Introduction

Childhood obesity is associated with several risk factors for chronic diseases, including heart disease in later life (Cole et al, 2000; WHO, 2000). This is especially important in many middle-income and developing countries, which continue to face the double burden of infection- and poverty-associated diseases, together with the emerging concerns of chronic diseases. South Africa (estimated population 40 million) has since its 1994 political transformation undergone simultaneous social, epidemiological and economic transitions. The province of KwaZulu-Natal has a population of approximately 9 million, 82% of whom are Zulu-speaking black South Africans and 60% of whom live in rural areas. The nutritional transition facing developing and middle-income countries also has important implications for preventive strategies to control chronic degenerative diseases (WHO, 2000; Popkin, 1994; Monyeki et al, 1999). The impact of the rapid process of socio-economic development during the 1990s on the nutritional status of South African children is poorly understood (May, 2000; Budlender, 2000). While it may be still too early for these dynamic developmental processes to substantially alter nutritional and other health status indicators, it is nonetheless important to establish the current prevalence of overweight and obesity in South African children and to investigate the changes in the nutritional status of children over time. Overweight and obesity are preventable conditions and early signs of increasing prevalence offer opportunities to develop preventive and promotive strategies (WHO, 2000; Bellizzi & Dietz, 1999).

The World Health Organization (WHO, 1995) has noted the inadequacy of the currently used National Center for Health Statistics (NCHS) growth curves, based on the anthropometric data gathered in the USA National Health and Nutrition Examination Survey I (NHANES I), and that new international reference curves are being developed. The International Obesity Task Force (IOTF) in 1999 had recommended that the adult cut-off points for BMI (body mass index) be used to develop internationally acceptable BMI centiles for children (Bellizzi & Dietz, 1999). The LMS method (a summary of the distribution of BMI at each age by its median, coefficient of variation plus measure of skewness) was used by Cole et al (2000). This was based on data obtained from six large nationally representative cross-sectional surveys on growth (from Brazil, The Netherlands, Singapore, the UK, Hong Kong and the USA), which were used to extrapolate backwards adult BMI cut-offs at 18 y of age, namely 25 kg/m² for overweight and for obesity, 30 kg/m² (Cole et al, 2000). Cole et al (while supporting the call for the monitoring of trends in childhood obesity because of their public health importance, has drawn attention to the difficulties in quantifying these trends and comparing them internationally. Their selected reference populations, however, were not ideal, and probably represent Western populations adequately but not developing countries. In particular, Cole et al commented on the lack of data from Africa, and called for further research to study BMI patterns among children in Africa.

In developing countries undergoing a nutritional transition a relationship between stunting and overweight has been postulated. Popkin et al (1996) showed that children in China, Russia and South Africa who were stunted were also at risk of being overweight (based on the WHO 85th percentile definition). A fourth country considered by Popkin et al, namely Brazil, did not show this effect. The relative risk of being overweight if stunted, ranged from 2.6 among children in South Africa to 7.7 amongst Russian children.

The study reported here investigated the nutritional status of South African children as measured by stunting and overweight/obesity, using both the National Center for Health Statistics (NCHS)-based WHO indicators, and the IOTF criteria, this study being one of the first studies of its kind from Africa. The purpose of this study was therefore to explore the relationship between under- and overnutrition among South African school children. The objectives were firstly to compare the magnitude of overweight and obesity using both the WHO and IOTF reference criteria and to investigate the prevalence of stunting, secondly to investigate the relationship between stunting and overweight, and thirdly to make recommendations to policy-makers, communities and health planners, to improve nutritional status in respect of the management of overweight from a public health perspective.

Methods

The study used cross-sectional anthropometric data from primary and secondary sources for measures of overweight, obesity and stunting, to serve as an internal South African baseline and to allow for comparisons of the prevalence of overweight internationally.

Primary data collection

A community-based cross sectional nutritional survey was undertaken in 1995 of rural primary school children in KwaZulu-Natal (KZN; one of nine provinces in South Africa). This formed part of a baseline survey for a randomized controlled trial to investigate the effects of anthelminthic treatment and micronutrient supplementation on the health status of rural school children, which has been previously reported (Jinabhai et al, 2001a,b). Anthropometric data were collected by health professionals, from all grade 3 pupils aged between 8 and 11 y from 11 schools in the Vulamehlo magisterial district (a rural district in the province of KwaZulu-Natal), giving a total sample of 802 children. The 11 schools were randomly selected from all 72 schools of this district (giving a 15% sampling rate).

