¹Department of Community Medicine, Mercer University School of Medicine, 1550 College Street, Macon, GA, USA

²Department of Health and Kinesiology, College of Health and Professional Studies, Georgia Southern University, Statesboro, GA, USA

Correspondence to: I S Okosun, Department of Community Medicine, Mercer University School of Medicine, 1550 College Street, Macon, GA 31207, USA.okosun_i@mercer.edu

PURPOSE: To determine whether white, black and hispanic young (17-39 y) and middle-aged (40-59 y) adults, and elderly (60-90 y) Americans have the same values of abdominal adiposity (estimated from waist circumference (WC) at the established levels of overweight (body mass index, BMI 25-29.9 kg/m²) and obesity (BMI30 kg/m²).

METHODS: Data (n=16,120) from the US Third National Health and Nutrition Survey were utilized. Age-adjusted linear regression analyses were used to estimate gender- and ethnic-specific WC values corresponding to overweight and obesity. Receiver operating characteristic (ROC) curves were also employed to determine the choices of WC values corresponding to the established BMI cut-off points. With ROC, gender- and ethnic-specific cut-off points producing the best combination of sensitivity and specificity were selected as optimal thresholds for WC values corresponding to the established BMI cut-off points.

RESULTS: WC values associated with the established BMI were lower in blacks and hispanics compared with whites. In men, the WC values that corresponded to overweight ranged from 89 to 106 cm, from 84 to 95 cm, and from 87 to 97 cm in whites, blacks and hispanics, respectively. The corresponding values for obesity ranged from 99 to 110 cm, from 96 to 107 cm, and from 97 to 108 cm. The WC values that corresponded to overweight in women ranged from 82 to 91 cm, from 81 in to 90 cm, and from 83 to 92 cm in whites, blacks and hispanics, respectively. The analogous values for obesity ranged from 94 to 101 cm, from 93 to 100 cm, and from 94 to 101 cm.

CONCLUSIONS: The lack of higher WC values in blacks (particularly women) and hispanics at the same levels of BMI for whites challenges previously held assumptions regarding the role of abdominal adiposity in cardiovascular disease experienced by non-whites. Defining the anthropometric variables that satisfactorily describe reasons for ethnic differences in cardiovascular disease is one of the challenges for future research.

International Journal of Obesity (2000) 24, 1279-1285

abdominal obesity; BMI; NHANES III; ROC; waist circumference

Introduction

The most widely used anthropometric measure of total and overall body adiposity is the body mass index (BMI). BMI is a height-adjusted weight that is associated with many cardiovascular diseases (CVD) including stroke, non-insulin dependent (type 2) diabetes mellitus and hypertension.^1,2,3 High BMI has also been shown to be associated with all-cause mortality in different population groups.^4,5 However, many studies report that abdominal or central obesity, which represents an aberrant fat accumulation, is equally as important as BMI, and may be a more potent adiposity phenotype and a better predictor for CVD or mortality in some population groups.^6,7

First described as gynoid and android body habitus some 40 years ago,⁸ the mechanism of abdominal fat distribution in CVD is not clear. Two possible mechanisms have been suggested for the mode of action of abdominal adiposity. First, through the highly sensitive lipolytic and hypertrophied visceral adipose tissues, elevated free fatty acids may induce insulin resistance.^9,10 Second, elevated free testosterone and reduced sex-hormone-binding globulin may promote increased abdominal adiposity and reduce fractional hepatic extraction of insulin.¹¹ Thus, the connection of adiposity with CVD may be due to enlarged visceral fat depots discharging free fatty acids into the portal and systemic circulation.^9,10,11

Adequate assessment of abdominal adiposity requires imaging techniques, such as computed tomography and magnetic resonance.^12,13 Due to the high costs and labor requirement, and possible risk of radiation, imaging techniques are often limited to laboratory settings. Hence, anthropometric surrogates are commonly employed. Waist circumference (WC) has been endorsed as the best anthropometric surrogate of abdominal adiposity.^12,14 WC is an aggregate measurement of the actual amount of total and abdominal fat accumulation and is a crucial correlate of abnormal metabolic syndromes found among obese and overweight subjects.^14,15 WC measurement is simple, requires only a flexible tape and, with appropriate training, measurement error is low due to large circumference. Indeed, many are now advocating WC as a valid anthropometric variable for health promotion and the basis for alerting those at risk of CVD.^16,17,18,19

