• dual-energy X-ray absorptiometry (DXA);
  • visceral obesity;
  • waist circumference


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations and strengths
  8. Conclusions
  9. Acknowledgements
  10. References


The accuracy of the use of anthropometrics to quantify visceral adipose tissue (VAT) in treated HIV-infected patients is unknown. We evaluated the predictive accuracy of waist circumference (WC) with and without dual-energy X-ray absorptiometry (DXA)-derived trunk : limb fat ratio [fat mass ratio (FMR)] as surrogates for VAT determined using computerized axial tomography (CT-determined VAT).


We performed a retrospective cohort analysis of treated HIV-infected male patients followed at the Modena HIV Clinic. We developed prediction equations for VAT using linear regression analysis and Spearman correlations. Receiver operating characteristic (ROC) analysis evaluated the accuracy of WC alone or with FMR at discrete VAT thresholds.


The 1500 Caucasian male patients had a median age of 45 years, body mass index (BMI) of 24, WC of 87 cm, VAT area of 127 cm2 and body fat percentage of 14%. The correlation between WC-predicted VAT and CT-VAT was 0.613, and this increased significantly if FMR was added. The WC-associated R2 of 0.35 increased to 0.51 if the prediction equation included WC plus FMR. The area under the ROC curve (AUC) using WC was 0.795−0.820 at all VAT thresholds. The positive predictive value (PPV) and negative predictive value (NPV) changed reciprocally at CT-VAT thresholds from 75 to 200 cm2 and ranged from 0.72 to 0.74, respectively, at a representative VAT of 125 cm2. Adding the FMR to the predictive equations increased the AUC in the range of 0.854−0.889 with the PPV and NPV increasing minimally, ranging from 0.780 to 0.821. Limits of precision were wide, especially at the highest CT-VAT levels, and varied from 24 to 68 cm2.


WC is a limited surrogate for CT-VAT in this population and DXA-derived parameters do not improve performance indices to a clinically relevant level. These findings should inform the applicability of WC to predict VAT in treated HIV-infected male patients.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations and strengths
  8. Conclusions
  9. Acknowledgements
  10. References

Long-term survival in HIV-infected patients has improved significantly since the introduction of highly active antiretroviral therapy (HAART). Nevertheless, overall survival rates remain 75–85% that of uninfected controls [1]. Currently, complications in treated patients are mostly attributable to non-AIDS-related events usually occurring more commonly in older uninfected persons, as well as specific hepatic, renal and malignant disorders. Metabolic complications, including visceral obesity, also occur commonly. Visceral obesity, observed soon after the introduction of HAART in the mid-1990s, still occurs in some treated patients [2]. In the general population, increased visceral adipose tissue (VAT) increases the risk of type II diabetes mellitus (T2DM), cardiovascular disease (CVD) and overall mortality. The independent contribution of VAT to mortality in treated HIV-infected patients highlights its clinical relevance [3].

VAT is reliably quantified by either single-slice CT or magnetic resonance imaging (MRI), most often at the L4−5 intervertebral level, which correlates strongly with total intra-abdominal fat mass. The need for costly equipment and CT-associated radiation exposure has resulted in neither criterion modality being recommended as a screening tool for visceral obesity. Dual-energy X-ray absorptiometry (DXA) scans, associated with minimal radiation exposure, quantify bone mineral content and also determine regional fat and lean tissue mass, although they cannot differentiate between subcutaneous adipose tissue (SAT) and VAT. DXA-derived trunk fat accounts for more than 80% of the variability of CT-determined VAT (CT-VAT) [4] and correlates strongly with CT-VAT or VAT measured using MRI (MRI-determined VAT) both in the general population and in HIV-infected subjects [5]. Furthermore, DXA-derived trunk fat predicts mortality [6] in the general population.

The limitations associated with criterion imaging modalities have led to the evaluation of anthropometric parameters, particularly waist circumference (WC), as surrogate markers of VAT. WC correlates with CT-VAT and MRI-VAT and predicts morbidity and mortality [7]. However, its ability to accurately predict specific VAT levels in the general population is uncertain [8]. Furthermore, the accuracy associated with using WC to predict VAT has not been systematically investigated in HIV-infected subjects.

The goals of this analysis were thus (1) to evaluate the accuracy and efficiency indices of WC compared with CT-VAT to predict VAT in treated HIV-infected male patients; and (2) to determine whether adding DXA-derived trunk fat measures improves the ability of WC to predict VAT.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations and strengths
  8. Conclusions
  9. Acknowledgements
  10. References

Study population

All seropositive male patients older than 18 years, on stable HAART, evaluated between 2005 and 2010 at the HIV Metabolic Clinic at the University of Modena and Reggio Emilia in Italy were eligible for inclusion in the study. Of the entire Clinic population of more than 2500 mostly Caucasian subjects, 66% are referred from other Italian sites, while the remainder receive their routine HIV care at the Clinic. All subjects have a standardized baseline evaluation consisting of comprehensive clinical, biological and imaging tests.


