Preferred clinical measures of central obesity for predicting mortality



To define the clinical measures of obesity that best predict all cause mortality and cardiovascular disease (CVD) mortality.

Design and Setting:

Eleven-year mortality follow-up of an Australian urban population sample of 9309 adults aged 20–69 years in 1989. Baseline measures of obesity included body mass index (BMI), waist circumference (WC), waist-to-stature ratio and the waist-to-hip ratio. The age-standardized hazard ratios for mortality were calculated for 1 s.d. above the mean for each measure of obesity using Cox regression analysis. We constructed receiver operator characteristic (ROC) curves to assess sensitivity and specificity of the measures and to identify approximate cut-points for the prediction of risk.


Waist-to-hip ratio was superior by magnitude and significance in predicting all cause mortality (male hazard ratio 1.25, P=0.003, female hazard ratio 1.24, P=0.003) and CVD mortality (male hazard ratio 1.62, P<0.001, female hazard ratio 1.59, P<0.001). Waist-to-stature ratio and WC were highly significant but less powerful predictors for CVD mortality. ROC analysis showed higher ‘area under the curve’ values for waist-related measures in males, with similar less marked trends in females. The ROC cut-points yielded values that corresponded to current promulgated criteria.


The waist-to-hip ratio is the preferred clinical measure of obesity for predicting all cause and CVD mortality. WC is a practical alternative. Waist-to-stature ratio is not more useful than WC alone.


Obesity is a rapidly growing threat to the health of populations worldwide (World Health Organisation, 1998). The factors that cause obesity will soon exceed cigarette smoking as preventable causes of death in the USA (Mokdad et al., 2004). Accurate and simple measures are essential to refine the descriptive and analytical epidemiology of this condition, and to evaluate the impact of interventions. One problem is that the commonly used measure for obesity is the body mass index (BMI) (World Health Organisation, 1998). BMI is a comparatively poor predictor of death, and very large population numbers (Calle et al., 1999), or meta-analyses (McGee, 2005), are required to demonstrate the relationships with increased total mortality, and with increased cardiovascular disease (CVD) and cancer death rates.

Recently the hazards of intra-abdominal or central obesity, using measures of the WC or the waist-to-hip ratio, have been emphasized as useful markers of the obesity-related health burden (Welborn et al., 2003). In a worldwide analysis of risk factors for acute myocardial infarction, the INTERHEART case–control study (Yusuf et al., 2005) reported that waist-to-hip ratio and WC were highly significant factors associated with this acute presentation of coronary heart disease (CHD), independent of other risk factors including BMI. The latest definitions of the metabolic syndrome as a risk factor for CVD and diabetes have central obesity and specifically WC as the core component (Executive summary, 2001; International Diabetes Federation, 2005). WC is a simple reproducible measure for intra-abdominal fat, but ethnic specific criteria are necessary (International Diabetes Federation, 2005).

There is a clear need for prospective studies of representative population samples that evaluate the various measures of obesity in relation to all-cause mortality as well as CVD endpoints. In this report we compare the use of the waist-to-hip ratio, the waist-to-stature ratio and WC with BMI in an Australian sample of adults, followed for 11 years.


In the 1989 National heart foundation risk factor prevalence study (Bennett and Magnus, 1994), Australian residents of capital cities from nine metropolitan centers, aged 20–69 years, were selected from electoral rolls by systematic probability sampling, using sex and 5-year age groups. Registration for voting is compulsory in Australia and the electoral rolls provide a convenient census of the national population. The city catchment areas were North Sydney, South Sydney, Melbourne, Brisbane, Adelaide, Perth, Hobart, Darwin and Canberra. Subjects were invited to attend a local survey centre after overnight fasting. Of 15,164 people selected, 2694 were no longer at the address or were absent during the study or in prison or had died. Of the 12,470 who confirmed contact, 9309 attended and completed the survey, a 75% response rate. The respondents were mainly Europid (93%), with a small proportion of Asians and Africans (5%), as determined by stated place of birth. A total of 5.2% males and 2.5% females gave a history of CHD and/or stroke. Details of the survey methods and laboratory techniques have been published (Boyle et al., 1993; Bennett and Magnus, 1994; Welborn et al., 2003).

Physical measurements in light clothing without shoes, included height measured to the nearest centimeter and weight to the nearest tenth of a kilogram (1 kg has been deducted from the recorded weight as an allowance for the clothing).

