• Body mass index (BMI);
  • receiver operating characteristics (ROC);
  • waist circumference (WC);
  • waist-to-height ratio (WHtR).


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Funding statement
  9. Conflict of Interest Statement
  10. Acknowledgements
  11. References

Our aim was to differentiate the screening potential of waist-to-height ratio (WHtR) and waist circumference (WC) for adult cardiometabolic risk in people of different nationalities and to compare both with body mass index (BMI). We undertook a systematic review and meta-analysis of studies that used receiver operating characteristics (ROC) curves for assessing the discriminatory power of anthropometric indices in distinguishing adults with hypertension, type-2 diabetes, dyslipidaemia, metabolic syndrome and general cardiovascular outcomes (CVD). Thirty one papers met the inclusion criteria. Using data on all outcomes, averaged within study group, WHtR had significantly greater discriminatory power compared with BMI. Compared with BMI, WC improved discrimination of adverse outcomes by 3% (P < 0.05) and WHtR improved discrimination by 4–5% over BMI (P < 0.01). Most importantly, statistical analysis of the within-study difference in AUC showed WHtR to be significantly better than WC for diabetes, hypertension, CVD and all outcomes (P < 0.005) in men and women.

For the first time, robust statistical evidence from studies involving more than 300 000 adults in several ethnic groups, shows the superiority of WHtR over WC and BMI for detecting cardiometabolic risk factors in both sexes. Waist-to-height ratio should therefore be considered as a screening tool.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Funding statement
  9. Conflict of Interest Statement
  10. Acknowledgements
  11. References

Body mass index (BMI) has been used as a proxy for obesity for many years, but, in recent years, indices of abdominal obesity (first waist–hip ratio [WHpR] and then waist circumference [WC]) have increasingly been associated with higher cardiometabolic risk in both cross-sectional and prospective studies. The use of waist-to-height ratio (WHtR) for detecting abdominal obesity, and health risks associated with it, was first proposed in the mid-1990s (1–3). Interest in the effectiveness of this measure is rising in both adults and children in many different ethnic groups and countries (4–10).

A meta-analysis published in 2008 (11) focused on hypertension (HT) and included data from 19 cross-sectional studies in the Asia–Pacific region. It concluded that ‘No anthropometric variable was systematically better than others at the discrimination of hypertension’. A further meta-analysis, looking at more risk factors but only including 10 papers published up to the end of 2006 (12), concluded that statistical evidence supported the superiority of measures of centralized obesity, especially WHtR, over BMI for detecting cardiovascular (cardiovascular disease [CVD]) risk in men and women.

A systematic review (9) included the evidence from 78 studies published from 1950 to mid-2009 and drew on evidence from prospective and cross-sectional studies, in adults and in children, which reported relationships between WHtR and either BMI, WC, or both, and outcomes of cardiometabolic risk. Although this narrative analysis was done on many more studies than its forerunners, it could only suggest that WHtR and WC were significant predictors of cardiometabolic outcomes more often than BMI because meta-analysis of all studies was not included.

The volume of papers reporting relevant data has increased substantially. We therefore took this opportunity to perform a full robust meta-analysis on the published data until 2010 to see which indicator of abdominal obesity should be recommended for screening purposes.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Funding statement
  9. Conflict of Interest Statement
  10. Acknowledgements
  11. References

Literature search strategy

A systematic review was conducted which considered papers in the Medline and EMBASE databases up until 31 August 2010. Search terms used a combination of keywords: body mass index or BMI, waist or wst circumference or WC, waist-to-height ratio or waist height or waist ht or WHtR or waist to stature ratio or wst stature or WSR or stature and girth. There were no language restrictions as long as an abstract in English was available. We identified additional studies through hand-searches of bibliographies from primary studies and review articles.

Study selection

Two reviewers (SG and MA) independently assessed the suitability of these retrieved articles for use in the meta-analysis using information supplied in the title and abstract in the first sift. The following inclusion/exclusion criteria were used.

Study inclusion criteria
  • • 
    Primary studies, either prospective or cross-sectional design.
  • • 
    Human subjects, male, female or mixed, adults, any ethnic group.
  • • 
    WHtR, and either BMI or WC, measured at least once.
  • • 
    Studies with a cardiometabolic risk factor or disease end point and presenting the relationship between the anthropometric indices and this end point (for diagnostics used to define end points, please see Table 1).
  • • 
    Studies reporting receiver operating characteristics (ROC) analyses.
Table 1.  Details of included studies
ReferencesAuthorsYear study conductedCountryEthnicityDesignSubjects – menSubjects – womenAge range (year) or mean ± SDRisk factors with definitions
  1. Full diagnostics available from the authors.