The survey measured height (cm) using a stadiometer. The stadiometer was placed on an even, uncarpeted area, and the child's shoes were removed but the child remained wearing light clothing. The child was positioned as follows, facing the fieldworker and with shoulders erect, arms relaxed at the side, feet flat and heels touching the rear of the stadiometer. The child looked straight ahead (Frankfort plane), the headpiece was slid down until it touched the crown of the head, measurement was made to the nearest 0.5 cm and the average of two measures was used. After removing sweaters, the child's weight (kg) was measured using a calibrated Masskot electronic scale, supplied by the KwaZulu-Natal Department of Health. The scale was placed on an even, uncarpeted area and was levelled with the aid of its in-built spirit level. The scale was switched on, fieldworkers waited for a zero indication and the stable indicator in the top left-hand corner of the display panel was checked and two measurements were recorded to ensure accuracy.

The health professionals were trained by the research team and the national fieldworkers were trained by the National Department of Health.

Secondary data collection

Anthropometric data from a national cross-sectional study, the South African National Primary Schools' Anthropometric (SANPS) Survey (Department of Health, 1994), was then used for comparison. This survey was undertaken by the Department of Health. Both the national and provincial data for KwaZulu-Natal reported here derive from this survey. In February 1994 more than 3300 South African schools participated in this survey. They represented all the education departments across all the geographic areas, and socio-economic and racial groups in South Africa. School health nurses collected anthropometric data on 105 000 school beginners. Ten schools from each magisterial district throughout the country were chosen based on population representation, and stratified according to the proportion of the population in each district. The sample of children was drawn from school lists provided by each Education Department—at each school 40 children (20 from each of the first and second grades, comprising 10 males and 10 females) were randomly selected. The birth date was recorded and the weight and height of each child measured by trained nurses. The final analysis included data for 97 790 children, representing 4.9% of the estimated 2 million children in those grades in 1994, aged from 4 to 12 y; data was provided nationally and provincially but not available per magisterial district nor split for urban and rural areas. In the South African educational system the relationship between the age of the children and their grade is complex, because of the apartheid education policies. As a result, in a particular grade there is a wide age range amongst the pupils present, owing to the fact that pupils enter schools at different ages. To allow for comparison with our primary data, only children aged between 8 and 11 y were selected for this analysis. In this age cohort nationally there were 24 391 children, and provincially for KwaZulu-Natal 4342 children.

Anthropometric measures from the primary Vulamehlo study and secondary data from the SANPS Survey were used to compare the prevalence of stunting (height for age Z scores), and overweight and obesity, by age and sex. The reference population used in calculating the Z-scores was that given by the NCHS (WHO,1995). For the purpose of this study mild stunting was defined as <-1 Z-score, moderate stunting as <-2 Z-scores, and severe stunting as <-3 Z-scores (height-for-age). With respect to overweight and obesity, calculations were based on both the WHO and the IOTF cut-off values (WHO, 1995; Cole et al, 2000). Comparisons were undertaken between the prevalence of the WHO 85th percentiles with the IOTF definition of overweight, and the WHO 95th percentiles with the IOTF definition of obesity. A risk ratio analysis was used to measure the association of being stunted and being overweight. A comparison of the South African prevalence with published international studies using IOTF definitions was undertaken to assess the levels of overnutrition amongst South African children.

Ethical approval for the study was provided by the Ethics Committee of the Faculty of Medicine, University of Natal. The South African Department of Health provided written permission for the analysis of the secondary data source. All data was analysed using SPSS version 10. Z-scores were calculated using the EPI-NUT component of the Epi-Info V 6.04 package. The kappa () coefficient of agreement was used to compare methods; this is a well-recognized method of agreement and it is used here to enable comparison with other published work related to IOTF cut-offs (Wang & Wang, 2000).

Results

The prevalence of stunting, overweight and obesity both in terms of the age and sex distribution, among primary school children, selected from schools at national, provincial (KZN) and district level was determined. Two sets of criteria were used to measure stunting and overweight/obesity prevalence—WHO and IOTF.

With respect to the Vulamehlo district from the eleven schools selected, 802 pupils from a total of 806 participated (four pupils were absent). This provided a 99.5% response rate for the primary data. In regard to the national data (which forms our secondary data source) a total of 105 000 children were identified, of which data for 97 790 was obtained, a 93.1% response rate. This information was not disaggregated by gender, race or province.

