ORIGINAL COMMUNICATION (2003) 57, 405 409 ß 2003 Nature Publishing Group All rights reserved 0954 3007/03 $25.00 www.nature.com/ejcn Differences in body composition between Singapore Chinese, Beijing Chinese and Dutch children P Deurenberg 1 *, M Deurenberg-Yap 2,3, LF Foo 2, G Schmidt 4 { and J Wang 5 1 Nutrition Consultant, Singapore and Visiting Professor, University Tor Vergata, Rome, Italy; 2 Research and Information Management, Health Promotion Board, Singapore; 3 National University of Singapore, Singapore; 4 Nanyang Technological University, Singapore; and 5 Institute of Nutrition and Food Hygiene, Beijing, People s Republic of China Objectives: To compare the relationship between body mass index (BMI) and body fat percentage (BF%) in children of different ethnic background. Design: Cross-sectional observational study. Settings: The study was performed in three different locations, Singapore, Beijing and Wageningen (The Netherlands). Subjects: In each centre 25 boys and 25 girls, aged 7 12 y, were selected. They were matched on age, sex and body height. Methods: Body weight and body height was measured following standardized procedures. The body mass index (BMI) was calculated as weight=height squared (kg=m 2 ). Body fat was measured by densitometry in Beijing and Wageningen and by dual energy X-ray absorptiometry (DXA) in Singapore. The DXA measurements in Singapore were validated against densitometry. Results: There were no significant differences in BF% or BMI within each gender group across the three study sites. However, after controlling for (non-significant) differences in age and BF%, the Singapore children had a lower (mean s.e.) BMI (15.6 0.3) than the Beijing 17.6 0.3) and Wageningen (16.9 0.3) children. For the same BMI, age and sex the Singapore children had a significant higher BF% (24.6 0.7) than the Beijing (19.2 0.8) and Wageningen (20.3 0.7) children. Conclusions: The study strongly suggests that the relationship between BF% and BMI (or weight and height) is different among children of different ethnic background. Consequently growth charts and BMI cut-off points for underweight, overweight and obesity in children may have to be ethnic-specific. (2003) 57, 405 409. doi:10.1038=sj.ejcn.1601569 Keywords: body composition; body fat percentage; body mass index; children; ethnic differences; Chinese; Caucasians; growth charts *Correspondence: Dr P Deurenberg, 135 Serangoon Avenue 3, no. 10-01, Chiltern Park, 556114 Singapore. E-mail: padeu@singnet.com.sg Guarantor: P Deurenberg. Contributors: Data were collected by PD, GS and JW. PD and MD-Y conducted the statistical analyses and wrote the draft paper. All authors contributed to the final version. { Present address: William Paterson University, New Jersey, USA. Received 14 January 2002; revised 31 May 2002; accepted 17 June 2002 Introduction Body fat is a normal constituent of the human body, its amount varying with age (Forbes, 1987). If the body contains excess body fat one is considered obese (Forbes, 1987; WHO, 1998). Body fat percentage can be measured in vivo using a variety of techniques (Forbes, 1987; Jebb & Elia, 1993). In laboratory studies reference methods like densitometry, deuterium oxide dilution or dual energy X-ray absorptiometry (DXA) are normally used to measure body fat percentage. In children some of these reference methods may be less suitable than in adults as the measurement may require intensive co-operation (as for densitometry) or the validity of the method is not sufficiently established in children. For population studies in adults the body mass index (BMI, kg=m 2 ) can be used as a surrogate measure for body fatness (Forbes, 1987, Deurenberg et al, 1991; Deurenberg- Yap et al, 2000). Adults are considered obese if their BMI exceeds 30 kg=m 2 (WHO, 1998). To classify children as overweight or obese, weight-for-height reference curves are normally used and many countries have developed their own growth charts. The World Health Organization (WHO, 1998) has suggested the use of international growth charts developed, for example, by the National Centre for Health Statistics (WHO, 1986). Recent studies suggest that the BMI
