A novel adiposity index as an integrated predictor of cardiometabolic disease morbidity and mortality
A novel adiposity index as an integrated predictor of cardiometabolic disease morbidity and mortality
Nam Hoon Kim
Sin Gon Kim
OPEN Published: xx xx xxxx We propose a new anthropometric index, weight-adjusted-waist index (WWI), to assess adiposity by standardizing waist circumference (WC) for weight. WWI, calculated as WC (cm) divided by the square root of weight (kg) (cm/?kg), was measured from 465,629 subjects in the Korean nationwide cohort (2008-2013). Cox regression analysis was used to compare WWI with BMI, WC, waist-to-height ratio (WHtR), and a body shape index (ABSI) for cardiometabolic morbidity and mortality risk in diagnostic and prognostic prediction models. For incident hypertension, type 2 diabetes and cardiovascular disease (CVD), BMI had the strongest predictive power, followed by WWI and WC. However, WWI showed the best predictive performance for CVD mortality. Also, a linear positive association between adiposity indices and cardiovascular and all-cause mortality was only shown in WWI and ABSI, not BMI, WC and WHtR which showed inverse J-shaped patterns. In the test of joint effects of each index, WWI combined with BMI was the strongest in both diagnostic and prognostic models. WWI is a unique adiposity index that shows linear positive association with both cardiometabolic morbidity and mortality. It also predicts incident cardiometabolic disease, cardiovascular and all-cause mortality risk with excellence in predictive power, especially when combined with BMI.
based on WC divided by its regression fit on weight and height so that ABSI is minimally correlated with weight,
height, and BMI. ABSI was found to be more closely associated with mortality risk than BMI and WC22. However,
WHtR also had a strong correlation with BMI, thus, it was not free from the influence of BMI, and had a weaker
association with mortality than BMI. In contrast, ABSI was uncorrelated with BMI, but several studies indicated
that it is a less reliable indicator of cardiometabolic risk factors than BMI or WC23,24.
Therefore, the limitations of traditional and newer anthropometric indicators suggest the need for an
integrated adiposity index to have an ability to predict both cardiometabolic disease morbidity and mortality. We
herein propose a new adiposity index termed the ?weight-adjusted-waist index? (WWI) that is a standardized
WC for weight, and aimed to validate the new index for the association with obesity-related disorders,
cardiovascular mortality and all-cause mortality.
Derivation of WWI. We wanted an adiposity index representing waist circumference, having weak
correlation with BMI to alleviate the obesity paradox of BMI for death, and having a negative correlation with height
to differentiate the effect of height on the same waist. This could be accomplished by adjusting waist only for
weight. When the BMI was first proposed, it was calculated by standardizing weight for height. Namely, the BMI
was obtained by regressing the logarithm of weight on the logarithm of height to remove the effect of height on
weight25 (the correlation between weight and height was 0.666, but that between BMI and height was as low as
0.072 in our data set). With the same concept, we proposed an index which standardizes waist circumference for
body weight by the least squared regression of the logarithm-transformed WC on the logarithm-transformed
weight as given by
ln(WC) = ?0 + ?1 ln(weight) + ?.
The estimated?1 was 0.494 (p-value < 0.0001) that is close to 0.5. Thus,In(WC) ? 0.5In(weight) = 1n wc
ln wc became an estimate of ?0 + ? so that wc is almost uncorrelated with weight (the correlationwewigahts
?0.025), by the characteristics of the least squared regression model24. We define the weight-adjusted-waist index
(WWI) as wc whose mean values were 10.0 (?0.63) for males and 10.1 (?0.86) for females.
Study population. The Korean National Health Insurance Cohort (NHIS) study (2002?2013) is a
population-based longitudinal study consisting of about one million Koreans, a representative 2.2% sample of the
national population data. The NHIS data are composed of demographic information, anthropometric measures
including body weight (kg) and height (m), medical and pharmacy records, health examination data, and death
records. Annually, 10?15% of the cohort population received health examinations. Since WC (cm) had been
measured from 2008, our data-set included only the subjects who had ever undergone health examinations from
2008 to 2013. More detailed information has been given in previous publications6,26.