(15)Aekplakorn, 20062000ThailandThaiCross-sectional2,0933,212≥35D (1), HT (1), Dyslip (1), MS (1)
(14)Aekplakorn, 20071985ThailandThaiProspective (17 years follow-up)2,536035–59CVD (1)
(16)Can, 20092003TurkeyTurkishCross-sectional5711,121≥18D (1), HT (2), Dyslip (1), MS (2), CVD (2)
(17)Cristo Rodriguez Perez, 20102000–5Canary IslandsSpanishCross-sectional2,9133,81643.1 ± 12.8 (m); 43.0 ± 12.9 (w)D (NR), HT (NR), Dyslip (NR), MS (NR), CVD (NR)
(18)Gracey, 20072003 and 2004AustraliaAboriginalCross-sectional14418618–88D (2), HT (2), Dyslip (8)
(19)He, 20092002China (National)ChineseCross-sectional23,98026,92518–79D (3)
(20)Ho, 20031995 and 1996Hong KongChineseCross-sectional1,4121,48325–74D (2), HT (3), Dyslip (2), CVD (3)
(21)Hsieh and Muto1996–1999TokyoJapaneseCross-sectional6,1412,13749.5 ± 8.9MS (3)
51.9 ± 9.0
(22)Haun, 20092000BrazilBrazilianCross-sectional39157730–74CVD (2)
(40)Koch, 20081997–1999ChileChileanProspective (8 years)6,7146,34042 ± 15 (m and w)D (1), HT (1), Dyslip (2), MS (5)
(23)Lee, 20082005–7KoreaKoreanCross-sectional577995>30D (4), HT (4), Dyslip (6)
(24)Li, Ford, 20101988–1994USA NationalNon-Hispanic White; non-Hispanic Black; and Mexican AmericansCross-sectional2,9943,283>20D (2)
(25)Li, 20091999–2001Australian Aborigines and Torres Strait islandersIndigenous AustralianCross-sectional881471>15D (5), HT (1), Dyslip (3)
(26)Lin, 20021998–2000TaiwanTaiwaneseCross-sectional26,35929,20437.3 ± 10.9 (m)D (6), HT (1), Dyslip (1), MS (1)
37.0 ± 11.1(w)
(27)Mansour, 20072005IraqIraqiCross-sectional6,6936,29345.6 ± 15.7 (m and w)D (2), HT (2)
(28)Mellati, 20092002–2003Zanjan, IranIranianCross-sectional1,3101,45821–75D (6), HT (6), Dyslip (7)
(29)Mirmiran, 2004Pre-2000?Tehran, IranIranianCross-sectional4,4496,07318–74D (6), HT (2), Dyslip (1), MS (1)
(30)Mombelli, 20092008ItalyItalianCross-sectional55255258 ± 13 (m)MS (4)
64 ± 11 (w)
(41)Page, 20091986USA Nurses studyAmerican mixedProspective (16 years)045,56340–65CVD (1)
(31)Paniagua, 20082006–2007ThailandThaiCross-sectional451940≥35D (4), HT (4), Dyslip (6)
(32)Park, 20092005Korea NationalKoreanCross-sectional2,3273,102>20D (1), HT (5), Dyslip (1)
(33)Pua, 20052003SingaporeChinese, Malay, IndianCross-sectional056618–68D (6), HT (1), Dyslip (4), MS (1)
(34)Sakurai, 20061996JapanJapaneseCross-sectional2,9351,62235–59HT (2)
(42)Sargeant, 20021993–1996JamaicaAfrican ancestryProspective (4 years)29043825–74D (2)
(5)Schneider, 20072003GermanyMainly CaucasianCross-sectional2,0163,36120–79D (7), Dyslip (5), MS (3)
(43)Shafiee, 2009Not statedIranIranianProspective (3.5 years)02,801>20D (3)
(35)Taylor, 2010Not statedNew Zealand(47% Maori, 53% WhiteCross-sectional4521,08717–82D (8), MS (4)
(36)Tseng, 20102001TaiwanTaiwaneseCross-sectional2,2802,40325–75D (1), HT (3), Dyslip (6)
(37)Wang, 20092007Beijing, ChinaChinese HanCross-sectional3,7046,39218–85MS (5)
(38)Welborn, 20071989AustraliaAustralianProspective (11 years)4,5084,69820–69CVD (1)
(39)Zhou, 20092006ChinaChineseCross-sectional13,55815,52130–100HT (2)
Study exclusion criteria
  • • 
    Literature reviews, intervention studies, abstracts from conference proceedings.
  • • 
    Studies in children and adolescents.

Once papers had been identified on the basis of information in the title and abstract, full papers were obtained for all relevant studies, either from journal archives or directly from the authors. These papers were then scrutinized by the two reviewers (SG and MA) independently to identify those with sufficient information for data extraction for the meta-analysis.

Meta-analysis of data from receiver operating characteristics curves

Meta-analysis was used to assess the overall power and precision of each of the three anthropometric indices in predicting the five disease outcomes or risk factors (diabetes, HT, dyslipidaemia, metabolic syndrome [MS] and CVD [including coronary heart disease (CHD) incidence/mortality and high CHD risk]). Data on area under the ROC curve (AUC) for each study (means, 95% confidence interval [CI], sample size) were extracted and entered into a database for analysis by software package Comprehensive Meta-analysis v2 (CMA: Biostat Inc., NJ, USA). Where papers gave values adjusted for covariates, these were used in preference to unadjusted values. Results for men and women were treated separately. Where studies reported results in two ethnic groups, these were treated separately (i.e. study group was used as the unit of analysis). Where studies reported individual health outcomes as well as composite risk scores, the analysis was restricted to the former. Data given for different age groups were combined to give an all-age mean. Random effects models were used, as most appropriate for studies from different populations, or where criteria for outcomes may differ between studies (13). Random effects analysis allows that the true effect size may vary from study to study. The schematic plots (forest plots) illustrate the AUC and confidence limits for each study and the weighted mean AUC over all studies, by index and by outcome. Differences in mean effect size (pooled AUC) between BMI and WC and between BMI and WHtR were assessed for significance using the Q statistic (heterogeneity). For outcomes in which WHtR was significantly different from BMI, we also investigated whether WHtR gave improved prediction over WC by running a separate meta-analysis of the within-study difference between these two indices (Z-test) assuming a correlation of 0.95 between WC and WHtR (14).


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Funding statement
  9. Conflict of Interest Statement
  10. Acknowledgements
  11. References

Study selection

Figure 1 shows the flow chart of study selection. The search produced 693 references, of which 304 were excluded for the reasons shown. From the 389 relevant papers, 77 abstracts were identified as potentially suitable because they contained ROC analysis of the relevant anthropometric indices and the relevant cardiometabolic outcomes. Full text was therefore scanned to ascertain whether sufficient data were available to input into the meta-analysis. Forty-nine of the 77 papers were excluded because they did not have the relevant data. In some instances, authors were contacted directly to supply missing information (e.g. obtaining CIs for the mean AUC values). Three other papers were identified as being relevant from cross referencing. Finally, 31 papers were deemed to be suitable for data extraction for the meta-analysis. Of these, 30 included women and 28 included men.


Figure 1. Flow chart of study selection. AUC, area under the curve; AUROC, area under the receiver operating characteristics curve; BMI, body mass index; CI, confidence interval; WC, waist circumference; WHtR, waist-to-height ratio.

