Browse the corpus

Strengths and limitations of this study This study provided evidence that anthropometric measures of central obesity are better predictors of cardiovascular disease (CVD) risk compared with general obesity measures in women. Central obesity measures add prognostic information on CVD risk in women above the measures of general obesity and should be considered for incorporation into the clinical assessment of CVD risk. Although this study is cross-sectional, it is a representative sample of the Australian female population. Only one set of baseline measurements is recorded for some risk variables but some important variables are measured twice. The predicted 10-year CVD risks are calculated using risk score models to stratify individuals against the treatment thresholds for various risk score models. Prospective data of CVD events were not used. Introduction In 2008, more than 200 million men and approximately 300 million women were obese.1 Overweight and obesity is one of the leading risk factors for mortality, estimated to account for 23% of the ischaemic heart disease burden.1 It results in the deterioration of the entire cardiovascular risk profile.2 3 Large prospective studies such as the Framingham Heart Study,4 the Nurses’ Health Study5 6 and the Buffalo Health Study7 have all shown that overweight and obesity are associated with increased cardiovascular disease (CVD) risk. Excess adipose tissue contributes to the cardiovascular and other risks associated with being overweight or obese.8

s such as the Framingham Heart Study,4 the Nurses’ Health Study5 6 and the Buffalo Health Study7 have all shown that overweight and obesity are associated with increased cardiovascular disease (CVD) risk. Excess adipose tissue contributes to the cardiovascular and other risks associated with being overweight or obese.8 The American Heart Association released a Scientific Statement emphasising the importance of assessing adiposity.8 New guidelines have also been released by the American College of Cardiology, American Heart Association Task Force on Practice Guidelines and The Obesity Society for the management of overweight and obesity in adults to prevent CVD.9 General and central obesity are associated with CVD risk.5 10–15 Currently used general and central obesity anthropometric measures for assessing adiposity-related risk include: body mass index (BMI; weight in kilograms divided by square of height in meters), waist circumference (WC), hip circumference (HC), waist-to-hip ratio (WHR; ratio of WC to HC), waist-to-stature ratio (WSR; ratio of WC to height) and body adiposity index16 (BAI; HC divided by height1.5, and subtracting 18 from the result). BMI or WC is most commonly used to measure body fatness.10

height in meters), waist circumference (WC), hip circumference (HC), waist-to-hip ratio (WHR; ratio of WC to HC), waist-to-stature ratio (WSR; ratio of WC to height) and body adiposity index16 (BAI; HC divided by height1.5, and subtracting 18 from the result). BMI or WC is most commonly used to measure body fatness.10 It is, however, unclear which anthropometric measurements are better correlated with CVD risk factors and CVD risk in women, considering adiposity is highly heterogeneous with age, sex and ethnic differences in body fat distribution.8 Previous studies have reported that BMI identified individuals at increased risk of CVD as effectively as WC.11 12 It has also been suggested that BMI is a better predictor of CVD than WC.13 Conversely, some studies reported that WC is a better indicator of CVD risk than BMI and WHR, in ethnically diverse groups.14 15 WC and WHR have also been identified as independent predictors of CVD risk but not BMI, accounting for conventional risk factors in the Framingham risk score model.17 More research is thus needed to ascertain which measures are better correlated with CVD risk factors and subsequent CVD risk in women.

verse groups.14 15 WC and WHR have also been identified as independent predictors of CVD risk but not BMI, accounting for conventional risk factors in the Framingham risk score model.17 More research is thus needed to ascertain which measures are better correlated with CVD risk factors and subsequent CVD risk in women. We aim to assess the associations between general and central obesity anthropometric measures with CVD risk factors, using a representative sample of 4487 women aged 20–69 years without heart disease, diabetes or stroke. The associations between these indices of obesity with predicted risk calculated from the Framingham risk score model for 10-year CVD incidence or death,18 SCORE risk chart for high-risk regions for 10-year CVD death,19 general CVD and simplified general CVD risk score models for 10-year CVD incidence and death20 were examined. To aid comparison between obesity indices, which are measured in different units, the incremental shift in CVD risk with 1 SD increment in each anthropometric measurement above the mean would be assessed. Finally, we determined which indices of obesity are most sensitive and specific for identifying women at increased 10-year CVD risk.

between obesity indices, which are measured in different units, the incremental shift in CVD risk with 1 SD increment in each anthropometric measurement above the mean would be assessed. Finally, we determined which indices of obesity are most sensitive and specific for identifying women at increased 10-year CVD risk. Methods Study cohort and measurements We selected 4487 women aged 20–69 years with no history of heart disease, diabetes or stroke from the population representative sample of 4727 women from the National Heart Foundation (NHF) Risk Factor Prevalence Study.21 Participants taking medications to lower their CVD risk factors were also excluded. The participants of the NHF study consisted of residents on the federal electoral rolls of December 1988 in North and South Sydney, Melbourne, Brisbane, Adelaide, Perth, Hobart, Darwin and Canberra in a systematic probability sampling by sex and 5-year age groups. Information on demographic characteristics was collected using a self-administered questionnaire, and conventional CVD risk variables recorded in this prevalence study include anthropometric measures, smoking status, systolic and diastolic blood pressure and lipid levels. Physical measurements of height (to the nearest centimetre), weight (to the nearest 10th of a kilogram) and waist and HC were collected according to standardised methodologies22 23 using two observers. The WC was measured from the front at the narrowest point between the rib cage and iliac crest after full expiration while the HC was measured from the side at the maximal extension of buttocks by one observer using a metal tape. A second observer recorded another set of measurements and ensured that the metal tape was kept strictly horizontal at all times. The mean of two measurements was taken at each site to the nearest centimetre. Participants were classified as non-smokers, previous smokers or current smokers.21 Mercury sphygmomanometers were used to record blood pressure levels on the right arm of seated participants 5 min apart.21 Two readings were taken and the average was used in the analysis. Fasting blood samples were also collected in EDTA tubes and dispatched to the central laboratory at the Division of Clinical Chemistry, Institute of Medical and Veterinary Science, Adelaide each week for lipid levels to be assayed.21

pants 5 min apart.21 Two readings were taken and the average was used in the analysis. Fasting blood samples were also collected in EDTA tubes and dispatched to the central laboratory at the Division of Clinical Chemistry, Institute of Medical and Veterinary Science, Adelaide each week for lipid levels to be assayed.21 Risk score models The Framingham 10-year predicted risk for CVD incidence or death was developed using data from the American Framingham Heart Study.18 Participants aged 30–74 years who were free of CVD and cancer were included in the model development. The 10-year risk for CVD incidence or death was calculated using these variables: age, sex, systolic blood pressure (SBP), diastolic blood pressure, total cholesterol level, high-density lipoprotein (HDL) cholesterol level, smoking status and diabetes status. The SCORE risk chart was developed by pooling 12 cohort studies to predict the 10-year CVD death risk in Europe. The cohorts consisted of participants aged 19–80 years with no previous history of heart attack.19 The SCORE model was derived from a much larger dataset than the Framingham, general CVD and simplified general CVD risk score models. Fewer variables were used in the calculation of the 10-year predicted CVD death risk with the SCORE risk chart for high-risk regions (Denmark, Finland and Norway),19 24 these included: age, sex, smoking status, mean total cholesterol level, mean HDL cholesterol level and mean SBP. The general CVD risk score model was also developed using data from the American Framingham Heart Study but using a larger cohort than the Framingham model.20 Individuals without CVD were used in the development of the general CVD risk score model.20 The simplified general CVD risk score model was developed similarly as the general CVD risk score model. It is, however, a simpler CVD risk prediction model which is calculated using non-laboratory predictors. Risk variables (age, SBP, current antihypertensive treatment, smoking status and diabetes status) were used in both models.20 The only difference is that, BMI is included in the simplified general CVD risk score model instead of total and HDL cholesterol which is used in the general CVD risk score model.

ry predictors. Risk variables (age, SBP, current antihypertensive treatment, smoking status and diabetes status) were used in both models.20 The only difference is that, BMI is included in the simplified general CVD risk score model instead of total and HDL cholesterol which is used in the general CVD risk score model. Statistical analysis The data on the representative sample of 4487 Australian women were described using mean±SD for continuous variables, while counts and percentages were used for categorical variables. Non-parametric Spearman's rank correlation was used to assess the associations between anthropometric measurements of obesity with CVD risk factors, and with the calculated 10-year predicted risks, due to the skewness in the distribution of some variables. Anthropometric measurements were also converted to z-scores (original value subtracted by the mean and result divided by the SD) to represent the number of SDs above and below the mean for each participant. Logistic regression was used to assess the effects of each standardised anthropometric measurement of being above the recommended treatment thresholds for various risk score models as a result of 1 SD increment above the mean for each anthropometric measure of obesity. ORs and associated 95% CIs represented the likelihood of being above the recommended treatment thresholds for the specific risk score models (20% for the Framingham risk score model for 10-year CVD incidence or death; 10% for SCORE risk chart for high-risk regions for 10-year CVD death; 10% and 20% for the general CVD and simplified general CVD risk score models for 10-year CVD incidence and death). The predictive ability of these anthropometric measures to identify individuals above and below the treatment thresholds was assessed using sensitivity, specificity and area under the receiver operating characteristic (ROC) curve. p Values of less than 0.05 were considered to be statistically significant. All statistical analyses were performed with IBM SPSS Statistics V.21.

