Ethics statement

The Australian Institute of Health Interim Ethics Committee provided ethical clearance for the 1989 third Risk Factor Prevalence Survey, after consultation with the Commonwealth Privacy Commissioner. Participation was entirely voluntary. Participants signed an informed consent form to indicate their willingness to participate in the survey [18]. The subsequent linkage and analyses of the survey data with the National Death Index were approved by the current Ethics Committee of the Australian Institute of Health and Welfare.

Study participants

The Australian third Risk Factor Prevalence Survey was conducted in 1989 by the National Heart Foundation (NHF) to provide a picture of the level of risk factors in the population of registered voters living in Australian capital cities [18]. The city catchment areas were Sydney (North and South), Melbourne, Brisbane, Adelaide, Perth, Hobart, Darwin and Canberra. From the initial age and sex stratified sample of 9279 Australian residents, 4175 males who were free of heart disease, diabetes or stroke at baseline were selected for this study. Respondents were mainly of European descent.

Risk factor measurements

At baseline, fasting serum lipid levels, systolic and diastolic blood pressure, and smoking habits were recorded [19]. The Framingham Risk Scores (FRS) for CHD and CVD death were calculated at baseline for all 4175 males, using the equations from Anderson's paper [20]. The FRS were calculated from the baseline data of subjects free of heart disease, diabetes or stroke, after accounting for age, gender, systolic blood pressure, total cholesterol, high density lipoprotein cholesterol and cigarette smoking.

Physical measurements were made in light clothing and without shoes. General obesity was assessed by BAI and BMI using standardised methods [21]. BAI, which correlates well with % body fat using DXA, was calculated with the formula [1]. Central obesity was assessed by WC and WHR, using two observers by standardised methods [21]–[23]. WC was measured from the front at the narrowest point between the rib cage and the iliac crest after full expiration and HC was assessed from the side at the maximal extension of the buttocks using a metal tape, with no compression of the skin. The mean was calculated from the measurements (to the nearest centimetre) of the two observers. Height was measured to the nearest centimetre and weight to the nearest 0.1 kilogram (kg), with 1 kg deducted for light street clothing.

Cardiovascular disease outcomes

The data on these subjects were linked with the National Death Index to determine the causes of death for 346 subjects who had died by December 31, 2004. Of the 4175 males (age 42.3±13.1 years), 88 died from CVD and 64 were from CHD. The causes of death were coded according to the 10th revision of the International Classification of Diseases (ICD-10): Codes I 00.0 to I 99.0 for CVD deaths and codes I 20.0 to I 25.9 for CHD deaths [24].

Statistical analysis

Spearman's rank correlation was used to assess the association between the measures of general and central obesity, and CVD risk while accounting for fasting serum lipid levels, and systolic and diastolic blood pressure. The data on the general and central obesity variables were also divided into tertiles (low, medium and high tertiles, respectively) for further analysis using regression techniques. Multivariable logistic regression was used to assess the effects of the measures of obesity on CVD and CHD mortality using the 1989 NHF cohort, before adjustment (crude) and after adjustment for FRS variables (age, systolic blood pressure, total cholesterol, high density lipoprotein cholesterol and cigarette smoking). The effects of obesity variables and FRS on CVD and CHD mortality were expressed as odds-ratios and associated 95% confidence intervals. P-values less than 0.05 were considered to be statistically significant. Data was analysed using IBM SPSS Statistics Version 19.

Key findings

This study showed that BAI, like BMI, predicted CVD and CHD mortality when unadjusted for age and cardiovascular risk factors. But after adjustment for FRS variables (age, systolic blood pressure, total cholesterol, high density lipoprotein cholesterol and cigarette smoking), the association became non-significant. It is of interest that BAI and BMI performed similarly. In this context, the American Heart Association advises that BMI should be used with another measure such as WC, as primary tools for assessing adiposity. Individuals with elevated BMI or a proportionally high WC for a given BMI should have other cardio-metabolic risk factors evaluated for risk stratification. BAI requires a mathematical calculation of some complexity, and this cannot be used conveniently in field studies. For all of these reasons, BAI cannot be justified as a measure of any utility [25].

In contrast, this study showed that measures of central obesity were clearly superior. Both WC and WHR predicted CVD and CHD mortality, but WHR showed the strongest risk ratios and was independent of FRS. WHR is preferred because it is free of ethnic bias whereas WC requires ethnic specific criteria [23], [26]. Both WC and WHR are very simple techniques that can be employed in remote locations without scales. We emphasise that the measurement techniques, although simple, require careful standardisation and quality control in terms of the sites of measurement. WC should be measured at the narrowest level between the ribs and hips after exhaling when viewed from the front. HC should be measured from the point of maximum buttock extent when viewed from the side. Two consecutive placements would be recorded for each site and to the nearest 1 centimetre using a non-stretchable tape on a horizontal plane without compression of the skin and the average value used. When extreme obesity exists with abdominal apron, the HC can be measured in supine position.

Comparison with other studies

Similar results were reported in other studies [10], [11], [27]–[29]. BMI, WC and weight were more consistently correlated with cardiometabolic disease risk factors, compared to BAI, in a cohort of 698 Mexican Americans [11]. Where significant correlations were observed, BAI reported similar or weaker correlations with cardiometabolic trait, compared to BMI [11]. A cross-sectional study on Spanish Caucasian adult workers found that BAI was less correlated with CVD and metabolic risk factors, compared to BMI, WC and waist-to-height ratio (WHtR) [27]. WHtR reported the best correlations [27]. In addition, BAI had lower discriminatory capacity in diagnosing metabolic syndrome, compared to BMI [27]. BAI was also less associated with CHD risk factors, compared to BMI and WC, in another study among adults in the 1988–1994 National Health and Nutrition Examination Survey (NHANES III) [29]. It was observed that BAI provided little additional information on risk factor levels above those provided by BMI [29]. Another cross-sectional study evaluating the predictive ability of BAI, and % body fat and CVD risk factors in a Chinese population found that BAI was not a better indicator of % body fat, hypertension, dyslipidaemia, metabolic syndrome and intima-media thickening of the common carotid arteries, compared to BMI and WC [28]. These results were consistent with another study which found that BAI was inferior to BMI in predicting lipids, blood pressure and other CVD risk factors [16]. In summary, BAI does not overcome the limitations of BMI and is possibly a poorer indicator of CVD risk than BMI [10], [15], [27], [29]. Central obesity measures are better indicators of CVD risk [12], [15].

Support for WHR as the superior index of obesity for cardiovascular risk assessment is seen in large cross-sectional studies including the INTERHEART study [30] and the Obesity in Asia collaboration study [23], [31], [32]. Similarly the Dallas Heart Study demonstrated that WHR clearly outperforms WC and BMI in estimating coronary calcium scores, although the cross-sectional analysis could not identify WHR as an independent predictor compared to conventional risk factors.

Strengths and limitations

This study provided evidence that BAI is inadequate for assessing body adiposity and its association with CVD. Measures of central obesity are better predictors of CVD compared to BAI and BMI, in men. In addition, this study was carried out using a representative sample of the Australian male population. Although there is only one set of baseline measurements recorded for some risk variables but variables including measures of obesity were measured twice.