406 might also allow a good estimation of body fatness in children (WHO, 1998; Rolland-Cachera et al, 1987), as long as age specific reference values are used. Cole et al (2000) recently published BMI references curves based on international data. Current literature indicate that the BMI cut-off value for obesity as recommended by WHO (1998) at 30 kg=m 2 for adults could be too high in some Asian population groups (Wang et al, 1994; Gurrici et al, 1998; Deurenberg et al, 1998; Gallagher et al, 2000; Deurenberg-Yap et al, 2000; He et al, 2001; Ko et al, 2001). For example, Indonesians (Guricci et al, 1998), Singaporeans (Deurenberg-Yap et al, 2000) and Hong Kong Chinese (Ko et al, 2001; He et al, 2001) have 5 7% more body fat percentage compared with Caucasians with the same BMI. This higher body fat percentage coincides with higher relative risks for cardiovascular risk factors at lower levels of BMI, providing more evidence for the need to lower the BMI cut-off value for obesity in these populations (Ko et al, 1999; Steering Committee, 2000; Deurenberg-Yap et al, 2001). However, northern (Beijing) Chinese were not found to have a higher body fat percentage for the same BMI as Caucasians (Deurenberg et al, 1997). Their more stocky body build, more comparable with Caucasians than with southern Chinese (Hong Kong, Singapore), makes them, in their BF%=BMI relationship, more comparable with Caucasians than with southern Chinese (Deurenberg et al, 1991), who generally have a more slender body build. It is currently not known whether also (southern) Chinese children have higher body fat compared with Caucasians children with the same body mass index (weight and height). The aim of this pilot study was to compare body composition (body fat percent) of age-matched boys and girls of three different ethnic groups: Singapore (southern) Chinese, Beijing (northern) Chinese and Dutch Caucasians. Subjects and methods In Singapore, data of a group of 50 children, 25 girls and 25 boys were made available for the purpose of this study by Nanyang Technological University. The children ranged in age from 7 to 12 y and were of Chinese ancestry. Their body composition was measured in 1998 in an ongoing body composition study. The Singapore children were matched for sex, age and height (closest value) with data from children in Beijing (People s Republic of China) and Wageningen (The Netherlands). The Beijing and Wageningen children were measured in 1995 and in 1994 respectively. Characteristics of the children are given in Table 1. At all three study sites the Medical Ethical Committees approved the study protocol. Body weight was measured in underwear or swimsuit to the nearest 0.1 kg using digital scales. Body height was measured without shoes, Frankfurt plane horizontally, to the nearest 0.1 cm using wall-mounted stadiometers. From weight and height the body mass index (BMI, kg=m 2 ) was calculated. In Singapore body fat was measured using a Hologic whole body dual energy X-ray absorptiometer (QDR-4500, Hologic, Waltham, MA, USA; software version V8.23a:5). In Beijing and in Wageningen body fat was measured using densitometry (underwater weighing). Residual lung volume was measured simultaneously using helium dilution. The procedures for underwater weighing in Beijing and Wageningen are comparable but different scales and respirometers were used. The method and procedure in each centre have been described in detail (Deurenberg et al, 1991; Wang & Deurenberg, 1996). From body density body fat was calculated using an age-specific equation (Deurenberg et al, 1990). In a subgroup of 22 Singapore children the validity of the DXA measurements was tested against densitometry using air-displacement plethysmography (BODPOD 1, Body Composition System, Life Measurement Instruments, Concord, CA, USA). In these 22 children mean body fat from DXA and mean body fat from BODPOD was 22.6 