International Journal of Obesity (1997) 21, 484±488 ß 1997 Stockton Press All rights reserved 0307±0565/97 $12.00 Body mass index variations by age and sex, and prevalence of overweight in Japanese adults M Yanai 1, A Kon 2, K Kumasaka 1 and K Kawano 1 1 Department of Clinical Pathology, Nihon University School of Medicine, 30-1, Oyaguchi-Kamimachi, Itabashiku, Tokyo 173, Japan; and 2 Sekishindo Health Care Center, 23-2, Wakita-Honcho, Kawagoe, Saitama 350-11, Japan OBJECTIVE: To determine the reference value of BMI for Japanese subjects and to estimate the prevalence of overweight based on this reference value. DESIGN: Epidemiological analysis with the LMS method, which provides a way of obtaining normalized BMI distributions. SUBJECTS: 7508 Japanese subjects aged 18±69 y in 1993. MEASUREMENTS: Height, age and body weight. CALCULATION: BMI was calculated and tables for percentiles of BMI were plotted against age and sex. Furthermore, the prevalence of overweight was estimated based on 85th percentile of BMI in the men and women 20±29 y of age, who were considered the reference group. RESULTS: The geometric mean BMI and the prevalence of overweight in men was highest in the 30±39 y age group. For women the maximum BMI and prevalence of overweight occurred in the decade 50±59 y. The cut-off points for overweight in this sample were 24.7 kg/m 2 for men and 22.6 kg/m 2 for women. These are considerably lower than the gures of 27.8 kg/m 2 and 27.3 kg/m 2 estimated for Americans. CONCLUSIONS: The prevalence of obesity in Japanese populations should be estimated using ethnic speci c values of BMI, rather than those drawn from Caucasians who tend to have higher BMI in each age group. The prevalence of overweight is increased as age increased in both sexes, especially in women. Keywords: body mass index; reference value; obesity; Japanese adults Introduction Obesity is an excess of body fat and is a risk factor for major chronic disease such as hypertension, heart disease, diabetes mellitus and some kinds of cancers. 1 However, the precise determination of the amount of body fat requires technically sophisticated methods that are available only in research laboratories. For public health studies and clinical practice, simple and convenient anthropometric measurements based on height, weight, and skin fold thickness are recommended. However, prevalence rates of overweight depend on the methods used as well as the cutoff points chosen. Many methods have been reported for de ning overweight or obesity. Body mass index (BMI), or Quetelet's index expressed as weight/height 2 has been widely used as a measure of overweight because it is easy to calculate and correlated with body fat as estimated by body water or densitometry. 2 The recommended BMI is 20±25 kg/m 2, according to the values based on data from the Metropolitan Life Insurance Company study in 1959. Furthermore, the National Institute of Health (NIH) indicated that males at a BMI of 27.8 kg/m 2 or more and females at a BMI of Correspondence: Dr M Yanai. Received 2 october 1996; revised 24 February 1997; accepted 26 February 1997 27.3 kg/m 2 or more, had 2.9 and 2.1 times higher risks of hypertension and hypercholesterolaemia, respectively. 1 Recently, Matsuzawa et al 3 found that a BMI of 22.2 kg/m 2 in men and 21.9 kg/m 2 in women was associated with the lowest morbidity in a middleaged Japanese population, and suggested that the ideal body weight for Japanese was 22 6 height (m) 2.It may be that the `ideal' BMI for Japanese is lower than that appropriate in Caucasian populations. Ethnic speci c standards for BMI should be calculated for the Japanese population. Compared to American population, Japanese may have longer trunks and shorter legs. 4 This study examines the BMI distribution across 10 y age intervals in a sample of Japanese individuals. The median BMI (50th percentile) is compared with results from other countries. Population-speci c cutoff points for overweight and severe overweight are derived from the 20±29 y of age reference group as has been reported for other populations. 5 Subjects and methods Subjects Data were obtained from the people who visited the Sekishindo Health Care Center (Kawagoe, Japan) for a medical check up without any symptoms and stu-