Clinical and anthropometric

All patients were clinically stable at the time of initial evaluation. WC was measured by a single operator at the narrowest point mid-way between the lowest rib and the iliac crest with the subject standing at the end of expiration and was calculated as the average of three measurements. Body mass index (BMI) was calculated as weight in kilograms divided by the square of the height in metres.

Laboratory analyses

Fasting biochemical evaluations including total cholesterol, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, triglyceride (TG), glucose and insulin levels were performed. Insulin resistance (IR) was calculated using the homeostasis model assessment equation [HOMA-IR = (fasting insulin in mU/mL × fasting glucose in mmol/L)/22.5].

Body composition imaging

Whole body, trunk, leg and arm fat and lean mass were quantified using DXA with a LUNAR DPX-MD (Lunar Corporation, Madison, WI) for scans before July 2007 and Hologic 4500 QDR Elite (Hologic Inc., Bedford, MA) for scans after that date. All LUNAR measurements were converted to Hologic equivalents using established crossover equations for statistical analysis. A single CT image at the level of the L4 vertebra was taken for quantification of VAT and abdominal SAT using a 64-multislice CT scanner (LightSpeed VTC; General Electric Medical System, Milwaukee, USA). Each CT image was analysed using a software application based on advantage for Windows 4.4 (General Electric Medical System). Two radiologists assessed CT images for VAT and abdominal SAT. Agreement between the operators for VAT measurement was calculated on a subset of 40 scans and demonstrated a high repeatability (r = 0.97; β = 0.98).

Study design

This was a retrospective cohort analysis using selected results derived from the routinely obtained tests. The presence of results for all parameters was not necessary for inclusion. We determined the number of subjects with missing primary outcome parameters, including WC-predicted, DXA-predicted and CT-VAT, and compared relevant descriptive features between subjects with missing results and those with complete data. This analysis was approved by the Ethics Committee of the University of Modena and Reggio Emilia and all patients gave informed consent.

Statistical analysis

Data were derived from the HIV Metabolic Clinic's database. To minimize the effects of outliers, we included data limited to those patients with VAT values between the 5th and 95th percentiles of all available VAT results. Basic descriptive, clinical, demographic and laboratory data are expressed as the median and 95% confidence interval (CI). A P-value < 0.05 was considered statistically significant. Analyses were performed using NCSS version 2007 (NCSS, Kaysville, UT).

Predictive equations for VAT were derived by linear regression analysis using WC only and WC plus the following DXA-derived trunk fat parameters: (1) absolute trunk fat mass (g); (2) ratio of trunk fat to extremity fat [fat mass ratio (FMR)] [9]; and (3) ratio of trunk fat to total body fat. Predictive equations adjusting for the effects of age, weight, BMI, HOMA, absolute CD4 T-helper cell count and TG concentration were also determined by linear regression. From the regression equations we derived the R2 (coefficient of determination, a measure of the goodness of the model's fit), as well as the root mean square error (RMSE) and 95% CIs.

Spearman correlations (rho values) were calculated between the measured WC and CT-VAT as well as between CT-VAT and various predicted VAT models.

We performed empirical receiver operating characteristic (ROC) analysis to evaluate the accuracy of using WC, either alone or combined with DXA-derived parameters, to predict VAT, wherein the actual condition variables were the VAT thresholds and the criterion variables were the predicted VATs. This analysis yielded the area under the ROC curve (AUC, the c-statistic), a single, quantifiable measure of overall accuracy, in addition to the associated sensitivity, specificity and Bayes' prevalence-corrected positive and negative predicted values (PPV and NPV, respectively) for each cut-point. The cut-point (discriminant point) selected corresponded to the maximum of the sum of the sensitivity and specificity. Significant differences between AUCs were determined by comparing the 95% CIs. Results of the ROC analysis, determined using the predictive equations derived by using both WC alone and WC plus a DXA-derived trunk fat measure, are described at discrete VAT thresholds ranging from 75 to 200 cm2. Despite the lack of published consensus guidelines, these thresholds were chosen based on general agreement concerning their association with the risk of clinical sequelae. Although a single-cut CT-VAT area < 100 cm2 is infrequently associated with complications, values of 125–140 cm2, and especially values > 175 cm2, are associated with increasing risk of metabolic complications and mortality, compared with < 100 cm2 [10-13].

The limits of prediction associated with the prediction equations for VAT were derived from regression analysis. The limits are defined as 1.96 × RMSE above and below the fitted regression line. In addition to results determined for the total cohort, the limits of prediction were obtained for each quartile of VAT: Q1 = VAT ≤ 87 cm2; Q2 = VAT > 87 cm2 and ≤ 125 cm2; Q3 = VAT > 125 cm2 and ≤ 176 cm2; Q4 = VAT > 176 cm2.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations and strengths
  8. Conclusions
  9. Acknowledgements
  10. References

Descriptive characteristics

Table 1 presents the HIV-related and metabolic, body composition and anthropometric characteristics of the 1500 subjects. The median age was 45 years. Results were available for 97.5% of subjects for WC, 90% for DXA-derived VAT and 99.5% for CT-VAT. There were no differences in relevant HIV-related, body composition, and metabolic parameters in patients without vs. with the availability of complete data (results not shown). Risk factors for HIV infection were heterosexual transmission in 55% of patients, transmission between men who have sex with men in 21% and injecting drug use in 24%. The median duration of known HIV infection was 14 years. Overall, subjects had achieved effective immune recovery, with a median CD4 T-helper cell count of 526 cells/μl at the time of initial assessment. Sixty per cent of subjects had an undetectable HIV viral load.

