Introduction. BD Cox 1, MJ Whichelow 1 and AT Prevost 2. Health, Cambridge, UK

Transcription

1 International Journal of Obesity (1998) 22, 966±973 ß 1998 Stockton Press All rights reserved 0307±0565/98 $ The development of cardiovascular disease in relation to anthropometric indices and in British adults BD Cox 1, MJ Whichelow 1 and AT Prevost 2 1 Department of Community Medicine and 2 Centre for Applied Medical Statistics, Department of Community Medicine, Institute of Public Health, Cambridge, UK OBJECTIVE: To examine the predictive ability of simple anthropometric indices for the development of cardiovascular disease (CVD) over seven years in British adults, and the in uence of on these associations. DESIGN: Longitudinal study of the development of CVD (morbidity or mortality) over a seven year period in a random strati ed sample of British adults who were respondents in the 1984 ±1985 Health and Lifestyle Survey (HALS1) and who were seen again in 1991±1992 (HALS2) or who had died by then. METHODS: Face-to-face interviews at HALS1 and HALS2 recorded socio-demographic, health and lifestyle details followed by measurements of height, weight, waist circumference and blood pressure (BP). The quintiles of body mass index (BMI), waist circumference, waist : height ratio (WHTR) and height were calculated for those aged between 35 and 75 y, at HALS1. Dates and causes of death were recorded. Logistic regression was used to estimate the odds ratios (OR) of developing CVD in 1284 men and 1570 women, aged 35±75 y who were free of known CVD, cancer and diabetes at HALS1. RESULTS: By HALS2, 316 respondents in the qualifying population had developed CVD, 114 of whom were dead. There were linear trends in the development of CVD (adjusted for age and smoking) for all the men from the lowest to the highest quintile of WHTR (P ˆ 0.034), but not for waist circumference (P ˆ 0.095), or BMI (P > 0.2). Excluding the hypertensive men increased the signi cance of the trend for WHTR (P ˆ 0.005) and waist circumference (P ˆ 0.027). The signi cant interactions with for WHTR (P < 0.001), waist circumference (P ˆ 0.006) and BMI (P ˆ 0.044) showed that there was an increasing incidence of CVD with increasing adiposity in non-hypertensive men but, in men with treated, although the overall incidence of CVD was higher, the relationship with adiposity was inverse. In the women, there were no signi cant linear trends for waist circumference, WHTR or BMI. Quintile estimates were more consistent with J-shaped curves with the lowest risk in the second quintile. Excluding the hypertensive women, increased the signi cance of these trends. In normotensive women, there was a signi cant quadratic trend (P ˆ 0.039) for the association between the incidence of CVD and the quintiles of waist circumference, but no associations for WHTR or BMI. For waist circumference there was weak evidence of an interaction with (P ˆ 0.053). CONCLUSIONS: For the men, indices involving waist circunference, particularly WHTR, had stronger linear associations with the log odds of CVD development than BMI. The interactions with were signi cant for WHTR, waist circunferenceand also BMI. In women, none of the indices was linearly associated with the log odds of CVD development, but there was a signi cant J-shaped curve for waist circumference and evidence of an interaction with. These results suggest that studies in which hypertensives are included, but in which possible interactions are overlooked, important hypertensive-speci c associations between anthropometric indices and CVD development may be masked. Men on anti-hypertensive medication with the lowest central adiposity, experienced higher short-term CVD risk than those with greater central adiposity. Keywords: cardiovascular disease; BMI; waist : height ratio; waist circumference; Introduction Many prospective studies in Europe and the USA, with duration of follow-up varying from 3±29 y have shown that obesity, as measured by body mass index (BMI) was associated with cardiovascular disease (CVD) in men 1±3 and women, 2±4 coronary heart disease (CHD) in men 2,3,5,6,8 and women, 2,3,7±9 and stroke in men 3,8,10,11 and women. 2,3,8 However, there is not complete Correspondence: Dr Brian Cox, Department of Community Medicine, Institute of Public Health, Robinson Way, Cambridge, CB2 2SR, UK Received 24 July 1997; revised 5 May 1998; accepted 15 May 1998 agreement as other workers found no relationship with CHD in men 10 or women, 12 or stroke in men 2,13,14 or women, 7 and some studies showed U or J shaped curves for women for CVD, 4 CHD 3,8 and stroke. 