Abdominal Obesity in the United States: Prevalence and Attributable Risk of Hypertension

Home , Abdominal obesity

Journal of Human Hypertension (1999) 13, 425–430  1999 Stockton Press. All rights reserved 0950-9240/99 $12.00 http://www.stockton-press.co.uk/jhh ORIGINAL ARTICLE Abdominal obesity in the United States: prevalence and attributable risk of hypertension

IS Okosun, TE Prewitt and RS Cooper Department of Preventive Medicine and Epidemiology, Loyola University Medical Center, Stritch School of Medicine, Maywood, IL, USA

Objective: The aim of this study was to determine the Results: The prevalences of abdominal obesity were prevalence of abdominal obesity and its impact on the 27.1%, 20.2% and 21.4% in White, Black and Hispanic risks of hypertension in the US adult population. men, respectively. The corresponding values in women Design and methods: Data from the third US National were 43.2%, 56.0% and 55.4%. Abdominal obesity was Health and Nutrition Examination Surveys, 1988–1994, found to be associated with a two to three-fold were utilised. Abdominal obesity was defined as waist increased risk of hypertension in this population. In circumference у102 cm in men and у88 cm in women. men, the attributable risk percent ranged from 20.9% in Hypertension was defined as mean diastolic blood Hispanics to 27.3% in Whites and in women ranged from pressure у90 mm Hg, systolic blood pressure у140 36.5% in Whites to 56.5% in Hispanics. We estimated mm Hg or current treatment with prescribed hyperten- that 24 million adult men and 40 million adult women sion medication. Prevalences of abdominal obesity were of Hispanic and non-Hispanic Black and White ethnicity estimated in non-Hispanic White, non-Hispanic Black were suffering from abdominal obesity. and Hispanic Americans. Gender-specific logistic Conclusions: In this population, hypertension appears regression analysis using empirical waist cut-off points to be associated with abdominal obesity. The estimates was used to determine the risks of hypertension. The of population attributable risks suggest that the risk of impact of abdominal adiposity on risk of hypertension hypertension could be potentially reduced if waist size was estimated from population-attributable risk were reduced to Ͻ102 cm in men and Ͻ88 cm in women. adjusting for age, current smoking and alcohol intake.

