Abdominal Obesity Defined As a Larger Than Expected Waist Girth Is

Home , Abdominal obesity

Journal of Human Hypertension (2001) 15, 307–312  2001 Nature Publishing Group All rights reserved 0950-9240/01 $15.00 www.nature.com/jhh ORIGINAL ARTICLE Abdominal obesity deﬁned as a larger than expected waist girth is associated with racial/ethnic differences in risk of hypertension

IS Okosun, S Choi, MM Dent, T Jobin and GEA Dever Department of Community Medicine, Mercer University School of Medicine 1550 College Street Macon, GA, USA

Objective: Waist circumference (WC) cut-points of Results: Relative to white, black race/ethnicity was у102 cm and у88 cm for men and women, respectively, associated with ෂ1.8 and ෂ2.7 greater risk of hyperten- representing abdominal obesity have been recom- sion in men and women, respectively, adjusting for mended for determining obesity related co-morbidities. abdominal obesity, age, smoking and alcohol consump- However, these cut-points carry the component of gen- tion. Having larger than expected waist girths were eralised obesity estimated by body mass index (BMI). associated with 1.58 and 1.39 increased risk of hyper- The aim of this investigation was to determine whether tension in black men and black women, respectively, abdominal obesity free of the inﬂuence of overall heavi- adjusting for confounders. Population attributable risks ness is associated with increased risk of hypertension of hypertension due to abdominal obesity were approxi- in a representative sample of white and black Amer- mately 24.9% and 15.9%, in black men and black icans. women, respectively. from the Third US National Conclusions: In Americans, hypertension is a public (11114 ؍ Methods: Data (n Health and Nutrition Examination Survey were used in health problem that is closely linked to abdominal adi- this investigation. Standardised residual values from posity. An important research challenge therefore is to the linear regression of WC on BMI were used to deﬁne determine the best way to regulate body weight under abdominal obesity status. The risk of hypertension conditions of food abundance. There is a need to clarify associated with abdominal obesity was estimated from how lifestyle habits promote large waist sizes leading the logistic regression model, adjusting for age, smok- to abdominal adiposity and associated cardiovascular ing and alcohol. We also estimated the public health disease in the US, particularly among black Americans. consequences of abdominal obesity from the popu- Journal of Human Hypertension (2001) 15, 307–312 lation attributable fraction of hypertension.