WC was measured at the narrowest point between ribs and hips after exhaling when viewed from the front (Boyle et al., 1993). Hip circumference was measured at the point of maximum extension of the buttocks when viewed from the side. Two consecutive placements were recorded for each site to the nearest 1 cm using a metal tape on a horizontal plane without compression of skin. The mean of the two values was used.

Statistical analysis

The identifier data collected from respondents in the 1989 survey were submitted to the Australian Institute of Health and Welfare (AHIW) for matching with the national death index. A record linking package (‘Automatch’) grouped date of birth, sex, full name, date of death. Additional matching used place of survey and place of death for inclusion in a weighted algorithm for identifying deceased survey subjects. Mortality was determined up to and including 31 December 2000.

Causes of death were coded according to the 9th (ICD-9) or 10th (ICD-10) revisions of the International Classification of Diseases (World Health Organisation 1977, 1998). ICD-9 codes 3900–4589 or ICD-10 codes I 00.0–I 99.9 were used for CVD deaths, and ICD-9 codes 4100–4499 or ICD-10 codes I 20.0–I 25.9 were used for CHD deaths.

Baseline measures were analyzed by Cox proportional hazards regression for survival time from the day of survey to day of death or to 31 December 2000. In the analysis of CVD deaths, the survival time for an individual was censored if the individual had died and the cause of death was other than CVD. The candidate variables were age-standardized using age and the square of age to account for linear and nonlinear relationships. We estimated hazard ratios, with 95% confidence intervals, for the associations between potential risk factors and total mortality and CVD deaths. The hazard ratios indicated the incremental risk of 1 s.d. above the mean for any risk factor.

We used receiver operator characteristic (ROC) curve analysis (SPSS version 14) (Haijian-Tilaki, 1997) to calculate the sensitivity and specificity for each obesity measure. Plotting sensitivity against 1−specificity allows comparison of area under the curve (AUC) as an indicator of the predictive power of the measure for total and CVD mortality. Also to identify optimal cut-off values for predicting mortality, the figure for the Youden index (Schisterman et al., 2005) was selected representing sensitivity +specificity −1.

Ethical approval

The Australian Institute of Health Interim Ethics Committee, after consultation with the Commonwealth Privacy Commissioner, provided ethical clearance for the 1989 survey. The linkage and analysis of the survey data with the National Death Index were approved by the current Ethics Committee of the Australian Institute of Health and Welfare.


The baseline characteristics and anthropometric measures of the survey subjects are shown in Table 1. The cardiovascular risk factors are reported as being representative of the Australian population and are described elsewhere (Boyle et al., 1993; Bennett and Magnus, 1994).

Table 1 Baseline characteristics of the cohort of 9206 adults aged 20–69 years in 1989.

During 11 years (1989–2000) there were 473 deaths (298 males, 175 females) including 137 CVD deaths (96 males, 41 females) and 336 non-CVD deaths (202 males, 134 females). Table 2 summarizes the calculated hazard ratios derived from the Cox regression analysis for each of the various measures of obesity in relation to the 11-year mortality.

Table 2 Age-standardized hazard rates for measures of obesity (BMI, WC, WSR, WHR) and 11-year mortality from all-causes, CVD mortality and non-CVD mortality

Age-standardized Cox regression analysis (Table 2)

All-cause mortality

For males and females the waist-to-hip ratio predicts all-cause mortality with a high level of significance. For females the WC and waist-to-stature ratio also predict all-cause mortality, but the hazard ratios and significance are less than those for the waist-to-hip ratio. BMI does not reach significance.

CVD mortality

All measures of obesity predict CVD deaths in this cohort, with the exception of BMI in females. The waist-to-hip ratio is by the far the superior predictor, in terms of the magnitude of the hazard ratios and their high statistical significance. WC and the waist-to-stature ratio do achieve significant but lesser hazard ratios with comparable values. These data are shown graphically in Figure 1.

Figure 1

Graphical representation of the hazard ratios for BMI, WC, waist statute ratio and WHR in relation to total mortality, CVD mortality and nonCVD mortality (actual data and significance levels shown in Table 2).

Non-cardiovascular mortality

The only significant obesity measure predicting non-CVD mortality in this cohort is BMI in males, where a low BMI predicts death with borderline significance (leanness is a risk factor in this context).

ROC curve analysis

Figure 2 shows the ROC characteristics of the obesity measures (plotted as sensitivity versus 1−specificity) for all-cause mortality and for cardiovascular deaths, as compared to the curves for waist, waist-to-stature ratio, and waist-to-hip ratio, all of which are congruent. This is confirmed by the AUC values (Table 3). In males the AUC for all waist-related indices are uniformly higher than that for BMI. Similar but less marked trends are evident for females.