8 Anthropometric indices measuring central adiposity ± waist circumference, waist : hip ratio (WHR), waist : height ratio (WHTR) and trunk skinfold thicknesses ± have also frequently been found to be more closely correlated with CVD than BMI. Higgins et al, 15 reported that in a 28-year follow-up study, a large waist circumference, a large WHTR and high BMI levels were related to CVD and CHD in men and women, and Kannel et al 2 showed that WHTR was related to CVD, CHD and stroke in both genders. Welin et al 13 found that both large waist circumference and

2 waist : hip (WHR), but not BMI values, were associated with an increased incidence of stroke in men, as did Walker et al 14 for WHR and stroke. Large WHR, but not BMI values have also been found to be related to CHD in women, 16,17 but not men. 17 Lapidus et al 7 showed that, adjusted for age only, WHR and BMI were both associated with CHD in women, but that the association was much stronger for the WHR. Most studies excluded those with pre-existing disease and adjusted for age and smoking. When also adjusting for other cardiovascular risk factors, such as diabetes and, the associations were attenuated. However, since diabetes and are also associated with obesity 18 and can be considered to be intermediaries 6,9,11,12,14 in the pathological process between obesity and CVD, such attenuation may be expected. However, other workers 3 reported an interaction between blood pressure (BP), BMI and CHD, which suggested that might not always have been an intermediary, and Cambien et al 19 found a decreasing risk of cardiovascular death with increasing BMI in hypertensive men. Short stature has been associated with cardiovascular mortality in women, 4,20±22 in studies where exclusions for pre-existing conditions were not mentioned, but the association became insigni cant when adjusted for lung function (FEV 1 ). 20 Short men have also been reported to be at more risk than taller men, 6,21 but not when adjustments for risk factors were made. 21 However, Kannam et al 21 found that there was a residual effect of height in women. From the Health and Lifestyle Survey of 1984± 1985 (HALS1), 23 close associations of elevated BP with measures of obesity, particularly central adiposity, have been identi ed in both men and women. 18 HALS1 23 and the follow-up study of 1991±1992 (HALS2) 24 provide the opportunity to examine the predictive power of several simple anthropometric indices for CVD over a relatively short period of time and to assess the impact of on any such associations. Methods This study observed 1284 men and 1570 women, aged 35±75 y, in whom valid anthropometric and BP measurements were made at HALS1 when they were not known to be suffering from cancer, diabetes or CVD and whose cardiovascular status at HALS2 was known because they were either re-surveyed or had died from CVD by then. HALS1 was carried out on a random strati ed sample of 9003 adults, resident in England, Scotland and Wales (a response rate of 77.6%). 23 They were selected from two electoral wards in each of 198 constituencies, chosen as being representative of the region. The sample compared well with the 1981 census in terms of age and gender structure. The CVD, and anthropometry respondents were seen in their own homes by a trained interviewer, who administered a structured questionnaire to elicit data about socio-demographic and background details, lifestyle (dietary, drinking, smoking, exercise and leisure habits), health attitudes and beliefs, current health status and past diseases. Consenting respondents (82.4%) were then visited by a nurse who carried out anthropometric, BP and lung function measurements and recorded details of any medication being taken. The survey was ` agged' with the NHS Central Register at Southport so that noti cation of dates and causes of death of all respondents who have died are received. Seven years later in 1991±1992 the survey was repeated (HALS2) on as many of the surviving respondents as could be traced and who were within the scope of the sampling areas. There were then 5352 subjects. 