International Journal of Obesity (1999) 23, 1136±1142 ß 1999 Stockton Press All rights reserved 0307±0565/99 $15.00 http://www.stockton-press.co.uk/ijo Prediction of hypertension, diabetes, dyslipidaemia or albuminuria using simple anthropometric indexes in Hong Kong Chinese GTC Ko 1 *, JCN Chan 1, CS Cockram 1 and J Woo 1 1 Department of Medicine and Therapeutics, Chinese University of Hong Kong, Prince of Wales Hospital, Shatin, Hong Kong OBJECTIVE: It is important to determine what values of simple anthropometric measurements are associated with the presence of adverse cardiovascular risk factors such as diabetes or hypertension to provide an indication for further detailed investigations. In this analysis, we aimed to assess which anthropometric cutoff values are best at predicting the likelihood of diabetes, hypertension, dyslipidaemia and albuminuria in Hong Kong Chinese. DESIGN AND SETTING: The data were obtained from a previously reported prevalence survey for glucose intolerance in a representative Hong Kong Chinese working population. SUBJECTS: 1513 subjects (910 men and 603 women) with mean age s.d. 37.5 9.2 y. MEASUREMENTS: We examined the likelihood ratios of having diabetes, hypertension, dyslipidaemia and albuminuria in subjects with various cutoff values of the four simple anthropometric indexes, namely, body mass index, waisthip ratio, waist circumference and the ratio of waist-to-height. RESULTS: We developed a nomogram to show the predictive values of different indexes for the cardiovascular risk factors using likelihood ratio analysis. Using Caucasian mean levels of the simple anthropometric indexes to predict diabetes or hypertension in Hong Kong Chinese gave a high likelihood ratio of 2:3:5. CONCLUSION: Higher levels of body mass index, waist-hip ratio, waist circumference and the ratio of waist-to-height are associated with risk of having diabetes mellitus or hypertension in Hong Kong Chinese as in Caucasians. However, the cutoff values of those anthropometric indexes to de ne obesity used in Caucasians may not be applicable to Chinese. Keywords: prediction; hypertension; diabetes; anthropometric indexes; Chinese Introduction Obesity is an important predictor of cardiovascular death 1,2 partly due to its close associations with increased prevalence of hypertension (HT), diabetes mellitus (DM) and dyslipidaemia. 3±7 Body mass index (BMI), waist-hip ratio (WHR) and waist circumference (WC) are all useful anthropometric indexes and provide important information on cardiovascular risks. However, the relative predictive values of these indexes for obesity and cardiovascular risk remain controversial. Some workers have shown that WC was more superior in predicting cardiovascular risks than other measurements. 8±10 More recently, the waist-to-height (WTH) ratio had been shown to have better predictive value than WC alone. 11 Apart from these associations between obesity indexes and cardiovascular risks, it is also important to de ne the cutoff values of an individual index to allow effective screening. For example, in Caucasians, obesity is often de ned as a BMI 27 ± 30 kg=m 2 in both men and women or a WC of 94 cm in men and 80 cm in women. 12 ± 14 However, these de nitions cannot be readily applied to Asians who often have smaller body frames than Caucasians. It is important from the public health perspective to determine what values of simple anthropometric measurements are associated with the presence of adverse cardiovascular risk factors such as diabetes, hypertension of dyslipidaemia, to provide an indication for further detailed investigations. In this analysis based on a population-based survey for cardiovascular risks in Hong Kong Chinese, 10 we examined the likelihood ratios (LR) of having DM, HT, dyslipidaemia and albuminuria in subjects with various cutoff values of the four simple anthropometric indexes, namely, BMI, WHR, WC and WTH. We aimed to assess the effect of different anthropometric cutoff on the risk of DM, HT, dyslipidaemia and albuminuria in Hong Kong Chinese. *Correspondence: Dr Gary TC Ko, Department of Medicine and Therapeutics, The Chinese University of Hong Kong, The Prince of Wales Hospital, Shatin, HK. E-mail: gtc_ko@hotmail.com Received 26 October 1999; revised 7 May 1999; accepted 14 May 1999 Patients and methods The data are obtained from a previously published survey for anthropometric indexes and cardiovascular