Figure 2

ROC curves for measures of obesity in males (upper box) and females (lower box). BMI is in red, WC is in orange, waist-to-hip ratio is in blue and waist-to-stature ratio is in green. The reference line is represented as the black dotted line.

Table 3 Receiver operator curve characteristics for measures of obesity in predicting all cause mortality and cardiovascular disease mortality, with area under the curve (95% confidence limits) and cut-points for optimal sensitivity and specificity

Using the ROC curves, cut-points for each obesity measure can identify the levels that show optimal sensitivity and specificity. Table 3 indicates that, in general, they are consistent in discriminating increased risk of mortality: BMI has quite low values for AUC and the sensitivities and Youden index are poor in comparison with all the waist-related measures and the cut-points show variability (from 24.7 to 27.1 kg/m2). For waist-related measures, the AUCs are remarkably similar, with cut-points that are comparable (waist 92–96 cm for males, 80 cm for females, waist-to-stature ratio 0.53–0.55 for males, 0.48–0.50 for females; waist-to-hip ratio 0.93 for males, 0.79–0.80 for females).


This 11-year mortality study of a national Australian sample confirms that the waist-to-hip ratio is a far superior measure than BMI in predicting all cause mortality and CVD mortality, with equivalent and highly significant hazard ratios in males and females. Waist-to-hip ratio is also better than waist-to-stature ratio and WC in terms of the magnitude and significance of predicting death. The ability of waist-to-hip ratio to predict all-causes of mortality in a population sample of this size is notable and is clearly mediated by its powerful association with CVD deaths.

We included ROC curve analyses as an additional means of comparison, and these showed equivalence of the three waist-related measures, all of which were superior to BMI. This form of analysis does not correct for age nor allow for survival times, and thus, does not provide as critical a comparison of the measures. The cut-points for maximal sensitivity and specificity are of interest. Our values were close to the promulgated values for WC 94 cm in males and 80 cm in females (Lean et al., 1995; World Health Organisation, 1998), for waist-to-stature ratios, 0.50 in males and females (Ho et al., 2003) and for waist-to-hip ratios is 0.90 for males and 0.80 for females (Larsson et al., 1992), but the BMI cut-points were more variable and generally higher than 25 kg/m2 (World Health Organisation, 1998; Calle et al., 1999; McGee, 2005).

We have reported elsewhere (Welborn et al., 2003) that the waist-to-hip ratio is the dominant risk factor predicting CHD and CVD mortality, independent of conventional risk factors including blood pressure and lipid levels. Diastolic blood pressure and triglyceride levels were additional predictors in males and cigarette smoking an independent predictor in females. A similar study of Finnish populations showed a superior independent association of waist-to-hip ratio with CVD deaths in males, whereas in females BMI and WC performed better (Hu et al., 2004).

Two other recent prospective studies of CVD have compared waist-to-hip ratio with WC. In an analysis of the HOPE Study of 8802 subjects with stable vascular disease followed for 2–5 years (Dagenais et al., 2005), waist-to-hip ratio showed hazard ratios for total mortality of 1.52 (P<0.001) WC hazard ratio of 1.17 (P<0.05); for CVD deaths, the waist-to-hip ratio hazard ratio was 1.24 (P<0.03) with WC being nonsignificant (3rd vs 1st tertile). The hazard ratios for myocardial infarction were similar: WC 1.23, waist-hip ratio (WHR) 1.20 (both P<0.01). In the Nurses Health Study using self-reported values, the 8-year independent risk ratios for CHD were 3.09 for waist-to-hip ratio and 2.69 for WC (5th vs 1st quintiles) (Rexrode et al., 1998).

The use of the waist-to-stature ratio may appear to be an attractive alternative to waist-to-hip ratio by providing a correction for body frame size using height that is more commonly and more conveniently measured than hip circumference. Strong support for this index has been cited by the Hong Kong Cardiovascular Risk Factor Prevalence Study, using cross-sectional correlations with CVD risk factors (Ho et al., 2003). Few prospective studies exist having any comprehensive end-point data. Nevertheless the waist-to-stature ratio provides a useful algorithm – that one's WC should not exceed half of one's height. Our ROC analyses suggested cut-points for WSR of 0.53–0.55 for males and 0.50–0.51 for females. In our study, waist-to-stature ratio performs at the same level or only slightly better than WC alone in the longitudinal prediction of mortality and is not a useful alternative clinically, nor in the epidemiological setting.