Keywords: abdominal obesity; hypertension; waist circumference

Introduction used and acknowledged as the best anthropometric alternative to WHR, CT and MRI. WC measures sub- The importance of gynoid and android body habitus cutaneous and intra-abdominal adiposity.9 Unlike (now regarded as abdominal or central adiposity) to WHR, WC is easier to interpret and better correlated several diseases was first described by Vague in 10 1 with visceral fat mass. Visceral adiposity is highly 1956. Subsequently, numerous epidemiological correlated with many metabolic complications of studies supporting a positive association between the insulin resistance syndrome, including hyperin- central or abdominal adiposity and cardiovascular 2–8 sulinaemia, hypercholesterolaemia, gluconaemia, diseases (CVD) risk factors have been published. hypertriglyceridaemia and high levels of low-den- While computed tomography (CT) and magnetic res- sity cholesterol.11,12 Compared to WHR, WC is onance imaging (MRI) remain the best methods for strongly correlated with total body adiposity estimating abdominal adiposity, they are imprac- assessed by body mass index (BMI).12 tical in large epidemiological studies because they There is no consensus on the WC cut-off points are laborious and expensive. Hence, anthropometric for abdominal adiposity. The two most cited refer- alternatives are often used in epidemiological stud- ences for WC cut-off points were proposed by Lean ies to determine abdominal obesity. The most com- et al13 and Lemieux et al.14 The generalisability of monly used anthropometric surrogate being the the proposed WC cut-off points of Lean et al13 and waist-to-hip ratio (WHR). Lemieux et al14 is limited on the grounds that the Waist circumference (WC) is increasingly being samples used to derive the cut-off points were from White populations of Scotland and Canada. Recently, the US National Heart, Lung, Blood Correspondence: Dr Ike S. Okosun, Department of Preventive Institute/National Institute of Health (NHLBI/NIH) Medicine and Epidemiology, Loyola University Medical Center, expert panel on the identification, evaluation, and Stritch School of Medicine, 2160 S. First Ave, Maywood, IL 60153, USA the treatment of overweight and obesity in adults Received 9 February 1999; Revised 24 March 1999; Accepted 5 proposed WC cut-off points of 102 cm or more and April 1999 88 cm or more for men and women, respectively.15 Abdominal adiposity and hypertension in US adults IS Okosun et al 426 These cut-off points were recommended in ident- Definition of terms ifying increased relative risk for the development of Hypertension was defined as mean diastolic blood obesity-associated risk factors for most adults with у 2 15 pressure (DBP) 90 mm Hg and systolic blood BMI of 25–34.9 kg/m . The NHLBI/NIH panel у made the recommendations based on a review of pressure (SBP) 140 mm Hg or current treatment with prescribed hypertension medication.21 published scientific literature in MEDLINE from Abdominal obesity was defined as WC у102 cm for 1980 to 1997 of topics identified as key to obesity у 15 evidence model. men and WC 88 cm for women. Alcohol was cat- egorised as 1 and 0, for current drinkers and non- To our knowledge, no studies assessing the prevalence of abdominal obesity have been carried out drinkers, respectively. Smoking was graded as 1 for using the NHLBI/NIH WC cut-off criteria on a US current smokers and 0 for non-smokers. adult population. Therefore, the present study was undertaken utilising the Third National Health and Statistical methods Nutrition Examination Survey (NHANES III) to esti- mate the prevalence of abdominal obesity in the US Statistical programmes available in SPSS version 8.0 adult population and determine the attributable for Windows and WesVarPC version 2.1 were util- 22,23 risks of hypertension to abdominal obesity. ised for these analyses. One-way analysis of variance was used to compare means of anthropometric variables across ethnic groups. Subjects and methods Prevalence estimates were weighted to account for cluster design and to represent the total civilian Study design non-institutionalised population of the US. The NHANES III was a stratified multistage cluster prob- prevalence of hypertension and abdominal obesity were age-adjusted by direct methods using the 1990 ability sample of the non-institutionalised civilian 24 US population groups examined in two phases US population census data. Gender-specific empirical WC (quartiles of WC, WC у102 for men between 1988 and 1994. The sampling and measure- у ment procedures have been extensively described and WC 88 for women) were used to compare odds elsewhere.16,17 Only subjects identified as non-His- of hypertension for Whites, Blacks and Hispanics panic White, non-Hispanic Black and Hispanic adjusting for age, smoking and alcohol intake. We used the estimates of the prevalence (PE) of abdomi- Americans aged 17–90 years were eligible for this у у investigation. The NHANES III protocol included a nal obesity (WC 102 cm in men and WC 88 cm in women) to calculate population attributable risks home interview followed by a physical examination 25 in a mobile examination centre. This study was (PAR%) as follows: restricted to subjects that were examined in the = + PAR% PE (OR-1)/[PE (OR-1)] 1*100 mobile examination centre and for whom blood pressures and anthropometric measurements were Odds ratio (OR) compares men and women with Ͻ Ͻ available, including weight, height and waist. WC 102 cm and 88 cm women, respectively with those with abdominal obesity, adjusting for age, smoking and alcohol intake in the logistic regression Anthropometric measurements model. The customary P-value of 0.05 was used to indicate statistical significance. Weight was measured in the upright position using a digital scale. Height was measured with a stadi- ometer. Waist measurement was made to the nearest Results 0.1 cm at minimal respiration at midpoint between Means and standard deviations of age and the the bottom of the rib cage and above the top of the anthropometric variables as well as the prevalence iliac crest. Description of measurement precision of hypertension are presented in Table 1. Overall, between technicians has been described else- 6760 Whites, 4713 Blacks and 4621 Hispanics were where.18 BMI was calculated as the measured weight eligible for this investigation. In both men and in kilograms divided by height in metres squared women, the mean BMI was higher than the World (kg/m2). Health Organisation (WHO) recommended value.26 Also, overall mean values of SBP were slightly Blood pressure measurements higher in White men, Black men and Black women than the clinically acceptable values. There were Detailed descriptions of blood pressure measure- significant ethnic differences for most variables. ment techniques have been previously docu- Among men, WC was highest in Whites and among mented.19,20 Briefly, examiners underwent rigorous women WC was highest in Blacks compared to their training on blood pressure measurement techniques. ethnic counterparts. Whites were significantly older Blood pressure measurements were taken in the sit- than Blacks and Hispanics (P Ͻ 0.05). Mean values ting position after 5 min of resting using the stan- for DBP and SBP for Blacks were higher than Whites dard mercury sphygmomanometer. Three blood and Hispanics. Age-adjusted prevalences of hyper- pressures were taken from each participant, with tension were higher in Blacks than Whites and His- intervals of 30–60 sec between cuff inflation. The panics (P Ͻ 0.01). average of the three blood pressure readings was Table 2 presents age-specific mean WC values for used in this analysis. men and women. Among men, the mean value of Abdominal adiposity and hypertension in US adults IS Okosun et al 427 Table 1 Basic characteristics of study population