Table 1. Descriptive characteristics and metabolic features
 Entire cohort
(n = 1500)
  1. BMI, body mass index; HDL, high-density lipoprotein cholesterol; HOMA, homeostasis model assessment; LDL, low-density lipoprotein cholesterol; TC, total cholesterol; TG, triglycerides; VAT, visceral adipose tissue; WC, waist circumference.

  2. *Values are median (95% confidence interval) unless otherwise stated.

HIV infection duration (months)170 (166, 176)*
Age (years)45 (45, 45)
CD4 count nadir (cells/μl)178 (168, 190)
CD4 count (cells/μl)526 (506, 541)
Undetectable HIV viral load (%)60
Glucose (mmol/L)5.2 (5.1, 5.2)
Insulin (pmol/L)95.8 (91.0, 100.7)
HOMA3.2 (3.0, 3.3)
TC (mmol/L)4.8 (4.7, 4.9)
LDL (mmol/L)2.9 (2.8, 2.9)
HDL (mmol/L)1.0 (1.0, 1.1)
TG (mmol/L)1.9 (1.8, 2.0)
Body fat (%)13.8 (13.4, 14.2)
Weight (kg)70 (70, 71)
BMI23.7 (23.5, 23.8)
WC (cm)87 (86, 87)
Trunk fat (g)5730 (5530, 5885)
Range1141–21 560
VAT (cm*)127 (123, 130)

The metabolic profile of the cohort was typical of patients on long-term HAART. Total cholesterol and LDL cholesterol were in the normal range, and triglycerides were mildly elevated above the normal range. Normal median fasting glucose and increased HOMA suggested the presence of insulin resistance. The BMI and WC were normal for sex and age. The DXA-derived total body fat percentage (median = 13.8%) was lower than in age-matched HIV-negative Caucasian men but consistent with that observed previously in treated HIV-infected patients [9]. The median VAT area was 127 cm2.

VAT vs. WC scatter plot

Figure 1 shows a scatter plot of CT-VAT (range 40−300 cm2) vs. WC. Spearman's correlation (rho) was 0.613 (P < 0.0001). The RMSE was 49.8 cm2 and the prediction limits (the fitted line ± 1.96 × RMSE) were ± 97.6 cm2.


Figure 1. Scatter plot of computerized axial tomography-determined visceral adipose tissue (CT-VAT) vs. waist circumference (WC). The centre line denotes the regression line of best fit (r = 0.613; P < 0001). The superior and inferior lines denote the limits of predictability and are equal to ± 1.96 × RMSE from the fitted line.

Download figure to PowerPoint

Predictive equations

The predictive equations for the measured CT-VAT, the respective R2, and the Spearman correlations (rho) between predicted VAT and CT-VAT are shown in Table 2. The equations were determined using individual parameters or combinations thereof. The R2 derived using WC alone accounted for 35% of the variability in CT-VAT. Similarly, the DXA-derived absolute amount of trunk fat alone accounted for 36% of the variability. The combination of WC plus the FMR gave the largest increase in R2. The addition of either age or BMI to WC plus the FMR increased the R2 by less than one percentage point (results not shown). The addition of any other single parameter, including weight, HOMA, CD4 count and TG concentration, or of any combination of these parameters, also did not significantly improve the ability to predict the VAT above that obtained by WC plus the FMR. Also, the correlation between predicted VAT and CT-VAT increased to the greatest extent when only WC and FMR were used to predict VAT.

Table 2. Predictive equations for visceral adipose tissue (VAT)*
Predictive VAT equationsnRhoR2RMSE (cm2)95% predictive limits (cm2)MAPE (%)
(WC + other variables)
  1. MAPE, mean absolute per cent error; rho, Spearman correlation; R2, coefficient of determination; RMSE, root mean square error; VAT, visceral adipose tissue; WC, waist circumference; 95% predictive limits = fitted line ± 1.96 × RMSE.

  2. *Predictive equations were derived using WC alone + WC with selected dual-energy X-ray absorptiometry (DXA)-derived fat mass variables and associated descriptive parameters including correlation with computerized axial tomography-determined visceral adipose tissue (CT-VAT). The significance of the rho for all tested correlations was P < 0.0001.

