Anthropometric indexes outperform bioelectrical impedance analysis-derived estimates of body composition in identification of metabolic abnormalities in morbid obesity

Surgery for Obesity and Related Diseases 9 (2013) 648 – 652

Original article

Anthropometric indexes outperform bioelectrical impedance analysis-derived estimates of body composition in identification of metabolic abnormalities in morbid obesity Verónica Perea, M.D.a, Amanda Jiménez, M.D.a, Lílliam Flores, M.D., Ph.D.a,b,c, Emilio Ortega, M.D., Ph.D.a,b,c, Maria J. Coves, M.D., Ph.D.a, and Josep Vidal, M.D., Ph.D.a,b,c,* b

a Obesity Unit, Department of Endocrinology and Diabetes, Hospital Clinic Universitari, Barcelona, Spain Centro de Investigación Biomédica en Red de Diabetes y Enfermedades Metabólicas Asociadas, Barcelona, Spain c Institut d’Investigacions Biomèdiques August Pi Sunyer, Barcelona, Spain Received February 14, 2012; accepted May 29, 2012

Abstract

Background: The validity of anthropometric indexes in ascertaining the body composition (BC) in morbidly obese (MO) subjects has been questioned. Our objective was to evaluate, in MO subjects, whether bioelectrical impedance analysis (BIA) of BC is more closely associated with the metabolic syndrome (MS) and insulin resistance (IR) than are classic anthropometric measurements. The setting was a university hospital. Methods: The association between anthropometric (body mass index, waist circumference [WC]) and BIA (total fat mass [FM] [percentage of FM], truncal FM, android FM) estimates of BC, MS, and IR was evaluated in 784 white MO subjects (212 men and 572 women). BIA estimates were calculated using equations specific for MO subjects developed by our own group and validated against dual energy x-ray absorptiometry. Results: The prevalence of the MS and IR was 78.6% and 88.6%, respectively. The body mass index was greater in women with the MS (P ⬍.001) or IR (P ⬍.001), and the WC was larger in subjects of both genders with the MS or IR (P ⬍.001). Moreover, the WC correlated significantly with all the MS components (P ⬍.05). In contrast, the percentage of FM, truncal FM, and android FM were significantly associated with the MS only in women. Stepwise logistic regression analysis demonstrated the WC as the only significant predictor of the MS or IR (both P ⬍.001). Furthermore, receiver operating curve analysis showed WC was the most accurate BC parameter for the identification of subjects with the MS (area under the curve, WC ⫽ .681, P ⬍.001) or IR (area under the curve, WC ⫽ .753, P ⬍.001). Conclusion: In MO subjects, the BIA-derived indexes of total and central adiposity were not better predictors of the MS or IR than were traditional anthropometric measurements. (Surg Obes Relat Dis 2013; 9:648–652.) © 2013 American Society for Metabolic and Bariatric Surgery. All rights reserved.

Keywords:

Morbid obesity; Body composition; Waist circumference; Bioelectrical impedance

Despite the prevalence of metabolic disturbances increasing in parallel with larger degrees of obesity [1,2], a significant proportion of morbidly obese (MO) subjects are not affected by metabolic abnormalities [3]. This MO, but

*Correspondence: Josep Vidal, M.D., Ph.D., Obesity Unit, Department of Endocrinology and Diabetes, Hospital Clínic Universitari, Villarroel 170, Barcelona 08036 Spain. E-mail: [email protected]

otherwise healthy, metabolic phenotype has been associated with insulin sensitivity in the normal range [4]. Importantly, the central body fat distribution, rather than the total fat mass (FM), has been linked to insulin resistance (IR) and the metabolic syndrome (MS) in MO subjects [5]. Body mass index (BMI) and waist circumference (WC) are the anthropometric parameters most commonly used in clinical practice to estimate the total body fat and abdominal adiposity, respectively. However, it could be argued that the

1550-7289/13/$ – see front matter © 2013 American Society for Metabolic and Bariatric Surgery. All rights reserved. http://dx.doi.org/10.1016/j.soard.2012.05.010