n Age (year) weight (kg) BMI (kg/m2) waist (cm) DBP [mm Hg] SBP (mm Hg) HTN (%)

Men White 3138 43.5 ± 17.3 82.6 ± 16.2 26.5 ± 4.7 95.6 ± 13.1 75.0 ± 11.8 122.6 ± 15.9 22.4 Black 2147 38.9 ± 16.3 81.8 ± 18.5 26.3 ± 5.4 90.7 ± 14.8 76.1 ± 14.0 125.2 ± 17.6 33.7 Hispanic 2321 35.0 ± 14.3 76.9 ± 14.5 26.6 ± 4.5 92.6 ± 12.4 73.2 ± 13.3 119.8 ± 14.8 23.3 All 7606 42.5 ± 17.2 82.2 ± 16.5 26.5 ± 4.8 94.9 ± 13.4 75.0 ± 12.2 122.7 ± 16.1 25.7 P-value Ͻ0.01 Ͻ0.01 Ͻ0.01 Ͻ0.01 Ͻ0.01 Ͻ0.01 Ͻ0.01

Women White 3622 45.3 ± 18.5 68.4 ± 16.4 25.9 ± 6.1 87.7 ± 15.1 69.3 ± 11.4 117.9 ± 19.9 21.7 Black 2566 40.0 ± 16.9 75.6 ± 19.8 28.4 ± 7.2 92.1 ± 16.7 70.9 ± 14.1 120.5 ± 22.0 31.0 Hispanic 2300 36.4 ± 15.0 68.3 ± 15.7 27.7 ± 6.1 90.4 ± 14.5 67.6 ± 12.1 113.9 ± 17.9 21.6 All 8488 44.2 ± 18.3 69.3 ± 17.0 26.3 ± 6.3 88.4 ± 15.3 69.4 ± 11.8 118.0 ± 20.1 22.3 P-value Ͻ0.01 Ͻ0.01 Ͻ0.01 Ͻ0.01 Ͻ0.01 Ͻ0.01 Ͻ0.01

DBP, diastolic blood pressure; SBP, systolic blood pressure; HTN, age-adjusted prevalence of hypertension deﬁned as DBP у90 mm Hg or SBP у140 mm Hg or current use of anthypertensive agents; P-values are from one-way analysis of variance comparing means across ethnic groups.

Table 2 Gender-speciﬁc mean of waist circumference (cm) by ethnicity and age-group

Age-group White Black Hispanic All

Men Ͻ20 82.4 ± 11.6 78.7 ± 12.5 83.6 ± 10.7 81.5 ± 11.8 20–29 88.2 ± 12.0 86.6 ± 14.7 87.9 ± 10.9 87.6 ± 12.5 30–39 94.6 ± 12.5 90.8 ± 14.6 93.7 ± 11.3 93.0 ± 13.0 40–49 96.7 ± 13.1 93.9 ± 13.5 99.3 ± 11.9 97.4 ± 13.1 50–59 100.7 ± 11.6 96.9 ± 14.8 100.7 ± 11.5 99.9 ± 12.6 60–69 102.6 ± 11.3 97.9 ± 13.2 107.4 ± 10.4 101.0 ± 11.7 70–79 100.6 ± 10.8 96.6 ± 13.0 97.8 ± 12.5 99.3 ± 11.7 80+ 100.0 ± 10.7 93.6 ± 11.7 96.5 ± 9.5 97.4 ± 10.4 All 97.2 ± 12.8 91.3 ± 15.0 94.0 ± 12.7 94.6 ± 13.6