o identify individuals above and below the treatment thresholds was assessed using sensitivity, specificity and area under the receiver operating characteristic (ROC) curve. p Values of less than 0.05 were considered to be statistically significant. All statistical analyses were performed with IBM SPSS Statistics V.21. Results The sample of 4487 women aged 20–69 years from the NHF Risk Factor Prevalence Study is a representative sample of the Australian female population, free of heart disease, diabetes and stroke. The characteristics of the sample are summarised in table 1. In addition to the conventional risk factors for CVD, all anthropometric measurements of general and central obesity were presented. Table 1 Characteristics of a representative Australian sample of 4487 women (aged 20–69 years) free of heart disease, diabetes and stroke

Results The sample of 4487 women aged 20–69 years from the NHF Risk Factor Prevalence Study is a representative sample of the Australian female population, free of heart disease, diabetes and stroke. The characteristics of the sample are summarised in table 1. In addition to the conventional risk factors for CVD, all anthropometric measurements of general and central obesity were presented. Table 1 Characteristics of a representative Australian sample of 4487 women (aged 20–69 years) free of heart disease, diabetes and stroke Variables Summary statistics Age (years), n (%) 20–29 840 (18.7) 30–39 1116 (24.9) 40–49 1139 (25.4) 50–59 743 (16.6) ≥60 649 (14.4) Ethnicity Australia 3329 (76.5%) UK and Ireland 416 (9.5%) Northern Europe 180 (4.1%) Southern Europe 234 (5.4%) Asia 195 (4.5%) Smoking status, n (%) Non-smoker 2652 (59.1) Previous smoker 880 (19.6) Current smoker 955 (21.3) SBP (mm Hg) 122.1±18.4 DBP (mm Hg) 75.7±10.8 Total cholesterol (mmol/L) 5.5±1.2 HDL cholesterol (mmol/L) 1.5±0.4 Ratio of total cholesterol to HDL cholesterol 3.9±1.3 BMI (kg/m2) 24.8±4.7 WC (cm) 76.2±11.1 HC (cm) 100.1±10.0 WHR 0.76±0.06 WSR 0.47±0.07 BAI (%) 30.6±5.4 BMI, body mass index; BAI, body adiposity index; DBP, diastolic blood pressure; HC, hip circumference; HDL cholesterol, high-density lipoprotein cholesterol; SBP, systolic blood pressure WC, waist circumference; WHR, waist-to-hip ratio; WSR, waist-to-stature ratio.

00.1±10.0 WHR 0.76±0.06 WSR 0.47±0.07 BAI (%) 30.6±5.4 BMI, body mass index; BAI, body adiposity index; DBP, diastolic blood pressure; HC, hip circumference; HDL cholesterol, high-density lipoprotein cholesterol; SBP, systolic blood pressure WC, waist circumference; WHR, waist-to-hip ratio; WSR, waist-to-stature ratio. The 10-year CVD risk of each participant in the sample was calculated using four risk score models. The frequency distribution of calculated risks is presented in table 2. Except for the Framingham model for CVD incidence, all other models predicted risks of less than 10% for at least 85% of the sample. The Framingham model for CVD incidence, general CVD model for CVD incidence and death and simplified general CVD model for CVD incidence and death predicted risk values across the entire range from 0% to greater than 40%. Table 2 Frequency distribution of 10-year predicted CVD incidence and mortality using various risk prediction models, in incremental risk categories of 10%

The 10-year CVD risk of each participant in the sample was calculated using four risk score models. The frequency distribution of calculated risks is presented in table 2. Except for the Framingham model for CVD incidence, all other models predicted risks of less than 10% for at least 85% of the sample. The Framingham model for CVD incidence, general CVD model for CVD incidence and death and simplified general CVD model for CVD incidence and death predicted risk values across the entire range from 0% to greater than 40%. Table 2 Frequency distribution of 10-year predicted CVD incidence and mortality using various risk prediction models, in incremental risk categories of 10% Risk categories 0–9% 10–19% 20–29% 30–39% ≥40% Framingham 10-year predicted risk for CVD incidence18 2936 (67.0%) 764 (17.4%) 417 (9.5%) 179 (4.1%) 89 (2.0%) Framingham 10-year predicted risk for CVD death18 4354 (99.3%) 29 (0.7%) 2 (0%) 0 (0%) 0 (0%) SCORE-HIGH 10-year predicted risk for CVD death19 4318 (98.5%) 53 (1.2%) 9 (0.2%) 4 (0.1%) 1 (0%) GCVD 10-year predicted risk for CVD incidence and death20 3738 (85.2%) 503 (11.5%) 109 (2.5%) 21 (0.5%) 14 (0.3%) SGCVD 10-year predicted risk for CVD incidence and death20 3809 (85.7%) 519 (11.7%) 90 (2.0%) 19 (0.4%) 9 (0.2%) Counts and percentages of women were presented. CVD, cardiovascular disease; GCVD, general cardiovascular disease risk score model; SCORE-HIGH, SCORE risk chart for high-risk regions; SGCVD, simplified general cardiovascular disease risk score model.

Risk categories 0–9% 10–19% 20–29% 30–39% ≥40% Framingham 10-year predicted risk for CVD incidence18 2936 (67.0%) 764 (17.4%) 417 (9.5%) 179 (4.1%) 89 (2.0%) Framingham 10-year predicted risk for CVD death18 4354 (99.3%) 29 (0.7%) 2 (0%) 0 (0%) 0 (0%) SCORE-HIGH 10-year predicted risk for CVD death19 4318 (98.5%) 53 (1.2%) 9 (0.2%) 4 (0.1%) 1 (0%) GCVD 10-year predicted risk for CVD incidence and death20 3738 (85.2%) 503 (11.5%) 109 (2.5%) 21 (0.5%) 14 (0.3%) SGCVD 10-year predicted risk for CVD incidence and death20 3809 (85.7%) 519 (11.7%) 90 (2.0%) 19 (0.4%) 9 (0.2%) Counts and percentages of women were presented. CVD, cardiovascular disease; GCVD, general cardiovascular disease risk score model; SCORE-HIGH, SCORE risk chart for high-risk regions; SGCVD, simplified general cardiovascular disease risk score model. Anthropometric measurements of obesity were positively correlated with age, SBP, total cholesterol and total cholesterol to HDL cholesterol ratio (all Spearman's r ≥0.195, p<0.001), with HC recording the lowest correlations. These obesity measures were negatively correlated with HDL cholesterol (all Spearman's r ≤–0.160, p<0.001). Measures of central obesity that included a measure of WC (WHR and WSR) generally recorded better correlations compared with measures of general obesity (BMI and BAI).

, with HC recording the lowest correlations. These obesity measures were negatively correlated with HDL cholesterol (all Spearman's r ≤–0.160, p<0.001). Measures of central obesity that included a measure of WC (WHR and WSR) generally recorded better correlations compared with measures of general obesity (BMI and BAI). The associations between anthropometric measurements of obesity and the 10-year predicted risks calculated using the four models are presented in table 3. All Spearman's rank correlations were statistically significant (p<0.0005). All anthropometric measures of central obesity (WC, WHR and WSR) generally had consistently higher correlations with the predicted risks calculated using the four CVD risk score models, as compared with measures of general obesity. Table 3 Non-parametric correlations between anthropometric measurements of general and central obesity and 10-year predicted risk of CVD incidence and mortality in 4487 women BMI WC HC WHR WSR BAI Framingham 10-year predicted risk for CVD incidence18 0.380 0.450 0.301 0.409 0.485 0.378 Framingham 10-year predicted risk for CVD death18 0.394 0.452 0.307 0.404 0.483 0.377 SCORE-HIGH 10-year predicted risk for CVD death19 0.309 0.381 0.253 0.348 0.419 0.338 GCVD 10-year predicted risk for CVD incidence and death20 0.385 0.452 0.307 0.405 0.487 0.383 SGCVD 10-year predicted risk for CVD incidence and death20 * 0.446 0.320 0.384 * * All Spearman's rank correlations significant at the p<0.0005 level.

10-year predicted risk for CVD death19 0.309 0.381 0.253 0.348 0.419 0.338 GCVD 10-year predicted risk for CVD incidence and death20 0.385 0.452 0.307 0.405 0.487 0.383 SGCVD 10-year predicted risk for CVD incidence and death20 * 0.446 0.320 0.384 * * All Spearman's rank correlations significant at the p<0.0005 level. *Correlation is not calculated for this obesity measure as it contains variables that are also used in the calculation of the simplified general CVD model. BMI, body mass index; BAI, body adiposity index; CVD, cardiovascular disease; GCVD, general cardiovascular disease risk score model; HC, hip circumference; WC, waist circumference; SCORE-HIGH, SCORE risk chart for high-risk regions; SGCVD, simplified general cardiovascular disease risk score model; WHR, waist-to-hip ratio; WSR, waist-to-stature ratio. Recommended treatment thresholds for the four CVD risk models were identified from a review of the literature. Table 4 presents the effects of 1 SD increment in each anthropometric measurement above the mean on the likelihood of being above the recommended thresholds or being indicated for treatment. All anthropometric measures of central obesity (WC, WHR and WSR) generally recorded higher ORs than general measures of obesity and they increased the likelihood of individuals being above the respective treatment thresholds. Table 4 OR and associated 95% CIs of being above the recommended treatment thresholds for various risk score models as a result of a 1 SD increment above the mean for each anthropometric measure of obesity