BC and metabolic abnormalities in MO / Surgery for Obesity and Related Diseases 9 (2013) 648 – 652

BMI should mainly be considered a measure of heaviness, because it cannot differentiate lean tissue mass from FM [6]. However, the difficulty in identifying external landmarks hampers the measurement of the WC in MO subjects. Thus, it could be argued that an unmet need exists of appropriate methods to estimate the body fat distribution in MO subjects in routine clinical practice. Bioelectrical impedance analysis (BIA) is a noninvasive, inexpensive, and easy method to evaluate body composition (BC). Nonetheless, its use in BC analysis in MO subjects has traditionally been questioned because of the demonstration that BIA devices tend to overestimate the free FM and underestimate the FM in obese populations [7,8]. To circumvent this problem, we developed a set of BIA equations validated against dual energy x-ray absorptiometry (DXA) allowing for good estimates of total FM and truncal, gynoid, and android FM in MO subjects [9]. Against this background, the primary aim of our study was to evaluate whether, in MO subjects, BIA-derived indexes of BC were more closely associated with the MS and IR than classic anthropometric measurements.

Methods A total of 784 white, MO subjects (212 men and 572 women) were consecutively recruited from the patients referred to our Obesity Unit for evaluation for bariatric surgery. The eligibility criteria included age ⱖ18 years and weight stability (⫾2 kg) for a 3-month period before enrollment. Patients with serious metabolic, cardiovascular, or endocrine diseases, as determined from the clinical history, were excluded (i.e., cancer, heart failure, unstable angina, myocardial infarction, pulmonary insufficiency, chronic kidney disease [Modification of Diet in Renal Disease ⬍60 mL/min], or impaired thyroid function). Additional exclusion criteria included pregnancy, breastfeeding, or the use of medications known to cause dehydration, electrolyte disturbances, or weight changes. For the analysis on the relationship between the MS and BC, we evaluated the whole cohort. Because hypoglycemic agents and insulin have been reported to interfere with the Homeostasis Model of Assessment–Insulin Resistance estimation of insulin sensitivity [10], diabetic patients who were taking hypoglycemic drugs were excluded from the analysis on the relationship between insulin sensitivity and the anthropometric measurements and BIA-derived indexes of BC. Anthropometric measurements The participants were weighed wearing light clothes and without shoes to the nearest .1 kg. Height was determined using a fixed wall stadiometer to the nearest .1 cm. The WC was measured to the nearest .5 cm, at the level of the iliac crest, with a standard flexible, inelastic measuring tape. The

649

BMI was calculated as the weight (kg) divided by the height in square meters. BIA estimation of total and segmentary FM The BIA measurements were performed using the Tanita BC 418 MA Segmental Body Composition Analyzer (Tanita, Tokyo, Japan) according to manufacturer’s manual and as previously described [9]. The Tanita BC 418 MA is a single frequency BIA device incorporating 8 electrodes. The total free FM, percentage of total FM, truncal FM, and android FM were estimated using equations specific for MO subjects that have been developed by our own group and validated against DXA [9]. Clinical and biochemical characteristics The demographic (age and gender), self-reported comorbidities (hypertension, type 2 diabetes, dyslipidemia), and the use of antidiabetic, antihypertensive, or lipid-lowering therapies were recorded. The blood pressure (BP), biochemical (lipid profile, fasting plasmatic glucose), and hormonal (insulin) parameters were measured as previously reported [11]. The MS was diagnosed according to the revised “Third Report of the Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III)” criteria [12]. The MS score was calculated as the sum of abnormal MS components in a particular patient. IR was categorized as present using the calculated Homeostasis Model of Assessment– Insulin Resistance [insulin (␮U/L) ⫻ glucose (mmol/L)/ 22.5], with the cutoff defining IR corresponding to the 80th percentile of the Homeostasis Model of Assessment–Insulin Resistance distribution in our reference normal weight population (2.94) [11]. Statistical analysis Statistical analysis was performed using SPSS, version 17.0 (SPSS, Chicago, IL). Because BC differs by gender, statistical analyses were performed separately for women and men or adjusting for gender. All continuous variables are presented as the mean ⫾ standard deviation, unless otherwise specified. Differences between groups were evaluated using the parametric test. Correlations between groups were performed using the Spearman correlation test adjusted for gender. Variables with P ⬍.05 on the ShapiroWilk test for normality were logarithmically transformed before correlation analysis. Stepwise logistic regression analysis was performed to model the relationship between the MS or IR and each of the obesity indexes, including gender and age as potential confounders. Because truncal FM and android FM correlated highly and showed co-linearity, they were not included simultaneously in our logistic regression models. Because the android FM resulted a better estimate of central adiposity than the truncal FM, the android FM was used as