The comparison of a new adiposity index with existing indices was done by model fit and predictive
accuracy of incident cardiometabolic diseases including hypertension and type 2 diabetes and of cardiovascular and
all-cause mortality. For the mortality prediction, we excluded subjects with pre-existing cancer or CVD before
2008 to avoid a possible confounding effect of those conditions on mortality. For each disease prediction, we also
excluded those with corresponding preexisting diseases including hypertension, type 2 diabetes and CVD before
2008 when the specific disease was the risk of interest. For diagnostic data setting, among total 468,981 subjects,
5,105 subjects were additionally deleted for all-cause death and 354 subjects for CVD death due to missing
information of explanatory variables. Accordingly, the total number of participants ranged from 425,917 to 465,627
depending on the study risks: 460,876 subjects with 5,469 deaths for all-cause death, 465,627 with 718 deaths for
CVD death, 425,917 with 40,334 incidents in 2008 for hypertension, 442,532 with 23,808 incidents in 2008 for
type 2 diabetes, 463,797 with 21,984 incidents in 2008 for CVD. The mean follow-up duration for all-cause death
was 5.61 (?0.45) years. For prognostic data setting, there were 167,203 subjects who had physical examinations at
year 2008. Among them, 2,954 all-cause deaths and 388 CVD deaths occurred until 2013. Prognostic prediction
models were applied to this data set for evaluating the effect of adiposity indices on death. The mean follow-up
for the prognostic model was 5.96 (0.383) years. All data from the NHIS cohort do not involve any personally
identifiable data such as name and personal ID. Thus, NHIS approved the cohort study without informed consent
from each person. This study was approved by the institutional review board of Korea University Anam Hospital
(IRB number: ED14188).
Identification of cause of death, disease status, and confounding variables. Causes of death were
classified by the International Classification of Disease, Tenth Revision (ICD-10). The data and causes of death
of each individual were recorded in their medical records by physicians, and all death records were included in
the NHIS data. CVD was identified by the disease codes (I20~I25 and I60-I69). Type 2 diabetes and
hypertension were identified by the disease codes (E11~E14, N083, I792, G590, G632, G990, H360, and M142 for type
2 diabetes and I10~I13 and I15 for hypertension) or laboratory data from health examination (fasting plasma
glucose level (?126 mg/dl) for type 2 diabetes, systolic blood pressure (?140 mmHg) or diastolic blood
pressure (?90 mmHg) for hypertension). Other laboratory variables included in the Cox regression models were
hemoglobin, alanine aminotransferase, and gamma-glutamyltransferase. History of smoking (current, former,
or never), alcohol consumption (?3 times/week,?2 times/week, or never), physical activity (?3 times/week,?2
times/week, or never), and socioeconomic status (SES, high 30%, middle 40%, or low 30%) and individual age
and sex were also included in the analysis.
BMI (kg2 )
Categorization of each adiposity index. The primary objective of this study was to compare the
association between adiposity indices and obesity-related diseases or mortality risk, as well we wanted to show the
association patterns or shapes. The established 4 or 5 categories of BMI is useful for clinical application, however,
those categorization was limited to display the association patterns. Thus, we used more detailed categorization
of BMI and corresponding categorization of each indices. BMI was categorized into 10 groups: <18.5, 18.5?20,
20?21.5, 21.5?23, 23?25, 25?26.5, 26.5?28, 28?30, 30?32.5, ?32.5 kg/m2, where the fifth group (
) was the
reference group. The 10 groups consisted of 3.99%, 8.21%, 13.50%, 17.71%, 24.53%, 13.87%, 8.65%, 5.81%, 2.67%,
and 1.06%, respectively. WC, WHR, ABSI, and WWI were also categorized into 10 groups so that the distribution
of the 10 groups was approximately the same as that of BMI for fair comparisons of prediction ability (Table?1).