−225.5755 + 4.1159 × WC14690.6130.34949.8± 97.635.1
59.5475 + 0.0123 × trunk fat (g)13520.36149.2± 96.434.1
−333.4988 + 45.2719 × trunk/peripheral fat + 4.4030 × WC13300.7140.51342.9± 84.228.2
−70.7668 + 0.0081 × trunk fat (g) + 1.7801 × WC13300.6580.38948.1± 94.233.0
−148.6258 + 2.7220 × % trunk fat/trunk mass + 2.6891 × WC13300.6470.38548.2± 94.533.1
−375.8361 + 2.4025 × % trunk fat/total body fat + 4.2162 × WC13300.7080.46844.9± 88.030.6

ROC analysis using both the WC variable alone and WC plus the FMR

Table 3 shows the AUC and performance characteristics (sensitivity, specificity, PPV and NPV) derived by ROC analysis. These values were determined at discrete CT-VAT thresholds using predictive equations that included the independent variable WC alone and WC plus the FMR. The PPV decreased as the CT-VAT threshold increased, although, as expected, the prevalence of CT-VAT at these higher thresholds decreased. The AUC increased significantly at all VAT thresholds using WC plus the FMR compared with using WC alone. However, this resulted in only a modest increase in the PPV and NPV.

Table 3. Receiver operating characteristic (ROC) analysis: computerized axial tomography-determined visceral adipose tissue (CT-VAT) vs. predicted VAT (n = 1330)a
CT-VAT threshold (cm2)CT-VAT prevalence (%)AUCdCut-point (cm2)SensitivitySpecificityPPVNPV
  1. AUC, area under the ROC curve; FMR, fat mass ratio; NPV, negative predictive value; PPV, positive predictive value.

  2. a

    The ROC analysis used the predictive equations derived using WC alone and WC plus the FMR. The CT-VAT prevalence was determined at the defined VAT thresholds. The selected cut-point (discriminate point) corresponds to the maximum of the sum of the sensitivity and specificity.

  3. b

    WC only, for all results in the column.

  4. c

    WC + FMR, for all results in the column.

  5. d

    P < 0.001 for all differences in AUC determined using WC+FMR compared with WC alone.


ROC analysis using a designated WC alone vs. a known WC plus the FMR

Table 4 shows the effects on the performance characteristics of using a designated WC of 95 cm to perform the ROC analyses. This WC was chosen as it predicts a VAT of 130 cm2, which is associated with CVD [10]. The results were derived using both the WC of 95 cm alone plus the combination of this WC and the cohort's median FMR. Results are shown at representative VAT thresholds of 125 and 175 cm2. The PPV and NPV results obtained by using WC plus FMR are represented as the difference in percentage points between the values derived using WC alone and those obtained by using WC plus the FMR. With a WC of 95 cm, the PPV increased compared with the value obtained using the cut-point at the maximum of sensitivity and specificity. The PPV was determined using both WC and WC plus the FMR. Tables 3 and 4 detail the comparisons at the VAT thresholds of 125 and 175 cm2. At both VAT thresholds the AUC derived using a WC of 95 cm plus the median FMR was significantly greater compared with that determined using WC alone. However, the resulting increases in PPV and NPV (+4.3 and +4.7 at the 125 cm2 threshold and +4.5 and +4.4 at the 175 cm2 threshold, respectively) were modest compared with using WC alone.

Table 4. Receiver operating characteristic (ROC) analysis using waist circumference (WC) = 95 cm plus the fat mass ratio (FMR)
  1. The table displays the values for the AUC, PPV and NPV obtained using a predetermined WC of 95 cm and the combination of this WC plus the cohort's median FMR. These values were determined for the entire cohort at the representative VAT thresholds of 125 and 175 cm2. For the PPV and the NPV results, the actual values (in bold within the PPV and NPV columns) are those obtained by using WC alone. The results for the PPV and NPV within the WC+FMR columns represent the difference in percentage points between the baseline values obtained at the predefined VAT levels and those obtained when WC plus the FMR were used.

  2. AUC, area under the ROC curve; FMR, fat mass ratio; NPV, negative predictive value; PPV, positive predictive value; ROC, receiver operating curve; VAT, visceral adipose tissue.

125 cm2      
Entire cohort0.8000.87484.9%+ 4.357.3%+ 4.7
175 cm2      
Entire cohort0.8020.86358.5%+ 4.583.2%+ 4.4

Analysis of predictive limits

The RMSE results are presented in Table 5 both for the entire cohort and as a function of CT-VAT quartiles. These were derived both from the regressions of CT-VAT on WC-VAT and of CT-VAT on WC plus FMR-predicted VAT. For the comparison of quartiles, the RMSE and the 95% prediction limits were greatest in Q4 (CT-VAT > 176 cm2), with no difference between Q1, Q2 and Q3. The RMSE for the total cohort exceeded that of Q4. The median age of subjects in Q4 was significantly greater than that of subjects in Q1, Q2 or Q3.

Table 5. Visceral adipose tissue (VAT) prediction equation-associated prediction limits for the entire cohort and by quartilesa
VAT quartilesRMSE (cm2)Prediction limits (cm2)Age [median (95% CI)]b
  1. CI, confidence interval; FMR, fat mass ratio; RMSE, root mean square error; VAT, visceral adipose tissue; WC, waist circumference.

  2. a

    The prediction limits, defined as 1.96 × RMSE, for the VAT prediction equations, determined using WC alone and WC + FMR, were derived by multiple linear regression analysis.

  3. b

    The median age for Q4 was significantly higher than that for the total cohort, Q1, Q2 and Q3; there was no difference in age among Q1, Q2 and Q3.