Keywords: waist circumference; abdominal obesity; ethnicity

Introduction ometry. However, these imaging methods are not practical in large-scale field epidemiology. They are Although the link between abdominal adiposity and laborious and expensive, and their use carries risk cardiovascular diseases (CVD) is well documented 1–8 of radiation. Hence, there is a need to establish a in the epidemiologic literature, there is no con- relatively cheap and risk-free surrogate marker of sensus on the most reliable anthropometric variable these imaging techniques. for abdominal adiposity. Ideally, the best method for The two most commonly employed anthropo- determining abdominal obesity is by imaging tech- metric surrogates for imaging techniques in niques such as computerised tomography, magnetic determining abdominal adiposity are ratios of waist- resonance imaging or dual energy X-ray absorpti- to-hip girths (WHR) and waist circumference (WC). WHR has many shortcomings, including inherent weakness as a ratio index9,10 and strong influence Correspondence: Ike S Okosun, PhD, MPH, FRIPHH, Department by pelvic structure. Also, WHR is an unsatisfactory of Community Medicine, Mercer University School of Medicine, measure of abdominal fat mass, especially in non- 1550 College Street Macon, Georgia, 31207, USA 11 E-mail: okosunFiȰmercer.edu obese individuals. Indeed, non-obese and obese Received 26 June 2000; revised 19 September 2000; accepted 30 subjects can have the same WHR values. Further- November 2000 more, the sensitivity of WHR in detecting changes Abdominal obesity and hypertension IS Okosun et al 308 in body adiposity is low because gluteal fat and sub- uals aged 17–90 yeas for whom weight, height, cutaneous abdominal fat can be decreased in some waist, diastolic blood pressure (DBP) and systolic subjects with weight loss.12 blood pressure (SBP) were obtained. Weight was WC has been endorsed as the best anthropometric measured at a standing position using a Toledo self- surrogate of abdominal adiposity.13,14 WC is an zeroing weight scale. Height was measured at an aggregate measurement of the actual amounts of upright position with a standiometer. Waist abdominal fat accumulation and is a crucial corre- measurement was made to the nearest 0.1 cm at the late of abnormal metabolic syndromes found among natural waist midpoint between the bottom of the obese and overweight subjects.15 Measurement of rib cage and above the top of the iliac crest. WC is simple and requires only a flexible tape meas- In NHANES III, three blood pressure measure- ure, furthermore, measurement error is low due to ments were obtained from each subject using a stan- large circumference. In addition, WC is more highly dard mercury sphygmomanometer at a 60-second correlated with visceral adipose tissue accumulation interval between inflation cuffs. The average of the than WHR.16 Visceral adiposity is the component of three readings was utilised for this analysis. Smok- body composition that is most highly associated ing and alcohol intake were assessed by self-report. with many metabolic abnormalities such as hyper- Smoking was categorised as 1 and 0, for current tension, glucose intolerance, hyperinsulinaemia, smokers and non-smokers, respectively. Alcohol use hypercholesterolaemia, hypertriglyceridaemia, and was graded as 1 for current drinkers and 0 for non- high levels of low-density lipoprotein cholesterol.17–20 current drinkers. For this analysis, hypertension was Indeed, many are now advocating WC as a valid defined as DBP у90 mm Hg, SBP у140 mm Hg or anthropometric variable for health promotion and current treatment with prescribed anti-hyperten- the basis for alerting those at risk of CVD.21,22 sion medication.25 In a recent publication, the United States National Institute Health (NIH) expert panels recommended у у the WC cut-points of 102 cm and 88 cm, in men Statistical analysis and women as values representing risks for obesity- related comorbidities.13 These cut-points were Statistical programmes available in SPSS version determined after the review of published data ger- 10.0 for Windows were utilised for this analysis.26 mane to abdominal obesity in various journals pub- Racial/ethnic differences for continuous and categ- lished between 1980 and 1997. However, because orical variables were assessed with the Student’s t- these recommended WC cut-points do not account test and chi-squared statistics, respectively. The for overall body heaviness, there is a need to rede- standardised residual values from the linear fine WC corresponding to abdominal obesity devoid regression of WC on BMI were utilised to establish of generalised obesity. The aims of this investigation abdominal obesity status. Abdominal obesity was were therefore to define abdominal obesity free of thus defined as WC larger than expected WC as pre- the influence of overall heaviness as determined by dicted from BMI.27 These predicted WC values cor- body mass index (BMI), and to investigate if such respond to WC adjusted for BMI. measure is associated with increased risk of hyper- Predicted WC values were derived from residuals tension in a representative sample of white and obtained from linear regression of WC on BMI.27 For black Americans. We also sought to examine if white men the residuals were obtained from the equ- abdominal obesity free of the influence of general- ation: WC = 2.49*BMI + 31.31. The corresponding ised obesity was associated with racial/ethnic differ- residuals for white women were obtained from WC ences in the risk of hypertension. = 2.21*BMI + 31.34. For black subjects, the residuals were obtained from WC = 2.56*BMI + 24.07 for men, and were obtained from WC = 2.11*BMI + 32.34 for Materials and methods women. The positive residuals represent abdominal 27 Data source adiposity adjusted for BMI. We estimated the risk of hypertension that was Data from the Third US National Health and associated with abdominal obesity by classifying the Nutrition Examination Survey (NHANES III), as pro- residuals from regression of WC on BMI.27 Negative vided by the National Center for Health Statistics, residual values (less than expected) were coded as were used in this investigation. The sampling and 0 and positive values (greater than expected) were measurement procedures have been described in coded as 1, and utilised as an independent variable detail by other investigators.23,24 Briefly, NHANES in the race/ethnicity and gender specific logistic III is a multistage probability sample of non-insti- regression model, adjusting for age, smoking and tutionalised US population groups defined and alcohol. Radical/ethnic differences in the risk of examined in two phases between 1988 and 1994. hypertension associated with abdominal obesity Only subjects identified as non-Hispanic white and were determined by comparing blacks with whites non-Hispanic black Americans were eligible for in gender-specific logistic linear regression models this investigation. adjusting for age, smoking and alcohol consump- This study was further confined to 11114 individ- tion. The prevalence of hypertension and abdominal