24 The anthropometric measurements were made with the subjects in indoor clothing and without shoes. Weight was measured with electronic scales (regularly calibrated and checked) to the nearest 100 g, with an allowance being made for what the nurse judged to be heavy or light indoor clothing. Prior to the survey, bundles of indoor clothing were weighed to identify the allowances to be made. Height was measured with a portable stadiometer, with the subject standing upright with the eye and ear aligned in the Frankfurt plane. Waist circumference was measured with a plastic tape mid-way between the lowest rib and the top of the hip. From the anthropometric measurements, the BMI and WHTR were calculated. BP was measured by oscillometry using an automatic Datascope `Accutorr', with the subject sitting comfortably and the arm supported at heart height. Four readings of systolic and diastolic pressures were recorded and the lowest used for analysis. The use of the Accutorr eliminated observer error, and the dif culties of hearing systolic and diastolic sounds in noisy domestic environments. The machines were originally evaluated by comparison with intra-arterial catheterization and regularly checked in use against a mercury manometer. Three different cuffs were available, the nurse choosing the most suitable one for the arm size. Details of any medication for, which could affect the BP, were recorded. The presence of CVD at HALS1 and HALS2 was assumed if the responses or medication used indicated that the subject had had a myocardial infarction, heart by-pass operation, angina pectoris or a stroke. Mortality from CVD was identi ed from ICD9 codes 390 to 459 on the death certi cates. Within the 35±75 y selected age range those excluded from this study were: 452 men and 618 women alive, but not seen at HALS2, and whose cardiovascular status was therefore not known, a further 103 men and 86 women who died of a non- CVD cause between HALS1 and HALS2, a further 30 men and 79 women who had cancer at HALS1, 24 men and 23 women who had diabetes at HALS1,

3 968 CVD, and anthropometry men and 105 women with pre-existing CVD at HALS1 and nally 11 men and 16 women in whom some of the anthropometric measurements could not be made due to missing limbs, pregnancy or an inability to stand up straight. The respondents were categorized according to their BP status as normotensive, medication-treated hypertensives and hypertensive by measurement, that is, not previously known to be hypertensive, but with a BP measured by the nurse at 160 mm Hg systolic or above and/or 95 mm Hg diastolic or above. The respondents were classi ed as non-smokers (those who had never regularly smoked as much as one cigarette, pipe or cigar a day for six months), light smokers (current regular smokers who smoked cigars, pipes or 15 cigarettes a day), heavy smokers (those currently smoking > 15 cigarettes a day) and exsmokers (those who use to smoke regularly). Statistical analyses Quintiles of the waist circumference, WHTR, BMI and height were calculated from the values of all the 35±75 y old subjects who had valid anthropometric measurements at HALS1 (2057 men and 2497 women). 23 Logistic regression models (SPSS) were used to estimate the seven year incidence of CVD (morbidity and/or mortality) across the quintiles of each anthropometric index. In order to test whether the quintile estimates could be summarised by a linear trend, the quintiles were replaced in the model by a linear term representing their continuous values. A quadratic term was then added to this model to test whether quintile estimates could be summarised by a J-shaped curve. In all analyses, HALS1 age, in ve year categories, and HALS1 smoking, as a categorical variable, were adjusted for, since the proportions of smokers, ex-smokers and non-smokers varied considerably across the quintiles of the anthropometric indices. Social class (non-manual and manual) was also entered, but found not to affect the results and was thereafter omitted. The likelihood ratio test statistic was compared to chi-squared distributions with one degree of freedom in all signi cance tests for linear and quadratic trends across the quintiles. In order to investigate the role of in any association between the anthropometric indices and CVD, the logistic regression analyses were carried out rst on all subjects. In a second set of analyses, BP measured at HALS1 was then entered as a covariate having three categories ± non-hypertensive, hypertensive detected at measurement and treated hypertensive. Interactions between and linear followed by quadratic trends across the quintiles, were then added