Central obesity in both adults 123 and children 456 is related to cardiovascular (CV) risk factors (hyperlipidemia, insulin resistance and hypertension). Waist circumference (WC) and WC-derived indices such as waist-to-hip ratio and waist-to-height ratio (WHTR) have been used as proxy measures of central obesity, the former being widely used in adults 78. There is growing evidence however that waist circumference (WC) alone correlates more strongly than waist-to-height ratio (WHTR) with CV risk, in adults 2 as well as in children 91011. Previous studies have shown that adults from a number of South and East Asian countries are predisposed to central obesity and insulin resistance 1213. This ethnic difference of increased body fat distribution in the trunk region has also been observed in adolescents 14 and prepubertal children 15 of mainly East Asian origin, highlighting the importance of ethnic-specific studies.

Standard WC cutoffs for high CV risk have been proposed for adults 161718 and are widely used, but measurement of WC is not yet common practice in growing children. Age-related percentiles have been reported for children from a number of different countries: Italy 19; Spain 20; Cyprus 21; United Kingdom 22; Canada 23; USA 24; Holland 25; Australia 26; and Iran 27 (Table 1), but, to our knowledge, the only published WC data in Chinese children in the English literature is our relatively small study of 6 to 12 year old children 11, drawn from the same population as the present study.

WHTR has been proposed as a convenient alternative measure for assessing central fatness in children, on the basis that it is relatively age-independent and that in normalising for growth it might obviate the need for age-related reference charts 282930. "Keep your waist circumference to less than half your height" (i.e. WHTR 0.5) is a seductively simple target 31. However WHTR percentiles have not been reported and the generality of its independence of ethnic and age differences is questionable.

In 1993 a growth survey of 25,000 Hong Kong children from birth to 18 y was undertaken and growth charts for weight, height, weight-for-height and body mass index were developed 32. WC was not measured. We now report a further growth study of 14,842 ethnic Chinese children aged 6 to 18 y undertaken in 2005/2006, using similar methods but including WC measurement to provide reference values for WC and WHTR.

A total of 14,842 children participated in the study (7472 boys and 7370 girls). Sex and age-specific mean weight, height, BMI, WC and WHTR were obtained (Table 2) and then smoothed to develop sex- and age-specific WC and WHTR percentiles (Tables 3 and 4 and Figures 1 and 2). WC was relatively larger in boys than girls and increased with age, though to a smaller extent in girls after age 14 y. WHTR was slightly larger in boys than girls and in both sexes decreased with age but only up to age 14 y and changed little further. Over the age range 14 to 18 y WHTR 0.5 corresponded to the 95^th percentile for boys and the 97^th percentile for girls (Figure 2). WC correlated more closely than WHTR with BMI (r = 0.93, 0.91 cf. 0.65, 0.59, for boys and girls respectively).

Comparison of the 10^th, 50^th and 90^th percentile curves of WC with the 1988 British and 2003/4 Iranian data showed that both Hong Kong boys and girls were closer to British children except that our 90^th percentile curve for boys was higher. Our 50^th and 90^th percentiles were lower than the Iranian curves (Figure 3). Comparison of WHTR data of Hong Kong boys and girls was made with the British data (the only dataset which provided WHTR for each year group) showed they are very similar with the mean differences ranging from only zero to 0.02 (Table 5).

Sex and age-specific WC and WHTR percentiles, developed from data collected during 2005/6 on a large and representative population of 14,842 Chinese children aged six to 18 y old, are presented. These charts are based on WC measured midway between lowest rib and superior iliac crest. Comparisons of WC data between different studies need to be undertaken with caution since WC may be measured at different sites and currently there is no agreement on which site is optimal 34. In children, WC has been measured at five different sites: (i) midway between lowest rib and superior iliac crest (as in the present study) 20222527; (ii) at the umbilical level 192128; (iii) at the narrowest point of the torso 23; (iv) at the level of the right upper iliac crest 24; and (v) at the level of 2 cm above the umbilicus 35. WC measures at these different sites in children have not been formally compared. WC measurements in adults taken at four different sites have been compared and shown to provide slightly different, albeit highly reproducible data, that each correlate significantly with total and central body fat 34. If WC is to become an important public health assessment tool of central obesity in both adults and children, international agreement about measurement site is required.