Women Ͻ20 78.0 ± 11.9 81.2 ± 16.6 81.6 ± 12.7 80.4 ± 14.0 20–29 80.8 ± 13.8 86.2 ± 15.2 85.0 ± 13.7 84.2 ± 14.5 30–39 86.0 ± 15.3 92.8 ± 17.0 91.7 ± 14.1 90.2 ± 15.9 40–49 89.1 ± 14.6 96.9 ± 16.3 95.6 ± 12.5 93.7 ± 15.1 50–59 94.2 ± 15.0 100.0 ± 15.3 96.5 ± 13.0 96.4 ± 14.9 60–69 94.4 ± 13.5 101.1 ± 14.8 98.7 ± 12.9 97.0 ± 13.9 70–79 94.3 ± 13.6 98.7 ± 13.4 96.9 ± 12.3 95.3 ± 13.5 80+ 91.7 ± 11.0 90.9 ± 15.6 90.6 ± 10.5 91.5 ± 11.2 All 89.5 ± 14.8 92.7 ± 16.8 91.2 ± 14.4 90.9 ± 15.4

WC increased progressively with increasing age The crude and age-adjusted prevalences of from 82 cm, 79 cm and 84 cm in the Ͻ20 year age abdominal obesity by ethnicity and gender are group to 103 cm, 98 cm and 107 cm in the 60–69 presented in Table 3. Among men, the age-adjusted year age group in White, Black and Hispanic Amer- prevalences of abdominal obesity were 27%, 20% icans, respectively. Similarly, the WC means in and 21% in White, Black and Hispanic Americans, women increased from 78 cm, 81 cm and 82 cm in respectively. The corresponding values in women Ͻ20 years old group to 94 cm, 101 cm and 99 cm in were 43%, 56% and 55% in White, Black and His- the 60–69 age group for White, Black and Hispanic, panic Americans, respectively. respectively. Overall, in both men and women, there We calculated ethnic specific trends in the risks was a progressive decrease in mean WC values with of hypertension from gender-specific empirical age after the sixth decade of life. quartiles of WC adjusting for age, smoking and alco- Figure 1 presents age-specific prevalence of hol intake (Table 4). A consistent trend of increasing abdominal obesity in the US adult population. As risk of hypertension with increasing WC emerged expected, there was a steady increase in the preva- for both men and women. The strongest associations lences of abdominal obesity with age in both men of increasing WC with hypertension were in the and women. The highest prevalence values for fourth quartiles for both men and women compared White men (53%), Black men (37%) and White with the first quartiles in each ethnic group. Com- women (69%) were recorded in the 60–69 age group. paring the fourth quartiles to the first quartiles, there The highest prevalence values for Hispanic men were about four to six-fold increased risks of hyper- (49%) and Black women (80%) were observed in the tension in men. Compared to the first quartiles, there 50–59 year age group, while 80% was recorded in were three to six-fold increased risk of hypertension the 70–79 year age group among Hispanic women. in the fourth quartile for women, as estimated from Abdominal adiposity and hypertension in US adults IS Okosun et al 428 in men and women, respectively. The population attributable risk percentages were higher in women than men (P Ͻ 0.01).