650

V. Perea et al. / Surgery for Obesity and Related Diseases 9 (2013) 648 – 652

the estimate for central fat. The areas under the curve (AUCs) in the receiver operating characteristic (ROC) curve were calculated for the relationship between the different indexes of BC and IR. Results The study population included mainly women (72.9%). Of the 784 patients, 708 (90.3%) presented with a BMI ⬎40 kg/m2, 73 (9.3%) with a BMI of 35.0 –34.9 kg/m2, and 3 (.38%) with a BMI of 30 –35 kg/m2. The average age and BMI was 44.1 ⫾ 11.0 years (range 18 – 69) and 46.7 ⫾ 6.3 kg/m2 (range 33.1– 82.1), respectively. The prevalence of the MS and IR in the entire cohort was 78.6% and 88.6%, respectively. Men were more likely to be diagnosed with the MS (85.3% of men and 76.0% of women; P ⫽ .004), and presented with a greater prevalence of IR (92.2% of men and 87.3% of women; P ⫽ .015). Despite being MO, 21.4% and 11.4% of the study cohort did not present with the MS or IR, respectively. Comparison of patients with and without the MS In both genders, the patients with the MS were older than those without the MS (Table 1). As expected from the diagnostic criteria for the MS, women and men with the MS presented with greater BP, plasma triglycerides, fasting plasma glucose, and hemoglobin A1c levels and lower plasma high-density lipoprotein cholesterol than those without the MS. Women with the MS presented with a greater BMI (P ⬍.001) than women without the MS. In contrast, the

BIA-derived estimate of total adiposity (percentage of FM) did not differ between the 2 groups (P ⫽ .404). The central body fat distribution indexes were larger in the women with the MS using both the classic anthropometric indexes (WC, P ⬍.001) and the BIA-derived anthropometric indexes (android FM, P ⬍.001; truncal FM, P ⫽ .001. In men, neither the BMI (P ⫽ .302) nor the BIA-derived total percentage of FM (P ⫽ .834) was significantly different between those with and without the MS. Men with the MS presented with a larger WC (P ⬍.001); however, the BIA-derived estimates of central adiposity (android FM and truncal FM) did not differ. Differences in the BC between groups remained significant even after adjusting for age (data not shown). Comparison between patients with and without IR Of the 784 patients, 165 (21%) self-reported a previous diagnosis of type 2 diabetes (Table 1). Of these 165 patients, 36 were controlling their diabetes with diet, 118 were taking hypoglycemic agents, and 47 required insulin therapy. Accordingly, 129 diabetic patients were excluded from the analysis. Compared with those with normal insulin sensitivity (NIS), those with IR of both genders presented with a worst metabolic profile (greater fasting plasmatic glucose, men P ⫽ .015, women P ⫽ .017; greater glycated hemoglobin, men P ⫽ .049, women P ⬍.001; greater triglyceride level, P ⬍.001 for both genders). The women with IR also presented with lower levels of high-density lipoprotein cholesterol (P ⫽ .002) than those with NIS. Age and BP distribution was not significantly different between those with IR and NIS in both genders.