Total cohort49.8197.6345 (45, 45)
Q1, VAT ≤ 87 cm212.0723.6643 (43, 44)
Q2, VAT > 87 and ≤ 125 cm210.6020.7844 (43, 45)
Q3, VAT > 125 and ≤ 176 cm214.2627.9545 (44, 45)
Q4, VAT > 176 and ≤ 303 cm234.5167.6447 (47, 48)
WC + FMR   
Total cohort42.9384.1445 (45, 45)
Q1, VAT ≤ 87 cm211.2522.0543 (43, 44)
Q2, VAT > 87 and ≤ 125 cm210.4720.5244 (43, 45)
Q3, VAT > 125 and ≤ 176 cm214.2127.8545 (44, 45)
Q4, VAT > 176 and ≤ 303 cm233.7266.1047 (47, 48)


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations and strengths
  8. Conclusions
  9. Acknowledgements
  10. References

Evaluation of abdominal obesity is recommended as part of routine care in the general population [14-16], as well as in HIV-infected patients [17]. Inherent to the perceived clinical value of this measurement is the assumption that surrogate markers accurately predict the actual amounts of VAT. However, in our cohort WC had limited accuracy in predicting different threshold amounts of VAT. This conclusion is based on the results which showed only a moderate correlation between WC and CT-VAT with a modest coefficient of determination (R2) for prediction of the WC-derived VAT, and wide limits of prediction in the regression equations.

The correlation coefficient of 0.613 between WC and CT-VAT is lower than the 0.63−0.82 usually reported in HIV-negative men [18-20]. Cofactors contributing to this variability include age, sex, degree of fatness, comorbidities and cohort size. In this cohort, the prevalence of VAT at all thresholds was lower than that published in comparable seronegative persons. This was also observed in a study comparing body composition between HIV-positive and HIV-negative men [21]. To our knowledge, no published study has directly compared the accuracy of anthropometrics to predict VAT in HIV-positive and HIV-negative subjects. Rankinen et al. assessed the accuracy of WC to predict VAT in a Caucasian population and compared it to CT-VAT [19]. In a male subgroup from their cohort, with a mean age of 54 years, the AUC associated with the ability of a WC of 94 cm to predict VAT ≥ 150 cm2 was 0.936 and the PPV and NPV were 0.801 and 0.864, respectively. In a subgroup of our HIV-infected cohort, with a similar mean age of 56 years, the ability of a WC of 95 cm to predict the same VAT was associated with an AUC of 0.759; the corresponding PPV and NPV were 0.834 and 0.556, respectively. The prevalence of VAT ≥ 150 cm2 in Rankinen et al.'s cohort and in the HIV-infected subjects was 52% and 35%, respectively. Although of similar race and age, the HIV-infected subjects had lower BMI, percentage body fat and WC. The lower prevalence of high VAT amounts in the seropositive men may have contributed to the inferior performance indices overall.

The WC-derived VAT accounted for 35% of the variability of the CT-VAT, with a large mean absolute per cent error of 35% (Table 2). In the general population, using age and anthropometric variables for the VAT predictive equations and comparing with either CT-VAT or MRI-VAT, an R2 of 55−90% is usually reported [22, 23], also usually associated with a large variability of the R2.

We found wide 95% prediction limits for the regression of WC-VAT or WC plus FMR-predicted VAT on CT-VAT, especially in subjects with VAT ≥ 176 cm2, suggesting that anthropometrics may be least reliable in subjects with very large VAT values. Accurately identifying this level of VAT is important to HIV-infected patients, as we previously reported that most patients with anthropometrically predicted visceral obesity have concurrently measured CT-VAT of 175 cm2 [24]. Wide limits of variability are commonly reported in the general population. In separate studies, Despres et al. and Rankinen et al. found a wide RMSE, 25−31% of mean VAT, and advised caution when using anthropometrics to predict VAT [18, 19]. Bertin and Glickman reported wide limits of agreement when using Bland−Altman analysis to compare anthropometrics plus DXA-VAT with CT-VAT [4, 20]. Kamel et al. found no correlation between WC and MRI-VAT and a weak correlation between DXA-VAT and MRI-VAT in obese subjects and advised caution in using either modality to predict VAT [25]. Van der Kooy et al. reported a large coefficient of variation in obese men, concluding that anthropometrics had limited predictive ability [26]. However, in nonobese men, Kamel et al. reported a strong correlation between WC plus DXA-VAT and MRI-VAT [27]. These analyses suggest that the reported wide limits of variability diminish the reliability of anthropometrics to predict VAT, particularly in obese subjects.

The AUCs derived by ROC analysis are consistent with an overall good level of accuracy. However, when WC alone was used to predict, for example, a VAT of at least 125 cm2, the associated PPV and NPV ranged from 72 to 74%, suggesting false-positive and false-negative rates of about 25%. Using a predetermined WC of 95 cm in the predictive equations resulted in a similar AUC and an increased PPV, with the NPV being, as expected, reciprocally lower at all VAT thresholds (Table 4). Similar relationships were observed at the VAT threshold of 175 cm2.