risk in Hong Kong Chinese working population. 10 There were 1513 subjects (910 men, 603 women). All subjects were employees from two worksites of a major public utility company and a regional hospital, including all occupational groups from managers to labourers. The distribution of occupational groups in these subjects was similar to that recorded in the Hong Kong Census (1991) and was representative of the Hong Kong working population. 10 The methodology has been described in detail previously. 10 In brief, all subjects attended their worksites after an overnight fast. Demographic data were documented and height and weight (measured to the nearest 0.1 kg) were measured with the subject in light clothing without shoes. Body mass index was calculated as the weight (kg) divided by the square of the height (m). Waist circumference was taken as the minimum circumference between the umbilicus and xiphoid process and measured to the nearest 0.5 cm. Hip circumference was measured as the maximum circumference around the buttocks posteriorly and the symphysis pubis anteriorly and measured to the nearest 0.5 cm. Waist-hip-ratio (WHR) and waist-to-height ratio (WTH) were then calculated. After at least 5 min of sitting, blood pressure (BP) was measured in the right arm by the same research nurse using a standard mercury sphygmomanometer. The Korotkoff sound V was taken as the diastolic BP. The mean value of two readings measured one minute apart was used. Blood was taken after a 12 h fast for measurement of plasma glucose (PG), total cholesterol (TC), fasting triglyceride (TG) and high-density lipoprotein cholesterol (HDL). All subjects underwent a 75-gram oral glucose tolerance test (OGTT). A random spot urine sample was collected for the measurement of urine albumin concentration (Ualb). The various laboratory assays have been described previously. 10 In the present analysis, the WHO criteria were used to diagnose diabetes. 15 Diabetes was de ned as a fasting PG 7.8 mmol=l and=or 2-h PG 11.1 mmol=l. 15 They are all type 2 diabetes in view of the absence of any history of diabetic ketoacidosis. Hypertension was de ned as a systolic BP 140 mmhg and=or diastolic BP 90 mmhg. 16 Dyslipidaemia was considered to be present if plasma total cholesterol 5.2 mmol=l and=or fasting triglyceride 2.3 mmol=l and=or HDL- C < 0.9 mmol=l. 17 Increased albuminuria was de- ned as a random spot urinary albumin concentration 20 mg=l. 18 Statistical analysis Statistical analysis was performed using the Statistical Package for Social Sciences (version 6.0) software on an IBM compatible computer. All results are expressed as mean s.d. or percentage where appropriate. The optimal sensitivity and speci city of using various cutoff values of BMI, WHR, WC and WTH to predict diabetes, hypertension, dyslipidaemia or albuminuria were examined by the receiver operating characteristic curve (ROC) analysis. 19,20 ROC curves were plotted using measures of sensitivity and speci- city based on various anthropometric cutoff values. The ROC curve analysis allows visual evaluation of the trade-offs between sensitivity and speci city associated with different values of the test result. 