Recommended treatment thresholds for the four CVD risk models were identified from a review of the literature. Table 4 presents the effects of 1 SD increment in each anthropometric measurement above the mean on the likelihood of being above the recommended thresholds or being indicated for treatment. All anthropometric measures of central obesity (WC, WHR and WSR) generally recorded higher ORs than general measures of obesity and they increased the likelihood of individuals being above the respective treatment thresholds. Table 4 OR and associated 95% CIs of being above the recommended treatment thresholds for various risk score models as a result of a 1 SD increment above the mean for each anthropometric measure of obesity BMI WC HC WHR WSR BAI Framingham 10-year predicted risk for CVD incidence (threshold=20%)25 26 1.71*** (1.59–1.85) 2.12*** (1.95–2.29) 1.55*** (1.44–1.68) 2.27*** (2.08–2.47) 2.35*** (2.17–2.56) 1.92*** (1.77–2.09) Framingham 10-year predicted risk for CVD death (threshold=20%)25 26 1.68 (0.98–2.87) 3.13* (1.30–7.54) 1.60* (1.04–2.46) 2.52* (1.09–5.83) 3.33* (1.32–8.39) 1.58* (1.05–2.36) SCORE-HIGH 10-year predicted risk for CVD death (threshold=10%)19 1.53*** (1.29–1.82) 1.91*** (1.59–2.29) 1.40*** (1.18–1.66) 2.01*** (1.66–2.42) 2.04*** (1.70–2.46) 1.58*** (1.31–1.90) GCVD 10-year predicted risk for CVD incidence and death (threshold=10%)27 1.72*** (1.59–1.86) 2.11*** (1.95–2.29) 1.57*** (1.45–1.70) 2.23*** (2.04–2.43) 2.34*** (2.15–2.55) 1.94*** (1.79–2.11) SGCVD 10-year predicted risk for CVD incidence and death (threshold=10%)27 † 2.16*** (1.99–2.34) 1.66*** (1.54–1.80) 2.16*** (1.98–2.35) † † GCVD 10-year predicted risk for CVD incidence and death (threshold=20%)20 28 1.64*** (1.44–1.86) 2.03*** (1.77–2.31) 1.52*** (1.33–1.74) 2.08*** (1.81–2.39) 2.15*** (1.88–2.45) 1.72*** (1.49–1.97) SGCVD 10-year predicted risk for CVD incidence and death (threshold=20%)20 28 † 2.26*** (1.96–2.60) 1.72*** (1.50–1.99) 2.11*** (1.82–2.45) † † *p<0.05, **p<0.01, ***p<0.001.

eath (threshold=20%)20 28 1.64*** (1.44–1.86) 2.03*** (1.77–2.31) 1.52*** (1.33–1.74) 2.08*** (1.81–2.39) 2.15*** (1.88–2.45) 1.72*** (1.49–1.97) SGCVD 10-year predicted risk for CVD incidence and death (threshold=20%)20 28 † 2.26*** (1.96–2.60) 1.72*** (1.50–1.99) 2.11*** (1.82–2.45) † † *p<0.05, **p<0.01, ***p<0.001. †OR is not calculated for this obesity measure as it contains variables that are also used in the calculation of the simplified general CVD model. BMI, body mass index; BAI, body adiposity index; CVD, cardiovascular disease; GCVD, general cardiovascular disease risk score model; HC, hip circumference; WC, waist circumference; SCORE-HIGH, SCORE risk chart for high-risk regions; SGCVD, simplified general cardiovascular disease risk score model; WHR, waist-to-hip ratio; WSR, waist-to-stature ratio. Anthropometric measurements of central obesity (WC, WHR and WSR) also recorded higher area under the ROC curves, higher sensitivity and specificity, than BMI in identifying women above and below the 20% treatment threshold for the Framingham model for 10-year CVD incidence (figure 1A) and general CVD model for 10-year CVD incidence and death (figure 1B). Although BMI is included in the simplified general CVD model, high area under the ROC curve (>0.76) are reported for WC and WHR (figure 1C), indicating the independent contribution of central obesity measurements as compared with general obesity measurement in predicting the increased risk of CVD.

nd death (figure 1B). Although BMI is included in the simplified general CVD model, high area under the ROC curve (>0.76) are reported for WC and WHR (figure 1C), indicating the independent contribution of central obesity measurements as compared with general obesity measurement in predicting the increased risk of CVD. Figure 1 ROC curves to compare the predictive ability of obesity measures for being above the 20% cut-off of three CVD models: (A) Framingham risk score model for 10-year CVD incidence; (B) general cardiovascular disease risk score model for 10-year CVD incidence and death; (C) simplified general cardiovascular disease risk score model for 10-year CVD incidence and death. #Area under the ROC curve is not calculated for this obesity measure as it contains height which is also used in the calculation of the simplified general CVD model. BMI, body mass index; CVD, cardiovascular disease; ROC, receiver operating characteristic; WC, waist circumference; WHR, waist-to-hip ratio; WSR, waist-to-stature ratio. Discussion Measures of obesity are generally not included in the prediction of CVD risk. BMI is the only measure of obesity currently included in CVD risk score models such as the simplified general CVD risk score model, as an alternative to total and HDL cholesterol level for ease of measurement and calculation,20 and in the QRISK score model.29

ty are generally not included in the prediction of CVD risk. BMI is the only measure of obesity currently included in CVD risk score models such as the simplified general CVD risk score model, as an alternative to total and HDL cholesterol level for ease of measurement and calculation,20 and in the QRISK score model.29 In our study, anthropometric measurements of central obesity (WC, WHR and WSR) were more strongly associated with conventional CVD risk factors and the 10-year predicted risk calculated using the Framingham risk score model, SCORE risk chart for high-risk regions, general CVD and simplified general CVD risk score model, compared with general measures of obesity. Central obesity measures also recorded higher ORs and increased the likelihood of being above the recommended treatment threshold of the respective models with 1 SD increase above the mean. Central obesity measures which incorporated the measure of WC also exhibited higher sensitivity and specificity than BMI. Although BMI is included in the calculation of the simplified general CVD model, high area under the ROC curves were reported for WC and WHR, thus confirming that anthropometric measures of central obesity independently and significantly predicts CVD risk that is not accounted for by the general obesity measure. Hence, BMI alone is insufficient to account for the association between obesity and CVD risk.

rea under the ROC curves were reported for WC and WHR, thus confirming that anthropometric measures of central obesity independently and significantly predicts CVD risk that is not accounted for by the general obesity measure. Hence, BMI alone is insufficient to account for the association between obesity and CVD risk. Consistent with our study findings, previous studies also reported stronger associations between central obesity measures and CVD risk. Higher standardised ORs adjusted for BMI were reported for WC and CVD, compared with BMI, in women from the International Day for the Evaluation of Abdominal Obesity (IDEA) study.30 31 An increase in WC was associated with being 4.25 times more likely of stroke and transient ischaemic attacks.32 Conversely, some studies reported that the association between BMI and CVD was similar to measures of central obesity.33 34

rom the International Day for the Evaluation of Abdominal Obesity (IDEA) study.30 31 An increase in WC was associated with being 4.25 times more likely of stroke and transient ischaemic attacks.32 Conversely, some studies reported that the association between BMI and CVD was similar to measures of central obesity.33 34 There are several possible explanations for our study findings that measures of central obesity are better predictors of CVD risk than BMI. Greater central obesity is associated with systemic inflammation which directly contributes to CVD risk.35 Hence, measures that account for the accumulation of excess abdominal fat would report stronger associations and are desirable for assessing adiposity. They would also be more accurate at indicating CVD risk and should be incorporated into CVD assessment.36–39 The addition of central obesity measures to BMI has also been shown to improve the accuracy of stratifying participants into lower and higher risk categories for mortality40 and provides incremental value in predicting CVD above and beyond that provided by general obesity measures.41–45 BMI is a flawed measure as it does not correctly identify individuals with excess body fat due to its inability to differentiate fat and fat-free mass and it does not account for the effect of age and ethnicity on body fat distribution.46–50 An increase in muscle or fat-free mass would, however, be reflected in the central obesity measures.

e as it does not correctly identify individuals with excess body fat due to its inability to differentiate fat and fat-free mass and it does not account for the effect of age and ethnicity on body fat distribution.46–50 An increase in muscle or fat-free mass would, however, be reflected in the central obesity measures. Among central obesity measures, we found their performance to be comparable in our study. It remains unclear which measurement should be incorporated into CVD risk score models. A collaborative analysis of 58 prospective studies, however, reported that measures of general and central obesity did not improve CVD risk assessment when information is available on SBP, diabetes and lipids.51 Overweight and obesity are, nevertheless, important in CVD prevention, with one of three fatal and one of seven non-fatal CVD cases attributable to it.34

udies, however, reported that measures of general and central obesity did not improve CVD risk assessment when information is available on SBP, diabetes and lipids.51 Overweight and obesity are, nevertheless, important in CVD prevention, with one of three fatal and one of seven non-fatal CVD cases attributable to it.34 Opinion remains divided as to which is a more appropriate measurement for assessing adiposity and its association with CVD risk.37 Some studies recommended the use of WC in clinical assessment and research studies.52 53 In a systematic review and meta-analysis study of Caucasians without CVD, WC was most highly correlated with all CVD risk factors, compared with BMI, WHR, WSR and body fat percentage, in women.52 In other studies, WC was also more closely associated with CVD risk factors than other measures of central obesity and BMI in women.54–57 The advantages of WC are: it is easy to measure and interpret and it is less prone to measurement and calculation error.53 Appropriate sex, age and ethnic-specific WC cut-points would need to be established.44 It would also be difficult to use WC in today's multicultural societies due to requirements for different cut-points.50

