Article Text

Download PDFPDF

Body mass index in Australian children: recent changes and relevance of ethnicity
  1. Judith Lynch,
  2. Xing Li Wang,
  3. David E L Wilcken
  1. Department of Cardiovascular Medicine, Ground Floor, South Wing, Edmund Blacket Building, The Prince of Wales Hospital, Randwick, NSW 2031, Australia
  1. Professor Wilcken email: d.wilcken{at}


AIM To determine changes over time in the body mass index (BMI) profile of Australian primary schoolchildren and to assess the effects of sex and ethnicity.

METHODS Height and weight were measured in 3645 children (1869 girls and 1596 boys), aged 5–12 years from 39 schools in southeastern Sydney during 1994–7. Levels in the four largest ethnic groups of the population were compared with those measured by the 1985 Australian Council on Health, Physical Education, and Recreation (ACHPER) survey.

RESULTS The study population was 59.9% white (north European), 8.5% Mediterranean white, 7.7% Asian, 7.7% other, and 16.2% mixed (mainly Asian-white (36%) and Arab-white (24.7%)). There were sufficient numbers in four groups for analysis and comparison with the ACHPER survey: Mediterranean white, other white, mixed ethnicity, and Asian children. The age and sex adjusted BMI was highest in Mediterranean white, then white, mixed race, and Asian children. There were minimal differences between sexes within each group. However, boys had an age and sex adjusted BMI 1.5–6.5% higher (mean, 3.9%) than in ACHPER in 1985, as did white girls (mean increase, 2.4%).

CONCLUSIONS BMI in southeastern Sydney schoolchildren is related strongly to ethnicity and age; in boys and white girls it is on average 3.9% and 2.4% greater, respectively, than that recorded in the 1985 ACHPER survey. We suggest that this 10 year increase reflects a general trend in developed countries. Because increased BMI in childhood tracks to adulthood and is then associated with adverse effects on health, these findings signal a need for prevention.

  • adiposity
  • vascular disease
  • dyslipidaemia
  • health promotion
View Full Text

Statistics from

Obesity is an increasing health problem in developed countries. Associations with dyslipidaemia, coronary disease, hypertension, type II diabetes, orthopaedic complications, and psychological disturbances are well documented and attest to its adverse consequences for individuals, and for national health budgets.1 It is now known that increased body weight in childhood tracks to adulthood and reflects subsequent adult values.2 3 In the course of an ongoing family based heart health education programme we recently reported data suggesting that weight relative to height, the body mass index (BMI), has increased over the past 10 years in Australian children.4 In view of the possible long term consequences of this we have undertaken an expanded study to characterise more fully the current profile of weight relative to height in Australian primary school aged children, specifically to explore sex and ethnicity differences. The results define these differences and provide evidence indicating that during the past 12 years there has been a significant increase in the number of overweight boys and a similar trend in girls.


Between the middle of 1994 and 1997 we approached 42 primary schools in the southeastern Sydney area and received permission from 39 to offer our risk factor screening programme (“the heart health education programme”) to all 5–12 year olds. These children were part of a larger group for whom lipid profiles have already been reported,5 and were the ones whose height and weight were also measured. We began recording height and weight midway through the programme.

Over the three years we assessed 3499 children, of whom we have complete information on 3465: 1596 boys (46.1%) and 1869 girls (53.9%) aged 5–12 years in government primary (31.6%), Catholic primary (37.6%), and private fee paying (30.8%) schools in the area. The ethnicity of each child was established initially from the consent form sent home to all parents. This was confirmed by asking the ethnic origin of each parent on a family history questionnaire, which was mailed to the home of each child directly after testing at school. This asked specifically the country of origin of each parent's family, and from this we determined the ethnicity of the child.

The risk factors screened for and the nature of the family history questionnaire have been detailed and the results reported elsewhere4-6; however, in this group we also measured the height and weight of each of the children tested. By dividing the weight (kg) by the square of the height (m) we calculated the BMI. The children's BMI values were compared with the data obtained from the 1985 Australian Council on Health, Physical Education, and Recreation (ACHPER) survey of over 8000 randomly selected Australian schoolchildren.7

To explore the possibility of any bias in our population introduced by assessing only children whose parents had consented to fingerprick blood lipid testing, as opposed to a random population, we measured the height and weight of all the children in the age range in nine consecutive participating schools. This allowed us to compare the BMIs of children whose parents had and had not given permission for fingerprick blood testing. A total of 796 children were included in this sample; 331 (41.6%) without blood testing (54% boys, 46% girls) and 465 (58.4%) with blood testing (47% boys and 53% girls). There were no BMI differences between the boys and girls of the two groups.

Our study was approved by the ethics committee of the University of New South Wales.