The measurement of WC has been popularized as a simple and practical tool to identify central obesity. It is clearly superior to BMI. Our data confirm this with WC predicting CVD mortality significantly and with superior discrimination of risk compared to BMI. The literature corroborates WC as a good correlate of abdominal visceral adipose tissue using computerized tomography and magnetic resonance imaging (Desprès et al., 2001). It is a useful indicator of weight reduction for clinical purposes (Van der Kooy et al., 1993). WC is independent of height, being a body dimension that includes only a small bone component of mass apart from the vertebrae (Han et al., 1997). Nevertheless, ethnic specific criteria are recommended to account for variable frame size (International Diabetes Federation, 2005).

The levels for WC that are said to indicate frank obesity (males 102 cm, females 88 cm) and overweight (males 94 cm, females 80 cm) were initially recommended from a cross-sectional analysis (Lean et al., 1995) of a small population sample. The latter have now been included in international definitions of the metabolic syndrome (Executive summary, 2001; International Diabetes Federation, 2005). Being a continuous variable, such cut-points are arbitrary. The true definition of action levels will ultimately depend on prospective population data with well defined end-points.

The hip circumference alone is usually associated with improved CVD outcomes (Heitmann et al., 2004). It is likely that hip circumference reflects an increased musculature and frame size, but also indicates gynecoid (or low risk) body fat. Measuring hip circumference in the clinical setting is an additional task. Other disincentives for using the hip circumference include the practical difficulty in the extremely obese, and aesthetic and cultural constraints in various communities. Nevertheless, the use of the hip circumference in the waist-to-hip ratio enhances the assessment of CVD risk in a similar manner to the use of the ratio of total cholesterol (atherogenic) to HDL cholesterol (anti-atherogenic).

Strong support for the waist-to-hip ratio as the best measure of obesity comes from the INTERHEART Study where there is a consistent graded association with acute myocardial infarction across all populations and sub-groups in a case–control study (Yusuf et al., 2005). The tertile cut-points for the waist-to-hip ratio study were 0.90 and 0.95 in men and 0.83 and 0.90 in women. The top vs the lowest tertiles gave an overall odds ratio for CHD of 2.2 and a population attributable risk of 34% after adjustment for age, sex and smoking.

It seems likely that the waist-to-hip ratio is the preferred measure of obesity for defining mortality. In our data set, only the waist-to-hip ratio predicts all cause mortality, whereas all three waist-related measures predict CVD mortality and all three are superior to BMI. The waist-to-hip ratio is the best measure of obesity to identify cardiovascular and CHD risk (independent of blood pressure and lipid levels (Welborn et al., 2003). Clearly, further prospective studies of homogeneous populations will consolidate this issue and will provide finite cut-points for clinical and public health intervention. There is an urgent requirement for standardized techniques for these measurements both in the clinical setting and in epidemiological studies (Molarius and Seidell, 1998).


  1. Bennett SA, Magnus P (1994). Trends in cardiovascular risk factors in Australia: results from the National Heart Foundation's Risk Factor Prevalence Study 1980–1989. Med J Aust 161, 519–527.

  2. Boyle CA, Dobson AJ, Egger G, Bennett SA (1993). Waist-to-hip ratios in Australia: a different picture of obesity. Aust J Nutr Diet 50, 57–64.

  3. Calle EF, Thun MJ, Petrelli JM, Rodriguez C, Heath CW (1999). Body mass index and mortality in a prospective cohort of US adults. N Engl J Med 341, 1097–1115.

  4. Dagenais GR, Yi Q, Mann JFE, Pogue J, Yusuf S (2005). Prognostic impact of body weight and abdominal obesity in women and men with cardiovascular disease. Am Heart J 149, 54–60.

  5. Desprès J-P, Lemieux I, Prud'homme D (2001). Treatment of obesity: need to focus on high-risk abdominally obese patients. BMJ 322, 716–720.

  6. Executive summary of the third report of the national cholesterol education program (2001). Expert panel on detection evaluation and treatment of high blood cholesterol in adults. JAMA 285, 2486–2497.

  7. Haijian-Tilaki KO (1997). Sampling Variability of nonparametric estimates of the areas under Receiver Operating Characteristic Curves: an update. Acade Radiol 4, 49–58.

  8. Han TS, Seidell JC, Currall JE, Morrison Ce, Deurenberg P, Lean ME (1997). The influences of height and age on waist circumference as an index of adiposity in adults. Int J Obes 21, 83–89.

