Hostname: page-component-848d4c4894-8kt4b Total loading time: 0 Render date: 2024-06-17T19:11:52.378Z Has data issue: false hasContentIssue false

Predictors of adolescent weight status and central obesity in rural South Africa

Published online by Cambridge University Press:  28 February 2011

Elizabeth W Kimani-Murage*
Affiliation:
MRC/Wits Rural Public Health and Health Transitions Research Unit (Agincourt), School of Public Health, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa African Population and Health Research Centre (APHRC), PO Box 10787, 00100 Nairobi, Kenya
Kathleen Kahn
Affiliation:
MRC/Wits Rural Public Health and Health Transitions Research Unit (Agincourt), School of Public Health, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa Umeå Centre for Global Health Research, Department of Public Health and Clinical Medicine, Umeå University, Sweden INDEPTH Network, Accra, Ghana
John M Pettifor
Affiliation:
MRC Mineral Metabolism Research Unit, Department of Paediatrics, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa
Stephen M Tollman
Affiliation:
MRC/Wits Rural Public Health and Health Transitions Research Unit (Agincourt), School of Public Health, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa Umeå Centre for Global Health Research, Department of Public Health and Clinical Medicine, Umeå University, Sweden INDEPTH Network, Accra, Ghana
Kerstin Klipstein-Grobusch
Affiliation:
School of Public Health, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa
Shane A Norris
Affiliation:
MRC Mineral Metabolism Research Unit, Department of Paediatrics, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa
*
*Corresponding author: Email lizmurage@gmail.com; ekimani@aphrc.org
Rights & Permissions [Opens in a new window]

Abstract

Objective

To investigate predictors of adolescent obesity in rural South Africa.

Design

Cross-sectional study. Height, weight and waist circumference were measured using standard procedures. Overweight and obesity in adolescents aged 10–17 years were assessed using the International Obesity Taskforce cut-offs, while the WHO adult cut-offs were used for participants aged 18–20 years. Waist-to-height ratio of >0·5 defined central obesity in those at Tanner stages 3–5. Linear and logistic regression analysis was used to evaluate risk factors.

Setting

Agincourt sub-district, rural South Africa.

Subjects

Participants (n 1848) were aged 10–20 years.

Results

Combined overweight and obesity was higher in girls (15 %) than boys (4 %), as was central obesity (15 % and 2 %, respectively). With regard to overweight/obesity, fourfold higher odds were observed for girls and twofold higher odds were observed for participants from households with the highest socio-economic status (SES). The odds for overweight/obesity were 40 % lower if the household head had not completed secondary level education. For central obesity, the odds increased 10 % for each unit increase in age; girls had sevenfold higher odds v. boys; post-pubertal participants had threefold higher odds v. pubertal participants; those with older mothers aged 50+ years had twofold higher odds v. those whose mothers were aged 35–49 years; those in highest SES households had twofold higher odds v. those in lowest SES households.

Conclusions

In rural South Africa, adolescent females are most at risk of obesity which increases with age and appears to be associated with higher SES. To intervene effectively, it is essential to understand how household factors influence food choice, diet and exercise.

Type
Research paper
Copyright
Copyright © The Authors 2011

Obesity is increasingly a public health concern in both high-income and low- to middle-income countries (LMIC), but the prevalence is expected to increase more rapidly in LMIC(Reference Popkin and Gordon-Larsen1, 2). Obesity in these countries is associated with a rise in chronic diseases such as hypertension, CHD, type 2 diabetes and certain types of cancer, which contribute substantially to the burden of disease(2). Childhood and adolescent obesity is of particular concern, as obese children and adolescents are more likely to become obese adults and hence are at a higher risk of metabolic diseases(Reference Bibbins-Domingo, Coxson and Pletcher3, Reference Dietz4). In South Africa, high levels of overweight/obesity have been documented among adults(5). There is also some evidence of high levels of overweight/obesity among children and adolescents(Reference Jinabhai, Reddy and Taylor6, Reference Labadarios, Swart and Maunder7). In a recent study in Agincourt, rural South Africa, we described patterns of overweight/obesity and central obesity among children and adolescents aged 1–20 years. The prevalence of combined overweight and obesity was high in adolescent girls, reaching a peak of 25 % at age 18 years, while central obesity was also high in adolescent girls, increasing with pubertal development and reaching a peak of 35 % in girls at Tanner stage 5(Reference Kimani-Murage, Kahn and Pettifor8).