Table 1 Comparison between morbidly obese subjects with or without MS or IR Variable

Men

Women

MS

Age (y) BMI (kg/m2) WC (cm) Total FM (%) A-FM (kg) T-FM (kg) SBP (mm Hg) DBP (mm Hg) HDL-C (mg/dL) Triglycerides (mg/dL) FPG (mg/dL)

IR

No (n ⫽ 31)

Yes (n ⫽ 181)

No (n ⫽ 13)

34.5 ⫾ 1.6 46.0 ⫾ .9 135.2 ⫾ 1.9 46.2 ⫾ .8 7.2 ⫾ 1.1 37.4 ⫾ 6.7 126.7 ⫾ 3.1 76.9 ⫾ 1.8 42 ⫾ 1.7 102 ⫾ 4.3

45.2 ⫾ .8* 47.3 ⫾ .5 141.6 ⫾ 1.0* 46.0 ⫾ .4 7.6 ⫾ 1.5 39.6 ⫾ 8.7 136.8 ⫾ 1.6* 82.8 ⫾ 1.3* 39 ⫾ .6* 192 ⫾ 10.4*

43.1 ⫾ 3.7 45.2 ⫾ 1.1 131.9 ⫾ 2.6 45.6 ⫾ .9 7.0 ⫾ 1.2 36.5 ⫾ 7.0 124.7 ⫾ 3.5 77.1 ⫾ 1.8 41.0 ⫾ 3.0 104.1 ⫾ 9.2

117.0 ⫾ 2.6*

94.2 ⫾ 4.3

94 ⫾ 1.3

MS Yes (n ⫽ 154)

IR

No (n ⫽ 137)

Yes (n ⫽ 435)

No (n ⫽ 62)

Yes (n ⫽ 426)

40.1 ⫾ .8 48.3 ⫾ .5 142.4 ⫾ 1.1† 46.7 ⫾ .4 7.6 ⫾ 1.5 39.6 ⫾ 8.7 134.7 ⫾ 1.7 82.3 ⫾ 1.3 39.7 ⫾ .6 154.7 ⫾ 5.7†

39.5 ⫾ 1.0 45.1 ⫾ .4 123.2 ⫾ 1.1 52.1 ⫾ .4 6.1 ⫾ 1.2 34.1 ⫾ 7.1 120.9 ⫾ 1.0 74.2 ⫾ .7 55 ⫾. 9 93 ⫾ 2.5

45.2 ⫾ .8* 47.3 ⫾ .5* 141.6 ⫾ 1.0* 46.0 ⫾ .4 6.8 ⫾ 1.5* 37.2 ⫾ 8.7* 136.8 ⫾ 1.6* 82.8 ⫾ 1.3* 39 ⫾ .6* 192 ⫾ 10.4*

43.5 ⫾ .5 43.5 ⫾ .5 120.4 ⫾ 1.8 55.9 ⫾ .8 5.5 ⫾ 1.2 30.2 ⫾ 6.5 125.9 ⫾ 1.0 77.5 ⫾ .7 53.3 ⫾ .9 98.4 ⫾ 4.7

42.7 ⫾ .5 47.1 ⫾ .3† 128.1 ⫾ .6† 58.3 ⫾ .3† 6.8 ⫾ 1.5† 37.3 ⫾ 8.3† 129.6 ⫾ .8 79.9 ⫾ .6 48.3 ⫾ .5† 132.5 ⫾ 3.3†

112.8.0 ⫾ 2.2†

101 ⫾ 8.0

117.0 ⫾ 2.6*

93.8 ⫾ 1.4

109.9 ⫾ 2.5†

MS ⫽ metabolic syndrome; IR ⫽ insulin resistance; BMI ⫽ body mass index; WC ⫽ waist circumference; FM ⫽ fat mass; A-FM ⫽ android FM; T-FM ⫽ truncal FM; SBP ⫽ systolic blood pressure; DBP ⫽ diastolic blood pressure; HDL-C ⫽ high-density lipoprotein cholesterol; FPG ⫽ fasting plasmatic glucose. Data expressed as mean ⫾ standard error. * P ⬍0.05, for comparison of subjects with or without IR within same gender. † P ⬍0.05 for comparison of subjects with or without MS within the same gender.