Statistical Analysis. Pearson?s partial correlation analysis was carried out among the anthropometric
indicators adjusting for age and sex. Each subject had multiple health examination records as he/she could have
undergone several health examinations during the study period of 2008?2013. To take into account the time-dependent
nature of health records and such multiple records of each subject in the analysis, we used the counting process
formulations in the Cox regression model. The counting process formulation is a data rearranging method based
on the time interval of health examination so that multiple health records of an individual can be represented by
multiple observations allocated to non-overlapping time intervals for different health examinations. The average
number of health records per participant was 2.14 (1.45) times. The Cox regression applied to these multiple
health records of an individual belongs to a diagnostic prediction model. We were also interested in how adiposity
indices measured at year 2008 as the base line contribute to the prediction ability of death for certain future time
periods. To do this, we applied prognostic prediction models to these data consisted of single heath record per
individual at 2008.
In the diagnostic prediction model, we computed likelihood ratio (LR) test statistics to measure partial
contribution of adiposity indices to model fit, and C statistics to examine overall measures of predictive accuracy
from the models with different adiposity indices as predictors associated with the risks of death and diseases. On
the other hand, in the prognostic prediction model, we computed LR, AUC (area under ROC) at three different
future time points denoted by year 1, year 3, and year 5 for the predictive accuracy at 1, 3, and 5 years later after
2008, and IAUC (integrated AUC) and C statistics for overall predictive accuracy of prognostic prediction
models. The larger the LR, the more contribution the adiposity index for model fit, and the closer the AUC, IAUC, or
C is to 1, the better predictive accuracy the model used a specific adiposity index. All methods were performed
in accordance with the ethical standards of the institutional and national research committee and with the 1964
Helsinki declaration. All statistical analyses were performed using SAS version 9.4.
Table?2 displays the mean of each adiposity index for the living and dead persons during 2008?2013. The total
number of cohort participants was 460,878 for the all-cause mortality analysis (227,598 men with 3,397 deaths
and 233,278 women with 2,072 deaths) and 465,627 in CVD mortality analysis (230,585 men with 410 CVD
deaths and 235,042 women with 308 deaths).
Mean BMI was unexpectedly higher in survivors than decedents regardless of sex in both analyses. This partly
explains the obesity paradox of BMI as illustrated in Fig.?1. On the other hand, the mean values of WHtR, ABSI,
and WWI were lower in survivors than decedents. We noted that mean BMI, WC and ABSI were higher in males
than in females, whereas mean WHtR and WWI were slightly lower in males than in females (Table?2).
Pearson?s partial correlation analysis was carried out for the each adiposity index (Table?3). BMI was strongly
correlated with WC and WHtR, weakly correlated with WWI, and negatively correlated with ABSI. WWI was
strongly correlated with ABSI (r = 0.898) and had stronger correlations with WC and WHtR than ABSI,
indicating that WWI is more representative for waist-related-indicators than ABSI. Height is not correlated with BMI
and ABSI, positively correlated with WC, and negatively correlated with WWI more than WHtR, meaning that
WWI differentiated the effect of height on the same WC.
Adiposity index and mortality risks. Diagnostic prediction model. Table?4 provides likelihood ratio test
statistics between diagnostic prediction models with and without specific adiposity indices as predictors and C
statistics for the diagnostic prediction models with different adiposity indices for all-cause and CVD mortality.
The regression models include all confounding variables other than the adiposity index: age, sex, systolic blood
pressure, fasting glucose, hemoglobin, alanine aminotransferase, gamma-glutamyltransferase, smoking, alcohol
consumption, physical activity level, and socioeconomic status. BMI showed the greatest contribution in
predicting all-cause mortality, and WWI showed the greatest contribution in predicting CVD mortality. According to
the C statistics, the diagnostic model including WWI distinguished the survival times of survivors and decedents
by 85% for all-cause mortality and 89% for CVD mortality. The C-statistics showed that all adiposity indices were
comparable; however, the order is the same as the goodness of fit statistics.