dents in Nihon University School of Commercial Sciences in 1993. The subjects from Sekishinndo Health Care Center generally live in the western part of Saitama prefecture, which is located in north of Tokyo, and the subjects from Nihon University School of Commercial Sciences also live in Tokyo area. The occupations of those from Sekishinndo Health Care Center varied widely from physical labourers such as security guards to sedentary occupations such as of ce-workers. Furthermore, many housewives were included in women. Any subjects with liver disease, renal disease, or heart disease of which the diagnosis was made by the physical examination, urine and blood analysis, were excluded. Seven thousand, ve hundred and eight people of 18 y of age or older were included in the study. They were 3905 males and 3603 females. The distributions by age and sex are shown in the Table 1. Pregnant women were not included in the study. Calculation of BMI Using standard methods, the standing height (H, m) was recorded without shoes and the body weight (BW, kg) was recorded in light-weight indoor clothing whose weight was subtracted from the measured weight. Then, BMI (kg/m 2 ) was calculated as follows: BMI ˆ BW=H 2 Statistical analysis As the distribution of BMI for Japanese may be very skewed, as is the case with the French population, 6 the Z-scores (or s.d. scores) and percentiles cannot be calculated from the mean and standard deviation assuming a normal distribution. So, the LMS method described by Cole 7 was used in order to smooth the data for each age and sex group. This method is based on the assumption that the data can be normalized by using a power transformation, which stretches one tail of the distribution and shrinks the other, removing the skewness. The median (equal to the mean after transformation) and the coef cient of variation are easily determined from the calculation of geometric mean, arithmetic mean, harmonic mean and their standard deviations. The LMS method also converts BMI values in individuals to Z-scores and percentiles. The formula to obtain the Z-score is given by Z ˆ Q=M L 1Š=LS where Z is the Z-score; Q, observed BMI; L, the Box- Cox power; M, the median, and S, the coef cient of variation for the individual's age and sex. The Z- score, once known can be converted to a percentile using normal distribution tables. The median BMI obtained in this study was compared to those in other studies of Caucasians which were already reported. Criteria of overweight Overweight was de ned as the sex-speci c 85th percentile values of BMI for males and females aged 20 through 29 y as used in the second National Health and Nutrition Examination Survey (NHANES II). 8 This age group has been used as the reference group based on the assumption that maximum stature has been reached and that increased weight at later ages is due to increased fat. Severe overweight was de ned as the 95th percentile. Results The geometric mean and standard deviation (s.d.) values of height, weight and BMI are shown in Table 1. Table 2 and Table 3 show the values of L and S, and the percentile value of BMI for each 10 y group of age in males and in females, respectively. The 50th percentile of BMI is always higher in males than in females for all ages. The greatest difference occurs in young adults. The differences in BMI with age are slightly different for men and women. The BMI for males was highest in the 30±39 y group and remained close to this level to the oldest group. The BMI for females was increased gradually with age and was the highest for the 50±59 y age group. The 85th and 95th percentiles of BMI for males from 20±29 y of age were 24.7 kg/m 2 and 27.2 kg/m 2, respectively, and those for females were 22.6 kg/m 2 and 24.5 kg/m 2, respectively. Figure 1 shows the patterns for the prevalence of overweight and severe overweight with age by sex. For males, the prevalence of overweight was the highest in thirties and decreased with each successive 10 y increment of 485 Table 1 Samle size (N) and the geometric mean ( s.d.) values of height in cm, weight in kg and body mass index (BMI) in kg/m 2 for Japanese adults by age and sex Male Female Age (years) N Height (cm) Weight (kg) BMI (kg/m 2 ) N Height (cm) Weight (kg) BMI (kg/m 2 ) 18±19 648 171.2 5.7 64.3 9.2 21.9 2.8 204 158.0 4.6 51.6 6.2 20.6 2.3 20±29 1304 171.9 5.6 64.9 9.2 22.0 2.9 828 158.7 5.1 51.2 5.9 20.3 2.1 30±39 519 170.3 5.5 68.0 10.6 23.4 3.4 662 157.1 5.0 52.7 7.1 21.4 2.7 40±49 817 168.1 6.0 65.3 9.1 23.1 2.9 988 154.9 5.2 53.3 7.0 22.2 2.7 50±59 481 165.4 5.8 63.3 8.1 23.1 2.7 790 153.0 4.8 53.6 6.9 22.9 2.7 60±69 136 162.4 5.3 59.7 7.2 22.6 2.4 131 150.9 4.9 51.7 7.3 22.7 2.8