Download figure to PowerPoint

Included papers

Table 1 shows details of these 31 papers. Of these, 26 reported cross-sectional studies (5,15–39), and five reported prospective studies (14,40–43), the longest with a follow-up of 17 years (14).

Studies were conducted between 1985 and 2008, in 18 different countries. Six countries were from Asia, two from the Middle East, two from Australasia, four from Europe, two from South America and one from the Caribbean. Fifteen studies included subjects of Asian ethnicity. The study population size ranged from less than 200 to over 45,000 participants with a total of 123,231 men and 182,620 women. The age limits for inclusion into each of the individual studies ranged from 18 to 100 years.

Outcomes were grouped into five broad categories: diabetes (D), HT, dyslipidaemia, MS and CVD outcomes (includes CHD and CVD outcomes and calculations of high coronary risk score). Papers used different diagnostic criteria for these outcomes, which are summarized in Table 1. A summary category ‘all outcomes’ was also created, with a mean AUC calculated as the mean of all measured outcomes within study group, (n = 33).


Table 2 shows the mean area under the curve (AUC) and lower and upper 95% CIs for two sexes, for three anthropometric indices, for five specific health outcomes and for a combined outcomes variable. Statistical significance (P values) of pooled AUCs for the two indices of abdominal obesity (WC and WHtR) compared with BMI (a proxy for total obesity) is also shown for all the health outcomes.

Table 2.  AUC values with 95% confidence intervals for anthropometric indices against health outcomes in men and women
 MeanLowerUpperP value for comparison with BMI
  1. Statistical test (Q statistic) for heterogeneity in effect sizes between indices (WC vs. BMI; WHtR vs. BMI).

  2. For more powerful test of the hypothesis that the difference between WC and WHtR equals zero, see Table 3.

  3. AUC, area under the curve; BMI, body mass index; CVD, cardiovascular disease; HT, hypertension; MS, metabolic syndrome; TG, triglyceride; WC, waist circumference; WHtR, waist-to-height ratio.

Men (n = 22 groups)
Women (n = 24 groups)
Men (n = 18 groups)
Women (n = 19 groups)
Dyslipidaemia/high TG
Men (n = 16 groups)
Women (n = 17 groups)
Men (n = 12 groups)
Women (n = 13 groups)
Men (n = 6 groups)
Women (n = 6 groups)
All outcomes (mean of measured outcomes for each study)
Men (n = 31 groups)
Women (n = 33 groups)

The implications of these results are reported in relation to the specific health outcome:


Twenty studies for men (5,15–20,23–29,31,32,35,36,40,42) and 22 for women (as men plus (33) and (43)) included diabetes as an outcome. As the studies by Li and McDermott (25) and Taylor et al. (35) each gave data on two ethnic groups, this made a total of 22 study groups for men and 24 for women.

As Table 2 shows, WHtR had the highest, and BMI had the lowest, pooled AUC for both men and women, while the AUC for WC was intermediate. Diabetes risk among women was correctly discriminated by BMI 70% of the time (i.e. mean AUC was 0.70 with 95% CIs at 0.67 and 0.73). In contrast, diabetes was correctly discriminated 74% of the time for WC (mean AUC = 0.74 [0.72, 0.76]) and 75% for WHtR (mean AUC = 0.75 [0.73, 0.77]). Among men, mean AUC for BMI was 0.66 (0.64, 0.69); for WC it was 0.70 (0.68, 0.72); and for WHtR it was 0.71 (0.69, 0.73).

Figures 2 and 3 show the Forest plots of AUC scores for diabetes risk in men and women, respectively. Random effects pooled area under the ROC curves (AUC) is shown for BMI, WC and WHtR (a, b and c). Horizontal lines represent the 95% CI; diamonds represent the overall estimates.


Figure 2. Discrimination of diabetes among men and women: pooled AUCs for BMI, waist circumference and waist-to-height ratio. (a) Discrimination of diabetes among men: pooled AUC for BMI. (b) Discrimination of diabetes among men: pooled AUC for waist circumference. (c) Discrimination of diabetes among men: pooled AUC for waist-to-height ratio. AUC, area under the curve; BMI, body mass index; M, male; MA, male Aborigine; ME, male Europeans; MM, male Maori; MT, male Torres Strait islanders.

Download figure to PowerPoint


Figure 3. (a) Discrimination of diabetes among women: pooled AUC for BMI. (b) Discrimination of diabetes among women: pooled AUC for waist circumference. (c) Discrimination of diabetes among women: pooled AUC for waist-to-height ratio. AUC, area under the curve; BMI, body mass index; F, female; FA, female Aborigine; FE, female Europeans; FM, female Maori; FT, female Torres Strait islanders.

Download figure to PowerPoint

Statistical comparison of the AUC scores for indices of abdominal obesity with BMI indicated that both WHtR (P = 0.001 and P < 0.001 in men and women, respectively) and WC (P < 0.05 for WC for both men and women) were significantly better at discriminating diabetes risk compared with BMI.


Eighteen study groups included HT as an outcome for men (15–18,20,23,25–29,31,32,34,36,39,40) and 19 for women (as men plus (33)).

Discrimination was 3–4% better with WHtR than BMI. Mean AUCs were 0.65 (0.63, 0.68) for BMI and 0.69 (0.67, 0.71) for WHtR (P = 0.047) among men and 0.73 (0.71, 0.76) and 0.69 (0.66, 0.73) among women (P = 0.06). Mean AUCs for WC were intermediate. Statistical significance of the indices of abdominal obesity against BMI at P < 0.05 was achieved by WHtR in men but not by WC for men or women.


Sixteen studies in men (5,15–18,20,23,25,26,28,29,31,32,36,40) and 17 in women (as men plus (33)) included dyslipidaemia or high triglyceride levels as an outcome.

Among men there was no significant difference in AUCs between indices (0.68–0.69). In women, BMI (mean AUC = 0.65) was a significantly poorer indicator than WHtR (AUC = 0.69; P = 0.047).

Metabolic syndrome

Among men in 12 study groups (5,15–17,21,26,29,30,35,37,40), the difference in AUCs between indices was not statistically significant. Among women (13 study groups including (33)), BMI was significantly poorer than WHtR (0.72 vs. 0.76; P = 0.047).