Journal of Human Hypertension Abdominal obesity and hypertension IS Okosun et al 309 obesity were age-adjusted by direct methods using the 1990 US population census data. To estimate the public health consequences of abdominal obesity, we calculated the population attributable fraction percentage (PAF%) of hypertension. The population attributable fraction per cent was estimated28 as:

− = PE (OR 1) PAF% + − * 100 1 PE (OR 1) where PE is the proportion of subjects who were abdominally obese, and OR is the odds ratio comparing subjects who had abdominal obesity with those who did not. Statistical adjustments were made for age, smoking and alcohol intake in the logistic regression model. PAF, sometimes called population attributable risk, is deﬁned as the proportion of a population’s disease rate that would not Figure 1 Illustration of residuals obtained from regression of occur if the risk factor(s) of interest had been waist girth on body mass index. absent.28

Results waist girths are larger than expected for BMI predicted values (measured WC minus circumference The basic characteristics of the study variables predicted from BMI). Individuals with these values stratified by gender and race/ethnicity are presented were defined as abdominally obese. Individuals in Table 1. Overall, white men and women were with residuals values of 0 or less were defined as older taller and tended to have lower DBP than their normal. black counterparts. Both whites and blacks had Plots from regression analysis between WC and higher levels of SBP than the clinically desirable BMI fitted by least square methods are presented for values. Black women were heavier than white men and women in Figures 2 and 3, respectively. women as determined by BMI. The prevalence of As expected, there were significant positive associ- hypertension was higher among black men and ations between WC and BMI. Overall, the slope of black women compared to their whites counterparts the regression of WC on BMI tended to be higher in (P Ͻ 0.001). The prevalence of abdominal obesity men compared to women. The R2 for regression of was higher among black men than white men (P Ͻ WC on BMI was higher in men than women, and 0.001). More black men and black women reported higher in blacks than whites. The R2 were 0.89, 0.79, having smoked at least five cigarettes in the past 0.84 and 0.82, for white men, white women, black week than their white counterparts (P Ͻ 0.001). men and black women, respectively. Figure 1 illustrates distribution of residuals To determine whether racial/ethnic differences in obtained from the regression of WC on BMI used in the risk of hypertension was associated with differ- defining abdominal obesity. Subjects with residual ences in abdominal obesity independent of age, values above the reference line of 0 are those whose smoking and alcohol consumption, dummy vari-

Table 1 Characteristics of studied population

Variables Men Women

White Black P value White Black P-value n (%) 3065 (27.5) 2078 (18.6) 3539 (31.8) 3065 (22.1) Age (yrs) 53.4 ± 20.6 41.9 ± 18.1 Ͻ0.001 52.1 ± 20.8 40.7 ± 17.5 Ͻ0.001 Height (cm) 175.3 ± 6.5 175.7 ± 7.2 0.060 151.5 ± 6.9 163.0 ± 6.4 Ͻ0.001 Weight (kg) 81.3 ± 15.7 81.3 ± 18.2 0.910 68.3 ± 16.1 76.0 ± 19.9 Ͻ0.001 Body mass index (kg/m2) 26.4 ± 4.6 26.3 ± 5.3 0.404 26.3 ± 5.9 28.6 ± 7.2 Ͻ0.001 Diastolic BP (mm Hg) 74.4 ± 12.0 76.4 ± 14.1 0.212 69.4 ± 12.1 71.2 ± 13.7 0.001 Systolic BP (mm Hg) 127.7 ± 19.2 127.1 ± 19.2 Ͻ0.001 123.3 ± 22.7 121.4 ± 22.4 Ͻ0.001 Smoking (%) 28.0 41.2 Ͻ0.001 24.9 30.3 Ͻ0.001 Alcohol intake (%) 36.8 33.1 Ͻ0.001 50.6 59.1 Ͻ0.001 Hypertension (%) 25.3 34.4 Ͻ0.001 21.3 31.1 Ͻ0.001 Abdominal obesity (%) 44.6 56.2 Ͻ0.001 49.6 48.1 0.128

P-value compares gender speciﬁc variables across racial/ethnic groups.