In the study, we compared Hong Kong children's WC with data from Britain and Iran where the method of measuring WC was the same as our study. The Iranian children's 50^th and 90^th percentile curves were much higher than those from Hong Kong and Britain. A previous report has shown that Iranian children have a high prevalence of high blood triglyceride (20%) and low HDL-C (28%) 6. In contrast, our unpublished data has shown much lower prevalence of these risk factors in Hong Kong children. The Iranian paediatric population may provide an important opportunity to further study the relation between central obesity and dyslipidaemia.

Our study challenges the proposal that WHTR is a relatively age-independent measure that could obviate the need for age and growth-related reference standards in children. Our data show that WHTR is not independent of age in children under 14 y old. Kahn and colleagues reported that WHTR appears to be independent of age and sex 36. Ashwell and colleagues have proposed the use of an age-independent universal cut-off for WHTR of 0.5 29. This is despite their finding that the overall mean WHTR (with WC measured at the same site as in the present study) of 8135 British children aged 5 to 16 y decreased with age from 0.47 to 0.42 in boys and from 0.46 to 0.41 in girls 3137. WHTR measures obtained in our study were slightly less than the 1988 British study. Our WHTR decreased with age up to 14 y and then remained almost constant (Table 1, Figure 2). WHTR has now been correlated with CV risk markers in children in at least six studies 62830353638 and different WHTR cut-offs for predicting high CV risks in children at different ages were reported in one of them 6. Ashwell's proposed 0.5 cut-off point for predicting CVS risk has yet to be tested by direct correlation with CV risk markers in children. Notably, few of our older children would be categorized as being at CV risk if this single 0.5 WHTR cut-off were used (Figure 2). This would conflict with our previous findings that a WC cut-off at the 85^th percentile for children under the age of 12 y and at the 76^th percentile for those aged 13 to 18 y would predict CV risk 1139. For a WHTR cut-off of 0.5 to become a universal predictor of CV risk in children, further direct confirmatory evidence from comparative correlations with CV risk factors in children of different age and ethnic origins will be required.

BMI is the most widely used index to define children's weight status. Our finding that WC correlates considerably more closely than does WHTR with BMI in children is unsurprising given that WC and BMI increase, whereas WHTR decreases, with age. CV risk factors found in our previous study of around 2500 children drawn from the same population correlated in descending order of strength with WC, BMI and WHTR 11. Others have reported that WC 28394041 and WHTR 283036 correlate more strongly than does BMI with CV risk factors in children. No difference in the ability of WHTR and BMI-for-age to predict CV risk was found in one study 38. These inconsistencies may reflect different methods of analysis, different ethnic populations and the relatively small differences between the related parameters.

A limitation of our study is that 7% of the primary and 10% of the secondary school students declined to participate. Reasons for non-participation were not recorded so it is not possible to determine whether non-participation resulted in any systematic bias.

Reference values and percentile curves for WC and WHTR of Chinese children and adolescents are provided in the study. Both WC and WHTR were age dependent. As a public health tool, WHTR does not on present evidence offer major advantages over WC for children. WC, measured at an agreed site and interpreted against appropriate established reference standards, remains the simplest clinical measure of central obesity for the prediction of CV risk in children. For standards based on WC and WC-derived parameters to be of general public health and research use for international comparisons, it will be important that a single defined site be agreed at which to measure WC.

The authors declare that they have no competing interests.

RYTS, EASN and AML prepared the proposal and supervised the study. HKS coordinated the study, assisted in the supervision of data collection, and took part in the statistical work. KCC took active part in the data analysing. JATY and CWLK helped in data collection and develop the protocol. PCN and TFF made substantial contributions to the conception of the study and revising the manuscript. All authors read and approved the final manuscript.