BC and metabolic abnormalities in MO / Surgery for Obesity and Related Diseases 9 (2013) 648 – 652

Women with IR presented with a greater BMI (P ⬍.001) and larger total FM (percentage of FM, P ⬍.001). Moreover, all variables describing the central fat distribution were significantly greater in the IR group (WC, P ⬍.001; android FM, P ⫽ .001; and truncal FM, P ⫽ .002). The BMI and total percentage of FM did not differ between the men with IR and those with NIS. The WC was larger in the men with IR than in the men with NIS (P ⫽ .007). In contrast, the android FM and truncal FM did not differ between the 2 groups. Differences in BC between the groups remained significant even after adjusting for age (data not shown). Relationship among BC indexes, cardiometabolic risk factors, and NIS The correlation coefficients between the anthropometric and BIA-derived indexes of BC and the metabolic risk factors are listed in Table 2. A significant correlation was found among the BMI, BP, and the MS score. The BIAderived total percentage of FM significantly correlated only with the diastolic BP. Neither the BMI nor the total percentage of FM correlated significantly with the lipid parameters or fasting plasmatic glucose level. Of the markers of central adiposity, the WC showed the strongest association with the cardiometabolic risk factors. The WC correlated significantly with all the components of the MS (BP, P ⬍.001; fasting plasma glucose, P ⫽ .047; triglycerides, P ⫽ .002; and high-density lipoprotein cholesterol, P ⫽ .008) and the MS score (P ⬍.001). Of the BIA-derived indexes of central adiposity, the android FM and truncal FM were positively correlated with systolic and diastolic BP (P ⬍.001). Android FM also correlated with the MS score (P ⫽ .006). Nonetheless, at variance with what was observed with the WC, we did not find a significant correlation between the BIA-derived central adiposity estimates and either the lipid profile or fasting plasma glucose level.

Table 2 Pearson’s partial correlation between components of metabolic syndrome and anthropometric- and BIA-derived indexes of BC adjusted by gender WC (cm) SBP (mm Hg) DBP (mm Hg) FPG (mg/dL) Triglycerides (mg/dL) HDL-C (mg/dL) MS score

BMI (kg/m2)

A-FM (kg)

T-FM (kg)

Total FM (%)

.270* .253* .090* .141*

.206* .241* ⫺.069 .045

.206* .234* ⫺.053 .028

.203* .230* ⫺.065 .009

.082 .136* ⫺.139 ⫺.056

⫺.119* .221*

⫺.072 .095*

⫺.074 .091*

⫺.059 .072

⫺.013 ⫺.027

BIA ⫽ bioelectrical impedance analysis; BC ⫽ body composition; WC ⫽ waist circumference; BMI ⫽ body mass index; A-FM ⫽ android FM; T-FM ⫽ truncal FM; FM ⫽ total fat mass; SBP ⫽ systolic blood pressure; DBP ⫽ diastolic blood pressure; FPG ⫽ fasting plasma glucose; HDL-C ⫽ high-density lipoprotein cholesterol; MS ⫽ metabolic syndrome. * P ⬍.05.

651

Table 3 Area under ROC curve for anthropometric- and BIA-derived indexes of BC in diagnosis of the MS and IR Variable

MS

2

BMI (kg/m ) Total FM (%) WC (cm) A-FM (kg) T-FM (kg)

IR

AUC 95% CI

P Value

AUC 95% CI

P Value

.596 .505 .681 .614 .604

.002 NS ⬍.001 ⬍.001 .001

.687 .619 .753 .731 .724

⬍.001 .003 ⬍.001 ⬍.001 ⬍.001

.573–.656 .442–.568 .624–.738 .556–.671 .546–.662

.619–.754 .553–.685 .680–.825 .661–.801 .658–.790

ROC ⫽ receiver operating characteristic; BIA ⫽ bioelectrical impedance analysis; BC ⫽ body composition; MS ⫽ metabolic syndrome; IR ⫽ insulin resistance; AUC ⫽ area under curve; CI ⫽ confidence interval; BMI ⫽ body mass index; FM: fat mass; WC ⫽ waist circumference; A-FM ⫽ android FM; T-FM ⫽ truncal FM.