The prevalence of overweight (85th percentile) and obesity (95th percentile) as measured by the WHO (NHANES I) cut-offs was observed to be higher than that measured using the IOTF method (² test of trend over gender and geographical group, 0.05<P<0.10, in all cases; Table 1). However, when using the coefficient to measure agreement across age, sex and group (national, provincial and rural district), good levels of agreement were observed between these two standards (>0.55 in all cases), although agreement did vary across age, sex and group ranging from 0.55 to 0.94. The levels of agreement for obesity were in all cases lower than those observed for overweight (both ranged between 0.55 and 1.0). An example of this was when considering across both sexes and for all ages—for obesity, nationally =0.61, provincially =0.65 and for the rural district =0.73; for overweight, nationally =0.80, provincially =0.75 and at the district level =0.80. When considering gender across both obesity and overweight and across each of the age and geographical groups, it is interesting to note that females have a higher coefficient than males. When considering overweight the levels of agreement nationally were =0.76 (males) and = 0.84 (females), provincially =0.70 (males) and =0.80 (females) and at district level, =0.55 (males) and =0.94 (females). All coefficients were highly significant (P<0.005).

Table 1 - Prevalence (%) of overweight and obesity (IOTF and WHO definitions) and stunting among South African primary school children, by age and sex.

Full table (41K)

There was also a marked trend across sex with females demonstrating a higher prevalence of both overweight and obesity across all the groups. With respect to overweight in this study, the overall total prevalence for IOTF cut-offs was 2.3, 2.8 and 4.1% nationally, provincially and in the rural district, and for the WHO definition it was 3.4, 4.5 and 6.0%, respectively. The national and provincial data included children from both urban and rural areas and all socio-economic strata, whereas the district data was of rural children from communities of low socio-economic status. The overall prevalence of obesity was found to be low by both definitions. For IOTF cut-offs the levels were 0.4, 0.5 and 0.6% nationally, provincially and at the rural district level, while for the WHO definition the corresponding values were 0.8, 1.0 and 0.9%. Clearly, the levels of both overweight and obesity using the WHO criteria were higher than the cut-offs defined by the IOTF criteria (however, P>0.10 in both cases).

For national and provincial levels of mild stunting, a prevalence range of 45–72% was observed, while for moderate stunting the range was 15–40%. At the rural district level the prevalence of mild stunting ranged from 31 to 100%, and moderate stunting from 3 to 25% amongst the different age cohorts. When considering severe stunting, the overall prevalence was found to be 4.9% nationally, 4.5% provincially in KwaZulu-Natal, and 0.6% in the Vulamehlo rural district.

Table 2 considers those overweight and obese children as defined by the WHO and IOTF cut-off points, and considers the prevalence of mild and moderate stunting amongst these groups. Both definitions identified high levels of mild stunting amongst those defined as overweight, for both sexes. These results reflect the national and provincial data sets, with the numbers being too small in the Vulamehlo district data to make any firm conclusions.

Table 2 - Prevalence of mild and moderate stunting in primary school children among those classified as overweight and obese by IOTF and WHO definitions.

Full table (29K)

A risk ratio analysis was carried out to examine the association between being moderately stunted and overweight, the analysis being undertaken for both definitions of overweight (WHO and IOTF). This compared the probability of being overweight in a stunted sub-sample of each group, relative to the probability of being overweight in a non-stunted sample. For each group under consideration at national, provincial and district level, no statistical excess in relative risk was observed. As an illustration of this when looking at the national data, over both sexes and applying the WHO definition of overweight, the following risk ratios were obtained: for age 8, RR=1.05 (95% CI 0.86–1.29); for age 9, RR=1.07 (95% CI 0.77–1.49); for age 10, RR=0.78 (95% CI 0.29–2.10); for age 11, RR=1.32 (95% CI 0.33–5.29; P>0.10 in all other cases). When considering risk ratios for mild levels of stunting, decreased ratios were observed for children in every age group in the national, provincial and district data sets but these did not reach statistical significance. As a further example, again considering the national data, over both sexes with the WHO definition, the following risk ratios were obtained: for age 8, RR=0.79 (95% CI 0.62–1.02); for age 9, RR=0.85 (95% CI 0.71–1.03); for age 10, RR=0.87 (95% CI 0.55–1.38); for age 11, RR=0.68 (95% CI 0.17–2.70; P>0.10 in all other cases).