Because BMI provided the strongest discriminatory power for all-cause mortality, and differs from other
measures in that it does not include waist measurements, we tested which measure was the most complementary
to BMI in predicting all-cause and CVD mortality. Table?4 shows that WWI is the best complementary indicator
of BMI for all-cause and CVD mortality. This implies that WWI possesses the most independent information
from BMI to describe the risk of all-cause and CVD mortality.
Figure?1 illustrates the hazard ratios of all-cause and CVD mortality by 10-group stratifications for each
adiposity index when each index was included alone in the model (marginal model, Supplementary Tables?1 and 2
for confidence intervals of hazard ratios), and together with BMI (joint model, Supplementary Fig.?1 and Tables?4
and 5). First of all, we observed that BMI was not positively associated with all-cause and cardiovascular
mortality, indicating the obesity paradox. Second, WC and WHtR were also not positively associated with all-cause
mortality in the marginal model as BMI did, whereas, in contrast, they were positively associated in the joint
model. These results originated from the high correlation between BMI and WC or WHtR, as shown in Table?3.
Third, WWI and ABSI were positively associated with all-cause and cardiovascular mortality for both marginal
and joint models.
Prognostic prediction model. Tables?5 and 6 provided LR, time-dependent AUC with its standard error,
IAUC that is average of AUCs for entire study time period, and C statistics from prognostic models with different
adiposity indices. It appeared that the BMI plus WWI model had the best performance in model fit and predictive
accuracy for the risks of all-cause and CVD mortality. This confirmed the result in diagnostic prediction model
where WWI was the best complement for BMI. Regardless of models, AUC was the largest at year 1, the smallest
at year 3, and between at other years because IAUC was in between. This implies that the predictive accuracy of
each model hovered around its respective level of IAUC until the censoring time (i.e., 2013). The models for the
risk of all-cause mortality had lower predictive accuracy than those for the risk of CVD mortality. The hazard
ratios of all-cause and CVD mortality by 10-group stratifications for each adiposity index in prognostic models
were essentially the same patterns as in diagnostic models.
Adiposity index and risk of cardiometabolic diseases. LR test statistics of diagnostic prediction mod
els for marginal and joint effects of adiposity indices on the risks of incident hypertension, type 2 diabetes and
CVD are displayed in the Table?6. When each of the indices were included solely in the model, BMI was the
best in predicting all the three types of cardiometabolic diseases, while WWI was the second best for incident
type 2 diabetes and CVD, and WC was the second best for hypertension. Nevertheless, when each indicator was
included together with BMI in the model, WWI as a complementary indicator of BMI was the best in predicting
all three types of cardiometabolic diseases.
Figure?2 and Supplementary Fig.?2 depict the hazard ratios of incident cardiometabolic diseases for WWI
and ABSI calculated by the marginal and joint models (Supplementary Tables?3 and 6 for confidence intervals of
hazard ratios). The trend of hazard ratios of ABSI clearly depends on whether it was adjusted for BMI or not. ABSI
was not positively associated with incident hypertension in the marginal model but was positively associated with
it in the joint model, implying that ABSI should be adjusted by BMI to evaluate its effect on hypertension. On the
other hand, WWI was positively associated with hypertension and hence was independent of BMI in measuring
the risk of hypertension. For type 2 diabetes and CVD, a similar pattern was observed (Supplementary Fig.?2).
In this study, we proposed a new adiposity index, WWI, as an integrated predictor of both cardiometabolic
disease morbidity and mortality. From the validation process, we proved that WWI has a good predictive ability for
both cardiometabolic morbidity and mortality in the Korean population. In addition, WWI has a positive
association with all the outcomes, which was not shown in BMI and WC.