Stepwise logistic regression analysis was used to evaluate the relative contribution of the total body fat estimates and central body fat estimates to the presence of the MS and IR. When the MS and IR were considered as dependent variables, the WC emerged as the only significant predictor (both, P ⬍.001 and P ⬍.001). In contrast, the BMI and BIA-derived total percentage of FM and android FM was not. Finally, the accuracy of the anthropometric and BIAderived indexes of BC in identifying subjects with the MS or IR was evaluated using ROC analysis (Table 3). As a whole, the accuracy for all the evaluated BC parameters (BMI, WC, total FM, android FM, truncal FM) was low. Nonetheless, the WC was the index associated with greatest AUC of the ROC curves for the MS and IR alike (MS, AUCWC ⫽ .681, P ⬍.001; IR, AUCWC ⫽ .753, P ⬍.001). Discussion Our data have shown that in MO subjects, the BIAderived indexes and anthropometric measurements of total and central adiposity tissue are associated with the MS and IR. However, the BIA-derived indexes of BC were not more closely associated with the MS and IR than classic anthropometric measurements. The prevalence of the MS and IR in our cohort was comparable to that previously reported in populations with a similar BMI distribution [3]. However, although the patients were MO, 14.7% and 7.8% of the men and 24.0% and 12.7% of the women did not present with the MS or IR, respectively. The anthropometric estimates of total body fat (BMI) and central body fat distribution (WC) were significantly different in women with or without the MS or IR. In contrast, a larger WC was the only anthropometric estimate significantly associated with an altered metabolic phenotype in men. Admittedly, the low number of men without the MS or IR in our series might partly account for the gender differences. However, the WC correlated significantly with all the components of the MS and the MS score, but the

652

V. Perea et al. / Surgery for Obesity and Related Diseases 9 (2013) 648 – 652

BMI was associated only with the BP. Furthermore, the WC outperformed the BMI for the identification of the MS and IR in the ROC analysis. Thus, our data are consistent with those from previous studies emphasizing the role of the central body fat as a determinant of cardiometabolic health in MO subjects [5]. All subjects in our study presented with a WC well above the gender-specific cutoffs defining an enlarged WC. Thus, it could be argued that the cutoff for the WC to define a population at cardiometabolic risk should be adjusted for the BMI [13]. Our study failed to demonstrate that the BIA-derived indexes of total or central body fat distribution are better than the corresponding anthropometric measurements in identifying MO subjects with an increased cardiometabolic risk. Although the BIA-derived indexes of central body fat distribution were greater in women with metabolic abnormalities, the association between the MS components and the central body fat distribution BIA-derived estimates was not better than that with the WC. It could be argued that the BIA-derived estimates of total and central body fat distribution are inaccurate for the BC analysis in MO subjects [7,8]. However, we overcame this limitation by using BIA equations that had been developed by our group specifically for the MO population and validated using DXA as the reference method [9]. In our study, we compared 2 approaches to BC analysis suitable for use in routine clinical practice. Results similar to those we have reported have been published comparing anthropometric measures with other methods of BC analysis, such as air displacement plethysmography and DXA [14 –16]. Using the percentage of FM measurement by air displacement plethysmography as the reference method, Bosy-Westhpal et al. [15] found that the WC was better for identifying those with the MS or IR. Similarly, applying ROC analysis, Soto-González et al. [16] found that DXA was not better than the anthropometric indexes for the identification of the MS in overweight or obese subjects. In contrast, changes in visceral adipose tissue as assessed using computed tomography during a 6-year period have been associated with changes in cardiometabolic risk factors [17]. However, cost, access, and radiation exposure make computed tomography unsuitable for routine use for the identification of subjects at increased metabolic risk [18]. Conclusion Our data have shown that in MO subjects, the BIAderived indexes of total and central adiposity are not better predictors of the MS or IR compared with the traditional anthropometric measurements. Although the WC was the factor more closely associated with increased cardiometabolic risk, its limited sensitivity in the ROC analysis suggests that other parameters of body fat distribution or function might help explain the association between morbid obesity and cardiometabolic disturbances.