Cardiovascular disease outcomes

Only seven studies reported outcomes or risk related to CVD. In three studies, this was myocardial infarction (fatal or non-fatal) (14,20,41), in one it was cardiovascular mortality (38), and three calculated high coronary risk (usually Framingham score) (16,17,22). One study was just on men (14) and one just on women (41) giving six study populations for each sex.

All were consistent in showing WHtR > WC > BMI in AUC values, although not all differences were statistically significant. WHtR was 9% better than BMI at predicting CVD risk in men (P = 0.007). Values for WC were not significantly better than BMI in men or women. The results were similar following exclusion of the studies using the risk score (16,17,22).

All outcomes

Using data on all outcomes, averaged within study group, WHtR had significantly greater discriminatory power compared with BMI. Compared with BMI, WC improved discrimination of adverse outcomes by 3% (P < 0.05), but WHtR improved discrimination by 4–5% (P < 0.01).

Consistency of results

For all five specific health outcomes, WHtR had better discriminatory power compared with BMI. WC gave values that were nearly as good as WHtR, but more often fell short of statistical significance.

This ranking order was consistent among both sexes, although there was no statistical difference between indices in men for dyslipidaemia and MS, and for women in HT.

Discrimination, as measured by AUCs, tended to be 3–4% higher for women than for men for diabetes and HT outcomes. However, there were no significant differences in AUC between studies in Asian and non-Asian populations.

Is waist-to-height ratio significantly better than waist circuference in predicting outcomes?

Because each individual was measured on each index, a more powerful way to address the hypothesis that WHtR is superior to WC is by calculating the difference in AUC between these two indices for each study and then to test this against the null hypothesis that the difference is zero.

Table 3 shows that WHtR gave significantly better discrimination than WC for all outcomes apart from MS in men (P = 0.6). WHtR gave superior discrimination over WC for CVD outcomes (P < 0.0001 for men; P = 0.002 for women), for diabetes and HT (P < 0.0001), for dyslipidaemia (P = 0.036 for men; P = 0.001 for women) and for all outcomes combined (P < 0.0001). The absolute difference in AUC for WHtR over WC was 1–2%.

Table 3.  Difference in AUC between WHtR and WC in men and women within studies for diabetes, dyslipidaemia, hypertension, CVD and all outcomes
  Improvement in AUC (WHtR > WC) (95% CI) 
Number of studiesMeanLower 95%Upper 95%P value
  1. Results show a separate meta-analysis of the difference between these two indices assuming a correlation of 0.95 (Aekplakorn et al., 2007 (14)). Effects were pooled using random effects model. P values show Z-test of null (two-tailed).

  2. AUC, area under the curve; CI, confidence interval; CVD, cardiovascular disease; WC, waist circumference; WHtR, waist-to-height ratio.

 Diabetes220.0160.0090.022P < 0.0001
 Hypertension180.0140.0080.019P < 0.0001
 Dyslipidaemia160.0050.0000.010P = 0.036
 Metabolic syndrome120.003−0.0100.016P = 0.662
 CVD60.0390.0290.050P < 0.0001
 All outcomes320.0120.0060.017P < 0.0001
 Diabetes240.0110.0070.015P < 0.0001
 Hypertension190.0140.0070.022P < 0.0001
 Dyslipidaemia170.0080.0030.013P = 0.001
 Metabolic syndrome130.0090.0000.017P = 0.04
 CVD60.0200.0080.033P = 0.002
 All outcomes330.0100.0070.013P < 0.0001


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Funding statement
  9. Conflict of Interest Statement
  10. Acknowledgements
  11. References

This robust meta-analysis, including data on more than 300,000 individuals from diverse populations across the world, confirms previous claims from smaller and less robust analyses that measures of abdominal obesity, especially WHtR, provide a superior tool for discriminating obesity-related cardiometabolic risk compared with BMI. The studies within the meta-analysis used several different diagnostics for defining cardiometabolic risk (see Table 1) but this did not influence our conclusions. Our ‘within studies’ comparison, which showed that WHtR was significantly superior to WC, is the first, to our knowledge, to report this result. The differences were consistent across different health outcomes and statistically significant in several comparisons, even though our test for statistical difference between pooled mean AUC values was conservative. The ‘rank’ order for AUC was always WHtR > WC > BMI.

Meta-analysis is a powerful tool to assess the totality of evidence now available and can overcome the problem of basing conclusions on individual papers. For instance, inconsistent results can be found with the different outcomes (e.g. (15,27)), and the indices often perform differently in different sexes, age groups and ethnic groups (e.g. see (25) for inconsistent performance and (35) for consistent performance). Our meta-analysis has demonstrated statistically that WHtR is superior to WC as well as being superior to BMI in the discrimination of cardiometabolic risk.

Justification for comparing three anthropometric indices

Many anthropometric indices have been used to predict cardiometabolic risk, but our systematic review process focused on BMI, WC and WHtR and did not include waist-to-thigh ratio (e.g. (24)), waist-to-hip ratio (WHpR) (e.g. (25)), conicity index or sagittal diameter, which feature in several papers. The first two have performed no better than WC in previous meta-analyses (12). The latter two, even though they have been reported as the best predictor in certain populations (22)(44), were excluded from our analysis because so few studies used them and because they would be less suitable for routine public health screening purposes.

It should also be noted that some authors use different terminology and abbreviations (e.g. waist–stature ratio) for what is essentially WHtR. Elsewhere, we have urged consistency in the use of waist-to-height ratio, abbreviated to WHtR, which is the most popular terminology (45).

Justification for restricting meta-analysis to studies with receiver operating characteristics analyses

The only other meta-analyses of anthropometric indices and cardiometabolic risk, to our knowledge, have focused on studies that included ROC analysis. One concluded that ‘no anthropometric index was systematically better than others at the discrimination of hypertension’(11) and the other concluded that ‘statistical evidence supports the superiority of measures of central obesity, especially waist-to-height ratio over BMI, for detecting cardiovascular risk factors in both men and women’(12).