Given the increasing prevalence of obesity and obesity-related disorders in the modern society, it is critical
to assess obesity and to identify individuals at risk of cardiometabolic diseases in clinical practice. Among all the
anthropometric measures of obesity, BMI has been the most widely-used anthropometric indicator due to its
simple calculation and good performance in predicting cardiometabolic disease risk. In our analysis, BMI was also
proven to have better discriminatory power especially for all-cause mortality and prediction ability for
hypertension, type 2 diabetes, and CVD than other indicators. However, along with some reports from other populations,
we observed a U-shaped pattern of association between BMI and all-cause, and CVD mortality in our previous
study of the Korean population6. We also identified that the ?obesity paradox? phenomenon was intensified in
that the BMI range of the lowest mortality has been shifted from 23?25 kg/m2 (overweight) to 25?29.9 kg/m2
(moderate obesity) during the last 10 years. These results partly suggested that BMI was limited as a true measure
of obesity. In addition, the U-shaped, or J-shaped pattern of the association has frequently been observed in the
Asian population whose BMI values were generally lower than those of Caucasians27?29. A previous study
provided some clues for this difference in that BMI is largely limited to assess for adiposity (fat mass) particularly
among individuals with BMI ? 30 kg/m2 30.
On the other hand, WC was proposed as an alternative measure of obesity, especially for central obesity. With
its closer association with visceral adiposity than BMI, it has been suggested as an indicator of metabolic obesity
in a lot of studies31,32. However, we also found an inverse association between WC and all-cause, CVD mortality
in our study population, indicating the limitation of WC as a BMI-dependent index.
These observations strongly required a more accurate, clinically applicable indicator for assessing obesity.
Therefore, we produced a novel adiposity index more closely associated with obesity-related disease morbidity
and mortality. When the BMI was first proposed, it was calculated by standardizing weight for height25. With the
same concept, we proposed an index which standardizes waist circumference for body weight. As a result, this
new indicator showed expected positive associations with all-cause and CVD mortality unlike BMI, WC, and
WHR, expected positive associations with hypertension, type 2 diabetes, and CVD unlike ABSI, and was the
best in predicting the risks of all-cause and CVD deaths and of cardiometabolic disease onsets when it was used
together with BMI.
There are some limitations to this study. Although our study focused on identifying a new adiposity index as
a predictor for the risks of death and diseases, the WWI was developed and validated with the same dataset and,
therefore, there is a risk of overfitting and optimism in evaluation of the predictive performance33. Hence, we need
to include some forms of internal and external validations for the predictive performance of WWI. The variability
of WC measurement possibly limits the reliability of the observed association between this WC-based index and
outcomes. We did not obtain information about inter-rater or inter-individual variability of WC measurements
because we used given data from the established cohort. However, measurement of WC in the Korean national
health check-up programs were conducted by trained instructors in each health check-up centers, and the
measurement protocol indicates that WC should be measured at the midpoint between the lower rib margin and the
iliac crest in the standing position. In addition, we were not able to test the direct relationship between WWI and
visceral or subcutaneous adipose tissue area because of lack of data. We also are not sure that this index would
be applicable to other population. Further studies using abdominal fat quantification by computed tomography
imaging, or for other population would provide the more useful information in this regard.
In conclusion, we propose a novel adiposity index, WWI, as a useful alternative marker of obesity and
obesity-related adverse health consequences. In addition, when WWI and BMI are combined, it has the best
performance for the prediction of cardiometabolic disease and mortality. It also involves a simple calculation and
easy interpretation. So far, the known measures of obesity do not predict both morbidity and mortality with
linear trends. Thus, this new index is expected to provide easy information of individuals at risk of cardiometabolic
disease and associated mortality with just one measure.