8.9 and 23.0 9.4% respectively. The difference of 0.4 4.4% was not significant (P ¼ 0.62). The two methods correlated highly (r ¼ 0.88, P > 0.0001) and Bland and Altman (1986) analysis showed that the difference between the two methods was not correlated with the level of body fatness (r ¼ 0.11, P ¼ 0.63) and not related to age (r ¼ 7 0.11, P ¼ 0.611). It is known from the literature that air displacement and underwater weighing give comparable results in children (Nuñez Table 1 Characteristics of the children in each study site a Girls (n ¼ 75) Boys (n ¼ 75) Singapore Beijing Wageningen Singapore Beijing Wageningen Mean s.d. Mean s.d. Mean s.d. Mean s.d. Mean s.d. Mean s.d. Age (y) 9.6 1.8 9.6 1.8 9.5 1.7 10.1 1.8 10.1 1.8 10.1 1.8 Height (cm) 135.2 11.6 138.0 10.9 137.6 10.2 138.0 12.1 139.9 11.3 143.6 12.4 Weight (kg) 29.5 8.6 33.3 10.9 31.0 6.4 31.7 10.6 35.9 11.7 35.1 8.5 Body mass index (kg/m 2 ) 15.8 2.7 17.2 3.9 16.2 1.4 16.2 3.3 17.9 3.6 16.8 1.6 Body fat percent 24.2 6.8 21.6 9.4 20.6 5.5 22.3 8.1 20.3 9.6 19.3 5.6 a There were no significant differences between boys and girls in each study site and no significant differences between study sites within each gender group.
et al, 1999; Demerath et al, 2002). Based on this information it is assumed that differences in body fat percentage between Singapore and the other two study sites are not due to methodological differences. Data were statistically analysed using SPSS for Windows, version 10.0.0 (1999, Chicago, IL, USA). Differences between boys and girls and differences between centres were compared using analyses of (co-)variance with Bonferroni post hoc analyses where applicable. Differences between variables were tested using the Student t-test. Study site (country) was coded as dummy variable, where for Singapore c1 ¼ 0 and c2 ¼ 1, for Beijing c1 ¼ 1 and c2 ¼ 0 and for Wageningen c1 ¼ 0 and c2 ¼ 0. Gender also was coded as a dummy variable, where female ¼ 0 and male ¼ 1. Stepwise multiple regression analysis (P in > 0.05) was performed using BF% as dependent variable and age, BMI, sex, c1 and c2 and interaction factors as independent variables (Kleinbaum et al, 1998). Values are given as mean s.d., unless otherwise indicated. Significance is set at P > 0.05. equation did not improve the explained variance (r 2 ) or the standard error of estimate (s.e.e.) of the prediction equation (results not shown). The stepwise regression analysis reveals that there are no differences in the BF%=BMI relationship between Wageningen and Beijing children. However, the Singapore children have for the same BMI 4.7% more body fat than the Beijing and Wageningen children (Table 2). Analysis of co-variance shows that for the same sex, age and BF% the BMI (mean s.e.) of Singapore children is significantly lower (P > 0.005) at 15.6 0.3 kg=m 2 than in Beijing (17.6 0.3 kg=m 2 ) and Wageningen children (16.9 0.3 kg=m 2 ). BMI (corrected) between Beijing and Wageningen children did not differ significantly (P ¼ 0.12). Figure 1 shows the mean ( s.e.) differences in body fat percentage between Singapore, Beijing and Wageningen boys and girls, after correcting (ANCOVA) for differences in age and BMI. From this figure it is clear that, despite a lower BMI, body fat percentage in Singapore children is higher compared with Beijing and Wageningen children. 407 Results Table 1 gives the characteristics of boys and girls for each study site. The age of the children ranged from 7 to 12 y in each study site. There were no significant differences in age, height, weight, BMI and BF% between boys and girls in each study site and within each gender group between the study sites. When, however, the data of boys and girls were combined, the BMI of the Singapore children was significantly lower (P ¼ 0.03) than the Beijing children, and body fat percentage in Singaporean children tended to be higher (P ¼ 0.09) than in the Wageningen children. The higher variation for weight, BMI and BF% in the two Chinese groups is notable compared with the Dutch children. Table 2 