Our intention was to use a similar strategy, i.e. focusing on those studies that had included ROC analysis to compare the discriminatory power not only of CVD risk factors but also CVD if possible. ROC analysis has emerged as a popular method for assessing the effectiveness of diagnostic tests measured on a continuous scale, independent of the cut-off point used. It has been widely used in medical imaging and radiology (46). The ROC curve is a plot of q = (sensitivity) vs. P = (1-specificity) for all possible threshold values. A value of 1 would suggest perfect (100% discrimination) while 0.5 (the diagonal) indicates discrimination that is no better than chance. In practice, most anthropometric measures only expect to assess cardiovascular risk factors with 60–70% accuracy and are thus best regarded as first-stage, or population-based, screening measures.

Bias and confounding of the systematic review and meta-analysis

Results of all meta-analyses need to be interpreted with caution concerning their potential bias, confounding and generalizability. Publication bias may exist because publishers and authors often favour publishing positive findings over negative ones. However, a specific check on bias (funnel plot) revealed no association between effect size and sample size; such an association is often used as an indicator of publication bias. Confounding is most likely to arise from variability in the definitions of different diagnostic criteria in the different studies, especially for a composite diagnosis such as MS. These measurement error effects would, however, tend to reduce the likelihood of a significant result.

Large studies omitted from meta-analysis

This decision to focus on studies including ROC analysis meant that many papers, which compared relative risk (RR) of different anthropometric indices with health outcomes, but did not include a full ROC analysis, have not been included in our meta-analysis. Notable among the papers that lacked ROC analysis have been some major prospective studies (47–51), some details of which are discussed below.

In the European Prospective Investigation into Cancer and Nutrition study, reporting on nearly 15,000 deaths out of more than 350,000 subjects in nine countries, RR of death from all causes in the highest as compared with the lowest quintile of the WHtR in the multivariable-adjusted model (including BMI) was 2.22 (95% CI, 1.94–2.55) among men and 2.03 (95% CI, 1.76–2.34) among women, whereas for WC it was 2.05 (95% CI, 1.80–2.33) for men and 1.78 (95% CI, 1.56–2.04) for women (47,48).

A prospective (11 years) study of body size and risk for stroke among more than 45,000 women below age 60 showed that, in contrast to BMI, several different measures of abdominal obesity (WHtR > WC > WHpR) are strong predictors of stroke in women (49).

The multivariable-adjusted hazard ratios (95% CIs) for total stroke in Zhang's study of risk of stroke in more than 67,000 Chinese women comparing the highest vs. lowest quintiles of these measurements were 1.71 (1.49–1.97), 1.77 (1.53–2.05) and 1.91 (1.61–2.27) for BMI, WC and WHtR (51).

Gelber's analysis of data from the 14 years follow-up of more than 16,000 men in the Physicians Health Study and the 5.5 years follow-up of more than 32,000 women in the Women's Health Study led the authors to conclude that ‘The WHtR demonstrated statistically the best model fit and strongest associations with CVD (50)’.

Overall, these large prospective studies showed the same ranking order for predictive ability of the anthropometric indices, i.e. WHtR > WC > BMI. Their authors all support the use of WHtR in their recommendations. We would encourage other researchers, if possible, to conduct ROC analysis in order to help comparability with other studies.

Recent studies published after the systematic review

Our systematic review included papers published until the end of August 2010. This field is now attracting enormous interest and most studies now measure BMI and WC, with many of them also including calculation of WHtR as this requires no further measurement. Discriminatory power is often measured using ROC analysis and by comparing AUCs and so many recent papers would meet our inclusion criteria. We have read these papers and from the authors' data and conclusions, we are confident that they would support the overall conclusions of our meta-analysis.

Comparison with previous meta-analyses

The Obesity in Asia Collaboration meta-analysis of raw data collated specifically in a large database (11) focused on HT and included data from more than 173,000 individuals in 19 cross-sectional studies in the Asia–Pacific region. It concluded that ‘No anthropometric variable was systematically better than others at the discrimination of hypertension’. However, based on their pooled AUC values, WHtR had the highest discriminatory capability of the four anthropometric indices in the studies (WHtR > WC >  WHpR > BMI), but this was not statistically significant.

Lee's meta-analysis of 10 published studies (12) included data on 88,000 individuals. They consistently found that the AUC values were ranked in this order: WHtR (highest), WC = WHpR and BMI (lowest) with statistical significance being shown between WHtR and BMI for diabetes and for HT in men.

With our increased number of published studies (33) in the meta-analysis and increased subject numbers (n = 305,851), we observed the same consistent ranking order. We found WHtR and WC to be statistically superior to BMI for identifying diabetes in both sexes. Statistical significance was achieved for WHtR, but not WC, in HT in men and women, in CVD in men, and in dyslipidaemia and MS in women. Furthermore our matched pair analysis (Table 3) showed, for the first time, that WHtR was a better discriminator than WC for diabetes, dyslipidaemia, HT and CVD in both sexes.

Waist-to-height ratio and waist circumference compared across a wide range of heights

Many have argued that WC alone is unsatisfactory because people with the same WC but different heights are unlikely to have the same cardiometabolic risk (52). However, it has not often been possible to prove that WHtR is better than WC in populations where there is not a wide range in height. The innovative component of our meta-analysis is not only that it includes a large number of subjects but also studies across different populations, including 15 of Asian ethnicity whose height tends to be shorter than other races (53). This is possibly why we are the first to show conclusively that WHtR is a better screening tool than WC. The effect size was similar in Asian and non-Asian groups so we do not consider the ethnic mix a limitation on the generalizability of our findings. The difference in AUC we have shown between WHtR and WC in Table 3 is modest but we believe it to be clinically significant. Other authors (54) have argued for clinical superiority of different CVD risk scores on the basis of AUC differences of similar size.