Statistical analysis

Values for measured variables are presented as means (SEM). Student's t tests were used for comparisons between groups and an ANOVA F test for more than two groups for quantitative variables. As expected, distributions of the BMI in the total population and in boys and girls were skewed. However, in the subgroups divided by age, sex, and ethnicity the BMIs were normally distributed, with no significant difference in the variances between groups. Thus, we used parametric comparisons for the subgroup analysis. We used a general factorial model of ANOVA to assess the main and interactive effects of age, sex, and ethnicity on the variance of BMI in children. Two tailed p values are reported for the significance of tests.



Because ethnicity might have a significant influence on children's BMI, we assessed the effects of age and sex on BMI in each ethnic group. The ethnic distribution of the pupils screened was 59.9% white (north European), 7.7% Asian, 8.5% Mediterranean white, and 16.2% mixed ethnicity, with the remaining 7.7% comprising other smaller ethnic groups; those of mixed ethnicity were principally Asian-white (36.0% of the mixed population) and Arab-white (24.7% of the mixed population). The remaining 7.7% among the groups contained small numbers of children of five different ethnicities, Arab (n = 99), black (n =  4), Indian (n =  46), Pacific Islander (n = 43) and South American (n = 70); these numbers did not provide sufficient power to detect the measured effects and were excluded. Therefore, we conducted our analyses among the four ethnic groups from the southeastern Sydney area for whom there were measurements in a sufficient number of subjects, and for whom there were data available for comparison from the ACHPER 1985 survey. These were the white (north European), Mediterranean white, Asian, and mixed ethnicity groups.

The age distribution of the studied population was from 5 to 13 years, with most aged 6–11. Age was calculated from the date of birth and rounded to the closest year; that is, < 0.5 was rounded down to the year below and ⩾ 0.5 was rounded up to the year above. There was a linear increase in BMI with age in both girls and boys for each ethnic group (table 1) and in each of the four ethnic groups age was a significant predictor of the BMI (p < 0.0001) in both boys and girls. However, except for white children (p = 0.027), there was no significant sex effect on BMI in each ethnic group after controlling for age as a covariate (p = 0.670 for Asians; p = 0.089 for Mediterraneans; p = 0.809 for those of mixed race). As shown in table2, the differences between girls and boys within each of the four ethnic groups were minimal.

Table 1

Relation between age and BMI (mean (SEM)) in four different ethnic groups

Table 2

Age adjusted BMI (mean (SEM)) according to ethnic origin


As seen in table 2, in both girls and boys among the four ethnic groups, those of Asian ethnicity had the lowest BMI and those of Mediterranean ethnicity the highest. The differences remained significant after controlling for age in the univariate analysis of variance model (F = 17.425, degrees of freedom (df) = 3, p = 0.0001 for boys; F = 7.544, df = 3, p = 0.0001 for girls).


We compared the BMI differences among our present population of children with those of the ACHPER 1985 population7 in girls and boys for each of these four ethnic groups (table 3). There were some differences in the distribution of ethnicities between the two populations, although this may be more apparent than real, and a result of differences in data collection techniques. The ACHPER survey recorded parental birthplace, which might not always reflect ethnicity; we determined ethnicity by asking parents for their family's country of origin (see methods). However, it is relevant that the percentage of children whose parents were Mediterranean was the same in both populations; in ACHPER it was 9.8% and in our population 8.6%. It is noteworthy too that there was a much higher proportion of Asian children in our study (7.5%) than in the ACHPER survey (1.5%). Nevertheless, the increase in BMI over the 10–12 year period was consistent. There was an average increase of 3.9% for all ethnic groups in boys but of only 1.2% in girls (table 3). Although age was a significant independent predictor of BMI in both populations, the difference in BMI between the current and ACHPER studies remained significant after controlling for age (table 3). Table 4 shows age related BMI for white boys and girls in the ACHPER study and our study. The BMI is consistently higher in our study population.

Table 3

Age adjusted BMI (mean (SEM)) according to ethnic origin and compared to those for ACHPER in 1985

Table 4

BMI (mean (SEM)) for white children compared to those for ACHPER in 1985 in each age category


The BMI findings for the diverse ethnic groups in our study demonstrate substantial differences occurring in a stable but cosmopolitan community setting. Previous studies comparing different communities have shown the same ethnic differences. In a comprehensive recent review of more than 66 000 healthy children from the USA aged 5–17 years (from nine studies) the centiles of BMI in each ethnic group were, from highest to lowest, Hispanic boys and girls, followed by all other white children, with Asian children having the lowest values.8 In adults also, European studies have consistently identified higher BMI values and obesity in southern versus northern European adults,9 with generally lower BMI values in Asians10 and larger BMI values in the Arab population.11 In each of our four groups of children there was a linear increase in BMI with age, consistent with the findings at the same age in other populations.8 12 Clearly, our overall findings result from the effects on growth of both genetic and environmental impacts, but with recent changes an effect of the latter.