to the model and were assessed by comparing the likelihood ratio test statistic to a chi-squared distribution with two degrees of freedom. The regression lines for quintiles of WHTR and CVD were obtained separately for the non-hypertensive, hypertensive at measurement and treated hypertensive men from the interaction model. A further set of analyses was performed on the nonhypertensive subjects alone. Results Table 1 shows the incidence of CVD over seven years for the sample with relation to the presence or absence of. The lowest incidence of morbidity and mortality was among the non-hypertensive women, with a somewhat higher rate among the non-hypertensive men. Men and women with treated had the highest rates of CVD, closely followed by men with measured, but previously undiagnosed,. Overall, there was a lower incidence of CVD amongst the women (8.8%) than the men (13.9%). Table 2 shows the range of values in the quintiles for each anthropometric index. The results of the logistic regression analyses, adjusting for age and smoking, are given in Table 3 for men and Table 4 for women. For each of the anthropometric indices the odds ratios (ORs) are shown compared to the quintile with the lowest incidence of CVD for all subjects, for all subjects adjusting for and for non-hypertensive respondents only. The test for interaction with is also shown. In men in the all-inclusive group (normotensive, hypertensive by measurement and treated hypertensives), WHTR (not adjusted for ) showed a signi cant positive linear association with Table 1 Development of cardiovascular disease (CVD) in respondents aged 35 ±75 y at the 1984± 1985 Health and Lifestyle Survey (HALS1) 23 in relation to blood pressure (BP) status at HALS1 Men Women No Measured Treated No Measured Treated n % n % n % n % n % n % Seven year status No CVD CVD alive CVD dead Total

4 Table 2 CVD, and anthropometry Anthropometric indices ± range of values in each quintile (Q) in respondents aged 35± 75 y 969 Men Women Q BMI Waist (cm) W/H BMI Waist (cm) W/H ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Table 3 Odds ratios (OR) for developing cardiovascular disease (CVD) (morbidity and/or mortality) by HALS2 24 in relation to anthropometric indices in men aged 35±75 y at HALS1, 23 adjusting for age and smoking All subjects not adjusting for All subjects adjusting for Hypertension by linear trend interaction Only non-hypertensive subjects (n ˆ1284) (n ˆ1284) (n ˆ1284) (n ˆ1072) OR (95% CI) OR (95% CI) Chi-squared OR (95% CI) Waist circumference Quintile w 2 ˆ (0.64± 2.13) 1.16 (0.63± 2.13) 1.80 (0.89± 3.61) (0.68± 2.13) 1.16 (0.65± 2.06) df ˆ (0.67± 2.74) (0.83± 2.47) 1.34 (0.78± 2.32) 2.35 (1.23± 4.51) (0.87± 2.59) 1.33 (0.77± 2.32) P ˆ (1.01± 3.95) Linear trend w 2 ˆ 2.78 P ˆ w 2 ˆ 1.29 P > 0.2 w 2 ˆ 4.91 P ˆ Waist/height ratio (WHTR) Quintile w 2 ˆ (0.59± 4.24) 1.07 (0.58± 1.97) 1.45 (0.70± 2.99) (0.67± 2.13) 1.17 (0.66± 2.09) df ˆ (0.77± 3.09) (0.81± 2.48) 1.34 (0.76± 2.34) 2.25 (1.15± 4.39) (0.97± 2.95) 1.50 (0.85± 2.65) P < (1.18± 4.71) Linear trend w 2 ˆ 4.52 P ˆ w 2 ˆ 2.67 P ˆ 0.10 w 2 ˆ 8.01 P ˆ Body mass index (BMI) Quintile w 2 ˆ (0.72± 2.25) 1.22 (0.71± 2.09) 1.54 (0.81± 2.93) (0.71± 2.16) 1.20 (0.68± 2.10) df ˆ (0.85± 3.24) (0.67± 2.07) 1.15 (0.65± 2.03) 1.86 (0.96± 3.59) (0.76± 2.27) 1.16 (0.66± 2.03) P ˆ (0.71± 2.91) Linear trend w 2 ˆ 0.59 P > 0.2 w 2 ˆ 0.10 P > 0.2 w 2 ˆ 1.40 P > 0.2 HALS2 ˆ 1991±1992 follow-up to HALS1; HALS1 ˆ 1984± 1985 Health and Lifestyle Survey; 95% CI ˆ 95% con dence interval. the development of CVD (P ˆ 0.034), but none of the ORs were signi cantly different from the reference value. Some evidence of increasing CVD incidence with increasing waist circumference was observed, although the test for linear association was not signi cant (P ˆ 0.095). In contrast, there was no indication of a trend for BMI. Adjusting for BP (in three categories) attenuated the associations, with the relationship for WHTR also becoming non-signi cant. When the non-hypertensives were examined, the positive trends for CVD with WHTR and waist circumference became stronger and signi cant, and the ORs for the two highest quintiles were signi cantly greater than that for the lowest quintile. The effect was monotonic, increasing from the lowest to the highest quintile exceeding, on average exceeding a 10% elevated risk of CVD per quintile. There were no signi cant associations for BMI and no signi cant quadratic trends for any