Biological plausibility for why waist-to-height ratio could be superior to waist circumference

Many authors have considered mechanisms to explain why measures of abdominal obesity are superior to BMI in predicting cardiometabolic risk. These invariably relate to the high metabolic and inflammatory activity of the visceral fat depots within the abdominal cavity (55) in comparison to subcutaneous depots in other parts of the body such as the gluteo-femoral region (56). This can give a plausible explanation for the superiority of measures of abdominal obesity, which reflect visceral obesity, over BMI in predicting metabolic risk, but why should WC divided by height be superior to WC alone? In general, height has usually been shown to have inverse associations with cardiometabolic morbidity and mortality (57) and this is probably because height, as well as having a major genetic component, can also reflect general early life exposures (58). A recent report from Chile (59) proposes that adverse environmental exposures in critical growth periods in early life ‘programme’ short stature and predisposition to abdominal adiposity, insulin resistance and other cardiometabolic risk factors in adult life, thus providing a biologically plausible way to explain the superiority of WHtR over WC and BMI. Further, the independent effect of height on cardiometabolic risk might not be the total explanation for the superiority of WHtR. Schneider et al. (60) found that short subjects in the DETECT study have higher levels of risk factors than tall subjects if grouped by WC, but not if grouped by WHtR, and they speculate that these differences cannot be attributed to height alone.

Practical considerations of screening tools: waist-to-height ratio is superior to waist circumference

In 2006, Franzosi (61) posed the question ‘Should we continue to use BMI as a cardiovascular risk factor?’ On the basis of the evidence in our meta-analysis and that of Lee et al. (12), we believe that it is definitely time to reconsider other simple screening tools for cardiometabolic risk.

The most effective screening measures must be practical as well as effective. BMI requires measures of weight and height, while WHtR requires measures of WC as well as height. Self-assessment of height is known to be more accurate than that of weight (62) and the measure of WC requires a simple tape measure rather than weighing apparatus. Although WC can be measured at different sites, it has been demonstrated that this does not alter its risk prediction (63). Importantly, WHtR offers advantages of a simple boundary value which could be used for men and women, and maybe children, of all ethnic groups (4). Analysis of suggested cut-off values from 34 analyses in 16 different papers showed that the mean of proposed boundary values for WHtR, weighted for study size, in men and women, respectively, was 0.52 and 0.53 for diabetes, 0.53 and 0.50 for CVD, 0.50 and 0.50 for HT outcomes, 0.49 and 0.49 for lipid outcomes, and 0.50 and 0.49 for MS outcomes. The mean proposed boundary value (the first cut-off level indicating risk) for WHtR was 0.5. Within these study populations, there were subjects with Caucasian, Asian, Afro-Caribbean and Central American ethnic backgrounds (9). We did not calculate a mean value for WHtR in the studies used in this meta-analysis from similar ethic backgrounds, but those authors who suggested a value invariably supported the value of WHtR 0.5. This value not only converts into the simple message of ‘Keep your waist circumference to less than half your height’, but also provides the first boundary value for increased risk on a public health tool – a chart of WC against height (64–66).


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Funding statement
  9. Conflict of Interest Statement
  10. Acknowledgements
  11. References

This systematic review and meta-analysis is the first to show that WHtR was a better predictor than WC for diabetes, dyslipidaemia, HT and CVD risk in both sexes in populations of various nationalities and ethnic groups.

By including data on more than 300,000 individuals from diverse populations across the world, it supports previous suggestions that measures of abdominal obesity provide superior tools for discriminating obesity-related cardiometabolic risk compared with BMI. Moreover, WHtR has better discriminatory power than WC.

Conflict of Interest Statement

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Funding statement
  9. Conflict of Interest Statement
  10. Acknowledgements
  11. References

The authors declare no conflict of interest. MA devised and copyrighted the Ashwell® Shape Chart which is distributed to health professionals on a non-profit making basis.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Funding statement
  9. Conflict of Interest Statement
  10. Acknowledgements
  11. References