Predictors of child and adolescent obesity have been described in various studies(Reference Griffiths, Rousham and Norris9, Reference Davison and Birch10) and seem to involve a complex array of influential domains: community, family/household, individual and genetic. In a literature review conducted in 2001, Davison and Birch(Reference Davison and Birch10) described an ecological model of predictors of child weight status. In this model, child characteristics interact with familial/household and societal/community factors to influence weight status in a child. The child's behavioural patterns including dietary intake, sedentary behaviours and physical activity predict his/her weight status. Furthermore, child characteristics including sex, age and genetic susceptibility to gaining weight moderate the impact of child behavioural patterns on the development of obesity. Family/household factors such as socio-economic factors, parenting styles, parental dietary patterns and sibling interactions influence a child's weight status through shaping the development of child behavioural patterns(Reference Griffiths, Rousham and Norris9, Reference Davison and Birch10). Community and wider societal and environmental factors influence the weight of the child through influencing the child's behavioural patterns and family factors(Reference Davison and Birch10).

There are limited studies describing the predictors of child and adolescent obesity in rural South Africa(Reference Kruger, Kruger and Macintyre11). The present study aimed to assess associations between child-, maternal-, household- and community-level factors and weight status and central obesity among adolescents aged 10–20 years, randomly selected from a rural South African setting.

Study setting and population

The study was conducted in the Agincourt sub-district of Mpumalanga Province in rural north-east South Africa, close to the country's border with Mozambique. It was nested within the Agincourt health and sociodemographic surveillance system (HDSS). The Agincourt HDSS is a multi-round prospective community study established in 1992, covering the entire Agincourt sub-district. Until 2007 when the site was extended, the HDSS followed some 70 000 people living in 11 500 households in twenty-one villages, about 30 % of whom were of Mozambican origin. The Mozambicans entered South Africa mainly as refugees in the early to mid-1980s following the civil war in Mozambique. Despite voluntary repatriation programmes, most elected to remain as so-called self-settled former refugees. They are also Tsonga-speaking, have much in common culturally and have integrated to large extent with the host South African population. Despite this, the Mozambicans remain a vulnerable subgroup within the broader community(Reference Hargreaves, Collinson and Kahn12, Reference Dolan, Tollman and Nkuna13).

While there have been government development initiatives, infrastructure remains limited in Agincourt. The area is dry with household plots too small to support subsistence farming. The poverty level is high(Reference Gelb14) and labour migration is widespread, involving up to 60 % of working age men and increasing numbers of women(Reference Collinson, Tollman and Kahn15, Reference Collinson, Tollman and Wolff16). The Agincourt HDSS and study area have been described in detail previously(Reference Kimani-Murage, Kahn and Pettifor8, Reference Kahn, Tollman and Collinson17).

Data and methods

We conducted a growth survey between April and July 2007. The survey targeted 4000 children and adolescents aged 1–20 years – 100 males and 100 females for each year of age – comprising about 12 % of the total Agincourt HDSS population within this age spectrum. Participants had lived in the study area at least 80 % of the time since birth, or since 1992 when enrolment in the Agincourt HDSS began. A total of 3511 children and adolescents (close to 80 % of the sample) participated in the study(Reference Kimani-Murage, Kahn and Pettifor8). All survey participants aged 10–20 years (n 1848) were included in the present analysis.

Measurements

Anthropometric measurements were carried out on all participants while pubertal assessment was obtained from those aged 9–20 years. Height was measured using a stadiometer (Holtain, Crymych, UK) calibrated in millimetres. Weight in kilograms (to one decimal point) was determined using a mechanical bathroom scale (Hanson; Bathroom Trends Limited, Epsom, UK), and waist circumference was measured in millimetres using an inelastic tape measure. All measurements were taken according to standard procedures(Reference Kimani-Murage, Kahn and Pettifor8, Reference Lohman, Roche and Martorell18). Adolescents self-rated their pubertal stage using the Tanner 5-point pubertal self-rating scale that reflects physical development based on external primary and secondary sex characteristics(Reference Tanner19). This self-rating scale has been validated for black South Africans(Reference Norris and Richter20). Genital development in boys and breast development in girls were used to define the stages in the present study. Data collection procedures are detailed in a previous publication(Reference Kimani-Murage, Kahn and Pettifor8).

Explanatory variables

Explanatory variables used in the study were obtained from the Agincourt HDSS. This involves systematic annual recording of all births, deaths and migration events occurring in the Agincourt sub-district since 1992. Additional data are collected as special census modules nested within the annual update rounds. These include education, child social grant uptake, union status and food security. An asset survey conducted in each household every two years provides a relative measure of household socio-economic status (SES)(Reference Kahn, Tollman and Collinson17).

Child-level factors included age, sex and pubertal development stage; maternal-level factors included mother's age, nationality, highest education level, marital/union status and co-residence with the child; household-level factors included age, sex and highest education level of the household head and his/her relationship to the child, household food security and household SES; area of residence (predominantly South African or Mozambican) was used as a proxy for community-level characteristics.