of the anthropometric indices. There were highly signi cant interactions between BP and the linear trend across the quintiles of waist circumference and WHTR (Table 3) indicating different relationships between the quintiles of the anthropometric indices and the development of CVD for normotensives, hypertensives by measurement and treated hypertensives. Thus the linear trends for the treated hypertensives, measured hypertensives and non-hypertensives, were signi cantly different from each other and this is demonstrated for WHTR in Figure 1. For non-hypertensive men, the risk of developing CVD increased linearly with increasing WHTR, whilst for the treated hypertensive men, there was a strong inverse association. Those treated hypertensives with the highest risk were in the lowest quintile of WHTR, with those in the highest quintile having a similar risk to the normotensives in that quintile. Those identi ed as hypertensive by measurement occupied an intermediate position. The patterns for waist circumference and BMI were similar. For women the pattern of results was different (Table 4) and the association of CVD incidence with the anthropometric indices was weaker than for the

5 970 CVD, and anthropometry Figure 1 Trends in the log odds of developing cardiovascular disease (CVD) in men, in non-hypertensive, hypertensives detected at measurement and treated hypertensives in relation to non-hypertensives in the lowest quintile of waist:height ratio (WHTR), age and smoking. men. There were no signi cant linear trends in any of the groups for waist circumference, WHTR or BMI. Since those in the second quintile had, consistently, the lowest incidence of CVD, quadratic terms were used to describe the trends. Only for waist circumference in the non-hypertensive women was the quadratic trend signi cant (P ˆ 0.039). The rst and fth quintiles of waist circumference had ORs signi cantly above that of the second quintile. For all the women, the ORs for the rst, fourth and fth quintiles of waist circumference were signi cantly greater than that of the second quintile, although the quadratic trend was not signi cant (P ˆ 0.10). There were no signi cantly different ORs or trends for any group for WHTR or BMI. Some evidence of an interaction between and the quadratic trend across waist was observed, although this was not signi cant (P ˆ 0.053). There was no indication of any associations, linear or quadratic, between height and CVD in either men or women (results not shown). There were signi cant independent associations with the development of CVD for both and current smoking habit in men, and for treated, but not smoking in women. These were independent of anthropometric associations as represented by WHTR in Table 5. Table 4 Odds ratios (OR) for developing cardiovascular disease (CVD) (morbidity and/or mortality) by HALS2 24 in relation to anthropometric indices in women aged 35 ±75 y at HALS1, 23 adjusting for age and smoking (reference quintile 2) All subjects not adjusting for All subjects adjusting for Hypertension by linear trend interaction Only non-hypertensive subjects (n ˆ1570) (n ˆ1570) (n ˆ1570) (n ˆ1364) OR (95% CI) OR (95% CI) Chi-squared OR (95% CI) Waist circumference Quintile (1.06±4.81) 2.38 (1.11±5.11) w 2 ˆ (1.02±5.28) ± 1.00 ± 1.00 ± (0.87± 3.74) 1.91 (0.92± 3.98) df ˆ (0.72± 3.68) (1.13± 4.59) 2.16 (1.07± 4.37) 1.48 (0.65± 3.38) (1.29± 5.12) 2.22 (1.10± 4.48) P ˆ (1.09± 5.43) Linear trend w 2 ˆ 2.72 P ˆ 0.10 w 2 ˆ 0.82 P > 0.2 w 2 ˆ 0.30 P > 0.2 Quadratic trend w 2 ˆ 1.98 P ˆ 0.16 w 2 ˆ 0.59 P > 0.2 w 2 ˆ 4.28 P ˆ Waist/height ratio (WHTR) Quintile (0.55±2.43) 1.08 (0.56±2.80) w 2 ˆ (0.56±2.86) ± 1.00 ± 1.00 ± (0.56± 2.17) 1.04 (0.60± 2.61) df ˆ (0.57± 2.52) (0.92± 3.15) 1.51 (0.70± 2.90) 1.37 (0.49± 2.37) (0.90± 3.14) 1.35 (0.73± 3.23) P > (0.69± 3.12) Linear trend w 2 ˆ 3.62 P ˆ w 2 ˆ 1.39 P > 0.2 w 2 ˆ 0.59 P > 0.2 Quadratic trend w 2 ˆ 0.18 P > 0.2 w 2 ˆ 0.00 P > 0.2 w 2 ˆ 0.25 P > 0.2 Body mass index (BMI) Quintile (0.90±3.51) 1.82 (0.92±3.63) w 2 ˆ (0.79±3.72) ± 1.00 ± 1.00 ± (0.94± 3.46) 1.81 (0.94± 3.51) df ˆ (0.85± 3.76) (0.76± 2.83) 1.40 (0.72± 2.73) 1.07 (0.49± 2.37) (0.82± 3.06) 1.33 (0.68± 2.61) P > (0.60± 3.07) Linear trend w 2 ˆ 0.01 P > 0.2 w 2 ˆ 0.30 P > 0.2 w 2 ˆ 0.35 P > 0.2 Quadratic trend w 2 ˆ 0.18 P > 0.2 w 2 ˆ 0.00 P > 0.2 w 2 ˆ 0.07 P > 0.2 HALS2 ˆ 1991±1992 follow-up to HALS1; HALS1 ˆ 1984±1985 Health and Lifestyle Survey; 95% CI ˆ 95% con dence interval.