19,20 The optimal sensitivity and speci city were the values yielding maximum sums from the ROC curves. The Student's t-test and chi-square tests were used for between group comparisons where appropriate. Ageadjusted partial correlation coef cients were used to test the associations amongst the four anthropometric indexes. The likelihood ratio (LR) was calculated to estimate the odds of having diabetes, hypertension, dyslipidaemia or albuminuria in subjects with different cutoff values of the anthropometric indexes. 21 LR is de ned as sensitivity=(1-speci city). The sensitivity and speci city of having diabetes etc at various anthropometric cutoff values were calculated using the whole population of 1513 subjects classi ed with reference to each cutoff level and hence the corresponding LR was derived. According to Simel et al, since LR refers to actual test results before disease status is known, they are more immediately useful to clinicians than sensitivity and speci city. 21 Multiple stepwise logistic regression analysis was used to examine the independent relationship between the four anthropometric indexes and the odds ratio of having DM, HT, dyslipidaemia or albuminuria. A P-value < 0.05 (2-tailed) was considered to be signi cant. Results Of the 1513 subjects, 910 were men and 603 were women. Their anthropometric indexes, prevalence of DM,HT, dyslipidaemia and albuminuria are summarised in Table 1. Men were older, had higher WC, WHR and prevalence of HT, dyslipidaemia and albuminuria than women. There were close associations amongst the four anthropometric indexes of BMI, WC,WHR, WTH, as shown by the age-adjusted partial correlation coef cients. (Table 2). Table 3 summarises the optimal cutoff values of various anthropometric indexes to predict DM, HT, dyslipidaemia or albuminuria using the ROC analysis. The optimal BMI cutoff to predict DM, HT, dyslipidaemia or albuminuria varied from 23.0 ± 24.3 kg=m 2 in both men and women. The optimal WHR cutoff varied from 0.87 ± 0.91 men and 0.80 ± 0.84 in women; the optimal WC cutoff varied from 80 ± 84 cm in men and 75 ± 78 cm in women; and the optimal WTH cutoff varied from 48 ± 51 cm in both men and women. 1137
1138 Table 1 Anthropometric indexes and clinical conditions of the 1513 subjects Total Men Women Variables (n ˆ1513) (n ˆ 910) (n=603) P-value Age, y 37.5 9.2 36.7 9.2 38.6 9.1 < 0.001 BMI, kg=m 2 23.3 3.2 23.4 3.0 23.3 3.5 < 0.556 Waist circumference, cm 78.5 8.5 80.8 7.8 74.9 8.3 < 0.001 WHR 0.85 0.07 0.87 0.05 0.80 0.06 < 0.001 WTH 48.4 5.2 48.4 4.9 48.4 5.7 < 0.817 Diabetes, n (%) 69(4.6) 46(5.1) 23(3.8) 0.257 Hypertension, n (%) 189(12.5) 157(17.3) 32(5.3) < 0.001 Dyslipidaemia, n (%) 728(48.1) 522(57.4) 206(34.2) < 0.001 Albuminuria, n (%) 123(8.1) 62(6.8) 61(10.1) 0.021 Data are mean s.d. or number of subjects, n(%) where appropriate. P-values are comparing men and women using Student t-test or w 2 test where appropriate. BMI; body mass index, WHR; waist-hip ratio. Table 2 Age-adjusted partial correlation coef cient among body mass index, waist-hip ratio, waist circumference and waist-to-height in the 1513 Chinese subjects. (a) men (n ˆ 910) and (b) women (n ˆ 603) BMI WHR WC WTH (a) WHR 0.626* ± 0.796* 0.840* WC 0.900* 0.796* ± 0.931* WTH 0.902* 0.840* 0.931* ± (b) WHR 0.548* ± 0.774* 0.786* WC 0.889* 0.774* ± 0.933* WTH 0.903* 0.786* 0.933* ± P-values: * < 0.001. BMI: body mass index, WHR: waist-hip ratio, WC; waist circumference, WTH: waist-to-height. Figures 1 ± 4 show the likelihood ratios of having DM, HT, dyslipidaemia or albuminuria in subjects with different values of these four anthropometric indexes. They represent nomograms in adult Hong Kong Chinese of working age on various anthropometric indexes in predicting risk of DM, HT, dyslipidaemia or albuminuria. Table 4 summarises the independent relationships between these four anthropometric indexes and the relative risk of having DM, HT, dyslipidaemia or albuminuria using multiple stepwise logistic regression analysis. WHR and=or WTH are independently associated with DM or HT while all four anthropometric indexes are associated with dyslipidaemia or albuminuria in either