By either definition when considering overweight the percentage of stunted males was greater than that of females for both mild and moderate stunting for both national and provincial data (0.01<P<0.10 in all cases). The numbers are too small to make any judgement for the district data. When considering obesity by either of the definitions, there appears to be no difference between the sexes when considering stunting at any level (P>0.10, in all cases). However, the percentage of mild and moderate stunting for those defined as obese is greater in all cases for the WHO classification as compared with that of the IOTF (nationally P<0.05 in all cases, provincially P>0.10 in all cases; numbers were too small at the district level).

Discussion

This is one of the first studies in Africa to explore the relationship between stunting and overweight children using two sets of definitions (WHO and IOTF) to measure nutritional status. This analysis of the relationship between stunting and overweight among children has important implications for policy-makers in developing and planning public health interventions, and for monitoring the nutritional transition.

A comparison was also undertaken of what could be considered as the more globally representative IOTF definition of overweight in comparing the prevalence of overweight in South African children with the internationally published data (Table 3). Comparison of these international studies using IOTF cut-offs suggests the following trends. The prevalence of overweight children in South Africa is low, and this may reflect the fact that the majority of the South African population, during the period under review, were in the lower socio-economic strata. In the other studies illustrated in Table 3 the prevalence of overweight in the UK and USA ranged from 9 to 31%. Those from China ranged from 0.2 to 11%. Prevalences in Russia and Spain are high, approaching those of UK and USA. In Trinidad and Tobago, when looking at three distinct race groups, African, Indian and mixed race descent, the pattern of overweight observed for this developing country appears similar to that of the UK, reflecting a developing country with upper middle income status, with a high reported per capita gross national product (United Nations Development Programme, 2000). South Africa, another developing country with upper middle income status based on its gross national product, does not have this pattern of overweight, however; this reflects the considerable disparities in income and nutritional status among its diverse population. For South Africa the prevalences of overweight observed appear comparable with those seen in rural China. However, in rural China boys are more likely to be overweight than girls, whereas in the South African data this appears to be reversed.

Table 3 - Comparison of the prevalence (%) of overweight (IOTF definition) from published data sources.

Full table (42K)

This study has found good measures of agreement between the WHO and IOTF definitions for overweight and obesity for South African children. However, there are variations observed nationally, provincially and at rural district level. High levels of mild stunting occurred uniformly across all children in South Africa, while there were lower levels of moderate stunting, which were not distributed uniformly, and very low levels of extreme stunting. Matching this nutritional profile of undernutrition, there were low levels of overnutrition as measured by overweight and obesity using both the WHO and IOTF cut-offs.

No clear measure of association was found between being stunted and being overweight according to WHO or IOTF definitions. However, when considering the relationship between factors such as stunting, overweight and obesity, caution must be applied as the WHO definition appears to be broader, including more stunted children in those defined as obese. Cole et al have also advised caution in the use of IOTF cut-offs for obesity since the centiles involved more extrapolation and produced greater variability than the centiles for overweight, concluding that the obesity cut-offs are imprecise.

The prevalence of overweight using IOTF is low relative to other developing and developed countries, suggesting that South Africa is at an early stage of any nutritional transition. In the nationally representative Demographic Health Survey, 18.4% of male urban Africans and 25.1% female urban Africans were found to be overweight in 1998 (Department of Health, 1999). Traditionally African countries including South Africa have focused on undernutrition and food security and have not taken sufficient cognisance of overnutrition. The results of the DHS survey clearly indicate that overweight and obesity are public health problems among adults and require interventions. To date there has been a lack of available information on overweight and obesity in children, particularly in marginalized rural populations. However, it is important to note that the data analysed in this study is more reflective of the health status of children born during the 1980s. It does not take cognisance of the rapid changes in socio-economic status which occurred in South Africa during the 1990s, which in turn may produce trends towards overnutrition.

Investigation of both undernutrition and overnutrition have important clinical and public health implications for planning Primary Health Care services and health promotion strategies to prevent chronic diseases in adulthood. Such investigations in developing countries require a shift in the focus from under-nutrition and food security, to also assessing trends in overweight and obesity (WHO, 2000). This study has found that the prevalence of overweight and obesity amongst a cohort of rural children undertaken in 1995 appeared higher than that of a national and provincial sample of school children (a mixture of urban and rural) undertaken in 1994. However, the prevalence in the 1990s of stunting would appear to be decreasing as compared to the previous decades. In the 1970s and 1980s in South Africa, a third of the black children below the age of 14 y were underweight and stunted for their age, and this increased to 60–70% in some urban (Chatsworth, Durban) and rural (Ciskei) areas (Hansen, 1989). Coexistence of stunting and overweight in school children may however reflect an aspect of the nutritional transition (a transition from a traditional African diet to a more Western-type diet), where the already stunted and underweight become both stunted and overweight, a double burden.