WC is highly correlated with visceral adipose tissue accumulation.^20,21,22 Visceral adiposity is the component of body composition that is most highly associated with many metabolic abnormalities such as hypertension, glucose intolerance, hyperinsulinemia, hypercholesterolemia, hypertriglyceridemia, and high levels of low-density lipoprotein cholesterol.^{1,23,24,25,26,27}

Despite numerous reports stating that the higher risks of CVD in non-whites compared with whites may be explained by their difference in adiposity phenotypes, only little information is available on the relationship between overall adiposity and abdominal adiposity in white, black and hispanic Americans. In this research, we took advantage of the third US National Health and Nutrition Examination Survey (NHANES III) to examine whether these ethnic groups have the same levels of abdominal fat accumulation at the established BMI cut-off points for overall body overweight and obesity. BMI of 25-29.9 kg/m² (overweight) and 30 kg/m² (obesity) have been endorsed my many international agencies as the levels of adiposity for increased risk of obesity related co-morbidities.^19,28

Methods

Study population

NHANES III was a multistage probability sample of non-institutionalized civilian US population groups examined between 1988 and 1994. The sampling and measurement procedures have been extensively described elsewhere.^29,30 Only subjects identified as non-hispanic white, non-hispanic black and hispanic Americans aged 17-90 y were eligible for this investigation. This study was also restricted to individuals for whom anthropometric measurements were available, including weight, height, hip, and waist.

Anthropometric measurements

Weight was measured in the upright position using a digital scale. Waist measurement was made to the nearest 0.1 cm at midpoint between the bottom of the rib cage and above the top of the iliac crest. Height was measured in meters to the nearest 0.1 cm using a stadiometer against a vertical wall with a rigid headboard using an inelastic tape measure. A description of measurement precision between technicians has been given elsewhere.³¹ BMI was calculated as the measured weight in kilograms divided by height in meters squared (kg/m²).

Statistical methods

Statistical programs available in SPSS³² (version 9.0) and SIMSTAT³³ (version 1.24) for Windows were utilized for these analyses. Means and standard deviations of age and anthropometric variables were computed and compared within gender and across ethnic group. Pearson's correlation analysis was utilized to quantify the univariate relationship between BMI and WC. The prevalence of abdominal adiposity was age-adjusted by direct methods using the 1990 US population census data. Linear regression analyses were performed to determine gender- and ethnic-specific WC values corresponding to BMI of 25-29.9 kg/m² and 30 kg/m² in young (17-39 y) and middle-aged (40-59 y) adults, and elderly (60-90 y) Americans. Further adjustment for age was made using gender and ethnic specific median value of age. In the linear regression, the full model was specified as:

BMI=₀+&beta₁(age)+β₂(black)+₃(hispanic)

+₄(waist)+₅(black´waist)

+₆(Hispanic´waist)

where is the intercept, and black´waist and hispanic´waist are interaction terms for black with WC, and hispanic with WC, respectively, created from dummy variables in which the WC values for non-whites were compared with whites. Hence, WC values corresponding to each established BMI cut-points were calculated as:

WC Value=

Gender- and ethnic-specific receiver operating characteristic (ROC) curves were also evaluated to determine WC values corresponding to BMI cut-off points.^34,35 With the ROC technique, comparison of sensitivity with the false-positive rate for BMI cut-off point was made over the entire range of WC values. WC value corresponding to an established BMI cut-off point was determined by interpolation from the point of intersect of the lines of specificity and sensitivity.³⁴ The point of intersect between lines of specificity and sensitivity represent the highest numbers of subjects with and without overweight or obesity.