6 CVD, and anthropometry Table 5 Odds of developing cardiovascular disease (CVD) over seven years, in relation to smoking and, age and waist : height ratio (WHTR) 971 Men Women (n ˆ1284) (n ˆ1570) OR (95% CI) OR (95% CI) Smoking habits (adjusting for ) Non-smokers Light smokers 1.79 (1.04±3.09) 1.43 (0.85±2.40) Heavy smokers 1.93 (1.12±3.32) 1.46 (0.79±2.71) Ex-smokers 1.25 (0.76±2.04) 0.84 (0.52±1.37) Trend w 2 ˆ 8.14 df ˆ 3 P < 0.05 w 2 ˆ 4.40 df ˆ 3 NS Blood pressure (adjusting for smoking) Non-hypertensive Hypertensive by measurement 1.68 (1.06±2.66) 1.47 (0.84±2.59) Treated hypertensive 1.79 (1.62±3.14) 2.95 (1.70±5.13) Trend w 2 ˆ 7.36 df ˆ 2 P < 0.05 w 2 ˆ df ˆ 2 P < OR ˆ odds ratio; 95% CI ˆ 95% con dence intervals; NS ˆ not signi cant. Discussion The present study, carried out on a random sample of British adults, who were followed up over the relatively short period of seven years, con rms ndings in other studies, of an association of central adiposity with cardiovascular disease. 2,7,10,12±15 Although the trends for BMI were in the same direction, they were not so strong or signi cant, nor did the ORs in any of the quintiles differ from others. These ndings are in accord with those of other workers, who could not nd a signi cant relationship between BMI and CHD 10,12 or stroke. 2,7,13,14 Had the follow-up period of the present study been longer, the association of CVD with BMI might have become stronger and achieved signi cance, since most of the reports of BMI as a risk for CVD had follow-up times longer than seven years 2,7,11,15 or a much larger cohort. 6,9,11 This particularly applies to the women where the low incidence of CVD resulted in relatively few cases (Table 1). Thus the present ndings indicate that, particularly in men, measurements of central adiposity are predictors of CVD in the short term and since other workers have also shown variously that waist circumference, WHR and WHTR are closely associated with CVD development over 12 years or more, 2,7,10,13,15 they are also important in the long term. Further support for this comes from examination of the data given by Kannel et al, 2 which shows that the difference between the lowest and highest quintiles of WHTR in the age adjusted rates of CVD, increase steadily over the 24 y of follow-up, being quite small at seven years of follow-up and 3±5 times as great at 24 y of follow-up. The observation that the relationship between CVD incidence and the anthropometric indices was J-shaped, rather than linear, in normotensive women, was in accord with the ndings of Selmer and Tverdal 3 who also showed, in Norwegian women, but not men, that cardiovascular mortality over 27 years of follow-up was J-shaped in relation to BMI. Similarly LaÈaÈraÈ and Rantakallio 4 found that cardiovascular mortality was lowest in women with a BMI in the rage 21±25 (equivalent to quintiles 2 plus 3 in the present study). In the men, the observation that there were signi cant interactions with (particularly in the indices re ecting central adiposity) showed that the relationships with CVD for those with and without were different. In fact, the present ndings suggest that for medication treated hypertensives, the risk of CVD is much greater for the thinner than for the more obese men. In a population of men aged 50±79 y, Barratt-Conner and Khaw 25 also found that non-obese hypertensive men had a signi cantly higher cardiovascular mortality rate than did obese hypertensive men, or lean or obese non-hypertensives. Many other studies have not adjusted for or considered hypertensives separately, presumably because of the view, as expounded by a number of authors, 6,9,11,12,14 that is an intermediary between obesity and CVD. Thus any interaction of in the relationship of CVD and anthropometric indices, will not have been revealed. However, the present ndings are consistent with those of Selmer and Tverdal, 3 who reported interactions between systolic BP, BMI and CHD mortality in the rst ve years of a 27 y follow-up in older men and women. Possible explanations for this, are suggested by Cambien et al 19 who described, in a large study, a decreasing risk of cardiovascular death amongst hypertensives with increasing BMI. They discounted any variation between obese and lean subjects in the measurement of BP, but postulated that, in view of the known association between and obesity, the more obese subjects might have been more vigorously treated than the lean hypertensives. However, in