486 Table 2 The Box-Cox power (L), its standard deviation (S) and BMI (kg/m 2 ) centiles for each ten-years group of age in males Percentile value (kg/m 2 ) Age (years) L S 2.5% 5% 15% 50% 85% 95% 97.5% 18±19 71.43 0.12 17.7 18.2 19.3 21.5 24.6 27.0 28.4 20±29 71.60 0.12 17.7 18.2 19.3 21.6 24.7 27.2 28.8 30±39 70.70 0.14 18.0 18.7 20.1 23.0 26.8 29.5 31.1 40±49 70.17 0.13 18.0 18.7 20.2 22.9 26.1 28.2 29.4 50±59 0.75 0.12 18.0 18.8 20.4 23.1 25.9 27.6 28.4 60±69 0.43 0.11 18.1 18.8 20.1 22.5 25.1 26.7 27.6 Table 3 The Box-Cox power, its standard deviation and BMI (kg/m 2 ) centiles for each ten-years group of age in females Percentile value (kg/m 2 ) Age (years) L S 2.5% 5% 15% 50% 85% 95% 97.5% 18±19 71.74 0.10 17.2 17.6 18.5 20.3 22.8 24.6 25.7 20±29 71.69 0.10 16.9 17.4 18.3 20.1 22.6 24.5 25.7 30±39 71.74 0.12 17.3 17.8 18.8 21.0 23.9 26.4 27.9 40±49 70.84 0.12 17.7 18.3 19.5 21.9 24.9 27.1 28.4 50±59 70.28 0.12 18.2 18.8 20.2 22.7 25.7 27.7 28.8 60±69 70.88 0.12 18.1 18.7 19.9 22.4 25.5 27.8 29.1 age up to sixties. On the other hand, for females, the prevalence of overweight tended to increase with each successive 10 y increment of age up to the fties. The patterns for the prevalence of severe overweight were similar to the prevalence of overweight in both sexes and parallel with the median BMI. Table 4 shows the comparisons of the BMI against age from countries already reported. 6,8,9 The BMI value at 20±39 y of age of Japanese was very similar to that of French population in both sexes. However, after 40 y of age the BMI in the French population was increased relative to the Japanese. It is noteworthy that BMI in American populations is much higher than that in Japanese at all ages. The BMI changes in Finns were similar to the American population but the median BMI was always lower than in the Americans. Discussion Figure 1 Prevalence of overweight (white bars; BMI > 24.7 kg/ m 2 for males and > 22.6 kg/m 2 for females) and severe overweight (dark bars; BMI > 27.2 kg/m 2 for males and > 24.5 kg/m 2 for females) among Japanese adults, aged 18±69 y. Many indices of relative weight have been reported in order to estimate obesity and the nutritional status. Keys et al 10 indicated that the index of the relative weight should have a low correlation with height and high correlation with the body fatness. Furthermore, they concluded that the BMI seemed preferable over other indices of relative weight on these grounds as well as on the simplicity of the calculation and in Table 4 The comparison of the 50th percentile of BMI by age in both sexes to French, 6 American 8 and Finnish populations 9 Male Females Age (Years) Japanese French American Finn Japanese French American Finn 20±29 21.6 22.2 24.9 23.5 20.1 20.3 24.1 22.6 30±39 23.0 23.2 26.1 25.3 21.0 21.1 26.4 24.4 40±49 22.9 24.4 27.3 25.8 21.9 22.2 26.7 26.5 50±59 23.1 25.3 27.6 25.9 22.7 23.6 28.5 27.9 60±69 22.5 25.3 26.9 25.8 22.4 24.4 27.3 28.6
contrast to percentage of average weight, the applicability to all populations. Similarly, Rolland-Cachera et al 11 also reported that body adiposity might be expressed independently of age and sex in children by using BMI, although the BMI is not independent of height among children. 12 Furthermore, Roche et al 13 reported that the BMI is also highly correlated with more direct measures of fatness. So, the BMI has been accepted widely for the estimation of body fatness. The present study illustrates the differences in BMI with age in Japanese adults. The problem might be raised that our results were not representative for the entire Japanese population because of using the subjects from only one region near Tokyo and from university students. Kuczmarski 14 indicated that the income, level of education, and occupation affects the prevalence of overweight. However, the height and body weight in our study are not so different from the data in the Ministry of Health and Welfare in Japan which are representative of the entire Japanese population. 15 Furthermore, it has been reported that more than 40% of Japanese adolescents go to university and their height and body weight are not so different from the remainder of the population. 16 We found that BMI increased less with age in men than in women, as also observed in Caucasian populations, although the change in BMI with age was less. 6,8,9 The changes in BMI throughout life re ect total fat body mass which is mainly centrally located in men and peripheral in women, 17 although it re ects the development of the lean body mass between the ages of 6 and 20 y and loss of lean body mass after the age 65 in both sexes. 18 An increase in the median BMI from the second decade (20±29 y) to fth decade (50±59 y) was apparently higher in the American population than in Japanese of both sexes. However, the differences of median BMI in the twenties between Japanese and Americans, is of interest. Further investigations are necessary to clarify two possible explanations: whether the lean body mass is more in American than in Japanese, or whether the younger American population already has more fat than the Japanese. Ethnic differences in BMI are apparent when comparing our data for Japanese and those for Caucasian populations, such as Americans and Finns. Furthermore, differences in BMI and prevalence of overweight are also observed between Japanese and Hispanics. 