The areas under each curve (AUCs) were calculated and bootstrapping procedures were used to estimate the 95% confidence intervals and to determine whether ROC curves were similar across ethnic groups.³⁵ The plausible values of AUC range between 0 and 1. AUC values of 0.5,<0.5 and>0.5 represent random, worse than random, and better than random, respectively, while the value of 1 indicate a perfect performance.³⁵

Results

The anthropometric characteristics of the participants are presented in Table 1. Overall, 16 120 subjects (6772 non-hispanic whites, 4720 non-hispanic blacks and 4628 hispanic Americans) were eligible for this investigation. There were significant ethnic differences for some variables, with whites tending to be older than their black and hispanic men counterparts. White men had greater waist girth than black and hispanic men (P<0.01). Among women, blacks had significantly higher mean weight, BMI and WC values than whites or hispanic counterparts (P<0.01).

Age-adjusted distribution of elevated WC by gender and ethnicity is presented in Figure 1. Overall, the prevalence of abdominal adiposity, defined as WC102 cm for men and 88 cm for women was 34.1%, 21.8% and 25.2% for white, black and hispanic American men, respectively. The corresponding values for white, black and hispanic American women were 50%, 57.8% and 57.4%, respectively. WC102 cm for men and 88 cm for women have recently been recommended by the National Institute of Health (NIH) expert panels for identifying increased relative risk for obesity co-morbidities.¹⁹

Age-adjusted Pearson's correlation analyses were performed to determine the degree of linear association of BMI with WC. BMI, as expected, very highly correlated with waist girth in men and women (r0.87; all P<0.01). Among men, BMI was more strongly correlated with WC in blacks (r=0.94) than whites (r=0.88) and hispanics (r=0.87) BMI was also more highly correlated with WC in black women (r=0.92) compared with their white (r=0.89) and hispanic (r=0.88) counterparts. The explained variances obtained from linear regression models for the prediction of WC from BMI were similar to the results of correlation analyses. The overall explained variances were high (r ²0.82) and tended to decrease with age in the three ethnic groups (data not shown). A similar stronger r ² value for BMI with WC was also observed in blacks than in whites or hispanics.

Linear regression-derived age-adjusted WC values corresponding to the established BMI cut-off points are shown in Table 2. Overall, with the exception of young women, white men and white women had consistently higher WC values at a given level of overall adiposity than their black and hispanic counterparts. Also, the estimated WC corresponding to overweight and obesity tended to increase with age. The model for men and women explained 87% and 83% of the variances in WC, respectively.

Table 2 summarizes the findings of linear regression analyses of WC values corresponding to BMI of 25-29.9 kg/m² and 30 kg/m² in young and middle-aged adults, and elderly. In the young adult men, WC values corresponding to BMI of 25 kg/m² ranged from 85 cm in blacks to 89 cm in whites. In the middle-aged adult men, the WC values ranged from 88 cm in blacks to 92 cm in whites, while in the elderly the values ranged from 92 cm in blacks to 106 cm in whites. The corresponding values for obesity in young adult men ranged from 99 cm in blacks to 103 cm in whites, from 103 cm in black to 106 cm in whites middle-aged and from 107 cm in black to 110 cm in white elderly. Among women, WC values corresponding to overweight ranged from 81 cm in black to 83 cm in hispanic young adults. Among the middle-aged adult women, the WC ranged from 83 cm in blacks to 86 cm in whites, while in the elderly the values ranged from 87 cm in blacks to 89 cm in whites. The analogous values for obesity in young adult women ranged from 93 cm in black to 96 cm in whites, from 95 cm in blacks to 98 cm in middle-aged whites and hispanics, and from 98 cm in black to 101 cm in elderly whites.

Table 3 presents the findings of ROC analyses of WC for BMI that jointly provided optimal sensitivity and specificity for overweight and obesity. In the young adult men, WC values that maximized sensitivity and specificity for overweight ranged from 84 cm in blacks to 89 cm in whites. In the middle-aged adult men, the WC values ranged from 91 cm in blacks to 95 cm in whites, while in the elderly the values ranged from 95 cm in blacks to 98 cm in whites. The analogous values for obesity in young adult men ranged from 96 cm in blacks to 99 cm in whites, from 102 cm in black to 105 cm in white middle-aged and from 103 cm in both black and hispanic to 105 cm in white elderly. Among women, WC values of 83 cm corresponded to overweight in young adults of the three ethnic groups. Among the middle-aged adult women, the WC ranged from 89 cm in blacks and whites to 90 cm in the hispanics, while in the elderly the values ranged from 90 cm in blacks to 92 cm in hispanics. WC values of 94 and 97 cm were respectively associated with obesity in young and middle-aged adult women of the three ethnic groups. In the elderly women, WC optimizing sensitivity and specificity for obesity ranged from 100 cm in blacks and whites to 101 cm in the hispanics.

