98)<0.00011.31 (0.76–2.27)0.33291.55 (1.25–1.93)<0.00011.67 (1.13–2.46)0.00940.0533
T2D3.49 (2–6.11)<0.00012.07 (0.73–5.87)0.16942.62 (1.84–3.75)<0.00012.25 (1.21–4.19)0.01080.2950
MetS10.73 (6.56–17.57)<0.00013.46 (1.70–7.06)0.00064.07 (2.92–5.67)<0.00011.75 (1.08–2.82)0.02270.0003

Additional analysis was conducted to adjust the DXA VAT regression analysis for effects of BMI and waist circumference, two common proxies for excess adiposity. After this adjustment, DXA VAT was still associated with an increased odds of type 2 diabetes in men (odds ratio: 2.25) and metabolic syndrome in men and women (odds ratio: 1.75 men; odds ratio: 3.46 women). In addition, effect sizes also remained significantly larger (P < 0.05) for high density lipoprotein (−4.85 mg/dL), triglycerides (LogTG = 0.27), and risk for metabolic syndrome (OR = 3.46) in women compared to men, and for diastolic blood pressure (2.11 mmHg) in men compared to women.


In this cross-sectional sample of Italian adult subjects, there was a strong association between DXA-derived measures of VAT using CoreScan and cardiometabolic risk factors. Despite strong correlations of BMI and waist circumference with DXA VAT, most odds ratios for VAT regressions adjusted for those variables remained significant. This may indicate that DXA VAT has an independent association with disease. This study, to our knowledge, is the first demonstration of the association of DXA VAT with clinical parameters.

Multiple investigators have reported cross-sectional associations between VAT and cardiometabolic disease indicators. While ORs between studies are not directly comparable, it is important to see similar trends of significance across multiple cohorts. The Japanese-American Community Diabetes Study demonstrated an odds ratio of 2.33 for hypertension [27]. While odds ratio in the current study achieved statistical significance, they are no higher than 1.69 for men or women. Similar to our study, Health ABC, a large study that measured VAT in older adults, reported significant odds ratio for metabolic syndrome for all gender and BMI groups [28]. While the sample size in the current study did not enable us to compute odds ratio for every BMI category separately, our overall odds ratio for metabolic syndrome were significant. Health ABC also reported on the odds ratio for type 2 diabetes in their sample of older adults. In Health ABC, after adjustment for BMI, the odds ratio for type 2 diabetes was 3.0 in women and 1.3 in men [5]. The odds ratio in women is higher than observed in the current study (2.07 after adjustment for BMI and waist circumference) but lower in men (2.25 after adjustments for BMI and waist circumference).

One of the largest studies to measure volumetric VAT has been the Framingham Heart Study, where VAT was measured in a subset of subjects (n = 3001) [18]. Because DXA VAT measured using CoreScan is a volumetric measurement of VAT, we would expect our results to be most comparable to these results. While our results are generally consistent with those demonstrated in the Framingham cohort, odds ratio for type 2 diabetes in men were stronger in our study (2.25 relative to 0.9) than previously observed. However, results for metabolic syndrome, for both men and women, were slightly stronger in the Framingham cohort than in this analysis (2.6 relative to 1.75 for men and 4.7 relative to 3.46 for women). This may be a result of our smaller sample size (n = 939) or maybe the result of a slightly different region over which VAT is quantified.

Despite numerous studies demonstrating the negative health effects associated with increased visceral adiposity, its widespread use has been limited by the lack of accessible, reliable methods to quantify VAT. This study is the first demonstration of the ability to use an automated DXA VAT algorithm to correlate with cardiometabolic disease indicators. Unlike previous attempts to estimate VAT from DXA, which relied on anthropometric measures such as waist circumference, the DXA VAT measured using CoreScan is applied to a standard DXA total body examination and generates a VAT volume over the DXA android region automatically upon completion of the DXA scan. Creation of an automated, simple method for VAT quantification, which has been shown to be both technically accurate and associated with clinically meaningful parameters, may be a catalyst for clinical adoption of a central adiposity measurement.

There are several limitations in this study. First, our study subjects were predominately White. Since several reports have indicated that the quantity of VAT as well as the relationship between levels of VAT accumulation and cardiometabolic disease is altered in specific ethnic groups, it is important to replicate these findings in diverse populations [8, 28]. Second, these data represent a cross-sectional population. As such, we cannot provide any causal link between visceral adiposity and cardiometabolic disease.

DXA VAT measured using CoreScan was associated with hypertension, dyslipidemia, metabolic syndrome, and type 2 diabetes in this cross-sectional analysis of Caucasian adults. The magnitude and direction of these findings was consistent with previously reported literature. Future studies should be conducted to expand these findings to different ethnic groups, as well as to determine the ability to DXA VAT to augment existing clinical indicators as a predictor of cardiometabolic disease development. The findings of this analysis provide strong cross-sectional evidence that DXA VAT is a clinically meaningful tool for VAT measurement.


MPR contributed to the design of the project, led the data analysis, and drafted/edited the manuscript. ALC contributed to the conception and design of the project, led the data collection, and reviewed the manuscript. JX researched and analyzed data, and reviewed the manuscript. WKW analyzed data and reviewed the manuscript. CT and LG contributed to collection of study data. YX reviewed/edited the manuscript. DLE contributed to the conception and design of the project, and reviewed/edited the manuscript. DLE is the guarantor of the work and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors would like to acknowledge Cindy E. Davis, GE Global Research Center, and Laura Stoltenberg, GE Healthcare, for their support and encouragement, and Katia Garlaschelli, Center for the Study of Atherosclerosis, for collecting the biochemical analyses from the PLIC study.