Data Availability Statement
The data are available for replication through approval and oversight by the Korean National Health Insurance
We thank all participants in the Korean Health Insurance Cohort study, as well as the National Health Insurance
Service who developed the NHIS-NSC (2002?2013) database (NHIS-2014-2-006). The views expressed in this
article are those of the authors and do not necessarily represent the official position of the department of Korean
National Health Insurance Service. This study was partly supported by a grant of the Korean Health Technology
R&D Project (HI14C2750), Ministry of Health & Welfare, Republic of Korea.
Y.P. and S.G.K. contributed to the idea, design of the research. Y.P. and T.Y.K. analyzed the data. N.H.K. and Y.P.
wrote the manuscript. All authors contributed to the review and revision of the manuscript.
Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-018-35073-4.
Competing Interests: The authors declare no competing interests.
Publisher?s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons license, and indicate if changes were made. The images or other third party material in this
article are included in the article?s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article?s Creative Commons license and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
1. Nguyen , N. T. , Magno , C. P. , Lane , K. T. , Hinojosa , M. W. & Lane , J. S. Association of hypertension, diabetes, dyslipidemia, and metabolic syndrome with obesity: findings from the National Health and Nutrition Examination Survey, 1999 to 2004 . J Am Coll Surg 207 , 928 - 934 ( 2008 ).
2. Pi-Sunyer , F. X. The obesity epidemic: pathophysiology and consequences of obesity . Obes Res 10 ( Suppl 2 ), 97s - 104s ( 2002 ).
3. Kim , C. S. et al. Prevalence, awareness, and management of obesity in Korea: data from the Korea national health and nutrition examination survey ( 1998 - 2011 ). Diabetes Metab J 38 , 35 - 43 ( 2014 ).
4. Colditz , G. A. et al. Weight as a risk factor for clinical diabetes in women . Am J Epidemiol 132 , 501 - 513 ( 1990 ).
5. Wilson, P. W., D'Agostino , R. B. , Sullivan , L. , Parise , H. & Kannel , W. B. Overweight and obesity as determinants of cardiovascular risk: the Framingham experience . Arch Intern Med 162 , 1867 - 1872 ( 2002 ).
6. Kim , N. H. et al. Body Mass Index and Mortality in the General Population and in Subjects with Chronic Disease in Korea: A Nationwide Cohort Study ( 2002 - 2010 ). PLoS One 10 , e0139924 ( 2015 ).
7. Uretsky , S. et al. Obesity paradox in patients with hypertension and coronary artery disease . Am J Med 120 , 863 - 870 ( 2007 ).
8. Carnethon , M. R. et al. Association of weight status with mortality in adults with incident diabetes . JAMA 308 , 581 - 590 ( 2012 ).
9. Hainer , V. & Aldhoon-Hainerova , I. Obesity paradox does exist . Diabetes Care 36 ( Suppl 2 ), S276 - 281 ( 2013 ).
10. Standl , E. , Erbach , M. & Schnell , O. Defending the con side: obesity paradox does not exist . Diabetes Care 36 ( Suppl 2 ), S282 - 286 ( 2013 ).
11. Jackson , A. S. et al. The effect of sex, age and race on estimating percentage body fat from body mass index: The Heritage Family Study . Int J Obes Relat Metab Disord 26 , 789 - 796 ( 2002 ).
12. Lam , B. C. , Koh , G. C. , Chen , C. , Wong , M. T. & Fallows , S. J. Comparison of Body Mass Index (BMI), Body Adiposity Index (BAI), Waist Circumference (WC), Waist-To-Hip Ratio (WHR) and Waist-To-Height Ratio (WHtR) as predictors of cardiovascular disease risk factors in an adult population in Singapore . PLoS One 10 , e0122985 ( 2015 ).
13. Okorodudu , D. O. et al. Diagnostic performance of body mass index to identify obesity as defined by body adiposity: a systematic review and meta-analysis . Int J Obes (Lond) 34 , 791 - 799 ( 2010 ).
14. Cornier , M. A . et al. Assessing adiposity: a scientific statement from the American Heart Association . Circulation 124 , 1996 - 2019 ( 2011 ).