The overall performance of the ROC curves for the established BMI cut-points was quantified by computing gender- and age-specific AUCs (Table 4). Overall, WC tended to perform better for obesity compared to underweight in all ethnic groups. In all ethnic groups, WC performed better than what was expected by random chance in classifying subjects to their true overall adiposity status (AUC>0.50). Also, WC tended to perform better for the established BMI cut-off points in young adults compared to middle-aged adults and the elderly.

We combined men and women separately and determined gender-specific WC values for overweight and obesity. As shown in Table 5, WC values of 91 and 86 cm corresponded to overweight in men and women, respectively, using linear regression models. WC values of 103 and 97 cm corresponded to obesity in men and women, respectively, using linear regression models. The corresponding values using ROC techniques were 92 cm (sensitivity 85%, specificity 86%) and 87 cm (sensitivity 87%, specificity 86%) for overweight, and 101 cm (sensitivity 90%, specificity 85%) and 97 cm (sensitivity 86%, specificity 88%) for obesity in men and women, respectively.

Gender- and ethnic-specific ROC curves of WC for obesity are shown in Figure 2. The ROC curves visually represent the relationship between sensitivity (true positive rate) and 1-specificity (false-positive rate) for each of the three ethnic groups over the entire range of WC. Like the observed curves for overweight (figure not shown), the AUCs associated with obesity in each ethnic group were significantly higher than what was expected by chance alone (P<0.01). Overall there were no statistically significant ethnic differences in the AUCs for males and for females of the three groups, indicating that WC performed reasonably well in identifying overweight and obese subjects in the three groups.

Discussion

Although a large body of epidemiological data suggests BMI to be as an important obesity phenotype in CVD, many investigators think aberrant fat localization, as seen in central or abdominal body regions, may be equally as potent as BMI for CVD.^{12,13,14,15,16,36} As suggested by some investigators, blacks have higher upper-body adiposity, based on subscapular skinfold thickness and waist-to-hip ratio (WHR) compared to whites,³⁶ hence blacks are more likely to have larger visceral fat deposits.³⁷ Some studies conclude that blacks propensity to metabolic aberrations may be a result of more excess visceral adiposity compared with whites. Others have postulated that the propensity for CVD among blacks may be due to the stronger correlation between abdominal adiposity and visceral adipose tissue in blacks than in whites.³⁷

Many studies investigating the association between metabolic abnormalities have used WHR as a surrogate measure of abdominal adiposity because of some assertions that WHR show a more potent association with CVD in some populations.³⁸ Indeed, WHR is associated with risk of CVD in several populations, independent of BMI, where WC does not remain independent. WC has been advocated in the guidelines for overweight and obese individuals;¹⁹ thus, in this group WHR would be a reasonable measure. Since obese and non obese subjects can have same WHR, it may reflect the fact that WHR can capture relative fat distribution more independently of BMI or total obesity. However, in this analysis we chose to use only WC to assess abdominal adiposity because of the inherent weakness of WHR as a ratio index,^39,40 and because it is strongly influenced by pelvic structure.⁴¹

Consistent with previous findings, the results of this investigation showed BMI to be more strongly correlated with WC in blacks than whites and hispanics.³⁷ Despite black men having similar mean weight and BMI as white men, and black women having significantly higher mean values of weight and BMI compared with white women, black men and black women were found to have lower WC values at given levels of overweight and obesity. hispanic men and women were also found to have significantly lower WC values than their white counterparts at the given BMI of 25-29.9 kg/m² and 30 kg/m². Our finding in relation to small WC girth in blacks is consistent with studies by Conway et al,³⁷ Albu et al ⁴² and Dowling and Pi-Sunyer,⁴³ who also found blacks to have less visceral adipose tissues than whites at similar BMI values: our result is also consistent with studies by Stevens et al,⁴⁴ who found the ratio of abdomen to midarm circumference was larger in white women than black women.