tages of WC are: it is easy to measure and interpret and it is less prone to measurement and calculation error.53 Appropriate sex, age and ethnic-specific WC cut-points would need to be established.44 It would also be difficult to use WC in today's multicultural societies due to requirements for different cut-points.50 The use of WHR is also supported as it is less strongly associated with BMI than WC, and is thus a more specific surrogate for fat distribution.40 A longitudinal population study on 1462 women from Sweden reported stronger relations between WHR and CVD endpoints, compared with BMI, WC and HC.58 These relations were mostly independent of age, BMI and either SBP, cholesterol level or smoking habit.58 In a meta-regression analysis of prospective studies, WHR was also more strongly associated with CVD compared with WC, although the difference was not significant.37 Another study reported that WHR was associated with CVD mortality but not WC in elderly women from the UK.59 Elevated WHR was also independently associated with a higher CVD risk in the Nurses’ Health Study and in the Swedish Women's Lifestyle and Health Cohort Study.45 60 Women with a WHR of ≥0.88 were 3.25 times more at risk of CHD compared with women with a WHR of <0.72 after adjusting for BMI and other CVD risk factors.45 Higher age-adjusted and sex-adjusted ORs were also reported with WHR and CHD and CVD mortality, compared with WC and BMI, in an Australian population without heart disease, diabetes or stroke.61 Similar results were presented in other studies. WHR reported the highest age standardised HRs in relation to CVD mortality, followed by WSR, WC and BMI in women.62 63 The advantages of WHR are: it has low measurement error, high precision and no bias over a wide range of ethnic groups.64 WHR, however, may not be suitable for assessing central obesity in the elderly65 due to laxity of abdominal muscles which would undermine the predictive value of abdominal circumferences.55 It is also more difficult to measure than WC.37 Despite its limitations, WHR has been recommended for incorporation into CVD risk assessment.37

may not be suitable for assessing central obesity in the elderly65 due to laxity of abdominal muscles which would undermine the predictive value of abdominal circumferences.55 It is also more difficult to measure than WC.37 Despite its limitations, WHR has been recommended for incorporation into CVD risk assessment.37 WSR is the least commonly used measure of central obesity. In a systematic review and meta-analysis study, WSR reported the weakest correlations with CVD risk factors, compared with BMI and other measures of central obesity,52 which is contrary to our study findings. In contrast, WSR was most highly correlated with CHD risk predicted using the Framingham model18 in women from England, compared with BMI, WC and WHR in another study.66 WSR, however, reported lower correlations than WC and BMI following adjustments for age.66 The advantage of WSR is that the same cut-point could be applied across a wide range of populations. A cut-off value of 0.5 indicates increased risk for men and women and people of different ethnic groups, and this value may also be used in children and adults, unlike WC which requires different cut-offs.67 68 More research is required to assess the association between WSR and CVD risk in women, in comparison with WC, WHR and BMI.

ue of 0.5 indicates increased risk for men and women and people of different ethnic groups, and this value may also be used in children and adults, unlike WC which requires different cut-offs.67 68 More research is required to assess the association between WSR and CVD risk in women, in comparison with WC, WHR and BMI. Our study has limitations. This study is cross sectional; however, it is a representative sample of the Australian female population. There is only one set of baseline measurements recorded for some risk variables but important variables including anthropometric measures of obesity are measured twice. Further, the 10-year CVD risks are calculated using risk score models to stratify individuals against the treatment thresholds of the various models, and are not prospective CVD events. Conclusions Central obesity is more strongly associated with CVD risk than general obesity. The deposition of adipose tissue is associated with systemic inflammation which has a direct effect on CVD risk. Therefore, increments in central obesity have a more detrimental effect on CVD risk compared with increments in general obesity.

entral obesity is more strongly associated with CVD risk than general obesity. The deposition of adipose tissue is associated with systemic inflammation which has a direct effect on CVD risk. Therefore, increments in central obesity have a more detrimental effect on CVD risk compared with increments in general obesity. When used alone, BMI is inadequate for identifying individuals at increased risk of CVD as it does not differentiate between fat and fat-free mass. On the other hand, anthropometric measurements of central obesity have higher sensitivity and specificity. These measures are also more sensitive to lifestyle modifications. An increase in muscle mass through diet and training would lead to changes in measures such as WC and WSR but little change might be indicated with BMI.69 It would be more useful to measure a patient's central obesity during clinical assessment to evaluate the effect of lifestyle changes in relation to CVD risk compared with BMI. Central obesity measures are also significant and independent predictors of CVD risk, accounting for additional risk above BMI. These measurements should be incorporated into CVD risk assessment, particularly when assessing the risk in women and the elderly.53 70–73 Future prospective studies are required to elucidate which anthropometric measurements of central obesity are better indicators or predictors of CVD risk.69 Studies measuring body fat distribution using CT or MRI are desirable to better understand the association between body fat distribution and mortality, but are costly.74

prospective studies are required to elucidate which anthropometric measurements of central obesity are better indicators or predictors of CVD risk.69 Studies measuring body fat distribution using CT or MRI are desirable to better understand the association between body fat distribution and mortality, but are costly.74 In conclusion, WC, WHR and WSR, or measures of central obesity that include a measurement of WC, should be considered for incorporation into the clinical assessment of CVD risk. Treatment of well-established CVD risk factors coupled with reducing overweight and obesity through lifestyle modifications would be an advisable goal in the primary prevention of CVD.4 It is equally important to maintain a healthy weight and to prevent central or abdominal obesity concurrently. Supplementary Material Author's manuscript Reviewer comments Curtin University provided educational support to LGHG through the Curtin International Postgraduate Research Scholarship. Contributors: LGHG was involved in drafting the manuscript, interpretation of the data and revising the manuscript critically for important intellectual content. SSD conceived the study, performed the analysis and data interpretation and revised the manuscript critically for important intellectual content. TAW participated in the study design, acquired the data and revised the manuscript critically for important intellectual content. All authors read and approved the final manuscript. Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Contributors: LGHG was involved in drafting the manuscript, interpretation of the data and revising the manuscript critically for important intellectual content. SSD conceived the study, performed the analysis and data interpretation and revised the manuscript critically for important intellectual content. TAW participated in the study design, acquired the data and revised the manuscript critically for important intellectual content. All authors read and approved the final manuscript. Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: None. Patient consent: Obtained. Ethics approval: Australian Institute of Health Interim Ethics Committee and Human Research Ethics Committee at Curtin University. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Strengths and limitations of this study This study confirms that ethnicity influences the association between anthropometric obesity measures and cardiovascular disease (CVD) risk. Central obesity measures such as waist circumference and waist-to-hip ratio are better indicators of CVD risk compared to body mass index across ethnic groups. The treatment threshold used for a risk score model affects the predictive ability of anthropometric obesity measures and the same cut point may not be suitable across ethnic groups. It is a cross-sectional study of the Australian female population in 1989 and these results require confirmation from prospective studies. Owing to a sample size of about 200 for the Asian population, different regions in Asia could not be compared. The CVD risk was estimated using risk score models in order to stratify individuals above and below the respective treatment thresholds and not actual CVD events. Introduction In Australia, approximately 63% of adults were overweight and obese in 2011–2012.1 The proportion of the Australian population who are overweight and obese is expected to increase to approximately 66% in the next 5 years.2 The National Health and Medical Research Council has developed Clinical Practice Guidelines for the Management of Overweight and Obesity for Adults, Adolescents and Children in Australia to provide guidance on assessing and managing obesity.3

and obese is expected to increase to approximately 66% in the next 5 years.2 The National Health and Medical Research Council has developed Clinical Practice Guidelines for the Management of Overweight and Obesity for Adults, Adolescents and Children in Australia to provide guidance on assessing and managing obesity.3 Overweight and obesity affects all socioeconomic groups in Australia, but it is more prevalent in some ethnic groups.4 5 Variations exist in the associations between excess weight and obesity-related conditions among different racial and ethnic groups. Ethnicity significantly affects the associations between anthropometric indices used to assess adiposity such as body mass index (BMI) and waist circumference (WC) and cardiovascular disease (CVD) risk factors.6 Previous epidemiological studies which assessed the associations between anthropometric indices of obesity and CVD were mostly conducted in Western societies.7 It is thus not clear which anthropometric obesity measures are more strongly associated with CVD risk in different ethnic groups.8 To address this, it is necessary to examine the relationship between anthropometric obesity measures and CVD risk by ethnicity and this has been proposed in previous studies as well.9–11 These fundamental issues need to be addressed in order to recommend effective weight management and disease prevention strategies to reduce the burden associated with overweight and obesity in all population groups.

c obesity measures and CVD risk by ethnicity and this has been proposed in previous studies as well.9–11 These fundamental issues need to be addressed in order to recommend effective weight management and disease prevention strategies to reduce the burden associated with overweight and obesity in all population groups. The objectives of this study were to determine whether the cross-sectional associations between anthropometric obesity measures (BMI, WC and waist-to-hip ratio (WHR)) and calculated 10-year CVD risk using the Framingham and general CVD risk score models are the same for women of Australian, UK and Ireland, North European, South European and Asian descent. This study would investigate which anthropometric obesity measure is most predictive at identifying women at increased CVD risk in each ethnic group.

ar CVD risk using the Framingham and general CVD risk score models are the same for women of Australian, UK and Ireland, North European, South European and Asian descent. This study would investigate which anthropometric obesity measure is most predictive at identifying women at increased CVD risk in each ethnic group. Methods Study participants Participants were selected from the third Risk Factor Prevalence Study12 conducted by the National Heart Foundation (NHF) of Australia in 1989. Residents on the federal electoral rolls of December 1988 in North and South Sydney, Melbourne, Brisbane, Adelaide, Perth, Hobart, Darwin and Canberra were recruited for the Risk Factor Prevalence Study by systemic probability sampling of sex and 5-year age groups. Complete data were available on 4727 women. Country of birth was used as a surrogate for ethnicity and grouped into regions.12 Most participants were of Australian, UK and Ireland, North European, South European or Asian descent (97%). We selected a representative sample of 4354 women aged 20–69 years with no history of heart disease, diabetes or stroke for analysis. There were 3329 Australian women, 416 women from the UK and Ireland, 180 Northern European women, 234 Southern European women and 195 Asian women. Further details have been described in the third Risk Factor Prevalence Study and in a previous study.12 13 Ethics statement Participation was entirely voluntary. Those who participated signed an informed consent form.12 Participant information was anonymised prior to analysis.