Journal of Human Hypertension Abdominal obesity and hypertension IS Okosun et al 310

Figure 2 Relationship between waist girth and body mass index Figure 3 Relationship between waist girth and body mass index in white and black American men. in white and black American women.

ables were used to compare blacks with whites in gender-specific logistic regression models (Table 2). In men and women, black race/ethnicity was asso- Table 2 Multiple logistic regression analysis of race/ethnicity and other factors with hypertension ciated with greater risk of hypertension relative to white, adjusting for abdominal obesity, age, smoking Variables Men Women and alcohol consumption. The risk for hypertension ෂ due to black race/ethnicity was 1.8 for men and OR (95% CI) OR (95% CI) ෂ2.7 for women. Increase in age was also associated with enhanced risk of hypertension. Abdominal 1.15 (0.98–1.35) 1.19 (1.00–1.43) We fitted gender and race/ethnic specific logistic obesity regression models for hypertension, adjusting for Age (yrs) 1.05 (1.04–1.06) 1.07 (1.06–1.08) age, smoking and alcohol intake (Table 3). In both Black 1.80 (1.54–2.11) 2.70 (2.07–3.28) Alcohol intake 1.08 (0.92–1.26) 1.02 (0.86–1.22) black men and black women, abdominal obesity was Smoking 0.91 (0.77–1.07) 0.70 (0.57–0.86) independently associated with increased risk of hypertension. Larger waist girths than expected OR, odds ratio; CI, confidence intervals. were associated with 58% and 39% increased risk of hypertension in black men and black women, respectively, adjusting for age, smoking and alcohol intake. There was no significant association between hypertension due to abdominal obesity were abdominal obesity and hypertension in white men approximately 24.9% and 15.9%, in black men and and white women. Population attributable risks of black women, respectively.

Journal of Human Hypertension Abdominal obesity and hypertension IS Okosun et al 311 Table 3 Race/ethnic and gender speciﬁc odds ratio and population attributable fraction of hypertension due to abdominal obesity

Variables Men Women

White Black White Black OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)

Abdominal obesity 0.96 (0.78–1.18) 1.58 (1.20–2.07) 1.08 (0.85–13.6) 1.39 (1.05–1.83) Age (yrs) 1.05 (1.04–1.06) 1.05 (1.04–1.06) 1.08 (1.07–1.09) 1.07 (1.06–1.08) Alcohol intake 1.15 (0.95–1.39) 1.00 (0.78–1.28) 1.22 (0.97–1.54) 0.82 (0.62–1.09) Smoking 0.82 (0.66–1.03) 0.99 (0.79–1.26) 0.53 (0.39–0.72) 0.86 (0.65–1.14) Population AF (%) — 24.9 — 15.9

OR, odds ratio from logistic regression analysis; CI, conﬁdence intervals; AF, attributable fraction.