The measurement of overweight and obesity has been the subject of much debate. The use of BMI for older children is based on age-dependent cut-offs using the NCHS data of USA children and is not globally representative. Considerable debate continues about the appropriateness of various cut-off points and the merits of the different measures of BMI, which depend on age, height, sex and pubertal status (Mulligan, 2000). Leung (2001) also questions the argument for using an international reference for growth assessment in childhood. However, the IOTF prevalence data, using cut-off points based on an internationally representative sample of children, provide a means of comparing data and are useful for monitoring trends. Cole et al (2000) noted that an important omission in the development of their internationally representative sample of BMI curves was data from Africa. This study explores the patterns of BMI in South African children as a contribution to the existing international information.

From an epidemiological perspective, it is important to establish the measure of agreement between prevalence rates measured by the WHO recommended 85th percentiles as cut-off points for overweight and the IOTF cut-off points. Wang and Wang (2000) compared the prevalence of overweight amongst children (aged 6–9 y) and adolescents (aged 10–18 y) using the WHO and the IOTF standards, based on data from China, Russia and the USA. Using the coefficient as a measure of agreement, they found excellent agreement (0.8) between the two standards, which they concluded are comparable, and further recommended the IOTF standard for international use. Our results support these findings. Kinra (2000), however, has noted that the IOTF guidelines are based on data from six countries, five of which have a gross national product above that of most African and Asian countries. Several authors have noted that the new IOTF cut-offs are not entirely globally representative, and especially lack data from developing and middle-income countries (Moreno, 2001; Kinra, 2000). This requires the IOTF database to include representative data from more developing countries as undertaken in this study, in an attempt to construct a valid, more representative international standard. Additional data sets from such countries would contribute to making the IOTF dataset more representative.

The IOTF has noted that populations with stunted children, especially in countries undergoing rapid nutrition transition, may have an altered relationship between BMI and adiposity. In this cohort of children this link was not evident. The SANPS survey included only school-going children, omitting those not attending school, who may be more likely to be from disadvantaged households with higher levels of undernutrition, disability and illnesses. South Africa has undergone major and rapid socio-economic development, which is continuing today. These changes are likely to alter the low prevalence of overweight and obesity documented, and impact on the epidemiological and nutritional status of children. The current data, however, makes it difficult to elucidate whether the South African population is undergoing a nutritional transition. Popkin et al (1996) in analysing the nutritional status of South African pre-school children (3–6 y) using the WHO NHANES III definition, found that the prevalence of overweight (85th centile) was 25.0% for males and 20.3% for females, and that of obesity (95th centile) was 7.4% for males and 4.1% for females. The fact that younger children born in the 1990s may have a higher prevalence of overweight than children born in the 1980s may suggest the early stages of a nutritional transition. All of these factors emphasize the need to understand the nature of the nutrition transition in South Africa.

In summary, the following features have been identified in this study: firstly, lower levels of stunting in the 1990s in relation to wasting and stunting prevalences in the 1980s, suggesting a decline in severe and overt malnutrition, although high levels of mild stunting continue to occur; secondly, very low levels of obesity were found by either definition and low levels of overweight, which are comparable to those of other developing countries such as China, and lower than that of many developed countries.

In terms of a public health response to overweight and obesity the WHO has recommended a range of preventive strategies. These should be universal, selective (ie directed at high-risk groups) or targeted (ie aimed at those with weight problems) and should be undertaken at family, school and primary care level, all with the aim of increasing physical activity and improving the quality of diet (WHO, 2000). The focus needs to be on education and behaviour change to control overweight. These public health interventions are required because expensive and technologically advanced treatment for obesity would overwhelm public health resources. Dramatic increases in obesity have been followed by epidemics of non-communicable diseases, such as cardiovascular disease. This has relevance for policymakers, health planners and communities to implement appropriate actions to prevent and control the emergence of these epidemics at an early stage.