7 972 CVD, and anthropometry the present study, there was no obvious difference in the prevalence of normalised BP between the leaner and fatter hypertensives. Another hypothesis put forward by Cambien et al, 19 based on the suggestion of Messerli et al, 26 was that cardiac output and peripheral resistance contribute to a different extent to the level of blood pressure in obese, overweight and lean hypertensive subjects, with the obese hypertensives having a higher cardiac output. The implication was, that elevated peripheral resistance, which for any one level of BP would be higher in lean hypertensives, is more hazardous than increased cardiac output in terms of CVD development. If the ndings of the present study are con rmed, it can be hypothesised that central adiposity, as identi ed by measures which include waist circumference, is an independent risk factor for CVD in non-hypertensive men. Hypertension itself in all circumstances is an independent risk factor and may act with obesity as an intermediary in the development of CVD, but in some obese subjects the may re ect a physiological response to the obesity, as demonstrated in the observations that weight (and waist) reduction is an effective treatment for. However in the leaner hypertensives, where weight reduction is not an option and where is not an intermediary with obesity, is, it appears, a more powerful predictor of CVD. From the point of view of the assessment of risk of obesity in the clinical or survey situation or by the public, waist circumference and WHTR are simple indices which can be measured accurately enough with any tape measure. For men, risk of CVD development (at least in the non-hypertensive) is positively related to size. Lean et al 27 have investigated the relationships of waist circumference to BMI and WHR, and also the sensitivity and speci city, of particularly, waist circumference in predicting cardiovascular risk factors such as hypercholesterolaemia, low levels of high density lipoprotein (HDL) levels and. Based on their ndings, they have recommended `action levels' for waist measurements for the reduction of CVD. 28 Waist reduction has been shown to be concomitant with weight loss 29 and weight reduction is thus recommended for men with a waist measurement > 94 cm (Quintiles 4 and 5 in the present study) and > 80 cm for women (Quintiles 4 and 5 in the present study). 30 For non-hypertensive men, these recommendations would t well with the present ndings where the risk of CVD is signi cantly raised in the top two quintiles of waist and WHTR. As far as women are concerned, the situation is more complex, since women with the smallest waists have as high a risk of CVD as those with increased central adiposity. Nevertheless the present ndings support the Lean et al 27 recommendations for an `action level' of an 80 cm waist circumference for women, since those above this level (Quintiles 4 and 5) have a signi cantly greater incidence of CVD. Conclusion The ndings from the present study show that anthropometric indices involving waist measurements, are more sensitive indicators of CVD risk over a short period of time than BMI. The interaction of with the anthropometric indices and CVD incidence reveals that the highest risk of CVD in men, is in medication treated hypertensives with the lowest central adiposity, and that for women, the relationships differed between hypertensives and non-hypertensives, but were weaker than for men. References 1 Wannamethee G, Shaper AG. Body weight and mortality in middle aged British men: impact of smoking. BMJ 1989; 299: 1497 ± 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 ± Selmer R, Tverdal A. Body mass index and cardiovascular mortality at different levels of blood pressure: a prospective study of Norwegian men and women. J Epidemiol Commun Health 1995; 49: 265 ± LaÈaÈraÈ E, Rantakallio P. Body size and mortality in women: a 29 year follow up of pregnant women in northern Finland. J Epidemiol Commun Health 1996; 50: 408 ± Imeson JD, Haines AP, Meade TW. Skinfold thickness, body mass index and ischaemic heart disease. J Epidemiol Commun Health 1989; 43: 223 ± Rimm EB, Stampfer MJ, Giovannucci E, Ascherio A, Spiegelman D, Colditz GA, Willett WC. Body size and fat distribution as predictors of coronary heart disease among middle-aged and older US men. Am J Epidemiol 1995; 141: 1117± Lapidus L, Bengtsson C, Larsson B, Pennert K, Rybo E, SjoÈstrom L. Distribution of adipose tissue and risk of cardiovascular disease and death: a 12 year follow up of participants in the population study of women in Gothenburg, Sweden. BMJ 1984; 289: 1257± Gar nkel L, Stellman SD. Mortality by relative weight and exercise. Cancer 1988; 62: 1844 ± Manson JE, Colditz GA, Stampfer MJ, Willett WC, Rosner B, Monson RR, Speizer FE, Hennekens CH. A prospective study of obesity and risk of coronary heart disease in women. N Engl J Med 1990; 322: 882± Larsson B, SvaÈrdsudd K, Welin L, Wilhelmsen L, BjoÈrntorp P, Tibbling G. Abdominal adipose tissue distribution, obesity, and risk of cardiovascular disease and death: 13 year follow-up of participants in the study of men born in BMJ 1984; 288: 1401± Shinton R, Shipley M, Rose G. Overweight and stroke in the Whitehall Study. J Epidemiol Commun Health 1991; 45: 138 ± Prineas RJ, Folsom AR, Kaye SA. Central adiposity and increased risk of coronary heart disease mortality in older women. Ann Epidemiol 1993; 3: 35 ± Welin L, SvaÈrdsudd K, Wilhelmsen L, Larsson B, Tibblin G. Analysis of risk factors for stroke in a cohort of men born in N Engl J Med 1987; 317: 521± Walker SP, Rimm EB, Ascherio A, Kawachi I, Stampfer MJ, Willett WC. Body size and fat distribution as predictors of stroke among US men. Am J Epidemiol 1996; 144: 1143 ±1150.