We thank Dr Lucy Browning for help with the early stages of the systematic review and Mr Michael Day for statistical advice.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Funding statement
  9. Conflict of Interest Statement
  10. Acknowledgements
  11. References
  • 1
    Lee JS, Aoki K, Kawakubo K, Gunji A. A study on indices of body fat distribution for screening for obesity. J Occup Health 1995; 37: 918.
  • 2
    Hsieh SD, Yoshinaga H. Abdominal fat distribution and coronary heart disease risk factors in men-waist/height ratio as a simple and useful predictor. Int J Obes Relat Metab Disord 1995; 19: 585589.
  • 3
    Ashwell M, Lejeune S, McPherson K. Ratio of waist circumference to height may be better indicator of need for weight management. Br Med J 1996; 312: 377.
  • 4
    Ashwell M, Hsieh SD. Six reasons why the waist-to-height ratio is a rapid and effective global indicator for health risks of obesity and how its use could simplify the international public health message on obesity. Int J Food Sci Nutr 2005; 56: 303307.
  • 5
    Schneider HJ, Glaesmer H, Klotsche J et al. Accuracy of anthropometric indicators of obesity to predict cardiovascular risk. J Clin Endocrinol Metab 2007; 92: 589594.
  • 6
    Freedman DS, Kahn HS, Mei Z et al. Relation of body mass index and waist-to-height ratio to cardiovascular disease risk factors in children and adolescents: the Bogalusa Heart Study. Am J Clin Nutr 2007; 86: 3340.
  • 7
    Nambiar S, Truby H, Abbott RA, Davies PS. Validating the waist-height ratio and developing centiles for use amongst children and adolescents. Acta Paediatr 2009; 98: 148152.
  • 8
    Ashwell M, Gibson S. Waist to height ratio is a simple and effective obesity screening tool for cardiovascular risk factors: analysis of data from the British National Diet and Nutrition Survey of adults aged 19 to 64 years. Obes Facts 2009; 2: 97103.
  • 9
    Browning LM, Hsieh SD, Ashwell M. A systematic review of waist-to-height ratio as a screening tool for the prediction of cardiovascular disease and diabetes: 0.5 could be a suitable global boundary value. Nutr Res Rev 2010; 23: 247269.
  • 10
    Goulding A, Taylor RW, Grant AM, Parnell WR, Wilson NC, Williams SM. Waist-to-height ratios in relation to BMI z-scores in three ethnic groups from a representative sample of New Zealand children aged 5–14 years. Int J Obes (Lond) 2010; 34: 11881190.
  • 11
    Obesity in Asia Collaboration. Is central obesity a better discriminator of the risk of hypertension than body mass index in ethnically diverse populations? [see comment]. J Hypertens 2008; 26: 169177.
  • 12
    Lee CMY, Huxley RR, Wildman RP, Woodward M. Indices of abdominal obesity are better discriminators of cardiovascular risk factors than BMI: a meta-analysis. J Clin Epidemiol 2008; 61: 646653.
  • 13
    Borenstein M, Hendges L, Higgins J, Rothenstein H. Introduction to Meta-Analysis. John Wiley & Sons Ltd: London, UK, 2009.
  • 14
    Aekplakorn W, Pakpeankitwatana V, Lee CMY et al. Abdominal obesity and coronary heart disease in Thai men. Obesity (Silver Spring) 2007; 15: 10361042.
  • 15
    Aekplakorn W, Kosulwat V, Suriyawongpaisal P. Obesity indices and cardiovascular risk factors in Thai adults. Int J Obes 2006; 30: 17821790.
  • 16
    Can AS, Bersot TP, Gonen M et al. Anthropometric indices and their relationship with cardiometabolic risk factors in a sample of Turkish adults. Public Health Nutr 2009; 12: 538546.
  • 17
    Cristo Rodriguez Perez MD, Cabrera De Leon A, Aguirre-Jaime A et al. The waist to height ratio as an index of cardiovascular risk and diabetes. Med Clin (Barc) 2010; 134: 386391.
  • 18
    Gracey M, Burke V, Martin DD, Johnston RJ, Jones T, Davis EA. Assessment of risks of ‘lifestyle’ diseases including cardiovascular disease and type 2 diabetes by anthropometry in remote Australian Aborigines. Asia Pac J Clin Nutr 2007; 16: 688697.
  • 19
    He Y, Zhai F, Ma G et al. Abdominal obesity and the prevalence of diabetes and intermediate hyperglycaemia in Chinese adults. Public Health Nutr 2009; 12: 10781084.
  • 20
    Ho SY, Lam TH, Janus ED. Waist to stature ratio is more strongly associated with cardiovascular risk factors than other simple anthropometric indices. Ann Epidemiol 2003; 13: 683691.
  • 21
    Hsieh SD, Muto T. Metabolic syndrome in Japanese men and women with special reference to the anthropometric criteria for the assessment of obesity: proposal to use the waist-to-height ratio. Prev Med 2006; 42: 135139.
  • 22
    Haun DR, Pitanga FJ, Lessa I. Waist-height ratio compared to other indicators of obesity as predictors of high coronary risk. Rev Assoc Med Bras 2009; 55: 705711.
  • 23
    Lee K, Song Y-M, Sung J. Which obesity indicators are better predictors of metabolic risk?: healthy twin study. Obesity (Silver Spring) 2008; 16: 834840.
  • 24
    Li C, Ford ES, Zhao G, Kahn HS, Mokdad AH. Waist-to-thigh ratio and diabetes among US adults: the Third National Health and Nutrition Examination Survey. Diabetes Res Clin Pract 2010; 89: 7987.
  • 25
    Li M, McDermott RA. Using anthropometric indices to predict cardio-metabolic risk factors in Australian indigenous populations. Diabetes Res Clin Pract 2009; 87: 401406.
  • 26
    Lin WY, Lee LT, Chen CY et al. Optimal cut-off values for obesity: using simple anthropometric indices to predict cardiovascular risk factors in Taiwan. Int J Obes Relat Metab Disord 2002; 26: 12321238.
  • 27
    Mansour AA, Al-Jazairi MI. Cut-off values for anthropometric variables that confer increased risk of type 2 diabetes mellitus and hypertension in Iraq. Arch Med Res 2007; 38: 253258.
  • 28
    Mellati AA, Mousavinasab SN, Sokhanvar S, Kazemi SA, Esmailli MH, Dinmohamadi H. Correlation of anthropometric indices with common cardiovascular risk factors in an urban adult population of Iran: data from Zanjan Healthy Heart Study. Asia Pac J Clin Nutr 2009; 18: 217225.
  • 29
    Mirmiran P, Esmaillzadeh A, Azizi F. Detection of cardiovascular risk factors by anthropometric measures in Tehranian adults: receiver operating characteristic (ROC) curve analysis. Eur J Clin Nutr 2004; 58: 11101118.
  • 30
    Mombelli G, Zanaboni AM, Gaito S, Sirtori CR. Waist-to-height ratio is a highly sensitive index for the metabolic syndrome in a Mediterranean population. Metab Syndr Relat Disord 2009; 7: 477484.
  • 31