We investigated whether a DXA-derived trunk fat parameter could improve the predictive accuracy of using WC alone. The addition of the FMR to WC resulted in the largest increase in predictive ability. The AUC increased significantly at all VAT thresholds if WC plus the FMR was used instead of WC alone (Tables 3 and 4). However, this resulted in only a minimal, clinically limited increase in the PPV and NPV (Table 4). Studies in the general population have also compared the predictive accuracy of using anthropometrics plus DXA trunk fat measurements with that achieved by using either parameter alone. These have shown either no improvement [28, 29] or only modest, possibly gender-related improvements associated with the use of combined modalities [20, 30-32].

Subjects with the highest VAT had the widest limits of variation and were also the oldest. VAT normally increases with age while SAT decreases. Although the higher prevalence of larger amounts of VAT improves predictive accuracy, the increase in VAT at any given WC is disproportionate in older subjects [33], altering overall accuracy [34]. The increased FMR occurring with age, described as age-related lipodystrophy [35], may be a factor in the decreased accuracy of using anthropometrics in elderly people [36]. The increased FMR occurring in some younger treated HIV-infected patients may therefore also contribute to the limited accuracy we found of using WC.

Few studies have evaluated the comparative accuracy of different methods to predict VAT in HIV-infected patients. Schwenk et al. found limited agreement by ROC analysis between clinically diagnosed abdominal lipohypertrophy and either WC or waist : hip ratio in an HIV-infected cohort [37]. Batterham et al. found significant differences using Bland−Altman analysis between DXA-derived and anthropometrically derived fat masses [38]. Lan et al. compared regional DXA-derived fat mass with MRI-VAT using Bland−Altman analysis and concluded that DXA-derived fat mass accurately reflected MRI-derived fat mass in epidemiological studies but urged cautioned when using DXA in clinical situations [39]. In the Fat Redistribution and Metabolic Change Cohort Study, Scherzer et al. reported good agreement between DXA- and MRI-derived trunk fat, although comparison with MRI-VAT was not reported [5]. Kotler et al. demonstrated that anthropometrics were poor surrogates for tracking VAT changes in studies of pharmacological VAT reduction [40].

In the general population, Jensen et al. demonstrated that only a single-cut CT at L4−5 along with a DXA-derived abdominal fat parameter correlated with total CT-measured abdominal fat and concluded that this combination was superior to either anthropometrics or CT alone [41]. Bonora et al. recommended that direct assessment of VAT with a criterion imaging modality should be used clinically [42].

Limitations and strengths

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations and strengths
  8. Conclusions
  9. Acknowledgements
  10. References

There are several limitations in these analyses. Two-thirds of enrolled subjects were nonrepresentative of the general, treated male HIV-infected population, because they were specifically referred for metabolic evaluation. The subjects were from an ethnically homogenous background, limiting the generalizability of these findings to other ethnic groups. Furthermore, it is unknown whether body composition is similar in Caucasians of different ethnic origins, given variable lifestyle and environmental parameters. We did not have access to HIV-negative controls. We will report findings in women separately. The strengths of this study relate to the uniform determination of study parameters in a large number of subjects, and to the confirmation of results using different statistical methods.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations and strengths
  8. Conclusions
  9. Acknowledgements
  10. References

Our findings indicate that WC is an unreliable surrogate of CT-VAT in treated HIV-infected Caucasian men. Combining WC with DXA-derived FMR improved the accuracy of predicting VAT to a clinically limited extent. Although the decision to rely on the accuracy of WC-VAT is a clinical one, clinicians should be aware of the inherent limitations of using WC to predict visceral adiposity in such patients. Modifications of DXA trunk fat measurement algorithms may lead to improved predictive ability of VAT [43, 44].


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations and strengths
  8. Conclusions
  9. Acknowledgements
  10. References

Conflicts of interest: JF has received a research grant from Theratechnologies and speaker fees from Merck-Serono, ViiV, Bristol-Myers Squibb, Gilead Sciences and Abbott, has served on the advisory board of Merck-Serono and Theratechnologies, and has been a consultant for Theratechnologies. DK has received grant support for clinical trials from Gilead, Vertex, Merck and Boehringer-Ingelheim, research grants from Genentech, and educational grants from Janssen, Merck, Genetech, Gilead and Vertex. GG has received research funding from and served as a consultant to Boehringer-Ingelheim, Bristol-Myers Squibb, Gilead Sciences, GlaxoSmithKline, Merck Sharp & Dohme, Theratechnologies and Tibotec. LR and SZ report no conflicts of interest.