men or women. An increase of the anthropometric indexes of one unit (1 kg=m 2 in BMI, 0.01 in WHR, 1 cm in WC or 1 in WTH) results in a 1.05- to 1.27-fold increased likelihood of having DM, HT, dyslipidaemia or albuminuria except in the case of dyslipidaemia in men, BMI gives a decreased likelihood. The proportion of variance (Nagelkerke Ð R 2 ) of the anthropometric indexes accounting for DM,HT, dyslipidaemia or albuminuria is relatively small Ð up to 14% in men and 26% in women. This is compatible with the fact that these diseases are heterogeneous and multifactorial, and obesity is only part of the underlying causes. Table 3 The optimal cutoff values of various anthropometric indexes to predict hypertension, diabetes, dyslipidaemia or albuminuria based on ROC analysis Men Women Anthropometric Clinical Optimal Optimal Optimal Optimal indexes outcomes Cut-off sensitivity, % specificity, % Cut-off sensitivity, % specificity, % Hypertension 23.8 61.5 63.0 24.1 64.5 64.5 Diabetes 24.3 66.5 65.5 24.3 66.5 65.5 Dyslipidaemia 23.0 57.5 56.4 23.2 58.0 58.0 Albuminuria 23.5 56.5 57.1 24.1 58.5 58.5 Waist-to-hip-ratio Hypertension 0.886 65.0 65.0 0.838 79.5 79.5 Diabetes 0.908 75.5 76.0 0.828 71.0 70.0 Dyslipidaemia 0.870 60.9 61.1 0.802 61.0 61.0 Albuminuria 0.890 61.3 64.0 0.804 57.0 57.0 Waist circumference, cm Hypertension 82.0 64.3 62.9 78.4 70.5 70.5 Diabetes 84.0 67.4 67.2 78.4 70.0 70.0 Dyslipidaemia 80.0 59.8 59.5 74.6 57.0 57.0 Albuminuria 81.4 56.5 56.5 75.6 58.0 58.0 Waist-to-height Hypertension 50.0 66.2 68.0 51.5 75.0 74.6 Diabetes 50.8 70.5 70.5 51.2 72.5 72.5 Dyslipidaemia 47.9 61.0 61.5 48.5 58.7 59.1 Albuminuria 49.5 59.7 60.3 48.7 55.5 55.5 ROC: receiver operating characteristic curve analysis.
1139 Figure 1 Likelihood ratio to have diabetes, hypertension, dyslipidaemia or albuminuria in various BMI cutoff values. Figure 3 Likelihood ratio to have diabetes, hypertension, dyslipidaemia or albuminuria in various waist-circumference cutoff values. Figure 2 Likelihood ratio to have diabetes, hypertension, dyslipidaemia or albuminuria in various waist-hip ratio cutoff values. Discussion Body mass index, WHR and WC have all been shown to be associated with cardiovascular risk factors such Figure 4 Likelihood ratio to have diabetes, hypertension, dyslipidaemia or albuminuria in various waist-to-height cutoff values. as blood pressure, plasma glucose, lipid pro les and insulin concentrations in Caucasians. 3,22,23 These relationships have also been demonstrated in Asians including Hong Kong Chinese. 6,10,24 Recently, the waist-to-height (WTH) ratio has been suggested to be another useful predictor of cardiovascular risk. 11
1140 Table 4 Multiple stepwise logistic regression analysis using body mass index (BMI), waist circumference (WC), waist-hip ratio (WHR) and ratio of waist-to-height (WTH) as independent viables accounting the risk of having diabetes, hypertension, dyslipidaemia and albuminuria. (a) Men (n ˆ 910) and (b) Women (n ˆ 603) Outcome Independent variables* Odds ratio (95% CI) (a) Diabetes mellitus (Nagelkerke - R 2 ˆ 0.109) WHR 1.16(1.10, 1.23) Hypertension (Nagelkerke - R 2 ˆ 0.140) WTH 1.18(1.14, 1.23) Dyslipidaemia (Nagelkerke - R 2 ˆ 0.118) BMI 0.80(0.72, 0.88) WTH 1.27(1.19, 1.35) Albuminuria (Nagelkerke - R 2 ˆ 0.044) WHR 1.10(1.05, 1.55) (b) Diabetes mellitus (Nagelkerke - R 2 ˆ 0.149) WTH 1.20(1.11, 1.29) Hypertension (Nagelkerke - R 2 ˆ 0.259) WHR 1.26(1.18, 1.35) Dyslipidaemia (Nagelkerke - R 2 ˆ 0.073) WHR 1.09(1.06, 1.13) Albuminuria (Nagelkerke - R 2 ˆ 0.038) WC 1.05(1.02, 1.09) *Variables not included means that they did not enter into the model. Diabetes: diagnosed according to WHO criteria. Hypertension: systolic BP 140 mmhg and=or diastolic BP 90 mmhg. Dyslipidaemia: total cholesterol 5.2 mmol=l and=or fasting triglyceride 2.3 mmol=l and=or HDL < 0.9 mmol=l. Albuminuria: spot sterile urine albumin concentration 20 mg=l. In the present analysis, we found close associations amongst all these four anthropometric indexes. Using multiple regression analysis, WHR and WTH were selected as the main predictors for DM and HT in both men and women. On the other hand, all four indexes provided useful information on dyslipidaemia and albuminuria. However, our results need to be interpreted with caution since the population is limited to those who are of working age. In the west, a BMI 27 kg=m 2 is often used to de ne obesity. Using this de nition, 33% of US adults are considered to be overweight. 