    Paniagua L, Lohsoonthorn V, Lertmaharit S, Jiamjarasrangsi W, Williams MA. Comparison of waist circumference, body mass index, percent body fat and other measure of adiposity in identifying cardiovascular disease risks among Thai adults. Obes Res Clin Pract 2008; 2: 215223.
  • 32
    Park SH, Choi SJ, Lee KS, Park HY. Waist circumference and waist-to-height ratio as predictors of cardiovascular disease risk in Korean adults. Circ J 2009; 73: 16431650.
  • 33
    Pua YH, Ong PH. Anthropometric indices as screening tools for cardiovascular risk factors in Singaporean women. Asia Pac J Clin Nutr 2005; 14: 7479.
  • 34
    Sakurai M, Miura K, Takamura T et al. Gender differences in the association between anthropometric indices of obesity and blood pressure in Japanese. Hypertens Res 2006; 29: 7580.
  • 35
    Taylor RW, Brooking L, Williams SM et al. Body mass index and waist circumference cutoffs to define obesity in indigenous New Zealanders. Am J Clin Nutr 2010; 92: 390397.
  • 36
    Tseng CH, Chong CK, Chan TT et al. Optimal anthropometric factor cutoffs for hyperglycemia, hypertension and dyslipidemia for the Taiwanese population. Atherosclerosis 2010; 210: 585589.
  • 37
    Wang JW, Hu DY, Sun YH et al. Obesity criteria for identifying metabolic risks. Asia Pac J Clin Nutr 2009; 18: 105113.
  • 38
    Welborn TA, Dhaliwal SS. Preferred clinical measures of central obesity for predicting mortality. Eur J Clin Nutr 2007; 61: 13731379.
  • 39
    Zhou Z, Hu D, Chen J. Association between obesity indices and blood pressure or hypertension: which index is the best? Public Health Nutr 2009; 12: 10611071.
  • 40
    Koch E, Romero T, Manríquez L et al. Waist to height ratio: a better predictor of cardiovascular risk factors and mortality in Chilean adults. Diagnostic nomograms from the San Francisco project. Rev Chil Cardiol 2008; 27: 2325.
  • 41
    Page JH, Rexrode KM, Hu F, Albert CM, Chae CU, Manson JE. Waist-height ratio as a predictor of coronary heart disease among women. Epidemiology 2009; 20: 361366.
  • 42
    Sargeant LA, Bennett FI, Forrester TE, Cooper RS, Wilks RJ. Predicting incident diabetes in Jamaica: the role of anthropometry. Obes Res 2002; 10: 792798.
  • 43
    Shafiee G, Hadaegh F, Azizi F. Comparison of waist-to-height ratio and body mass index for prediction of type 2 diabetes mellitus risk in women: Tehran lipid and glucose study. Iran J Endocrinol Metab 2009; 11: 1724.
  • 44
    Duarte Pimentel G, Portero-McLellan KC, Maesta N, Corrente JE, Burini RC. Accuracy of sagittal abdominal diameter as predictor of abdominal fat among Brazilian adults: a comparation with waist circumference. Nutr Hosp 2010; 25: 656661.
  • 45
    Ashwell M, Browning L. The increasing importance of waist-to-height ratio to assess cardiometabolic risk: a plea for consistent terminology. Open Obes J 2011; 3: 7077.
  • 46
    Metz C. Some practical issues of experimental design and data analysis in radiological ROC studies. Invest Radiol 1989; 24: 234245.
  • 47
    Pischon T, Boeing H, Hoffmann K et al. General and abdominal adiposity and risk of death in Europe. N Engl J Med 2008; 359: 21052120.
  • 48
    Gaglione MM. Obesity and risk of death. N Engl J Med 2009; 360: 1042; author reply 1043–4.
  • 49
    Lu M, Ye W, Adami HO, Weiderpass E. Prospective study of body size and risk for stroke amongst women below age 60. J Intern Med 2006; 260: 442450.
  • 50
    Gelber RP, Gaziano JM, Orav EJ, Manson JE, Buring JE, Kurth T. Measures of obesity and cardiovascular risk among men and women. J Am Coll Cardiol 2008; 52: 605615.
  • 51
    Zhang X, Shu XO, Gao YT, Yang G, Li H, Zheng W. General and abdominal adiposity and risk of stroke in Chinese women. Stroke 2009; 40: 10981104.
  • 52
    Hsieh S, Yoshinaga H. Do people with similar waist circumference share similar health risks irrespective of height? Tohoku J Exp Med 1999; 188: 5560.
  • 53
    Parikh RM, Joshi SR, Menon PS, Shah NS. Index of central obesity – a novel parameter. Med Hypotheses 2007; 68: 12721275.
  • 54
    Collins GS, Altman DG. An independent and external validation of QRISK2 cardiovascular disease risk score: a prospective open cohort study. BMJ 2010; 340: c2442.
  • 55
    Kuk JL, Katzmarzyk PT, Nichaman MZ, Church TS, Blair SN, Ross R. Visceral fat is an independent predictor of all-cause mortality in men. Obesity (Silver Spring) 2006; 14: 336341.
  • 56
    Manolopoulos KN, Karpe F, Frayn KN. Gluteofemoral body fat as a determinant of metabolic health. Int J Obes (Lond) 2010; 34: 949959.
  • 57
    Langenberg C, Shipley MJ, Batty GD, Marmot MG. Adult socioeconomic position and the association between height and coronary heart disease mortality: findings from 33 years of follow-up in the Whitehall Study. Am J Public Health 2005; 95: 628632.
  • 58
    Barker DJ, Osmond C, Golding J. Height and mortality in the counties of England and Wales. Ann Hum Biol 1990; 17: 16.
  • 59
    Koch E, Romero T, Romero CX et al. Early life and adult socioeconomic influences on mortality risk: preliminary report of a ‘pauper rich’ paradox in a Chilean adult cohort. Ann Epidemiol 2010; 20: 487492.
  • 60
    Schneider HJ, Klotsche J, Silber S, Stalla GK, Wittchen HU. Measuring abdominal obesity: effects of height on distribution of cardiometabolic risk factors risk using waist circumference and waist-to-height ratio. Diabetes Care 2011; 34: e7.
  • 61
    Franzosi MG. Should we continue to use BMI as a cardiovascular risk factor? Lancet 2006; 368: 624625.
  • 62
    Bolton-Smith C, Woodward M, Tunstall-Pedoe H, Morrison C. Accuracy of the estimated prevalence of obesity from self reported height and weight in an adult Scottish population. J Epidemiol Community Health 2000; 54: 143148.
  • 63
    Ross R, Berentzen T, Bradshaw AJ et al. Does the relationship between waist circumference, morbidity and mortality depend on measurement protocol for waist circumference? Obes Rev 2008; 9: 312325.
  • 64
    Ashwell M. The Ashwell Shape Chart – a public health approach to the metabolic risks of obesity. Int J Obes 1998; 22: 213.
  • 65
    Ashwell M. Charts based on body mass index and waist-to-height ratio to assess the health risks of obesity: a review. Open Obes J 2011; 3: 7884.
  • 66
    Ashwell M. Shape: the waist-to-height ratio is a good, simple screening tool for cardiometabolic risk. Nutr Today 2011; 46: 8589.