15. Pouliot , M. C. et al. Waist circumference and abdominal sagittal diameter: best simple anthropometric indexes of abdominal visceral adipose tissue accumulation and related cardiovascular risk in men and women . Am J Cardiol 73 , 460 - 468 ( 1994 ).
16. Nyamdorj , R. et al. BMI compared with central obesity indicators in relation to diabetes and hypertension in Asians . Obesity (Silver Spring) 16 , 1622 - 1635 ( 2008 ).
17. Pischon , T. et al. General and abdominal adiposity and risk of death in Europe . N Engl J Med 359 , 2105 - 2120 ( 2008 ).
18. Clark , A. L. , Fonarow , G. C. & Horwich , T. B. Waist circumference, body mass index, and survival in systolic heart failure: the obesity paradox revisited . J Card Fail 17 , 374 - 380 ( 2011 ).
19. Zeller , M. et al. Relation between body mass index, waist circumference, and death after acute myocardial infarction . Circulation 118 , 482 - 490 ( 2008 ).
20. Browning , L. M. , Hsieh , S. D. & 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 23 , 247 - 269 ( 2010 ).
21. Ashwell , M. , Gunn , P. & Gibson , S. Waist-to-height ratio is a better screening tool than waist circumference and BMI for adult cardiometabolic risk factors: systematic review and meta-analysis . Obes Rev 13 , 275 - 286 ( 2012 ).
22. Krakauer , N. Y. & Krakauer , J. C. A new body shape index predicts mortality hazard independently of body mass index . PLoS One 7 , e39504 ( 2012 ).
23. Wang , H. et al. Comparison of anthropometric indices for predicting the risk of metabolic syndrome and its components in Chinese adults: a prospective, longitudinal study . BMJ open 7 , e016062 ( 2017 ).
24. Cheung , Y. B. ?A Body Shape Index? in middle-age and older Indonesian population: scaling exponents and association with incident hypertension . PloS One 9 , e85421 ( 2014 ).
25. Benn , R. T. Some mathematical properties of weight-for-height indices used as measures of adiposity . Br J Prev Soc Med 25 , 42 - 50 ( 1971 ).
26. Lee , J. , Lee , J. S. , Park , S. H. , Shin , S. A. & Kim , K. Cohort Profile: The National Health Insurance Service-National Sample Cohort (NHIS-NSC ) , South Korea. Int J Epidemiol 46 , e15 ( 2017 ).
27. Jee , S. H. et al. Body-mass index and mortality in Korean men and women . N Engl J Med 355 , 779 - 787 ( 2006 ).
28. Chen , Y. et al. Association between body mass index and cardiovascular disease mortality in east Asians and south Asians: pooled analysis of prospective data from the Asia Cohort Consortium . BMJ 347 , f5446 ( 2013 ).
29. Parr , C. L. et al. Body-mass index and cancer mortality in the Asia-Pacific Cohort Studies Collaboration: pooled analyses of 424,519 participants . Lancet Oncol 11 , 741 - 752 ( 2010 ).
30. Romero-Corral , A. et al. Accuracy of body mass index in diagnosing obesity in the adult general population . Int J Obes (Lond) 32 , 959 - 966 ( 2008 ).
31. Janssen , I. , Katzmarzyk , P. T. & Ross , R. Waist circumference and not body mass index explains obesity-related health risk . Am J Clin Nutr 79 , 379 - 384 ( 2004 ).
32. Klein , S. et al. Waist Circumference and Cardiometabolic Risk: a Consensus Statement from Shaping America's Health: Association for Weight Management and Obesity Prevention; NAASO, the Obesity Society; the American Society for Nutrition; and the American Diabetes Association . Obesity (Silver Spring, Md) 15 , 1061 - 1067 ( 2007 ).
33. Moons , K. G. et al. Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): explanation and elaboration . Ann Intern Med 162 , W1 - 73 ( 2015 ).