12 However, in Hong Kong Chinese of working age, only 11.6% have a BMI 27 kg=m 2. 10 If BMI 30 kg=m 2 is used to de ne obesity, 8 ± 15% of Caucasians 13,25,26 but only 2.2 ± 4.8% of Hong Kong Chinese will be considered to be obese. 10 Similarly, the mean BMI in UK subjects is 26.0 kg=m 2 for men and 26.3 kg=m 2 for women. 14 This is compared to 23.4 kg=m 2 and 23.3 kg=m 2 in Hong Kong Chinese men and women respectively. In one recent UK study conducted by Lean et al, 14 the mean WHR and WC in men were 0.93 and 93.3 cm, and in women, 0.80 and 82.0 cm, respectively. This is compared to a mean value of 0.87 for WHR and 90.8 cm for WC in Chinese men and 0.80 and 74.9 cm respectively for Chinese women, as shown in this study and our previous analysis. 10 Using a WHR of 0.95 for men and 0.80 for women as cutoff values as suggested by Lean et al, 37.6% of men and 48.2% of women in the UK are centrally obese. 14 This is compared to 9.3% of men and 48.1% of women in our study using the same de nitions. 10 Similarly, using a WC of 94 cm for men and 80 cm for women as cutoff values according to Lean et al, 46.7% of men and 50.9% of women in the UK, 14 as compared to 4.5% of men and 25.5% of women in our study are centrally obese. 10 Using these comparisons, it is obvious that if Western de nitions of obesity are used in Asian populations, the latter are often considered to be non-obese. Yet, there is now a wealth of data showing that the prevalence of DM and HT are reaching epidemic proportions in Asians. In migrant studies, the prevalence of these conditions is often higher in Asian than indigenous populations living in western countries, despite less degree of obesity. 27 In 1990, the prevalence of DM in Hong Kong was reported to be 4.5% using the 1985 WHO criteria. 28 In 1996, this has increased to 9.5% and 9.8% in Hong Kong Chinese men and women respectively. 29 This prevalence can be compared to that of 5 ± 10% as reported in Europe and the USA. 15 The recent NHANES III data indicated that diabetes (diagnosed and undiagnosed combined) affects 7.8% of adults 20 y of age in the US. 30 As in most studies, obesity is a strong predictor for DM and related disease such as HT, dyslipidaemia and albuminuria in Hong Kong Chinese. 6,10,31 The use of a simple anthropometric index to identify high risk subjects will be of value in the primary prevention of these prevalent but preventable and treatable conditions. In this analysis, we used the ROC analysis to de ne the optimal cutoff values of these four anthropometric indexes, all of which can be easily measured, to predict DM, HT, dyslipidaemia or albuminuria. We also developed a nomogram to show the predictive values of different indexes for these cardiovascular risk factors using likelihood ratio analysis. As shown in gures 1 to 4, if we choose a LR of 2.5 or above to identify high-risk subjects, in the case of diabetes, the corresponding cutoff values of BMI was 29.5 kg=m 2 for men and 25.5 kg=m 2 for women. The corresponding values of HT were 26.5 kg=m 2 for men and 25.3 kg=m 2 for women. Using the mean BMI level of 26 kg=m 2 derived from normal Caucasian population 14 to predict DM or HT in Hong Kong Chinese gave a LR of 2 in men and 3 in women. In other words, `normal' cutoff values in Western countries when applied to Chinese still pose a signi cant association with DM or HT. The optimal BMI cutoff values of 23 ± 24 kg=m 2 in either men or women (as shown in Table 3) correspond to a LR of around 1.5. Normal or low risk of DM or HT in Hong Kong Chinese is associated with a much lower BMI cutoff than in Caucasians. On the other hand, the present analysis showed that using Caucasian mean levels of WHR (0.93 for men and 0.80 for women) to predict DM or HT in Hong