We emphasise that the urban population we surveyed contains a socioeconomic mix of privileged and less privileged families; and most of the “private” schools were Catholic parish schools. Moreover there were no BMI differences between individual schools after controlling for ethnicity. It seems likely therefore that our findings are representative of the national urban pattern. Indeed, they might underestimate the change because, paradoxically, there are data consistent with adiposity being more common among the less privileged in developed countries, both in children13 and in adults.14 Surveys of other communities should be undertaken.

Because of the size and ethnic diversity of the population studied by the ACHPER group, we compared our current age and sex adjusted findings with those of the ACHPER survey of 1985 to explore possible time related BMI changes; the results show an increased BMI in boys, with a similar trend in girls during the intervening period. This was present in all ethnic subgroups and was significant for white boys and girls, for whom the numbers were large, and also for Mediterranean boys. Lack of significance in the other ethnic subgroups might be a reflection of the smaller numbers in each. It is likely, however, that an increase in BMI among primary school aged children began about 10 years before the ACHPER 1985 survey. Analyses of surveys of Queensland schoolchildren undertaken in 1911, 1950, and 1976 showed no significant change in BMI from 1911 to 1950, but a definite small increase in 1976.15 Our findings are consistent with an acceleration of this effect. It is now well documented that weight patterns in childhood reflect those in adulthood, with obesity tracking particularly strongly.1-3 We have shown previously that levels of apolipoprotein B (apo B) are significantly increased in overweight children,4 and that circulating apo B is correlated strongly with ethnicity in the cosmopolitan population of southeastern Sydney.5 Apo B is the carrier protein for the atherogenic low density lipoprotein (LDL) cholesterol, and circulating concentrations reflect LDL cholesterol concentrations16and cardiovascular risk.17 Indeed, in our earlier study, increased apo B concentrations in children were predictive of an increased risk of coronary vascular events in older family members—even across a generation gap,6 effects likely to be both genetically and environmentally mediated. However, the rapidity of the development of the present changes indicates that lifestyle alterations affecting diet and exercise levels are the likely cause. An inverse relation between physical activity and BMI has been documented for Belgian children,18 and also for children in Singapore, in whom an increase in BMI was also associated with an adverse lipid profile,19 as we have shown for Australian children.4

In relation to our findings, it is important to acknowledge the limitations in the use of BMI as an index of increased body weight in growing children.20-22 Total body fat measurement by magnetic resonance imaging and measurements of triceps and subscapular skinfolds provide more accurate estimations, but are suited more for use in small scale studies and clinical settings.23 24Despite these limitations, the extensive study of Lazarus and colleagues reveals a close association between BMI and adiposity in Australian children25; and in Busselton (Western Australia), Kelly et al found that most children who were heavier than their non-obese peers at age 6 years or over became overweight in adulthood.26 Furthermore, in German children of the same age as those in our study, the BMI provided a good estimate of total body fat measured by magnetic resonance imaging27; and in Italian children the BMI was also strongly associated with total body fat.28 There is a consistent relation between BMI and adverse lipid profiles in children, which is highly relevant to adult onset disease.2 4 6 16 29 Further to the data we have provided from Australian families on atherogenic lipid profiles in children,6 there are now recently reported important postmortem findings. Among children and young adults who died suddenly, mainly from trauma, and for whom antemortem risk factor profiles were available, Berenson and colleagues have documented strong correlations between BMI and the presence of fibrous plaques (p < 0.01) and fatty streaks (p < 0.0001) in coronary arteries.29 Therefore, we suggest that, despite its limitations, documentation of the BMI provides information relevant to health outcomes and also an index of adiposity that facilitates comparisons between populations and ethnic groups within them.

In summary, our study identifies current ethnic and age related differences in BMI among primary school aged Sydney children, which underscore the need to control for these when analysing the present cosmopolitan Australian population. The data are also consistent with an increase in weight relative to height in Sydney primary school aged boys and girls over the past 10–15 years. Because increased BMI and obesity tracks from childhood to adulthood, with potentially adverse long term health effects, these findings signal a need for prevention. The results of our study also emphasise a recent editorial comment that there is now a “clear case for redoubling efforts to promote population wide good health practices in children and young adults that may help to decrease the risk of future cardiovascular disease”.30


The ACHPER data were obtained from the Australian health and fitness survey (1985), jointly funded by the Commonwealth Department of Sport Recreation and Tourism, the Commonwealth Schools Commission, the Commonwealth Department of Health and the National Heart Foundation of Australia. We also wish to thank our heart health programme dieticians M Meischke, R Scott, and K Lobley for their assistance in collecting height and weight data.


View Abstract

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Archives this month