Author contributions

JF, DK and LR: study conception and design; LR: statistical analysis; GG and SZ: research and data acquisition; JF, LR, DK and GG: data analysis and interpretation; JF, LR and DK: drafting of the manuscript; JF, LR, DK, SZ and GG: critical revision of the manuscript. All authors read and approved the final version of the manuscript.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations and strengths
  8. Conclusions
  9. Acknowledgements
  10. References
  • 1
    Nakagawa F, Lodwick RK, Smith CJ et al. Projected life expectancy of people with HIV according to timing of diagnosis. AIDS 2012; 26: 335343.
  • 2
    Falutz J. HIV infection, body composition changes and related metabolic complications: contributing factors and evolving management strategies. Curr Opin Clin Nutr Metab Care 2011; 14: 255260.
  • 3
    Scherzer R, Heymsfield SB, Lee D et al. Decreased limb muscle and increased central adiposity are associated with 5-year all-cause mortality in HIV infection. AIDS 2011; 25: 14051414.
  • 4
    Glickman SG, Marn CS, Supiano MA, Dengel DR. Validity and reliability of dual-energy X-ray absorptiometry for the assessment of abdominal adiposity. J Appl Physiol 2004; 97: 509514.
  • 5
    Scherzer R, Shen W, Bacchetti P et al. Comparison of dual-energy X-ray absorptiometry and magnetic resonance imaging-measured adipose tissue depots in HIV-infected and control subjects. Am J Clin Nutr 2008; 88: 10881096.
  • 6
    Krakauer JC, Franklin B, Kleerekoper M, Karlsson M, Levine JA. Body composition profiles derived from dual-energy X-ray absorptiometry, total body scan, and mortality. Prev Cardiol 2004; 7: 109115.
  • 7
    Jacobs EJ, Newton CC, Wang Y et al. Waist circumference and all-cause mortality in a large US cohort. Arch Intern Med 2010; 170: 12931301.
  • 8
    Janssen I, Heymsfield SB, Ross R. Application of simple anthropometry in the assessment of health risk: implications for the Canadian Physical Activity, Fitness and Lifestyle Appraisal. Can J Appl Physiol 2002; 27: 396414.
  • 9
    Bonnet E, Delpierre C, Sommet A et al. Total body composition by DXA of 241 HIV-negative men and 162 HIV-infected men: proposal of reference values for defining lipodystrophy. J Clin Densitom 2005; 8: 287292.
  • 10
    Lemieux S, Prud'homme D, Bouchard C, Tremblay A, Despres JP. A single threshold value of waist girth identifies normal-weight and overweight subjects with excess visceral adipose tissue. Am J Clin Nutr 1996; 64: 685693.
  • 11
    Hunter GR, Snyder SW, Kekes-Szabo T, Nicholson C, Berland L. Intra-abdominal adipose tissue values associated with risk of possessing elevated blood lipids and blood pressure. Obes Res 1994; 2: 563568.
  • 12
    Marques MD, Santos RD, Parga JR et al. Relation between visceral fat and coronary artery disease evaluated by multidetector computed tomography. Atherosclerosis 2010; 209: 481486.
  • 13
    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 2006; 14: 336341.
  • 14
    Cornier MA, Despres JP, Davis N et al. Assessing adiposity: a scientific statement from the American Heart Association. Circulation 2011; 124: 19962019.
  • 15
    Klein S, Allison DB, Heymsfield SB et al. Waist circumference and cardiometabolic risk: a consensus statement from shaping America's health: Association for Weight Management and Obesity Prevention; NAASO, the Obesity Society; the American Society for Nutrition; and the American Diabetes Association. Diabetes Care 2007; 30: 16471652.
  • 16
    Balkau B, Deanfield JE, Despres JP et al. International Day for the Evaluation of Abdominal Obesity (IDEA): a study of waist circumference, cardiovascular disease, and diabetes mellitus in 168,000 primary care patients in 63 countries. Circulation 2007; 116: 19421951.
  • 17
    Lundgren JD, Battegay M, Behrens G et al. European AIDS Clinical Society (EACS) guidelines on the prevention and management of metabolic diseases in HIV. HIV Med 2008; 9: 7281.
  • 18
    Despres JP, Prud'homme D, Pouliot MC, Tremblay A, Bouchard C. Estimation of deep abdominal adipose-tissue accumulation from simple anthropometric measurements in men. Am J Clin Nutr 1991; 54: 471477.
  • 19
    Rankinen T, Kim SY, Perusse L, Despres JP, Bouchard C. The prediction of abdominal visceral fat level from body composition and anthropometry: ROC analysis. Int J Obes Relat Metab Disord 1999; 23: 801809.
  • 20
    Bertin E, Marcus C, Ruiz JC, Eschard JP, Leutenegger M. Measurement of visceral adipose tissue by DXA combined with anthropometry in obese humans. Int J Obes Relat Metab Disord 2000; 24: 263270.
  • 21
    Bacchetti P, Gripshover B, Grunfeld C et al. Fat distribution in men with HIV infection. J Acquir Immune Defic Syndr 2005; 40: 121131.
  • 22