Kong Chinese gave a LR of 3 ± 3.5 in men and a LR of 2 in women. Similarly, using the mean levels of WC in Caucasian population (93 cm for men and 83 cm for women) to predict DM or HT in Hong Kong Chinese gave a LR of 2 ± 3.5 in both men and women. So the `normal' levels of anthropometric indexes in Caucasians may not be applicable to Asian population. Much larger scale community-based survey is needed in different ethnic groups to de ne their own normal cutoff. The present study suggests that a BMI of 24 kg=m 2 in either sex, a WHR of 0.88 in men and 0.80 in women, a WC of 82 cm in men and 76 cm in women, and a WTH of 50 in either sex may be more `appropriate' to be regarded as cutoff in Hong Kong Chinese of working age. Nevertheless, prospective studies are needed to examine the relationship between different level of obesity and cardiovascular risk in Asian population In this analysis, dyslipidaemia and albuminuria did not show as good an association with the anthropometric indexes as hypertension and diabetes. In accordance with this, blood pressure and plasma glucose measurement is usually the initial screening, followed by plasma lipid levels and urine albumin, on cardiovascular risk in asymptomatic obese subjects. Concerning which of the four anthropometric indexes is `better' than the others, the BMI and WHR are widely used and there is ample evidence, including the present analysis, suggesting that BMI and WHR are closely associated with various cardiovascular risk factors. However, it also seems that all four indexes are useful and appears to complement one another. Conclusion Higher levels of body mass index, waist-hip ratio, waist circumference and the ratio of waist-to-height are associated with risk of having diabetes mellitus or hypertension in Hong Kong Chinese as in Caucasians. However the cutoff values of those anthropometric indexes to de ne obesity used in Caucasians may not be applicable to Chinese. Larger scale communitybased study is needed to clarify the discrepancy. References 1 Jousilahti P, Tuomilehto J, Vartianinen E, Pekkanen J, Puska P. Body weight, cardiovascular risk factors, and coronary mortality. 15-year follow-up of middle-aged men and women in eastern Finland. Circulation 1996, 93: 1372 ± 1379. 2 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. 3 Higgins M, Kannel W, Garrison R, Pinsky J, Stokes J. Hazards of obesity Ð the Framingham experience. Acta Med Scand 1988, 723 (Suppl): S23 ± S36. 4 Haffner SM, Stern MP, Hazuda HP, Pugh J, Patterson JK. Do upper-body and centralized adiposity measure different aspects of regional body-fat distribution? relationship to noninsulin-dependent diabetes mellitus, lipid, and lipoproteins. Diabetes 1987, 36 43 ± 51. 5 Srinivasan SR, Bao W, Wattingney WA, Berenson GS. Adolescent overweight is associated with adult overweight and related multiple cardiovascular risk factors: the Bogalusa Heart Study. Metabolism 1996, 45: 235 ± 240. 6 Woo J, Ho SC, Chan SG, Sham A, Yuen YK, Masarei JL. Lipid pro le in the Chinese elderlies: comparison with younger age groups and relationship with some cardiovascular risk factors and presence of diseases. Cardiology 1993, 83: 407 ± 414. 7 Ko GTC, Chan JCN, Cockram CS. The association between dyslipidaemia and obesity in Chinese men after adjustment for insulin resistance. Antherosclerosis 1998, 138: 153 ± 161. 8 Pouliot MC, Despres JP, Lemieux S, Moorjani S, Bouchard C, Tremblay A, Nadeau A Lupien PJ. Waist circumference and abdominal sagittal diameter: best simple anthropometric indexes of abdominal visceral adipose tissue accumulation and related cardiovascular risk in men and women. Am J Cardiol 1994, 73: 460 ± 468. 