19 This may result from the differences in customary food intake. About 30 y ago, the main source of proteins for the Japanese was sh and vegetables, and Japanese did not consume much animal fat. Recently, the dietary habits of the Japanese population have changed. 16 Nowadays, younger Japanese are eating more meat, 20 although the older population still prefer sh. 21 This suggests that BMI could change to the American or European pattern, in which BMI and body fat mass increased more rapidly to middle age. Recently, the average life span in Japanese has been the longest in the world. 22 However, the change in diet may lead to shortening of the life span. In fact, the prevalence of diabetes mellitus (especially non-insulin dependent) and ischaemic heart disease are increasing year by year. 15 Finally, cutoff points for overweight in our data are based on percentile values. However, cutoff points are never absolute, rather they are relative standards that must have some biologic meaning. 19 Prospective studies will be required to determine levels of body mass index associated with the lowest morbidity and mortality in the Japanese population. It is concluded that (i) the reference level for BMI should be determined speci cally for the Japanese population, (ii) BMI changes less with age in Japanese males compared to females, and (iii) the prevalence of overweight in the Japanese adults increased as age increased in both sexes, especially in women. References 1 National Institutes of Health Consensus Development Conference Statement. Health implications of obesity. Ann Intern Med 1985; 103: 1073±1077. 2 Garrow JS, Webster J. Quetelet's index (W/H 2 ) as a measure of fatness. Int J Obes 1985; 9: 147±153. 3 Matsuzawa Y, Tokunaga K, Kotani K, Keno Y, Kobayashi T, Tarui S. Simple estimation of ideal body weight from body mass index with the lowest morbidity. Diab Res Clin Prac 1990; 10: S159±S164. 4 Statistic Bureau, Management and Coordination Agency. Japan Statistical Yearbook. 46th edn, 1997. Japan Statistical Association: Tokyo, 1996. 5 Dawson D. Ethnic differences in female overweight: Data from the 1985 National Health Interview Survey. Am J Public Health 1988; 78: 1326±1329. 6 Rolland-Cachera MF, Cole TJ, Sempe M, Tichet J, Rossignol C, Charraud A. Body mass index variations: centiles from birth to 87 years. Eur J Clin Nutr 1991; 45: 13±21. 7 Cole TJ. The LMS method for constructing normalized growth standards. Eur J Clin Nutr 1990; 44: 45±60. 8 Kuczmarski RJ, Flegal KM, Campbell SM, Johnson CL. Increasing prevalence of overweight among US adultsðthe National Health and Nutrition Examination Surveys, 1960± 1991. J Am Med Assoc 1994; 272: 205±211. 9 Rissanen A, HelioÈvaara M, Aromaa A. Overweight and anthropometric changes in adulthood: a prospective study of 17000 Finns. Int J Obes 1988; 12: 391±401. 10 Keys A, Fidanza F, Karvonen MJ, Kimura N, Taylor HL. Indices of relative weight and obesity. J Chron Dis 1972; 25: 329±343. 11 Rolland-Cachera MF, Sempe M, Guilloud-Bataille M, Patois E, PeÂquignot-Guggenbuhl F, Fautrad V. Adiposity indices in children. Am J Clin Nutr 1982; 36: 178±184. 12 Michielutte R, Diseker RA, Corbett WT, Schey HM, Uredea JR. The relationship between weight-height indices and the triceps skinfold measure among children age 5 to 12. Am J Public Health 1984; 74: 604±606. 13 Roche AF, Siervogel RM, Chumlea WC, Webb P. Grading body fatness from limited anthropometric data. Am J Clin Nutr 1981; 34: 2831±2838. 14 Kuczmarski RJ. Prevalence of overweight and weight gain in the Untied States. Am J Clin Nutr 1992; 55: 495S±502S. 15 Statistics and Information Department, Minister's Secretariat, Ministry of Health and Welfare. Health and Welfare Statistics in Japan, 1994. Health and Welfare Statistics Association: Tokyo, 1995. 16 Health Service Bureau, Ministry of Health and Welfare. National Nutrition Survey of Japan, Annual Report, 1996. Daiichi Shuppan: Tokyo, 1996. 487
488 17 Cronk CE, Roche AF. Race- and sex-speci c reference data for triceps and subscapular skinfolds and weight/stature 2. Am J Clin Nutr 1982; 35: 347±354. 18 Forbes GB, Reina JC. Adult lean body mass declines with age: Some longitudinal observations. Metabolism 1970; 19: 653± 663. 19 Lopez LM, MaÃsse B. Comparison of body mass indexes and cutoff points for estimating the prevalence of overweight in Hispanic women. J Am Diet Assoc 1992; 92: 1343±1347. 20 Sugano M. Characteristics of fats in Japanese diets and current recommendations. Lipids 1996; 31 (Suppl): 283± 286. 21 Shibata H, Nagai H, Haga H, Yasumura S, Suzuki T, Suyama Y. Nutrition for the Japanese elderly. Nutr Health 1992; 8: 165±175. 22 Department of Economic and Social Information and Policy Analysis. Demographic yearbook, 46th issue, 1994. United Nations: New York, 1996.