Discussion Previous studies have demonstrated abdominal obesity as an important correlate of cardiovascular disease risk factors.3–8 Traditionally, evidence for the strong relationship between abdominal obesity and CVD had often been assessed by WHR. How- ever, although WHR reflects central fat distribution it is a less than ideal measure of visceral fat accumulation.11–14 As a ratio, it is difficult to interpret bio- logically because a change in body adiposity or weight do not necessarily result in a change in WHR.27 Also, ratios have limitations in statistical modelling since their use can introduce spurious correlations among the ratios and other variables.28,29 WHR is also profoundly affected by pelvic structure and is an imperfect measure of abdominal adiposity, particularly in lean individuals.30,31 Recent epidemiological data now describe WC as a better predictor of CVD than WHR.9 This is the first investigation utilising NHANES III to determine the prevalence of abdominal obesity. NHANES III represents the best available data since the sampling scheme was representative and national in scope. The training programme and qual- ity control measures instituted in NHANES III give an added credence to the data. This is also the first investigation to use the new proposed WC cut-off point for obesity recommended by the NHLBI/NIH expert panel on the identification, evaluation, and the treatment of overweight and obesity in adults.15 Figure 1 Age-specific prevalence of abdominal obesity in the US Although our data are cross-sectional and caus- adult population. ality cannot be established, we observed a precipi- tous increase in prevalences of abdominal obesity Table 3 Prevalence of abdominal adiposity by gender and eth- with age in men and women until the sixth decade nicity of life. The overall prevalence of abdominal obesity was two times as high in women compared to men. Ethnicity Crude-prevalence Age-adjusted prevalence Abdominal obesity was found to be associated with two to three times greater risk of hypertension. Our men women men women study showed a disproportionate burden of abdominal obesity in Whites, Blacks and Hispanic men. White 28.9 43.6 27.1 43.2 White men had age-adjusted prevalence of abdomi- Black 20.3 56.1 20.2 56.1 Hispanic 21.5 54.7 21.4 55.4 nal obesity of 27.1% which was 34% higher than that noted in Black men (20.2%) and 30% higher Overall 27.5 45.8 26.5 45.3 than that noted for Hispanic men (21.4%). In Values are percentages. women, Blacks (56.0%) and Hispanic (55.4%) had approximately the same age-adjusted prevalence values that were 19% higher than that observed in age, smoking and alcohol adjusted odds ratios Whites. Our analysis suggests that about 24 million obtained from gender-specific logistic regression adult men and 40 million adult women of Hispanic models. and non-Hispanic Black and White ethnicity were The impact of abdominal obesity was further suffering from abdominal obesity. In this population investigated across ethnic group and gender 27.3%, 21% and 20.9% hypertension might have (Table 5). In comparison with men with WC been prevented if WC were below 102 cm in White, Ͻ102 cm, a WC of 102 cm or more was associated Black and Hispanic American men, respectively. with increased risk of hypertension that stabilised The analogous values in women were 36.5%, 38.4% at 2 in men. In women, a WC of 88 cm or more was and 56.5% White, Black and Hispanic Americans, associated with risks of hypertension that ranged respectively, if WC were below 88 cm. from 2.1 in Black women to 3.4 in Hispanic women. In conclusion, data from these national surveys Overall, abdominal obesity was associated with two- document that hypertension is associated with fold and three-fold increased risks of hypertension abdominal obesity in both adult men and women Abdominal adiposity and hypertension in US adults IS Okosun et al 429 Table 4 Odds ratio for hypertension by gender-specific quartile of waist circumference and ethnicity

Men Women

mean ± s.d. OR [95% CI] mean ± s.d. OR [95% CI]

White I 79.9 ± 4.2 1.00 Reference 72.7 ± 4.2 1.00 Reference II 89.9 ± 2.7 2.11 2.10–2.12 87.5 ± 3.0 1.45 1.44–1.46 III 98.4 ± 2.6 3.21 3.20–3.22 95.0 ± 3.2 2.62 2.61–2.63 IV 112.0 ± 8.5 5.67 5.66–5.69 110.7 ± 9.0 6.01 6.00–6.03

Black I 76.8 ± 5.1 1.00 Reference 72.3 ± 4.5 1.00 Reference II 89.5 ± 2.6 1.62 1.61–1.63 84.6 ± 3.1 1.55 1.54–1.56 III 98.1 ± 2.6 2.26 2.25–2.27 95.2 ± 3.2 1.75 1.74–1.76 IV 113.3 ± 10.6 3.53 3.51–3.54 113.1 ± 10.5 3.53 3.50–3.55