Methods Study participants Participants were selected from the third Risk Factor Prevalence Study12 conducted by the National Heart Foundation (NHF) of Australia in 1989. Residents on the federal electoral rolls of December 1988 in North and South Sydney, Melbourne, Brisbane, Adelaide, Perth, Hobart, Darwin and Canberra were recruited for the Risk Factor Prevalence Study by systemic probability sampling of sex and 5-year age groups. Complete data were available on 4727 women. Country of birth was used as a surrogate for ethnicity and grouped into regions.12 Most participants were of Australian, UK and Ireland, North European, South European or Asian descent (97%). We selected a representative sample of 4354 women aged 20–69 years with no history of heart disease, diabetes or stroke for analysis. There were 3329 Australian women, 416 women from the UK and Ireland, 180 Northern European women, 234 Southern European women and 195 Asian women. Further details have been described in the third Risk Factor Prevalence Study and in a previous study.12 13 Ethics statement Participation was entirely voluntary. Those who participated signed an informed consent form.12 Participant information was anonymised prior to analysis. Anthropometry A single record of height (to the nearest centimetre) and weight (to the nearest 10th of a kilogram) was taken in light summer clothes without shoes. BMI was calculated based on weight in kilograms divided by square of height in metres. Waist and hip circumferences were measured according to standardised methodologies by trained anthropometrists.14 15 The WC was measured from the front at the narrowest point between the rib cage and iliac crest after full expiration while the hip circumference (HC) was measured from the side at the maximal extension of buttocks by one observer using a metal tape. A second observer recorded another set of measurements and ensured that the metal tape was kept strictly horizontal at all times. The mean of two measurements was taken at each site to the nearest centimetre. The WHR was calculated based on WC divided by the HC. Information on demographic characteristics, medical conditions and smoking behaviour were collected. Mercury sphygmomanometers were used to record blood pressure levels on the right arm of seated participants 5 min apart.12 Two readings were taken and the average was used in the analysis. Fasting blood samples were collected in ethylene diamine tetraacetic acid tubes and despatched to the central laboratory at the Division of Clinical Chemistry, Institute of Medical and Veterinary Science, Adelaide each week for cholesterol levels to be assayed.

dings were taken and the average was used in the analysis. Fasting blood samples were collected in ethylene diamine tetraacetic acid tubes and despatched to the central laboratory at the Division of Clinical Chemistry, Institute of Medical and Veterinary Science, Adelaide each week for cholesterol levels to be assayed. Risk score models The Framingham risk score model16 predicts the 10-year CVD incidence. It was developed from the American Framingham Heart Study using participants aged 30–74 years who were free of CVD and cancer. Risk variables used to calculate the 10-year risk include: age, sex, systolic blood pressure (SBP), diastolic blood pressure, total cholesterol level, high-density lipoprotein (HDL) cholesterol level, smoking status, diabetes status and ECG-left ventricular hypertrophy (ECG-LVH).16 The most commonly used treatment threshold for the Framingham model was 20%17; this denotes that an individual who has a risk score of more than 20% is considered to be at increased risk of experiencing a CVD event within the next 10 years and should be targeted for treatment.

-left ventricular hypertrophy (ECG-LVH).16 The most commonly used treatment threshold for the Framingham model was 20%17; this denotes that an individual who has a risk score of more than 20% is considered to be at increased risk of experiencing a CVD event within the next 10 years and should be targeted for treatment. Although the general CVD risk score model for predicting the 10-year CVD incidence and death was also developed based on data from the American Framingham Heart Study, it was developed from a larger cohort and consisted of participants without CVD only.18 The general CVD risk score model contains these variables: age, total cholesterol level, HDL cholesterol level, SBP, current antihypertensive treatment, smoking status and diabetes status.18 Treatment thresholds of 10% and 20% were reported for this model.18 19

r cohort and consisted of participants without CVD only.18 The general CVD risk score model contains these variables: age, total cholesterol level, HDL cholesterol level, SBP, current antihypertensive treatment, smoking status and diabetes status.18 Treatment thresholds of 10% and 20% were reported for this model.18 19 Statistical analysis Demographic and clinical characteristics of the sample were described using mean±SD for continuous variables, while counts (percentages) were used for categorical variables. Comparisons between means of continuous variables were conducted using Analysis of Variance, with age as a covariate and with Bonferroni adjustment for multiple comparisons. Means with different superscripts were significantly different at the 5% level of significance. Non-parametric Spearman's rank correlation was used to assess the associations between BMI, WC and WHR and the 10-year predicted CVD risk calculated using Framingham and general CVD risk score models by ethnicity, due to the skewness in the distribution of risk variables. These measures were also converted to z-scores (original value subtracted by the mean and the result divided by the SD) to represent the number of SDs above and below the mean of each anthropometric obesity measure for each individual. Logistic regression was used to assess the effects of each standardised obesity measure of being above the recommended treatment threshold for the respective risk score models (10 and 20%), as a result of a 1 SD increment above the mean of each measure of obesity, by ethnicity. These effects were represented using ORs and associated 95% CIs. The predictive ability of these anthropometric obesity measures to identify individuals from different ethnic groups above the treatment threshold of 20% for the Framingham model for 10-year CVD incidence, and 10 and 20% for the general CVD risk score model for 10-year CVD incidence and death was assessed using the area under the receiver operating characteristic (ROC) curve. Ethnic-specific cut-off values of the anthropometric obesity measures and associated level of specificity to predict increased risk of CVD at 70% and 80% sensitivity were also presented. p Values of less than 0.05 were considered to be statistically significant. All statistical analyses were performed with IBM SPSS Statistics V.21.

ic cut-off values of the anthropometric obesity measures and associated level of specificity to predict increased risk of CVD at 70% and 80% sensitivity were also presented. p Values of less than 0.05 were considered to be statistically significant. All statistical analyses were performed with IBM SPSS Statistics V.21. Results The demographic and clinical characteristics of the multiethnic sample of 4354 women without heart disease, diabetes or stroke are presented in table 1. Southern European women generally had higher BMI, WC and WHR compared to other ethnic groups, and Asian women had lower anthropometric obesity measures. Table 1 Characteristics of the sample of 4354 women without heart disease, diabetes or stroke by ethnicity

Results The demographic and clinical characteristics of the multiethnic sample of 4354 women without heart disease, diabetes or stroke are presented in table 1. Southern European women generally had higher BMI, WC and WHR compared to other ethnic groups, and Asian women had lower anthropometric obesity measures. Table 1 Characteristics of the sample of 4354 women without heart disease, diabetes or stroke by ethnicity Statistics Australia UK and Ireland Northern Europe Southern Europe Asia Count N 3329 416 180 234 195 Age (years) Mean±SD 41.9±13.5 45.7±12.5 49.0±11.7 47.8±10.6 40.5±10.9 Current smoker (Yes) n (%) 751 (22.6) 91 (21.9) 39 (21.7) 32 (13.7) 19 (9.7) Weight (kg) Mean±SD 65.4±12.6a 65.2±12.0a 66.5±12.6a 66.9±11.8a 58.6±11.6b Height (cm) Mean±SD 162.8±6.0a 162.3±6.2a 161.9±6.2a 156.8±6.1b 156.7±5.7b BMI (kg/m2) Mean±SD 24.7±4.8b 24.7±4.2b,c 25.4±4.6b,d 27.2±4.4a 23.8±4.3c,d WC (cm) Mean±SD 75.9±11.0b 76.2±10.5b 78.4±11.9b 81.2±11.0a 73.9±10.4b WHR Mean±SD 0.76±0.06c 0.76±0.06c 0.77±0.07b,c 0.79±0.06a 0.77±0.06a,b SBP (mm Hg) Mean±SD 122±18a 123±18b,c 126±19a,b,c 127±19a,b 116±19c HDL (mmol/L) Mean±SD 1.5±0.4a 1.5±0.4a 1.5±0.4a 1.4±0.3b 1.4±0.4a,b TC (mmol/L) Mean±SD 5.4±1.1 5.6±1.2 5.7±1.3 5.7±1.1 5.2±1.0 Ratio: HDL to TC Mean±SD 3.9±1.3b 4.0±1.4a,b 4.0±1.4b 4.3±1.4a 3.9±1.2a,b a,b,c,d Means with different superscripts were significantly different at the 5% level of significance, after adjusting for age. BMI, body mass index; HDL, high-density lipoprotein cholesterol; SBP, systolic blood pressure; TC, total cholesterol; WC, waist circumference; WHR, waist-to-hip ratio.