  9. Heitmann BL, Frederiksen P, Lissner L (2004). Hip circumference and cardiovascular morbidity and mortality in men and women. Obes Res 12, 482–487.

  10. Ho S-Y, Lam T-H, Janus ED, Hong Kong Cardiovascular Risk Factor Prevalence Study Steering Committee (2003). Waist-to-stature ratio is more strongly associated with cardiovascular risk factors than other simple anthropometric indices. Ann Epidemiol 13, 683–691.

  11. Hu G, Tuomilehto J, Silventoinen K, Barengo N, Jousilahti P (2004). Joint effects of physical activity, body mass index, waist circumference and waist-to-hip ratio with the risk of cardiovascular disease among middle-aged Finnish men and women. Eur Hea J 25, 2212–2219.

  12. International Diabetes Federation (2005). The IDF consensus worldwide definition of the metabolic syndrome Available from http://www.idf.orgaccessed June 2005.

  13. Larsson B, Bengtsson C, Bjorntorp P, Lapidus L, Sjostrom L, Svardsudd K et al. (1992). Is abdominal body fat distribution a major explanation for the sex difference in the incidence of myocardial infarction? The study of men born in 1913 and the study of women, Goteborg, Sweden. Am J Epidemiol 135, 266–273.

  14. Lean MEJ, Han TS, Morrison CE (1995). Waist circumference as a measure for indicating need for weight management. BMJ 311, 158–161.

  15. McGee DL, the Diverse Populations Collaboration (2005). Body mass index and mortality: a meta-analysis based on personal level data from 26 observational studies. Ann Epidemiol 15, 87–97.

  16. Mokdad AH, Marks JS, Stroup DF, Gerberding JL (2004). Actual causes of death in the United States. JAMA 291, 1238–1245.

  17. Molarius A, Seidell JC (1998). Selection of anthropometric indicators for classification of abdominal fatness – a critical review. Int J Obes 22, 719–727.

  18. Rexrode KM, Carey VJ, Hennekeus CH, Walters EE, Colditz GA, Stampfer MJ et al. (1998). Abdominal adiposity and coronary heart disease in women. JAMA 280, 1843–1848.

  19. Schisterman Ef, Perkins NJ, Liu A, Bondell H (2005). Optimal cut-point and its corresponding Youden Index to discriminate individuals using pooled blood samples. Epidemiology 16, 73–81.

  20. Van der Kooy, Leenan R, Seidell JC, Deurenberg P, Hautvast GAG (1993). Effect of a weight cycle on visceral fat accumulation. Am J Clin Nutr 58, 853–857.

  21. Welborn TA, Dhaliwal SS, Bennett SA (2003). Waist-hip ratio is the dominant risk factor predicting cardiovascular deaths in Australia. Med J Aust 179, 580–585.

  22. World Health Organisation (1977). International classification of diseases Manual of the international statistical classification of diseases, injuries, and causes of death, 9th revision (ICD-9). Geneva:WHO.

  23. World Health Organisation (1998). International classification of diseases. Manual of the international statistical classification of diseases, injuries, and causes of death, 10th revision (ICD-10). Geneva:WHO.

  24. World Health Organisation (1998). Obesity: preventing and managing the global epidemic. Report of a WHO consultation on obesity. WHO: Geneva.

  25. Yusuf S, Hawken S, Öunpuu S, On behalf of the INTERHEART Study Investigators (2005). Obesity and the risk of myocardial infarction in 27 000 participants from 52 countries: a case–control study. Lancet 366, 1640–1649.

Download references


We are indebted to the National Heart Foundation of Australia for permission to use data from the 1989 Risk Factor Prevalence Survey and to the officers of the Australian Institute of Health and Welfare who linked the survey data to the National Death Index; Stan Bennett, Gabrielle Hodgson, Robert Vanderhoek, Paul Jelfs, Tracy Dixon and John Harding. The initial study was supported by Healthway, the Western Australian Health Promotion Foundation, and this analysis was funded with the assistance of a grant-in-aid provided by Merck, Sharp and Dohme (Australia) Pty Ltd.

Dr Welborn reports receiving consulting fees and conference support from Abbott Australasia, Roche Products Pty Ltd and Sanofi–Aventis.

Author information

Correspondence to T A Welborn.

Additional information

Guarantor: TA Welborn.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Welborn, T., Dhaliwal, S. Preferred clinical measures of central obesity for predicting mortality. Eur J Clin Nutr 61, 1373–1379 (2007).

Download citation


  • obesity
  • waist- to-hip ratio
  • cardiovascular disease
  • mortality

Further reading