    Seidell JC, Oosterlee A, Thijssen MA et al. Assessment of intra-abdominal and subcutaneous abdominal fat: relation between anthropometry and computed tomography. Am J Clin Nutr 1987; 45: 713.
  • 23
    Janssen I, Heymsfield SB, Allison DB, Kotler DP, Ross R. Body mass index and waist circumference independently contribute to the prediction of nonabdominal, abdominal subcutaneous, and visceral fat. Am J Clin Nutr 2002; 75: 683688.
  • 24
    Falutz J, Rosenthall K, Kotler D, Zona S, Guaraldi G. Surrogate markers of visceral adipose tissue (VAT) in HIV-infected males: how accurate are they? 13th International Workshop on Adverse Drug Reactions and Co-morbidities in HIV Vol 16 (Suppl 2). Rome, Italy, Antivir Ther 2011: A18.
  • 25
    Kamel EG, McNeill G, Van Wijk MC. Usefulness of anthropometry and DXA in predicting intra-abdominal fat in obese men and women. Obes Res 2000; 8: 3642.
  • 26
    van der Kooy K, Leenen R, Seidell JC, Deurenberg P, Visser M. Abdominal diameters as indicators of visceral fat: comparison between magnetic resonance imaging and anthropometry. Br J Nutr 1993; 70: 4758.
  • 27
    Kamel EG, McNeill G, Han TS et al. Measurement of abdominal fat by magnetic resonance imaging, dual-energy X-ray absorptiometry and anthropometry in non-obese men and women. Int J Obes Relat Metab Disord 1999; 23: 686692.
  • 28
    Micklesfield LK, Evans J, Norris SA et al. Dual-energy X-ray absorptiometry and anthropometric estimates of visceral fat in Black and White South African Women. Obesity (Silver Spring) 2010; 18: 619624.
  • 29
    Gradmark AM, Rydh A, Renstrom F et al. Computed tomography-based validation of abdominal adiposity measurements from ultrasonography, dual-energy X-ray absorptiometry and anthropometry. Br J Nutr 2010; 104: 582588.
  • 30
    Hill AM, LaForgia J, Coates AM, Buckley JD, Howe PR. Estimating abdominal adipose tissue with DXA and anthropometry. Obesity (Silver Spring) 2007; 15: 504510.
  • 31
    Treuth MS, Hunter GR, Kekes-Szabo T. Estimating intraabdominal adipose tissue in women by dual-energy X-ray absorptiometry. Am J Clin Nutr 1995; 62: 527532.
  • 32
    Svendsen OL, Hassager C, Bergmann I, Christiansen C. Measurement of abdominal and intra-abdominal fat in postmenopausal women by dual energy X-ray absorptiometry and anthropometry: comparison with computerized tomography. Int J Obes Relat Metab Disord 1993; 17: 4551.
  • 33
    Stevens J, Katz EG, Huxley RR. Associations between gender, age and waist circumference. Eur J Clin Nutr 2010; 64: 615.
  • 34
    Kuk JL, Lee S, Heymsfield SB, Ross R. Waist circumference and abdominal adipose tissue distribution: influence of age and sex. Am J Clin Nutr 2005; 81: 13301334.
  • 35
    Gavi S, Feiner JJ, Melendez MM, Mynarcik DC, Gelato MC, McNurlan MA. Limb fat to trunk fat ratio in elderly persons is a strong determinant of insulin resistance and adiponectin levels. J Gerontol A Biol Sci Med Sci 2007; 62: 9971001.
  • 36
    Kuk JL, Saunders TJ, Davidson LE, Ross R. Age-related changes in total and regional fat distribution. Ageing Res Rev 2009; 8: 339348.
  • 37
    Schwenk A, Breuer P, Kremer G, Ward L. Clinical assessment of HIV-associated lipodystrophy syndrome: bioelectrical impedance analysis, anthropometry and clinical scores. Clin Nutr 2001; 20: 243249.
  • 38
    Batterham MJ, Garsia R, Greenop P. Measurement of body composition in people with HIV/AIDS: a comparison of bioelectrical impedance and skinfold anthropometry with dual-energy x-ray absorptiometry. J Am Diet Assoc 1999; 99: 11091111.
  • 39
    Lan SJ, Engelson E, Agin D et al. Validation of dual-energy X-ray absorptiometry in the assessment of change in fat compartments, compared to measurement by magnetic resonance imaging, in HIV-infected adults. Int J Body Composit Res 2003; 1: 3743.
  • 40
    Kotler D, Thompson M, Chang P, Gertner JM, Muurahainen N. Anthropometric measures are poor surrogate markers for monitoring change in abdominal adiposity in HIV patients treated with growth hormone. 4th International Workshop on Adverse Drug Reactions and Lipodystrophy in HIV. San Diego, CA, 2002.
  • 41
    Jensen MD, Kanaley JA, Reed JE, Sheedy PF. Measurement of abdominal and visceral fat with computed tomography and dual-energy x-ray absorptiometry. Am J Clin Nutr 1995; 61: 274278.
  • 42
    Bonora E, Micciolo R, Ghiatas AA et al. Is it possible to derive a reliable estimate of human visceral and subcutaneous abdominal adipose tissue from simple anthropometric measurements? Metabolism 1995; 44: 16171625.
  • 43
    Kaul S, Rothney MP, Peters DM et al. Dual-energy X-ray absorptiometry for quantification of visceral fat. Obesity 2012; 20: 13131318.
  • 44
    Micklesfield LK, Goedecke JH, Punyanitya M, Wilson KE, Kelly TL. Dual-energy X-ray performs as well as clinical computed tomography for the measurement of visceral fat. Obesity (Silver Spring) 2012; 20: 11091114.