provides the regression coefficients ( s.e.) of the stepwise multiple regression analysis between BF% and BMI, sex, and country (as dummy variable). There was no interaction between sex and BMI, sex and country and country and BMI. Age did not enter in the regression equation, as its P in value was 0.07. If forced into the equation the regression coefficient (s.e.) was 7 0.47 (0.26). Forcing age into the Figure 1 Body fat percentage (mean s.e.) in Singapore, Beijing and Wageningen girls and boys for the same age and body mass index (corrected for age and body mass using analysis of covariance). Table 2 Stepwise multiple regression of body fat percent as dependent variable a BMI C2 Sex Constant s.e.e. Step Mean s.e. Mean s.e. Mean s.e. Mean s.e. (%) r 2 1 1.737 0.161 7 7.6 2.7 5.8 0.44 2 1.853 0.152 4.6 0.9 7 11.1 2.6 5.4 0.52 3 1.896 0.149 4.7 0.9 7 2.6 0.9 7 10.6 2.6 5.3 0.55 a Variables offered in the equation: BMI, age, sex, country, interaction between BMI and sex, BMI and country, country and sex. BMI, body mass index; country Singapore c1 ¼ 0andc2¼ 1, Beijing c1 ¼ 1andc2¼ 0, Wageningen c1 ¼ 0 and c2 ¼ 0; sex females ¼ 0, males ¼ 1; s.e.e., standard error of estimate, r 2, explained variance.
408 Discussion The children from each study site are by no means representative of the children in their respective countries, nor do they represent Chinese or Caucasian in general. However, in all study sites the children were not specially selected for their degree of obesity or leanness. Data from Singapore children were used to age-match children from Beijing and Wageningen. Height and not weight was used as additional matching criterion to allow for maximum variation in weight and thus apparent body fatness. The matching resulted in three groups of children that did not differ in age, height, weight, body mass index and body fat percentage within each gender group. Also, in each study site there were no differences between boys and girls in height, weight, BMI and body fatness. This is to be expected at this age range. Differences in body composition, especially in body fatness between boys and girls start to appear at onset of puberty (Forbes, 1987; Deurenberg et al, 1990). The matching for age and height resulted in a data set that showed a clearly higher variation in body weight, body mass index and body fat percentage in the Chinese children. This higher variation reflects what can be observed in Singapore and Beijing; there are very lean but also very obese children, whereas in The Netherlands (Wageningen) extremes are not so common. The higher variation (standard deviation) might be a reason why differences in weight, BMI and body fat percentage across the study sites are not statistically significant, despite apparently lower mean values for weight and BMI and higher mean body fat percentage values in Singapore children. Despite the lack of statistical significance, the crude data in Table 1 suggest that Singapore children have a higher body fat percentage at a lower BMI. Stepwise multiple regression taking age, sex and study site into account showed that there are no significant differences in the BMI=BF% relationship between Beijing and Wageningen children, whereas the Singapore children have a significant higher BF% for the same BMI compared to Beijing and Wageningen children. No interaction was found between BMI and sex, showing that, as found in other studies (Deurenberg et al, 1991; Guricci et al, 1998; Deurenberg-Yap et al, 2000), the relationship between BF% and BMI has a similar slope for males and females. Also, there was no interaction between BMI and country and between country and sex, indicating that the relationship is equal in the three study sites, except that the intercept is higher in Singapore Chinese children. Figure 1 shows the corrected BF% (for the same age and BMI) for boys and girls in the three study sites. Differences in body composition in children of different ethnic groups have been reported earlier. For example in comparing American black, white and