Discussion of blood pressure change in non-obese individuals; this aspect of our findings warrants further investi- Although the literature linking abdominal obesity gation. estimated from WC to diseases is voluminous, only This is the first investigation utilising NHANES III few studies have accounted for the fatness carried to determine the association of abdominal obesity by generalised adiposity. Thus, the lack of associ- estimated by larger than expected WC. The strength ation of abdominal obesity to racial/ethnic differ- of this investigation lies on NHANES III data since ences in the risk of hypertension as reported by the sampling scheme was representative and many investigators may be due to the lack of appro- national in scope. The training programme and qual- priate adjustment for generalised obesity. With the 27 ity control measures instituted in NHANES III pro- exception of Han et al that investigated the associ- vide an added credibility to the data. However, some ation of larger than predicted WC and hip circumfer- limitations must be taken into account in the ence with lifestyle factors, our study is the only one interpretation of results from this study. First, bias that we are aware of that has specifically investi- due to survey non-response and missing values for gated the role of abdominal obesity free from the some variables cannot be ruled out. However, pre- influence of generalised adiposity as determined vious studies of National Health and Nutrition from regression residuals on the risk of hyperten- Examination Surveys have shown little bias due to sion. To our knowledge this is also the only study non-response.31 Second, the validity and reliability that has determined whether abdominal obesity of the measure of abdominal obesity as used in this defined as WC larger than expected is associated study against imaging techniques warrants further with racial/ethnic differences in risk of hyperten- investigation. sion. In this representative sample of American men and women, larger than expected waist girth was Conclusion associated with increased risk for hypertension In summary, abdominal adiposity is an important, among black men and black women. The risks as independent, and specific risk factor for hyperten- estimated from odds ratios were approximately 58% sion in black Americans. In blacks, hypertension is and 39%, for black men and black women, respect- a public health problem that is closely linked to ively. Relative to white, black race/ethnicity was abdominal adiposity. An important research chal- associated with 80% increased risk of hypertension, lenge therefore is to determine the best method to adjusting for age, smoking and alcohol intake. The regulate body weight under conditions of food abun- corresponding value for women was 170%. The esti- dance and sedentarism particularly among black mates of population attributable risks suggest that Americans.32,33 There is a need to clarify how life- the risk of hypertension could be potentially style habits promote large waist sizes leading to reduced if waist size were reduced to expected or abdominal adiposity and associated CVD in the US. less than expected sizes in black men and women. The result of this investigation shows that the slope of the regression of WC on BMI tends to be Acknowledgements higher in men compared to women. This obser- Data from the NHANES III were obtained from the vation is supported by previous studies that showed US National Center for Health Statistics. the slope of total fatness determined with BMI with blood pressure was higher in men than women.29,30 Although the differential levels of slopes may be References attributable to differences in body components at 1 Okosun IS, Prewitt TE, Cooper RS. Abdominal obesity different levels of BMI for men and women, this in the United States: prevalence and attributable risk could also suggest that the change in absolute fat of hypertension. J Hum Hypertens 1999; 13: 425–430. mass might be less significant in the determination 2 Okosun IS et al. Abdominal adiposity in six popu-