Statistics Australia UK and Ireland Northern Europe Southern Europe Asia Count N 3329 416 180 234 195 Age (years) Mean±SD 41.9±13.5 45.7±12.5 49.0±11.7 47.8±10.6 40.5±10.9 Current smoker (Yes) n (%) 751 (22.6) 91 (21.9) 39 (21.7) 32 (13.7) 19 (9.7) Weight (kg) Mean±SD 65.4±12.6a 65.2±12.0a 66.5±12.6a 66.9±11.8a 58.6±11.6b Height (cm) Mean±SD 162.8±6.0a 162.3±6.2a 161.9±6.2a 156.8±6.1b 156.7±5.7b BMI (kg/m2) Mean±SD 24.7±4.8b 24.7±4.2b,c 25.4±4.6b,d 27.2±4.4a 23.8±4.3c,d WC (cm) Mean±SD 75.9±11.0b 76.2±10.5b 78.4±11.9b 81.2±11.0a 73.9±10.4b WHR Mean±SD 0.76±0.06c 0.76±0.06c 0.77±0.07b,c 0.79±0.06a 0.77±0.06a,b SBP (mm Hg) Mean±SD 122±18a 123±18b,c 126±19a,b,c 127±19a,b 116±19c HDL (mmol/L) Mean±SD 1.5±0.4a 1.5±0.4a 1.5±0.4a 1.4±0.3b 1.4±0.4a,b TC (mmol/L) Mean±SD 5.4±1.1 5.6±1.2 5.7±1.3 5.7±1.1 5.2±1.0 Ratio: HDL to TC Mean±SD 3.9±1.3b 4.0±1.4a,b 4.0±1.4b 4.3±1.4a 3.9±1.2a,b a,b,c,d Means with different superscripts were significantly different at the 5% level of significance, after adjusting for age. BMI, body mass index; HDL, high-density lipoprotein cholesterol; SBP, systolic blood pressure; TC, total cholesterol; WC, waist circumference; WHR, waist-to-hip ratio. All Spearman's rank correlations were statistically significant (p<0.0005). Overall, WC appeared to have a stronger association with the 10-year predicted risk calculated using the general CVD and Framingham risk score models across all ethnic groups except in European women (table 2). BMI appeared to be more associated with CVD risk calculated using both models in Northern European women, while WHR was more associated with the predicted risk in Southern European women.

risk calculated using the general CVD and Framingham risk score models across all ethnic groups except in European women (table 2). BMI appeared to be more associated with CVD risk calculated using both models in Northern European women, while WHR was more associated with the predicted risk in Southern European women. Table 2 Non-parametric correlations between anthropometric measurements of general and central obesity and 10-year predicted risk of CVD incidence and mortality by ethnicity in 4354 women Ethnicity BMI WC WHR General CVD 10-year predicted risk for CVD incidence and death Australia 0.372 0.443 0.402 UK and Ireland 0.360 0.406 0.365 Northern Europe 0.504 0.462 0.435 Southern Europe 0.356 0.479 0.485 Asia 0.306 0.396 0.308 Overall 0.384 0.451 0.408 Framingham 10-year predicted risk for CVD incidence Australia 0.366 0.440 0.405 UK and Ireland 0.349 0.399 0.361 Northern Europe 0.500 0.464 0.445 Southern Europe 0.358 0.483 0.491 Asia 0.311 0.402 0.308 Overall 0.380 0.449 0.412 All Spearman's rank correlations significant at the p<0.0005 level. BMI, body mass index; WC, waist circumference; WHR, waist-to-hip ratio.

Ethnicity BMI WC WHR General CVD 10-year predicted risk for CVD incidence and death Australia 0.372 0.443 0.402 UK and Ireland 0.360 0.406 0.365 Northern Europe 0.504 0.462 0.435 Southern Europe 0.356 0.479 0.485 Asia 0.306 0.396 0.308 Overall 0.384 0.451 0.408 Framingham 10-year predicted risk for CVD incidence Australia 0.366 0.440 0.405 UK and Ireland 0.349 0.399 0.361 Northern Europe 0.500 0.464 0.445 Southern Europe 0.358 0.483 0.491 Asia 0.311 0.402 0.308 Overall 0.380 0.449 0.412 All Spearman's rank correlations significant at the p<0.0005 level. BMI, body mass index; WC, waist circumference; WHR, waist-to-hip ratio. The recommended treatment thresholds for the general CVD risk score model at 10% and 20%, and the Framingham risk score model at 20%, were identified from a review of the literature. Table 3 presents the effects of a 1 SD increment in BMI, WC and WHR above the mean on the likelihood of being above the recommended threshold in each ethnic group. Increase in anthropometric measurements was generally associated with an increased likelihood of being above the treatment thresholds for all models. A 1 SD change in all obesity measures in Asian women did not have a significant effect on the CVD risk as calculated using the general CVD model both at the 10% and 20% threshold. BMI was not effective in predicting the likelihood of being above the treatment threshold across all models for Southern European women.

ls. A 1 SD change in all obesity measures in Asian women did not have a significant effect on the CVD risk as calculated using the general CVD model both at the 10% and 20% threshold. BMI was not effective in predicting the likelihood of being above the treatment threshold across all models for Southern European women. Table 3 ORs and associated 95% CIs of being above the recommended treatment threshold for various risk score models as a result of a 1 SD increment above the mean for each anthropometric measure of obesity by ethnicity

ls. A 1 SD change in all obesity measures in Asian women did not have a significant effect on the CVD risk as calculated using the general CVD model both at the 10% and 20% threshold. BMI was not effective in predicting the likelihood of being above the treatment threshold across all models for Southern European women. Table 3 ORs and associated 95% CIs of being above the recommended treatment threshold for various risk score models as a result of a 1 SD increment above the mean for each anthropometric measure of obesity by ethnicity Ethnicity BMI WC WHR General CVD 10-year predicted risk for CVD incidence and death (threshold=10%)19 Australia 1.69*** (1.55 to 1.85) 2.16*** (1.96 to 2.38) 2.36*** (2.13 to 2.62) UK and Ireland 1.71*** (1.29 to 2.25) 1.86*** (1.42 to 2.43) 2.09*** (1.58 to 2.75) Northern Europe 2.50*** (1.67 to 3.74) 2.28*** (1.61 to 3.24) 2.23*** (1.55 to 3.21) Southern Europe 1.37 (0.97 to 1.94) 1.64** (1.18 to 2.28) 1.89** (1.32 to 2.70) Asia 1.14 (0.62 to 2.09) 1.57 (0.97 to 2.56) 1.48 (0.88 to 2.47) General CVD 10-year predicted risk for CVD incidence and death (threshold=20%)18 20 Australia 1.65*** (1.43 to 1.91) 2.07*** (1.78 to 2.41) 2.11*** (1.80 to 2.47) UK and Ireland 1.12 (0.64 to 1.96) 1.22 (0.73 to 2.05) 1.68* (1.05 to 2.69) Northern Europe 2.60** (1.44 to 4.70) 2.76*** (1.58 to 4.80) 3.23*** (1.74 to 5.97) Southern Europe 1.17 (0.58 to 2.35) 1.77 (0.96 to 3.28) 2.15* (1.11 to 4.18) Asia 0.96 (0.19 to 4.94) 1.15 (0.29 to 4.57) 0.71 (0.13 to 3.92) Framingham 10-year predicted risk for CVD incidence (threshold=20%)21 22 Australia 1.67*** (1.52 to 1.82) 2.13*** (1.94 to 2.34) 2.37*** (2.14 to 2.63) UK and Ireland 1.71*** (1.30 to 2.25) 1.88*** (1.45 to 2.45) 2.16*** (1.64 to 2.85) Northern Europe 2.55*** (1.70 to 3.85) 2.27*** (1.59 to 3.23) 2.33*** (1.60 to 3.40) Southern Europe 1.32 (0.94 to 1.84) 1.67** (1.21 to 2.30) 2.07*** (1.45 to 2.95) Asia 1.65# (0.99 to 2.76) 1.89** (1.20 to 2.97) 1.63* (1.02 to 2.61) *p<0.05, **p<0.01, ***p<0.001, #p=0.054.

.88*** (1.45 to 2.45) 2.16*** (1.64 to 2.85) Northern Europe 2.55*** (1.70 to 3.85) 2.27*** (1.59 to 3.23) 2.33*** (1.60 to 3.40) Southern Europe 1.32 (0.94 to 1.84) 1.67** (1.21 to 2.30) 2.07*** (1.45 to 2.95) Asia 1.65# (0.99 to 2.76) 1.89** (1.20 to 2.97) 1.63* (1.02 to 2.61) *p<0.05, **p<0.01, ***p<0.001, #p=0.054. BMI, body mass index; WC, waist circumference; WHR, waist-to-hip ratio. Table 4 summarises the results in table 3 by presenting only statistically significant anthropometric obesity measures which increase the likelihood of individuals being above the treatment threshold, with measures of obesity ordered corresponding to ORs, from the highest to the lowest. WHR generally recorded higher ORs than WC and BMI and increased the likelihood of individuals of different ethnicity being above the respective treatment thresholds of the respective models. Only BMI presented higher ORs and increased the likelihood of Northern European women being indicated for treatment based on the predicted risk calculated from the general CVD model at the 10% threshold but not 20% threshold and Framingham model at the 20% threshold. WC recorded higher ORs in Asian women using the Framingham model at the 20% threshold. Table 4 Significant anthropometric measurements of general and central obesity and 10-year predicted risk of CVD incidence and mortality by ethnicity

Table 4 summarises the results in table 3 by presenting only statistically significant anthropometric obesity measures which increase the likelihood of individuals being above the treatment threshold, with measures of obesity ordered corresponding to ORs, from the highest to the lowest. WHR generally recorded higher ORs than WC and BMI and increased the likelihood of individuals of different ethnicity being above the respective treatment thresholds of the respective models. Only BMI presented higher ORs and increased the likelihood of Northern European women being indicated for treatment based on the predicted risk calculated from the general CVD model at the 10% threshold but not 20% threshold and Framingham model at the 20% threshold. WC recorded higher ORs in Asian women using the Framingham model at the 20% threshold. Table 4 Significant anthropometric measurements of general and central obesity and 10-year predicted risk of CVD incidence and mortality by ethnicity Ethnicity General CVD (threshold=10%) General CVD (threshold=20%) Framingham (threshold=20%) OR criterion Australia WHR, WC, BMI WHR, WC, BMI WHR, WC, BMI UK and Ireland WHR, WC, BMI WHR WHR, WC, BMI Northern Europe BMI, WC, WHR WHR, WC, BMI BMI, WHR, WC Southern Europe WHR, WC WHR WHR, WC Asia – – WC, WHR, BMI* *p=0.054. Each cell represents statistically significant anthropometric measures of obesity ordered corresponding to ORs, from the highest to lowest. BMI, body mass index; WC, waist circumference; WHR, waist-to-hip ratio.