Hispanic children, Ellis et al (Ellis, 1997; Ellis et al, 1997) reported higher body fat percentage values in Hispanic boys and girls, also after correcting for body size. It is possible that the different methodologies used for body fat measurements are responsible for the observed differences across the study sites, but it seems unlikely that a difference as much as 4.7 percentage points body fat could be due to differences in methodology alone. Moreover, comparison between DXA and densitometry using Bland and Altman (1986) analysis showed that the methods are comparable (see method section). BODPOD measurements were not available for all children as there was lack of compliance with this methodology. This pilot study offers no explanation for the differences found and it can only be speculated that the reason might be the same as found in an earlier study in adults. The higher body fat percentage at low BMI in adult Singapore Chinese could be explained by differences in body build, more specifically differences in relative leg length and differences in slenderness compared with Beijing Chinese and Dutch Caucasians (Deurenberg et al, 1999). This effect of body build has been confirmed in two other studies (Guricci et al, 1998; Deurenberg-Yap et al, 2001). Recently we also observed that Singaporean adolescents have higher skinfold thickness (biceps, triceps, subscapular and supra iliac) compared with Caucasians (Dutch) adolescents, despite having a lower BMI (results not published). Another possible explanation would be differences in pubertal status, but that information was not available in Beijing and Singaporean children. As for Singapore adults a higher body fat percentage at a lower BMI (compared with Caucasians) is a strong argument for lowering the BMI cut-off point for overweight and obesity (Deurenberg-Yap et al, 2000; Deurenberg, 2001), a different relationship between weight and height (body mass index) and body fat in children is an argument against universal growth charts for children. At the moment in Singapore a nationwide anthropometric study is being carried out in order to get up-to-date growth charts (the current growth charts are more than 10 y old). In a sub sample study, in-depth measurements on body composition using reference methods and predictive methods are taken. This will result in specific body composition data that will facilitate the early detection and prevention of overweight and obesity in Singapore school children, in which the prevalence of overweight and obesity is currently as high as 12% (Ministry of Health Annual Reoport, 2000). Conclusion This pilot study shows that, as found earlier in Singaporean adults, Singaporean children also have a higher body fat for the same weight and height (body mass index) than children of the same age from Beijing and The Netherlands. Acknowledgements Nestlé Foundation sponsored the body composition studies in Beijing. The body composition measurements in Wageningen were part of various projects on children, carried out at the Wageningen University. The authors are grateful to the children who participated in the studies.
References Bland JM & Altman DG (1986): Statistical methods for assessing agreement between two methods of clinical measurements. Lancet i, 307 310. Cole TJ, Bellizzi MC, Flegal KM & Dietz WH (2000): Establishing a standard definition for child overweight and obesity worldwide: International survey. Br. Med. J. 320, 1240 1243. Demerath EW, Guo SS, Chumlea WC, Towne B, Roche AF & Siervogel RM (2002): Comparison of percent body fat estimates using air displacement plethysmography and hydrodensitometry in adults and children. Int. J. Obes. Relat. Metab. Disord. 