In a study by Lean et al,¹⁸ WC94 cm for men and 80 cm for women were found to correspond to a BMI of 25 kg/m² or more, while WC102 cm for men and 88 cm for women were found to correspond to a BMI of 30 kg/m² or greater. Although it is generally known that for a given waist girth, there is selective accumulation of visceral tissue with age, Lean et al ¹⁸ did not consider age in the determination of WC values corresponding to obesity. In addition, Lean et al ⁴⁵ included the criterion of high WHR (WHR>0.95 in men and >0.85 in women) to estimate WC values corresponding to these BMI cut-off points. Indeed, these WC values were designed for health promotion to give optimal enrichment to individuals in need of weight management because of overweight/obesity or because of central fat distribution. It is therefore not surprising that the WC values from our study are different. The studies by Lean and colleagues¹⁸ were also restricted to white populations from The Netherlands and Canada, and generalizability of the results is highly controversial.

The data presented for this analysis have the advantage of being large and national in scope, making the results more generalizable to the overall US population. The training program and quality control measures instituted in NHANES III give an added credence to the data. The observed differences in WC values corresponding to the established BMI cut-points in this study between regression and ROC methods were not surprising. With the regression method, WC is assumed to be a linear function of BMI, where as ROC procedures are based on the best combination of sensitivity and specificity.

Our results showed that the regression and ROC models fitted the data reasonably well, as indicated by r ²82%, and sensitivity and specificity 83%, respectively. However, the decision on what cut-off points to use must not be based on model performance alone. Other additional factors, such as the overall impact on health due to intervention (eg weight reduction), potential burden on health services if a low cut-off point is employed, and the benefit of alerting people developing a weight problem, must be considered in determining what WC value to recommend.

Conclusion

In summary, the lack of higher values of WC at similar BMI cut-off points in non-whites compared with whites defies previously held notions regarding the role of abdominal adiposity in CVD experienced by non-whites. Future research should include determining whether other anomalous fat distributions or patterning are more relevant to ethnic differences in CVD beyond those that are provided by traditionally used anthropometic variables such as BMI and WC.

Acknowledgements

Data from the NHANES III were obtained from the US National Center for Health Statistics.

1 Hirose H, Saito I, Tsujioka M, Mori M, Kawabe H, Saruta T. The obese gene product, leptin: possible role in obesity-related hypertension in adolescents. J Hypertens 1998; 16: 2007-2012. MEDLINE

2 Kaufman JS, Asuzu MC, Mufunda J, Forrester T, Wilks R, Luke A, Long AE, Cooper RS. Relationship between blood pressure and body mass index in lean populations. Hypertension 1997; 30: 1511-1516. MEDLINE

3 Forrester T, Wilks R, Bennett F, McFarlane-Anderson N, McGee D, Cooper R, Fraser H. Obesity in the Caribbean. Ciba Found Symp 1996; 201: 17-26 [Discussion 26-31, 32-36.], MEDLINE

4 Seccareccia F, Lanti M, Menotti A, Scanga M. Role of body mass index in the prediction of all cause mortality in over 62,000 men and women. The Italian RIFLE Pooling Project. Risk factor and life expectancy. J Epidemiol Community Health 1998; 52: 20-26. MEDLINE

5 Seidell JC, Verschuren WM, van Leer EM, Kromhout D. Overweight, underweight, and mortality. A prospective study of 48, 287 men and women. Arch Intern Med 1996; 156: 958-963. MEDLINE

6 Prineas RJ, Folsom AR, Kaye SA. Central adiposity and increased risk of coronary artery disease mortality in older women. Ann Epidemiol 1993; 3: 35-41. MEDLINE

7 Casassus P, Fontbonne A, Thibult N, Ducimetiere P, Richard JL, Claude JR, Warnet JM, Rosselin G, Eschwege E. Upper-body fat distribution: a hyperinsulinemia-independent predictor of coronary heart disease mortality. The Paris Prospective Study. Arterioscler Thromb 1992; 12: 1387-1392. MEDLINE

8 Vague J. The degree of masculine differentiation of obesities: a factor determining predisposition to diabetes, atherosclerosis, gout, and uric calculous disease. 1956. J Clin Nutr 1956; 4: 20-34.