9 Seidell JC, Cigolini M, Charzewska J, Ellsinger BM, Deslypere JP, Cruz A. Fat distribution in European men: a comparison of anthropometric measurements in relation to cardiovascular risk factors. Int J Obes Relat Metab Disord 1992, 16: 17 ± 22. 10 Ko GTC, Chan JCN, 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 Obesity. 1997; 21 995 ± 1001. 11 Ashwell M, Lejeune S, McPherson K. Ratio of waist circumference to height may be better indicator of need for weight management. Br Med J1996; 312 377. 12 Kuczmarski RJ, Flegal KM, Campbell SM, Johnson CL. Increasing prevalence of obesity among US adults. The National Health and Nutrition Examination Surveys, 1960 to 1991. JAMA 1994, 272: 205 ± 211. 13 Reeder BA, Angal A, Ledoux M, Rabkin SW, Young TK, Sweet LE. Obesity and its relation to cardiovascular disease risk factors in Canadian adults. Can Med Assoc J 1992, 146: 2009 ± 2019. 14 Lean MEJ, Han TS, Morrison CE. Waist circumference as a measure for indicating need for weight measurement. Br Med J 1995, 311: 158 ± 161. 15 World Health Organisation. Diabetes Mellitus. Report of a WHO Study Group. Technical Report Series 727 World Health Organisation: Geneva, 1985. 16 American Heart Association Science Advisory Guide to Primary Prevention of Cardiovascular Diseases: A Statement for Healthcare Professionals from the Task Force on Risk Reduction. Circulation 1997, 95: 2329 ± 2331. 17 American Diabetes Association. Management of dyslipidaemia in adults with diabetes. Diabetes Care 1998, 21 (Suppl): S36 ± S39. 18 Schwab SJ, Dunn FL, Feinglos MN. Screening for microalbuminuria. Diabetes Care 1992, 15: 1581 ± 1584. 19 Van der Schouw YT, Verbeek AL, Ruijs JH. ROC curves for the initial assessment of new diagnostic tests. Fam Pract 1992, 9: 506 ± 511. 20 Grzybowski M, Younger JG. Statistical methodology: III. Receiver operating characteristic (ROC) curves. Acad Emerg Med1997; 4 818 ± 826. 21 Simel DL, Samsa GP, Matchar DB. Likelihood ratios with con dence: sample size estimation for diagnostic test studies. J Clin Epidemiol 1991, 44: 763 ± 770. 22 Kannel WB, Cupples LA, Ramaswami R, Stokes J, Kreger BE, Higgins M. Regional obesity and risk of cardiovascular disease; the Framingham Study. J Clin Epidemiol 1991, 44: 183 ± 190. 23 Bjorntorp P. The associations between obesity, adipose tissue distribution and disease. Acta Med Scand 1988, 723 (Suppl): 121 ± 134. 24 Masuda T, Imai K, Komiya S. Relationship of anthropometric indices of body fat to cardiovascular risk in Japanese women. Ann Physiol Anthropol 1933, 12: 135 ± 144. 1141
1142 25 Millar WJ, Stephens T. The prevalence of overweight and obesity in Britain, Canada, and United States. Am J Public Health 1987, 77: 38 ± 41. 26 National Heart Foundation of Australia. Risk factor prevalence study. Report No. 3, 1990. 27 Zimmet P. Challenges in diabetes epidemiology from West to the rest. Diabetes Care 1992, 15: 232 ± 252. 28 Cockram CS, Woo J, Lau E, Chan JCN, Chan AY, Lau J, Swaminathan R, Donnan SP: The prevalence of diabetes mellitus and impaired glucose tolerance among Hong Kong Chinese adults of working age. Diabetes Res Clin Pract 1993, 21: 67 ± 73. 29 Janus ED. Hong Kong Cardiovascular Risk Factor Prevalence Study 1995 ± 1996. Hong Kong 1997. 30 Harris MI, Flegal KM, Cowie CC, Eberhardt MS, Goldstein DE, Lettle RR, Wiedmeyer HM, Byrd-Holt DD. Prevalence of diabetes, impaired fasting glucose, and impaired glucose tolerance in U.S. adults: the Third National and Nutrition Examination Survey, 1988 ± 1994. Diabetes Care 1998, 21: 518 ± 524. 31 Chan JCN, Cheung JCK, Lau EMC, Woo J, Swaminathan R, Cockram CS. The Metabolic Syndrome in Hong Kong Chinese Ð the inter-relationships amongst its components analysed by structural equation modeling. Diabetes Care 1996, 19: 953 ± 959.