Ethnicity General CVD (threshold=10%) General CVD (threshold=20%) Framingham (threshold=20%) OR criterion Australia WHR, WC, BMI WHR, WC, BMI WHR, WC, BMI UK and Ireland WHR, WC, BMI WHR WHR, WC, BMI Northern Europe BMI, WC, WHR WHR, WC, BMI BMI, WHR, WC Southern Europe WHR, WC WHR WHR, WC Asia – – WC, WHR, BMI* *p=0.054. Each cell represents statistically significant anthropometric measures of obesity ordered corresponding to ORs, from the highest to lowest. BMI, body mass index; WC, waist circumference; WHR, waist-to-hip ratio. Higher area under the ROC curve, sensitivity and specificity were recorded with WHR in predicting the 10-year CVD risk calculated using the general CVD and Framingham risk score models across most ethnic groups (table 5). The highest area under the ROC curve and specificity value at 80% sensitivity for WHR were 0.866 and 84.9% for Northern European women with the general CVD model at the 20% threshold. Table 5 Area under the curve and cut points for anthropometric measurements of general and central obesity to predict increased risk of CVD using risk score models at different thresholds for various levels of sensitivity and specificity by ethnicity

Higher area under the ROC curve, sensitivity and specificity were recorded with WHR in predicting the 10-year CVD risk calculated using the general CVD and Framingham risk score models across most ethnic groups (table 5). The highest area under the ROC curve and specificity value at 80% sensitivity for WHR were 0.866 and 84.9% for Northern European women with the general CVD model at the 20% threshold. Table 5 Area under the curve and cut points for anthropometric measurements of general and central obesity to predict increased risk of CVD using risk score models at different thresholds for various levels of sensitivity and specificity by ethnicity AUC Sensitivity=70% Sensitivity=80% General CVD 10-year predicted risk for CVD incidence and death (threshold=10%) Australia BMI 0.691 (0.666 to 0.716) 24.2 (60.1%) 23.0 (46.1%) WC 0.750 (0.727 to 0.772) 77.3 (69.6%) 74.3 (57.9%) WHR 0.759 (0.736 to 0.783) 0.77 (70.1%) 0.75 (58.0%) UK and Ireland BMI 0.655 (0.584 to 0.726) 23.7 (50.6%) 22.8 (41.2%) WC 0.676 (0.611 to 0.741) 75.3 (58.5%) 73.3 (51.2%) WHR 0.729 (0.671 to 0.787) 0.77 (65.6%) 0.75 (52.4%) Northern Europe BMI 0.770 (0.695 to 0.845) 25.8 (71.4%) 24.4 (58.7%) WC 0.761 (0.682 to 0.840) 77.8 (66.7%) 75.3 (57.1%) WHR 0.730 (0.642 to 0.817) 0.77 (59.5%) 0.75 (50.8%) Southern Europe BMI 0.618 (0.536 to 0.699) 26.5 (52.8%) 25.5 (44.9%) WC 0.686 (0.604 to 0.768) 81.8 (62.4%) 78.8 (53.4%) WHR 0.702 (0.619 to 0.785) 0.80 (61.8%) 0.79 (57.9%) Asia BMI 0.564 (0.411 to 0.717) 21.9 (38.2%) 21.8 (37.6%) WC 0.651 (0.524 to 0.778) 73.3 (60.0%) 71.8 (52.4%) WHR 0.614 (0.490 to 0.739) 0.76 (56.5%) 0.76 (54.7%) General CVD 10-year predicted risk for CVD incidence and death (threshold=20%) Australia BMI 0.725 (0.677 to 0.772) 25.5 (68.8%) 24.3 (58.1%) WC 0.782 (0.743 to 0.821) 79.8 (72.3%) 77.8 (66.4%) WHR 0.784 (0.745 to 0.823) 0.79 (76.3%) 0.77 (65.7%) UK and Ireland BMI 0.550 (0.414 to 0.685) 23.0 (40.2%) 21.7 (25.4%) WC 0.589 (0.472 to 0.706) 74.8 (52.2%) 73.8 (48.3%) WHR 0.682 (0.572 to 0.791) 0.77 (61.3%) 0.75 (47.3%) Northern Europe BMI 0.818 (0.727 to 0.908) 28.7 (82.4%) 26.3 (67.3%) WC 0.861 (0.785 to 0.936) 85.3 (81.1%) 84.3 (79.2%) WHR 0.866 (0.784 to 0.947) 0.84 (86.8%) 0.83 (84.9%) Southern Europe BMI 0.578 (0.437 to 0.719) 26.8 (51.9%) 26.7 (50.9%) WC 0.711 (0.562 to 0.859) 84.8 (69.6%) 84.8 (69.6%) WHR 0.725 (0.553 to 0.897) 0.80 (62.1%) 0.79 (55.6%) Asia BMI 0.555 (0.303 to 0.807) 25.4 (73.1%) 21.9 (37.9%) WC 0.630 (0.466 to 0.795) 78.3 (73.6%) 71.8 (50.5%) WHR 0.440 (0.306 to 0.573) 0.76 (52.2%) 0.74 (35.7%) Framingham 10-year predicted risk for CVD incidence (threshold=20%) Australia BMI 0.682 (0.657 to 0.707) 24.0 (57.9%) 22.9 (43.8%) WC 0.745 (0.723 to 0.768) 76.8 (67.5%) 73.8 (55.8%) WHR 0.759 (0.736 to 0.781) 0.77 (69.7%) 0.75 (58.1%) UK and Ireland BMI 0.656 (0.586 to 0.726) 23.7 (50.6%) 22.5 (37.5%

.2%) 0.74 (35.7%) Framingham 10-year predicted risk for CVD incidence (threshold=20%) Australia BMI 0.682 (0.657 to 0.707) 24.0 (57.9%) 22.9 (43.8%) WC 0.745 (0.723 to 0.768) 76.8 (67.5%) 73.8 (55.8%) WHR 0.759 (0.736 to 0.781) 0.77 (69.7%) 0.75 (58.1%) UK and Ireland BMI 0.656 (0.586 to 0.726) 23.7 (50.6%) 22.5 (37.5% ) WC 0.682 (0.620 to 0.745) 75.3 (58.6%) 73.3 (51.8%) WHR 0.735 (0.679 to 0.791) 0.77 (65.8%) 0.75 (54.2%) Northern Europe BMI 0.783 (0.710 to 0.856) 26.3 (75.2%) 24.9 (65.1%) WC 0.770 (0.691 to 0.850) 78.8 (71.3%) 76.3 (60.5%) WHR 0.742 (0.652 to 0.832) 0.77 (62.8%) 0.75 (51.2%) Southern Europe BMI 0.597 (0.514 to 0.680) 25.8 (47.1%) 25.1 (40.1%) WC 0.680 (0.601 to 0.760) 80.8 (57.6%) 78.3 (53.5%) WHR 0.711 (0.633 to 0.789) 0.79 (61.6%) 0.78 (51.7%) Asia BMI 0.647 (0.524 to 0.770) 23.5 (55.1%) 21.9 (39.5%) WC 0.688 (0.586 to 0.790) 73.3 (60.5%) 71.8 (53.3%) WHR 0.645 (0.530 to 0.759) 0.76 (56.9%) 0.75 (44.3%) AUC, area under the curve; BMI, body mass index; WC, waist circumference; WHR, waist-to-hip ratio.

difficulties in accurately measuring the HC of severely obese individuals (BMI≥40 kg/m2).13 Measurements may be made in the supine position to overcome this problem.13 In clinical settings, it may be more feasible to assess adiposity using WC while WHR could be measured in research studies as it is more informative.30 Although WHR was the best anthropometric obesity measure in relation to identifying individuals at increased CVD risk, this did not apply to Northern European women. BMI was a better indicator of CVD risk using the general CVD risk score model at the 10% threshold but not 20% threshold and the Framingham risk score model at the 20% threshold, with higher correlations, higher ORs, higher area under the ROC curves, sensitivity and specificity values presented, compared with WHR. At the 20% threshold of the general CVD risk score model, WHR was the better predictor of CVD risk in Northern European women compared to BMI. This indicates that the predictive ability of anthropometric measures of obesity varies with the treatment thresholds used for the respective risk score models, and that the same cut point may not be suitable across ethnic groups.