Journal of Human Hypertension Abdominal obesity and hypertension IS Okosun et al 312 lations of West African descent: prevalence and popu- atherogenic lipoprotein subfractions. Int J Obes Relat lation attributable fraction of hypertension. Obes Res Metab Disord 1998; 22: 432–439. 1999; 7: 453–462. 19 Lemieux S et al. Anthropometric correlates to changes 3 Molarius A et al. Varying sensitivity of waist action in visceral adipose tissue over 7 years in women. Int levels to identify subjects with overweight or obesity J Obes Relat Metab Disord 1996; 20: 618–624. in 19 populations of the WHO MONICA Project. J Clin 20 Haffner SM et al. Prospective analysis of the insulin- Epidemiol 1999; 52: 1213–1224. resistance syndrome (syndrome X). Diabetes 1992; 41: 4 Steptoe A, Cropley M, Griffith J, Joekes K. The influ- 715–722. ence of abdominal obesity and chronic work stress on 21 Han TS, Feskens EJ, Lean ME, Seidell JC. Associations ambulatory blood pressure in men and women. Int J of body composition with Type 2 diabetes mellitus. Obes Relat Metab Disord 1999; 23: 1184–1191. Diabet Med 1998; 15: 129–135. 5 Sichieri R, Siqueira KS, Moura AS. Obesity and 22 Seidell JC, Han TS, Feskens EJ, Lean ME. Narrow hips abdominal fatness associated with undernutrition and broad waist circumferences independently con- early in life in a survey in Rio de Janeiro. Int J Obes tribute to increased risk of non-insulin-dependent dia- Relat Metab Disord 2000; 24: 614–618. betes mellitus. J Intern Med 1997; 242: 401–406. 6 al-Asfoor DH, al-Lawati JA, Mohammed AJ. Body fat 23 Ezzati TM, et al. Sample design: Third National Health distribution and the risk of non-insulin-dependent dia- and Nutrition Examination Survey. Vital Helath Stat betes mellitus in the Omani population. East Mediterr 2 1992; 113: 1–35. Health J 1999; 5: 14–20. 24 National Center for Health Statistics. National Health 7 Lee ZS et al. Obesity is the key determinant of cardio- and Nutrition Examination Survey III: data collection vascular risk factors in the Hong Kong Chinese popu- forms. Hyattsville, National Center for Health Stat- lation: cross-sectional clinic-based study. Hong Kong istics: MD, USA, 1991. Med J 2000; 6: 13–23. 25 Joint National Committee on Detection, Evaluation 8 Gonalez CA et al. Factors associated with the accumu- and Treatment of High Blood Pressure. The 1988 Joint lation of abdominal fat estimated with anthropometric National Committee on Detection, Evaluation and indexes. Med Clin (Barc) 2000; 114: 401–406. Treatment of High Blood Pressure. Arch Intern Med 9 Kronmal RA. Spurious correlation and the fallacy of 1988; 148: 1023–1038. the ratio standard revisited. JR Statis Soc A 1993; 156: 26 Norusis MJ. SPSS 10.0 for Windows Release 10.0. 379–392. SPSS Inc. Chicago, 2000. 10 Allison DB, et al. Statistical considerations regarding 27 Han TS, Bijnen FCH, Lean MEJ, Seidell JC. Separate the use of ratios to adjust data. Int J Obes Relat Metab association of waist and hip circumference with life- Disord 1995; 19: 644–652. style factors. Int J Epidemiol 1998; 27: 422–430. 11 Goodman-Gruen D, Barrett-Connor E. Sex differences 28 Kleinbaum DG, Kupper LL, Morgenstern H. Epidemi- in measures of body fat and body distribution in the ologic Research. Principles and Quantitative Methods. elderly. Am J Epidemiol 1996; 9: 898–906. Livetime Learning Pub, 1982. 12 Caan B et al. Changes in measurements of body fat dis- 29 Luke A et al. Relation between body mass index and tribution accompanying weight change. Int J Obes body fat in black population samples from Nigeria, Relat Metab Disord 1994; 18: 397–404. Jamaica, and the United States. Am J Epidemiol 1997; 13 National Institute of Health/National Heart Lungs and 145: 620–628. Blood Clinical Guidelines on the Identification, Evalu- 30 Kaufman JS et al. Relationship between blood pressure ation, and Treatment of Overweight and Obesity in and body mass index in lean populations. Hyperten- Adults. The Evidence Report. Obesity Res 1998; 6: sion 1997; 30: 1511–1516. 51S–209S. 31 Landis JR, Lepkowski JM, Eklund SA, Stehouwer SA. 14 WHO 1990. Diet, Nutrition and the Prevention of A statistical methodology for analyzing data from a Chronic Disease. World Health Organization, Geneva complex survey: the first National Health and (Technical Report Series 797). Nutrition Examination Survey. Vital Health Stat 2 15 Kissebah AH et al. Relation of body fat distribution to 1982; 92: 1–5. metabolic complications of obesity. J Clin Endocrinol 32 Crespo CJ et al. Prevalence of physical inactivity and Metab 1982; 54: 254–260. its relation to social class in U.S. adults: results from 16 Molarius A, Seidell JC. Selection of anthropometric the Third National Health and Nutrition Examination indicators for classification of abdominal fatness – a Survey, 1988–1994. Med Sci Sports Exerc 1999; 31: critical review. Int J Obes Relat Metab Disord 1998; 22: 1821–1827. 719–727. 33 Crespo CJ et al. Race/ethnicity, social class and their 17 Ko GT et al. Simple anthropometric indexes and car- relation to physical inactivity during leisure time: diovascular risk factors in Chinese. Int J Obes Relat results from the Third National Health and Nutrition Metab Disord 1997; 21: 995–1001. Examination Survey 1988–1994. Am J Prev Med 2000; 18 Sattar N et al. Associations of indices of adiposity with 18: 46–53.

Journal of Human Hypertension