Hispanic I 77.8 ± 4.7 1.00 Reference 72.7 ± 4.2 1.00 Reference II 89.7 ± 2.7 1.99 1.97–2.02 84.6 ± 3.1 0.66 0.65–0.67 III 98.3 ± 2.6 3.39 3.35–3.43 95.0 ± 3.2 2.05 2.02–2.09 IV 112.3 ± 9.0 5.15 5.10–5.21 110.6 ± 9.0 2.75 2.71–2.79

OR, odds ratio from logistic regression model adjusted for age, current smoking and alcohol use; hypertension was deﬁned as DBP у90 mm Hg or SBP у140 mm Hg or current use of antihypertensive agents.

Table 5 Odds ratio and population attributable risk for hypertension by gender and ethnicity

Men Women

1 2 PE OR [95% CI] PAR% PE OR [95% CI] PAR%

White 28.9 2.30 2.29–2.31 27.3 43.6 2.32 3.31–3.33 36.5 Black 20.3 2.31 2.29–2.31 21.0 56.1 2.11 2.10–2.12 38.4 Hispanic 21.5 2.23 2.22–2.24 20.9 54.7 3.38 3.35–3.41 55.6 All 27.5 2.23 2.22–2.23 25.3 45.8 3.28 3.27–3.28 51.1