8 15 Higgins M, Kannel W, Garrison R, Pinsky J, Stokes J. Hazards of obesity ± the Framingham experience. Acta Med Scand 1988; 723 (Suppl): 23± Folsom AR, Prineas RJ, Kaye SA, Soler JT. Body fat distribution and self-reported prevalence of, heart attack, and other heart disease in older women. Int J Epidemiol 1989; 18: 361± Hartz A, Grubb B, Wild R, van Nort JJ, Kuhn E, Freedman D, Rimm A. The association of waist hip ratio and angiographically determined coronary heart disease. Int J Obes 1990; 14: 657± Cox BD, Whichelow MJ, Ashwell M, Prevost AT, Lejeune SRE. Association of anthropometric indices with elevated blood pressure in British adults. Int J Obes 1997; 21: 674± Cambien F, Chretien JM, Ducimetiere P, Guize L, Richard JL. Is the relationship between blood pressure and cardiovascular risk dependent on body mass index? Am J Epidemiol 1985; 122: 434 ± Walker M, Shaper AG, Phillips AN, Cook DG. Short stature, lung function and risk of heart attack. Int J Epidemiol 1989; 18: 602 ± Kannam JP, Levy D, Larson M, Wilson PWF. Short stature and risk for mortality and cardiovascular disease events: The Framingham Heart Study. Circulation 1994; 90: 2241± Rich-Edwards JW, Manson JE, Stampfer MJ, Colditz GA, Willett WC, Rosner B, Speizer FE, Hennekens CH. Height and the risk of cardiovascular disease in women. Am J Epidemiol 1995; 142: 909±917. CVD, and anthropometry 23 Cox BD, Blaxter M, Buckle ALJ, Fenner NP, Golding JF, Gore M, Huppert FA, Nickson J, Roth M, Stark J, Wadsworth MEJ, Whichelow MJ. The Health and Lifestyle Survey: Preliminary report of a nationwide survey of the physical and mental health, attitudes and lifestyle of a random sample of 9003 British adults. Health Promotion Research Trust: Cambridge, Cox BD, Huppert FA, Whichelow MJ. The Health and Lifestyle Survey: Seven Years On. The Dartmouth Publishing Co Ltd.: Aldershot, Barrett-Connor E, Khaw K-T. Is more benign when associated with obesity? Circulation 1985; 72: 53 ± Messerli FH, Christie B, DeCarvalho JGR. Obesity and essential. Haemodynamics, intravascular volume, sodium excretion, and plasma renin activity. Arch Intern Med 1981; 141: 81± Lean MEJ, Han TS, Morrison CE. Waist circumference as a measure for indicating need for weight management. BMJ 1995; 311: 158± Han TS, van Leer EM, Seidell JC, Lean MEJ. Waist circumference action levels in the identi cation of cardiovascular risk factors: prevalence study in a random sample. BMJ 1995; 311: 1041± Egger G, Bolton A, O'Neill M, Freeman D. Effectiveness of an abdominal obesity reduction programme in men: the GutBuster `waist loss' programme. Int J Obes 1996; 20: 227± Han TS, Richmond P, Avenell A, Lean MEJ. Waist circumference reduction and cardiovascular bene ts during weight loss in women. Int J Obes 1997; 21: 127±