Disclosures The authors have no commercial associations that might be a conflict of interest in relation to this article. References [1] Chan JM, Rimm EB, Colditz GA, Stampfer MJ, Willett WC. Obesity, fat distribution, and weight gain as risk factors for clinical diabetes in men. Diabetes Care 1994;17:961–9. [2] Rexrode KM, Hennekens CH, Willett WC, et al. A prospective study of body mass index, weight change and risk of stroke in women. JAMA 1997;21:1539 – 45. [3] Residori L, García-Lorda P, Flancbaum L, Pi-Sunyer FX, Laferrère B. Prevalence of co-morbidities in obese patients before bariatric surgery: effect of race. Obes Surg 2003;13:333– 40. [4] Soverini V, Moscatiello S, Villanova N, Ragni E, Di Domizio S, Marcheseni G. Metabolic syndrome and insulin resistance in subjects with morbid obesity. Obes Surg 2010;20:295–301. [5] Ledoux S, Coupaye M, Essig M, et al. Traditional anthropometric parameters still predict metabolic disorders in women with severe obesity. Obesity (Silver Spring) 2010;18:1026 –32. [6] Dulloo AG, Jacquet J, Solinas G, Montani J-P, Schutz Y. Body composition phenotypes in pathways to obesity and metabolic syndrome. Int J Obes 2010;34:S4 –S17. [7] Neovius M, Hemmingson E, Freyschuss B, Udde J. Bioelectrical impedance underestimates total and truncal fatness in abdominally obese women. Obesity 2006;14:1731– 8. [8] Peteyjohns IR, Brinkworth GD, Buckley JD, Noakes M, Clifton PM. Comparison of three BIA methods with DXA in overweight and obese men. Obesity 2006;14:2064 –70. [9] Jiménez A, Omaña W, Flores L, et al. Prediction of whole body and segmental body composition by bioelectrical impedance in morbidly obese subjects. Obes Surg 2012;22:587–93. [10] Wallace TM, Levy JC, Matthews DR. Use and abuse of HOMA modeling. Diabetes Care 2004;27:1487–95. [11] Vidal J, Morinigo R, Codoceo VH, Casamitjana R, Pellitero S, Gomis R. The importance of diagnostic criteria in the association between metabolic syndrome and cardiovascular disease in obese subjects. Int J Obes 2005;29:668 –74. [12] Grundy SM, Cleeman JI, Daniels SR, et al. Diagnosis and management of the metabolic syndrome: an American Heart Association/ National Heart, Lung, and Blood Institute scientific statement. Circulation 2005;112:2735–52. [13] Ardern CL, Janssen I, Ross R, Katzmarzyk PT. Development of health-related waist circumference thresholds within BMI categories. Obes Res 2004;12:1094 –103. [14] Sun Q, Van Dam RM, Spiegelman D, Heymsfield SB, Willett WC, Hu FB. Comparison of dual-energy x-ray absorptiometric and anthropometric measures of adiposity in relation to adiposity-related biologic factors. Am J Epidemiol 2010;172:1442–54. [15] Bosy-Westhpal A, Geisler C, Onur S, et al. Value of body fat mass vs anthropometric obesity indices in the assessment of metabolic risk factors. Int J Obes 2006;30:475– 83. [16] Soto-González A, Bellido D, Buño MM, et al. Predictors of the metabolic syndrome and correlation with computed axial tomography. Nutrition 2007;37:36 – 45. [17] Rhéaume C, Arsenault BJ, Dumas MP, et al. Contributions of cardiorespiratory fitness and visceral adiposity to six-year changes in cardiometabolic risk markers in apparently healthy men and women. J Clin Endocrinol Metab 2011;96:1462– 8. [18] Evans J, Micklesfield L, Jennings C, et al. Diagnostic ability of obesity measures to identify metabolic risk factors in South African women. Metab Syndr Relat Disord 2011;9:353– 60.