3.5%) WHR 0.711 (0.633 to 0.789) 0.79 (61.6%) 0.78 (51.7%) Asia BMI 0.647 (0.524 to 0.770) 23.5 (55.1%) 21.9 (39.5%) WC 0.688 (0.586 to 0.790) 73.3 (60.5%) 71.8 (53.3%) WHR 0.645 (0.530 to 0.759) 0.76 (56.9%) 0.75 (44.3%) AUC, area under the curve; BMI, body mass index; WC, waist circumference; WHR, waist-to-hip ratio. In Northern European women, BMI was a better predictor of CVD risk calculated using the general CVD risk score model at the 10% threshold but not 20% threshold and the Framingham risk score model at the 20% threshold, compared with WC and WHR. WHR, however, was the better indicator of CVD risk using the general CVD risk score model with a 20% threshold in Northern European women. In Asian women, WC reported a consistently higher area under the ROC curve, sensitivity and specificity across all CVD models and thresholds. The area under the ROC curve values ranged from 0.630 to 0.688 and specificity values ranged from 50.5% to 53.3% at 80% sensitivity in Asian women. The cut-off values for BMI, WC and WHR are also presented in table 5. A WHR value of 0.75 would indicate increased CVD risk for women from Australia and the UK and Ireland, while a value of 0.78 would indicate increased risk for Southern European women. In Asian women, a WC of 71.8 cm would indicate increased CVD risk. A BMI of 24.4 kg/m2 would indicate increased risk in Northern European women. The diagnostic abilities of the various anthropometric obesity measures to identify women as being above the threshold and hence identified for treatment vary according to ethnic groups.

men, a WC of 71.8 cm would indicate increased CVD risk. A BMI of 24.4 kg/m2 would indicate increased risk in Northern European women. The diagnostic abilities of the various anthropometric obesity measures to identify women as being above the threshold and hence identified for treatment vary according to ethnic groups. Discussion Our study found anthropometric measures of central obesity (WC and WHR) to be better indicators of CVD risk as they measure ectopic body fat (fat stored in the abdominal region), which is associated with decreased glucose tolerance, reduced insulin sensitivity, adverse lipid profiles and other metabolic abnormalities, which are risk factors for CVD and diabetes.8 Stronger associations were also reported between WC and the 10-year predicted CVD risk calculated using the general CVD and Framingham risk score models compared with BMI and WHR across most ethnic groups, while WHR recorded higher ORs than WC and BMI and increased the likelihood of women being above the respective treatment thresholds of the models. WHR also presented higher area under the ROC curve, sensitivity and specificity values. Our findings are consistent with previous studies which have shown that WC and WHR, measures of central adiposity, are superior to BMI in predicting CVD and other obesity-related risk.23–26 WC has already been incorporated in the diagnosis of the metabolic syndrome, a cluster of risk factors for CVD and diabetes.27

es. Our findings are consistent with previous studies which have shown that WC and WHR, measures of central adiposity, are superior to BMI in predicting CVD and other obesity-related risk.23–26 WC has already been incorporated in the diagnosis of the metabolic syndrome, a cluster of risk factors for CVD and diabetes.27 WHR should also be incorporated into CVD risk assessment. Our study provided evidence that WHR is a better diagnostic predictor of CVD than BMI and WC. It is also suitable for assessing adiposity and CVD risk in multiethnic cohorts as it has low measurement error, high precision and no bias over a wide range of ethnic groups.13 Equivalence tests across ethnic groups showed WHR to be independent of ethnicity.13 Similar cut-off values for WHR could also be applied across ethnic groups; a value of 0.75 and 0.78 would indicate increased CVD risk for women of Australia and UK and Ireland, and of Southern Europe descent, respectively. A study conducted on Latin Americans, non-Hispanic Whites and Blacks and Hispanics to estimate the accuracy and optimal cut points for BMI, WC and WHR also found that a cut point of 0.91 for WHR and 94 cm for WC could be used among women of different ethnicity to identify those at high coronary heart disease (CHD) risk.28 WHR also reported the highest area under the ROC curve across all ethnic groups, ranging from 0.75 to 0.82.28 It was also the most accurate measure to screen for high CHD risk individuals.28 Another large case–control study of markers of obesity and myocardial infarction confirmed that WHR is a stronger indicator of myocardial infarction than BMI and increased the population attributable risk of obesity by more than threefold in all ethnic groups.29 The superiority of WHR over BMI and WC in predicting CVD risk is also demonstrated in prospective studies.23 30–33

d myocardial infarction confirmed that WHR is a stronger indicator of myocardial infarction than BMI and increased the population attributable risk of obesity by more than threefold in all ethnic groups.29 The superiority of WHR over BMI and WC in predicting CVD risk is also demonstrated in prospective studies.23 30–33 The measurement of WHR, however, may pose some challenges. For example, it may be inappropriate to measure HC in certain cultures, but this can be overcome with same sex observers.13 Some studies reported that WHR is imprecise while others reported that it is a precise measure.13 30 34 35 The differing results could be related to the rigour of the techniques used; standardised techniques need to be adopted when measuring WHR.13 A study which evaluated the precision of measuring WHR, WC and HC with comparison across ethnic groups using data from the third Risk Factor Prevalence Study found that the coefficients of variation were 0.91% for WHR, 0.78% for WC and 0.57% for HC, less than 1%, indicating good precision in females.13 The measurement error was 0.02 for WHR, 1.66 cm for WC and 1.59 cm for HC between two successive measurements in females.13 In addition, the absolute difference between two WHR measurements for females was not significantly associated with the size of the participants.13 WHR is not suitable for assessing central adiposity in the elderly36 due to laxity of their abdominal muscles, which would undermine the predictive value of abdominal circumferences.37 In addition, WHR may remain constant during weight change and is not suitable for monitoring weight loss.38 Finally, there are technical difficulties in accurately measuring the HC of severely obese individuals (BMI≥40 kg/m2).13 Measurements may be made in the supine position to overcome this problem.13 In clinical settings, it may be more feasible to assess adiposity using WC while WHR could be measured in research studies as it is more informative.30

del, WHR was the better predictor of CVD risk in Northern European women compared to BMI. This indicates that the predictive ability of anthropometric measures of obesity varies with the treatment thresholds used for the respective risk score models, and that the same cut point may not be suitable across ethnic groups. WC was a better predictor of CVD risk among Asian women compared to BMI and WHR. This was consistent with the results of another cross-sectional population-based survey study on Chinese people which reported that WC is the best predictor of CVD risk factors in women.39 It was also the best marker of risk in a 6-year prospective study.40 A small increment in WC predicted a substantial increase in CHD risk in the Chinese population.40 It has been suggested that ethnicity influences specific fat depots, possibly explaining the relationship between ethnicity, adiposity and CVD risk.41 A lower WC cut-off for Asians is necessary to avoid underestimating the population at risk.41 42

redicted a substantial increase in CHD risk in the Chinese population.40 It has been suggested that ethnicity influences specific fat depots, possibly explaining the relationship between ethnicity, adiposity and CVD risk.41 A lower WC cut-off for Asians is necessary to avoid underestimating the population at risk.41 42 Our study has limitations. It is a cross-sectional study of the Australian female population in 1989 and these results require confirmation from prospective studies. In addition, it is limited by the use of country of birth as a proxy for ethnicity.43 Although country of birth is a good proxy for ethnicity in older age minority groups and is of intrinsic interest in distinguishing environmental and genetic differences, it is no longer an appropriate proxy as it does not consider the diversity of country of origin of the individual. The measurement or assignment of ethnicity is difficult and the way forward is possibly to enable people to identify themselves.44 Owing to a sample size of about 200 for the Asian population, different regions in Asia could not be compared. Menopausal status, which is associated with increased central obesity, has not been assessed in our study and has not been incorporated into these risk score models.45 Further, the CVD risk was estimated using risk score models in order to stratify individuals above and below the respective treatment thresholds and not actual CVD events. Only the Framingham and general CVD risk score models were assessed in our study. Other risk score models were excluded either because they could not be determined due to the requirement for variables not being assessed in our study (QRISK)46 or due to the low number of participants above the respective recommended treatment thresholds (SCORE).47 Finally, the 10-year CVD risk for young adults is very rarely elevated, even in the presence of significant risk factors.48

determined due to the requirement for variables not being assessed in our study (QRISK)46 or due to the low number of participants above the respective recommended treatment thresholds (SCORE).47 Finally, the 10-year CVD risk for young adults is very rarely elevated, even in the presence of significant risk factors.48 Conclusions Our study confirms that ethnicity influences the association between anthropometric obesity measures and CVD risk. Central obesity measures such as WC and WHR are better indicators of CVD risk compared to BMI across ethnic groups. WHR is the best anthropometric measure for predicting CVD risk in women except Northern European and Asian women. The treatment threshold used for a risk score model affects the predictive ability of anthropometric obesity measures and the same cut point may not be suitable across ethnic groups. It is important to incorporate ethnicity in CVD risk assessment. Prevention and treatment efforts should be tailored to meet the needs of each ethnic group.49 Ethnic-specific CVD prevention strategies need to be developed to promote healthy eating and physical activity to curtail obesity. Continued population-based prospective research is necessary to elucidate the link between obesity and CVD by ethnicity.

efforts should be tailored to meet the needs of each ethnic group.49 Ethnic-specific CVD prevention strategies need to be developed to promote healthy eating and physical activity to curtail obesity. Continued population-based prospective research is necessary to elucidate the link between obesity and CVD by ethnicity. Supplementary Material Author's manuscript Reviewer comments Contributors: LGHG was involved in drafting the manuscript, performing the analysis, interpretation of data and revising the manuscript critically for important intellectual content. SSD conceived the study, performed the analysis and data interpretation and revised the manuscript critically for important intellectual content. TAW participated in the study design, acquired the data and revised the manuscript critically for important intellectual content. All authors read and approved the final manuscript. Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: Curtin University provided educational support to LGHG through the Curtin International Postgraduate Research Scholarship. Patient consent: Obtained. Ethics approval: We have ethics approval for the use of the National Heart Foundation data from the Australian Institute of Health Interim Ethics Committee, after consultation with the Commonwealth Privacy Commissioner, and approval from the Human Research Ethics Committee at Curtin University. This study was carried out in accordance with the Declaration of Helsinki.

se of the National Heart Foundation data from the Australian Institute of Health Interim Ethics Committee, after consultation with the Commonwealth Privacy Commissioner, and approval from the Human Research Ethics Committee at Curtin University. This study was carried out in accordance with the Declaration of Helsinki. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.