26, 389 397. Deurenberg P (2001): Universal cut-off BMI points for obesity are not appropriate. Br. J. Nutr. 85, 135 136. Deurenberg P, Pieters JJL & Hautvast JGAJ (1990): The assessment of the body fat percentage by skinfold thickness measurements in childhood and young adolescence. Br. J. Nutr. 63, 293 303. Deurenberg P, Weststrate JA & Seidell JC (1991): Body mass index as a measure of body fatness: age and sex specific prediction formulas. Br. J. Nutr. 65, 105 114. Deurenberg P, Ge K, Hautvast JGAJ & Wang J (1997): Body mass index as predictor for body fat: comparison between Chinese and Dutch adult subjects. Asia Pacific J. Clin. Nutr. 6, 102 105. Deurenberg P, Yap M & Van Staveren WA (1998): Body mass index and percent body fat: a meta analysis among different ethnic groups. Int. J. Obes. Relat. Metab. Disord. 22, 1164 1171. Deurenberg P, Deurenberg-Yap M, Wang J, Lin Fu Po & Schmidt G (1999): The impact of body build on the relationship between body mass index and body fat percent. Int. J. Obes. Relat. Meab. Disord. 23, 537 542. Deurenberg-Yap M, Schmidt G, Staveren WA & Deurenberg P (2000): The paradox of low body mass index and high body fat percent among Chinese, Malays and Indians in Singapore. Int. J. Obes. Relat. Metab. Disord. 24, 1011 1017. Deurenberg-Yap M, Chew SK, Lin V, Tan BY, van Staveren WA & Deurenberg P (2001): Relationships between indices of obesity and its co-morbidities among Chinese, Malays and Indians in Singapore. Int. J. Obes. Relat. Metab. Disord. 25, 1554 1562. Ellis KJ (1997a): Body composition of a young multi-ethnic male population. Am. J. Clin. Nutr. 66, 1323 1331. Ellis KJ, Abrams SA & Wong WW (1997a): Body composition of a young multi-ethnic female population. Am. J. Clin. Nutr. 65, 724 731. Forbes GB (1987): Human Body Composition. New York: Springer Verlag. Gallagher D, Heymsfield SB, Heo M, Jebb SA, Murgatroyd PR & Sakamoto Y (2000): Healthy percentage body fat ranges: an approach for developing guidelines based on body mass index. Am. J. Clin. Nutr. 72, 694 701. Guricci S, Hartriyanti Y, Hautvast JGAJ & Deurenberg P (1998): Relationship between body fat and body mass index: differences between Indonesians and Dutch Caucasians. Eur. J. Clin. Nutr. 52, 779 783. He M, Tan KCB, Li ETS & Kung AWC (2001): Body fat determination by dual energy X-ray absorptiometry and its relation to body mass index and waist circumference in Hong Kong Chinese. Int. J. Obes. Relat. Metab. Disord. 25, 748 752. Jebb SA & Elia M (1993): Techniques for the measurement of body composition: a practical guide. Int. J. Obes. 17, 611 621. Kleinbaum DG, Kupper LL, Muller KE & Nizam A (1998): Applied Regression Analysis and Other Multivariate Methods, 3rd edn. Pacific Grove: Duxbury Press. Ko GTC, Chan JC, Cockram CS & Woo J (1999): Prediction of hypertension, diabetes, dyslipidaemia or albuminuria using simple anthropometric indexes in Hong Kong Chinese. Int. J. Obes. Relat. Metab. Disord. 23, 1136 1142. Ko GTC, Tang J, Chan JCN, Wu MMF, Wai HPS & Chen R (2001): Lower BMI cut-off value to define obesity in Hong Kong Chinese: an analysis based on body fat assessment by bioelectrical impedance. Br. J. Nutr. 85, 239 242. Nuñez C, Kovera AJ, Pietrobelli A, Heshka S, Horlich M, Kehayais JJ, Wang ZM & Heysmfield SB (1999): Body composition in children and adults by air displacement plethysmography. Eur. J. Clin. Nutr. 53, 382 387. Rolland-Cachera MF, Deheeger M, Avon P, Guilloud-Bataille M, Patois E & Sempe M (1987): Tracking adiposity patterns from 1 month to adulthood. Ann. Hum. Biol. 14, 219 222. Steering Committee (2000): The Asia perspective: redefining obesity and its treatment. Melbourne: International Diabetes Institute. Wang J & Deurenberg P (1996): The validity of predicted body composition in Chinese adults from anthropometry and bioelectrical impedance in comparison with densitometry. Br. J. Nutr. 76, 175 182. Wang J, Thornton JC, Russell M, Burastero S, Heymsfield SB & Pierson RN (1994): Asians have lower BMI (BMI) but higher percent body fat than do Whites: comparisons of anthropometric measurements. Am. J. Clin. Nutr. 60, 23 28. WHO (1986): Use and interpretation of anthropometric indicators of nutritional status. Bull. WHO 64, 929 941. WHO (1998): Obesity: preventing and managing the global epidemic. Report on a WHO Consultation on Obesity, Geneva, 3 5 June, 1997, WHO=NUT=NCD=98.1. Geneva: WHO. 409