Further cross-sectional studies are thus required to establish whether the prevalence of overweight and obesity is increasing in South Africa, and whether there is any link with stunting. Longitudinal and intervention studies are required to establish the trajectory of any possible nutrition transition in terms of geography, culture and socio-economic status, in order to identify and control risk factors and assess the differential impact of stunting and overweight on children.

References

1. Bellizzi, MC & Dietz, WH (1999). Workshop on childhood obesity: summary of the discussion. Am. J. Clin. Nutr., 70, 173–175.
2. Budlender, D (2000). Human development. In:Poverty and Inequality in South Africa: Meeting the Challenge, ed. J MayLondon: David Philip and Zed Press
3. Chinn, S & Rona, RJ (2001). Prevalence and trends in overweight and obesity in three sectional studies of British children, 1974–94. Br. Med. J., 322, 24–26. | Article |
4. 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, 1–6.
5. Department of Health (1994). Anthropometric survey in primary schools in the RSA, Sub-Directorate Statistical AnalysisPretoria: DOH
6. Department of Health (1999). South Africa Demographic and Health Survey 1998. Preliminary report, Department of Health, Pretoria: DOH
7. Gulliford, MC, Mahabir, D, Rocke, B, Chinn, S & Rona, R (2001). Overweight, obesity and skinfold thickness of children of African or Indian descent in Trinidad and Tobago. Int. J. Epidemiol., 30, 989–998. | PubMed |
8. Hansen, J (1989). Food and nutrition policy with relation to poverty: the child malnutrition problem in South Africa. In:Uprooting Poverty, the South African Challenge. Report for the Second Carnegie Inquiry into Poverty and Development in Southern Africa, ed. F Wilson & M RampheleJohannesburg: David Philip Publishers
9. Hesketh, T & Ding, QJ (2000). Young Chinese people in Hong Kong are not representative of those in China. (Letter to Editor.). Br. Med. J., 321, 1158
10. Jinabhai, CC, Taylor, M, Coutsoudis, A, Coovadia, HM, Tomkins, AM & Sullivan, KR (2001). A health and nutritional profile of rural school children in KwaZulu-Natal, South Africa. Ann. Trop. Paediatr., 21, 50–58.
11. Jinabhai, CC, Taylor, M, Coutsoudis, A, Coovadia, HM, Tomkins, AM & Sullivan, KR (2001). A randomised controlled trial of the effect of anthelminthic treatment and micronutrient fortification on health status and school performance of rural primary school children. Ann. Trop. Paediatr., 21, 319–333.
12. Kinra, S (2000). Reanalysis suggests questionable validity of new definition during puberty. (Letter to Editor.). bmj.com, 29 May
13. Leung, SSF (2000). Establishing a standard definition for child overweight and obesity worldwide. (Letter to Editor.). bmj.com, 20 August
14. May, J(ed.) (2000). Poverty and Inequality in South Africa: Meeting the Challenge, London: David Philip and Zed Press
15. Monyeki, KD, van Lenthe, FJ & Steyn, NP (1999). Obesity: does it occur in African children in a rural community in South Africa. Int. J. Epidemiol., 28, 287–292.
16. Moreno, LA, Fleta, J, Sarria, A & Bueno, (2001). Prevalence and trends in childhood obesity in Zaragoza (Spain). (Letter to the Editor.). bmj.com, 2 February
17. Mulligan, J (2000). Body Mass Index is harder to interpret in children than adults. (Letter to Editor). Br. Med. J., 321, 1158
18. Popkin, B (1994). The nutrition transition in low-income countries: an emerging crisis. Nutr. Rev., 52, 285–298. | PubMed | ISI | ChemPort |
19. Popkin, BM, Richards, MK & Montiero, CA (1996). Stunting is associated with overweight in children of four nations that are undergoing the nutrition transition. J. Nutr., 126, 3009–3016. | PubMed | ChemPort |
20. United Nations Development Programme (2000). Human Development Indicators, www.undp.org/hdr/2000/english/book/back1.pdf
21. Wang, Y & Wang, J (2000). Standard definition of child overweight and obesity worldwide. (Letter to Editor.). Br. Med. J., 321, 1158
22. WHO (1995). Physical status: the use and interpretation of anthropometry, Report of a WHO Expert Committee. Technical Report Series 854Geneva: WHO
23. WHO (2000). Obesity: preventing and managing the global epidemic, Report of a WHO Consultation. World Health Organization, Technical Report Series 894 Geneva: WHO