9 Salans LB, Knittle JL, Hirsch J. The role of adipose cell size and adipose tissue insulin sensitivity in the carbohydrate intolerance of human obesity. J Clin Invest 1968; 47: 153-165.

10 Bjorntorp P. Fatty acids, hyperinsulinemia, and insulin resistance: which comes first? Curr Opin Lipid 1994; 5: 166-174.

11 Evans DJ, Murray R, Kissebah AH. Relationship between skeletal muscle insulin resistance, insulin-mediated glucose disposal, and insulin binding: effects of obesity and body fat topography. J Clin Invest 1984; 74: 1515-1525. MEDLINE

12 Ross R, Shaw KD, Martel Y, de Guise J, Avruch L. Adipose tissue distribution measured by magnetic resonance imaging in obese women. Am J Clin Nutr 1993; 57: 470-475. MEDLINE

13 Seidell JC, Bjorntorp P, Sjostrom L, Sannerstedt R, Krotkiewski M, Kvist H. Regional distribution of muscle and fat mess in men¾new insight into the risk of abdominal obesity using computed tomography. Int J Obes 1989; 13: 289-303. MEDLINE

14 Rexrode KM, Carey VJ, Hennekens CH, Walters EE, Colditz GA, Stampfer MJ, Willett WC, Manson JE. Abdominal adiposity and coronary heart disease in women. JAMA 1998; 280: 1843-1848. MEDLINE

15 Kissebah AH, Vydelingum N, Murray R, Evans DJ, Hartz AJ, Kalkhoff RK. Relation of body fat distribution to metabolic complications of obesity. J Clin Endocrinol Metab 1982; 54: 254-260. MEDLINE

16 Han TS, Kelly IE, Walsh K, Greene RM, Lean ME. Relationship between volumes and areas from single transverse scans of intra-abdominal fat measured by magnetic resonance imaging. Int J Obes Relat Metab Disord 1997; 21: 1161-1166. MEDLINE

17 Hans TS, Feskens EJ, Lean ME, Seidell JC. Associations of body composition with Type 2 diabetes mellitus. Diabetic Med 1998; 15: 129-135. MEDLINE

18 Lean ME, Han TS, Morrison CE. Waist circumference as a measure for indicating need for weight management. BMJ 1995; 311: 158-161. MEDLINE

19 National Institute of Health/National Heart Lungs and Blood Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults. The evidence report. Obes Res 1998; 6: 51S-209S. MEDLINE

20 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: 685-693. MEDLINE

21 Han TS, McNeill G, Baras P, Forester MA. Waist circumference predicts intra-abdominal fat better than waist∶hip ratio in women. Proc Nutr Soc 1995; 54: 152A (Abstr.).

22 Lemieux S, Prud'homme D, Tremblay A, Bouchard C, Despres JP. Anthropometric correlates to changes in visceral adipose tissue over 7 y in women. Int J Obes Relat Metab Disord 1996; 20: 618-624. MEDLINE

23 Haffner SM, Valdez RA, Hazed HP, Mitchell BD, Morales PA, Stern MP. Prospective analysis of the insulin-resistance syndrome (syndrome X). Diabetes 1992; 41: 715-722. MEDLINE

24 Gus M, Moreira LB, Pimentel M, Gleisener AL, Moraes RS, Fuchs FD. Association of various measurements of obesity and the prevalence of hypertension. Arq Bras Cardiol 1998; 2: 111-114.