OR, odds ratio from logistic regression models, adjusted for age, smoking and alcohol use, comparing men with waist circumference у Ͻ у Ͻ 1 2 102 cm with men 102 cm and comparing women with waist circumference 88 cm with women 88 cm; PE and PE are prevalences of abdominal adiposity in men and women, respectively; PAR%, population attributable risk of hypertension defined as: PE (OR- + у у 1)/[PE(OR-1)] 1 * 100; hypertension was defined as DBP 90 mm Hg or SBP 140 mm Hg or current use of antihypertensive agents. and in the three US ethnic groups. Given the pervas- cumference as a screening tool for cardiovascular risk ive sedentary lifestyles and consumption of high cal- factors: evaluation of receiver operating characteristics orie foods in the US,32 one can reasonably assume (ROC). Obesity Res 1996, 4: 533–547. higher prevalence of cardiovascular diseases in the 4 Han TS, Feskens EJ, Lean ME, Seidell JC. Associations future. There is an urgent need to describe a public of body composition with Type 2 diabetes mellitus. health strategy to adequately enhance behavioural Diabet Med 1998; 15: 129–135. 5 Wei M, Gaskill SP, Haffner SM, Stern MP. Waist cir- transformation necessary for total and abdominal cumference as the best predictor of noninsulin depen- obesity reduction in this population. dent diabetes mellitus (NIDDM) compared to body mass index, waist/hip ratio over other anthropometric Acknowledgements measurements in Mexican Americans: A 7-year Pro- spective Study. Obesity Res 1997; 5: 16–23. Data from the NHANES III were obtained from the 6 James RW, Brulhart-Meynet MC, Lehmann T, Golay A. US National Center for Health Statistics. This work Lipoprotein distribution and composition in obesity: was supported in part by a grant from the National their association with central adiposity. Int J Obes Heart, Lung, and Blood Institute (HL 45508). Relat Metab Disord 1997; 21: 1115–1120. 7 Hunter GR et al. Fat distribution, physical activity, and cardiovascular risk factors. Med Sci Sports Exerc 1997; References 29: 362–369. 1 Vague J. The degree of masculine differentiation of 8 Gus M et al. Association of various measurements of obesitities: a factor determining predisposition to dia- obesity and the prevalence of hypertension. Arq Bras betes, atherosclerosis, gout, and uric calculous disease. Cardiol 1998; 70: 111–114. 1956. Am J Clin Nutr 1956; 4: 20–34. 9 Molarius A, Seidell JC. Selection of anthropometric 2 Okosun IS et al. Association of waist circumference indicators for classification of abdominal fatness-a with risks of hypertension and type 2 diabetes in critical review. Int J Obes Relat Metab Disord 1998; 22: Nigerians, Jamaicans and African-Americans. Diabetes 719–727. Care 1998; 21: 1836–1842. 10 Pouliot MC et al. Waist circumference and abdominal 3 Han TS, Van Leer EM, Seidell JC, Lean ME. Waist cir- sagittal diameter: best simple anthropometric indexes Abdominal adiposity and hypertension in US adults IS Okosun et al 430 of abdominal visceral adipose tissue accumulation and 21 WHO. Expert Committee. Arterial Hypertension Tech- related cardiovascular risk in men and women. Am J nical Report Series no 628. WHO: Geneva Switzer- Cardiol 1994; 73: 460–468. land, 1978. 11 Ko GT et al. Simple anthropometric indexes and car- 22 Norusis MJ. SPSS 7.5 for Windows Release 7.5.1. SPSS diovascular risk factors in Chinese. Int J Obes Relat Inc. Chicago, 1996. Metab Disord 1997; 21: 995–1001. 23 Wes Var VarPC. Westat, Inc. Rockville MD, 1997. 12 Sattar N et al. Associations of indices of adiposity with 24 King H, Rewers M. WHO Ad Hoc Diabetes Reporting atherogenic lipoprotein subfractions. Int J Obes Relat Group: Global estimates for prevalence of diabetes Metab Disord 1998; 22: 432–439. mellitus and impaired glucose tolerance in adults. Dia- 13 Lean ME, Han TS, Morrison CE. Waist circumsference betes Care 1993; 16: 157–177. as a measure for indicating need for weight manage- 25 Kleinbaum DG, Kupper LL, Morgenstern H. Epidemi- ment. BMJ 1995; 311: 158–161. ologic Research. Principles and Quantitative Methods. 14 Lemieux S et al. Anthropometric correlates to changes Livetime Learning Pub, 1982. in visceral adipose tissue over 7 years in women. Int 26 Physical status: the use and interpretation of anthro- J Obes Relat Metab Disord 1996; 20: 618–624. pometry. Report of a WHO expert committee. World 15 National Institute of Health/National Heart Lungs and Health Organization, Geneva. WHO Technical Report Blood Clinical Guidelines on the identification, evalu- Series 8541995 1995, pp 368–369. ation, and treatment of overweight and obesity in 27 Bjorntorp P. Abdominal fat distribution and disease: adults. The Evidence Report. Obesity Res 1998; 6: an overview of epidemiological data. Ann Med 1992; 51S–209S. 24: 15–18. 16 Ezzati TM et al. Sample design: Third National Health 28 Kronmal RA. Spurious correlation and the fallacy of and Nutrition Examination Survey. Vital Health Stat the ratio standard revisited. JR Statist Soc A 1993; 156: 2. 1992; 113: 1–35. 17 National Center for Health Statistics. National Health 379–392. and Nutrition Examination Survey III: Data collection 29 Allison DB et al. Statistical considerations regarding forms. National Center for Health Statistics: Hyatts- the use of ratios to adjust data. Int J Obes Relat Metab ville, MD, 1991. Disord 1995; 19: 644–652. 18 Lohman TG, Roche AF, Martorell R (eds). Anthropo- 30 Goodman-Gruen D, Barrett-Connor E. Sex differences metric Standardization Reference Manual. Human in measures of body fat and body distribution in the Kinetics Book: Champaign, IL, 1988. elderly. Am J Epidemiol 1996; 143: 898–906. 19 Burt VL et al. Trends in the prevalence, awareness, 31 Taylor RW et al. Body mass index, waist girth, and treatment, and control of hypertension in the adult US waist-to-hip ratio as indexes of total and regional adi- population. Data from the health examination surveys, posity in women: evaluation using receiver operating 1960 to 1991. Hypertension 1995; 26: 60–69. characteristic curves. Am J Clin Nutr 1998; 67: 44–49. 20 Burt VL et al. Prevalence of hypertension in the US 32 Crespo CJ, Keteyian SJ, Heath GW, Sempos CT. Leis- adult population. Results from the Third National ure-time physical activity among US adults. Results Health and Nutrition Examination Survey, 1988–1991. from the Third National Health and Nutrition Examin- Hypertension 1995; 25: 305–313. ation Survey. Arch Intern Med 1996; 156: 93–98.