25 Sattar N, Tan CE, Han TS, Forster L, Lean ME, Shepherd J, Packard CJ. Associations of indices of adiposity with atherogenic lipoprotein subfractions. Int J Obes Relat Metab Disord 1998; 22: 432-439. MEDLINE

26 Ko GT, Chan JC, Woo J, Lau E, Yeung VT, Chow CC, Wai HP, Li JK, So WY, Cockram CS. Simple Anthropometric indexes and cardiovascular risk factors in Chinese. Int J Obes Relat Metab Disord 1997; 21: 995-1001. MEDLINE

27 Despres J-P. Abdominal obesity as an important component of insulin-resistance syndrome. Nutrition 1993; 9: 452-459. MEDLINE

28 WHO 1990. Diet nutrition and the prevention of chronic disease. Technical Report Series 797. World Health Organization: Geneva.,

29 Ezzati Tm, Massey JT, Waksberg J, Chu A, Maurer KR. Sample design: Third National Health and Nutrition Examination Survey. Vital Health Stat 1992; 2: 113.

30 National Center for Health Statistics. National Health and Nutrition Examination Survey III: data collection forms. National Center for Health Statistics: Hyattsville, MD, 1991,

31 Lohman TG, Roche AF, Martorell R (eds). Anthropometric Standardization Reference Manual. Human Kinetics Book; Champaign, IL, 1988,

32 Norusis MJ. SPSS 9.0 for Windows Release 9.0. SPSS Inc.: Chicago, IL, 1999,

33 SIMSTAT for Windows version 1.24. Provalis Research Corp. Montreal, 1998,

34 Mossman D. Resampling techniques in the analysis on non-binomial ROC data. Med Decision Making 1995; 15: 358-366.

35 Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristics (ROC) curve. Radiology 1982; 143: 29-36. MEDLINE

36 Lovejoy JC, de la Bretonne JA, Klemperer M, Tulley R. Abdominal fat distribution and metabolic risk factors: effects of race. Metabolism 1996; 45: 1119-1124. MEDLINE

37 Conway JM, Chanetsa FF, Wang P. Intraabdominal adipose tissue and anthropometric surrogates in African American women with upper- and lower-body obesity. Am J Clin Nutr 1997; 66: 1345-1351. MEDLINE

38 Megnien JL, Denarie N, Cocaul M, Simon A, Levenson J. Predictive value of waist-to-hip ratio on cardiovascular risk events. Int J Obes Relat Metab Disord 1999; 23: 90-97. MEDLINE

39 Kronmal RA. Spurious correlation and the fallacy of the ratio standard revisited. JR Stat Soc A 1993; 156: 379-392.

40 Allison DB, Paultre F, Goran MI, Poehlman ET, Heymsfield SB. Statistical considerations regarding the use of ratios to adjust data. Int J Obes Relat Metab Disord 1995; 19: 644-652. MEDLINE

41 Goodman-Gruen D, Barrett-Connor E. Sex differences in measures of body fat and body distribution in the elderly. Am J Epidemiol 1996; 143: 898-906. MEDLINE

42 Albu JB, Murphy L, Frager DH, Johnson JA, Pi-Sunyer FX. Visceral fat and race-dependent health risks in obese nondiabetic premenopausal women. Diabetes 1997; 46: 456-462. MEDLINE

43 Dowling HJ, Pi-Sunyer FX. Race-dependent health risk of upper body obesity. Diabetes 1993; 42: 537-543. MEDLINE

44 Stevens J, Plankey MW, Keil JE, Rust PF, Tyroler HA, Davis CE. Black women have smaller abdominal births than white women of the same relative weight. J Clin Epidemiol 1994; 47: 495-499. MEDLINE

45 Lean ME, Han TS, Deurenberg P. Predicting body composition by densitometry from simple anthropometric measurements. Am J Clin Nutr 1996; 63: 4-14. MEDLINE

Figure 1 Distribution of waist girth in white, black and hispanic men and women.

Figure 2 Receiver operating characteristic (ROC) curves comparing waist girth values corresponding to obesity (BMI30 kg/m²) in whites, blacks and hispanics.

Table 1 Descriptive analysis of study participants

Table 2 Linear regression derived age- and ethnic-specific waist circumference values corresponding to BMI cut-off points

Table 3 ROC derived age- and ethnic-specific waist circumference values corresponding to BMI cut-off points

Table 4 Areas under the ROC curves of waist circumference for BMI stratified by age

Table 5 Gender-specific waist circumference values corre-sponding to overweight and obesity