Definitions of the explanatory variables are provided in Table 1. Pubertal stage was constructed with Tanner stage 1, 2–4 and 5 defined as pre-pubertal, pubertal and post-pubertal, respectively. Food insecurity was defined as reporting not having enough food to eat in either the last month or the last year, whereas food-secure households were those households reporting sufficient to eat both in the last month and last year. Household wealth index was constructed from household assets(Reference Filmer and Pritchett21) including type and size of dwelling; water and sanitation facilities; electricity; modern assets such as a fridge and television; transport assets such as a car; communication assets such as a telephone; and livestock such as cattle. A composite score for the wealth indicator was constructed. Household wealth tertiles were generated from the composite SES score using STATA's xtile command and labelled as ‘lowest’ (lowest third), ‘medium’ and ‘highest’ (highest third).

Table 1 Distribution of combined overweight and obesityFootnote * for adolescents aged 10–20 years (n 1848) and central obesityFootnote for adolescents in Tanner stages 3–5 (n 1110) by various covariates, Agincourt sub-district, South Africa, 2007

HHH, household head.

* International Obesity Taskforce criteria(Reference Cole, Bellizzi and Flegal23) for those <18 years and BMI ≥ 25 kg/m2 for those ≥18 years.

Waist-to-height ratio of >0·5(Reference Ashwell24).

Outcome measures

Outcome measures included BMI-for-age Z-score, waist-to-height ratio Z-score, combined overweight and obesity, and central obesity. BMI-for-age Z-scores were generated using the WHO 2007 reference, using the WHO AnthroPlus software (WHO, Geneva, Switzerland)(22). Age- and sex-specific waist-to-height ratio Z-scores were sample specific. Overweight and obesity in children aged 10–17 years were determined using the absolute age- and sex-specific cut-offs for BMI recommended by the International Obesity Taskforce(Reference Cole, Bellizzi and Flegal23). These are defined to pass through BMI of 25 and 30 kg/m2 at 18 years for overweight and obesity, respectively. For adolescents aged 18–20 years, adult cut-off points for BMI of ≥25 and ≥30 kg/m2 for overweight and obesity, respectively, were used. Overweight and obesity were combined in the analysis owing to the limited number of participants who were classified as obese. Waist-to-height ratio cut-off of >0·5 was used to determine the presence of central obesity in both boys and girls in Tanner stages 3–5(Reference Ashwell24).

Underweight status was determined among the participants to assess its influence on the outcomes. This was defined by the 5th percentile using the WHO 2007 reference for BMI-for-age Z-scores(22, 25). About 4 % of the study participants were classified as underweight. However most of those underweight were in lower ages and were in pubertal stages 1 and 2, indicating that they were still growing. Exclusion of these children altered the results only marginally; thus they were retained in the final analysis.

Ethical clearance

For the growth study, signed informed consent was obtained from the parent/caregiver for children and adolescents aged less than 18 years and from the adolescents aged 18–20 years themselves, while assent was also obtained from adolescents aged less than 18 years. Verbal informed consent is routinely obtained from participants in the Agincourt HDSS. Ethical clearances for the growth survey ( M070244) and the Agincourt HDSS ( M960720) were granted separately by the University of the Witwatersrand Committee for Research on Human Subjects (Medical).

Statistical analysis

Data analysis was undertaken using the STATA statistical software package version 10·0 (StataCorp LP, College Station, TX, USA). Student's t test was used to determine difference in means, while the χ 2 test was used to determine difference in proportions. Univariate and multiple linear and logistic regression analyses were conducted with outcome variables (BMI-for-age Z-score, waist-to-height ratio Z-score, overweight/obesity, central obesity) and explanatory variables as described above, respectively. Missing values for each variable were allocated an independent category in the regression models to maintain all participants in the analysis. The variables affected included: mother's age (9 %), nationality (9 %) and education (21 %); household head's education (11 %) and relationship with the child (<1 %); food security (4 %) and SES (2 %). Only explanatory variables significantly associated with the outcome variable at the 10 % significance level in the univariate analysis were included in the multiple regression analysis. A two-sided P value of <0·05 was considered statistically significant.

Results

The study sample comprised 903 (49·6 %) boys and 945 (50·5 %) girls aged 10–20 years; mean age was 14·6 (95 % CI 14·43, 14·83) years. There was no significant difference in mean age by sex.

Distribution of overweight/obesity and central obesity by explanatory variables

The prevalence of combined overweight and obesity among participants aged 10–20 years was 10 %, while that of central obesity among participants in Tanner stages 3–5 was also 10 %. The following categories had significantly higher prevalence of either combined overweight and obesity or central obesity: 15- to 20-year-olds v. 10- to 14-year-olds; girls v. boys; pubertal and post-pubertal v. pre-pubertal; those with older mothers; those with mothers of South African origin v. those with mothers of Mozambican origin or those who lived in villages inhabited predominantly by people of South African origin; those with the most educated household heads; and those from the highest SES households (all P < 0·05; Table 1).

Multivariate analysis

Predictors of BMI-for-age Z-scores, waist-for-height ratio Z-scores, overweight/obesity and central obesity are presented in Tables 2 and 3. Only variables that were significantly associated with the outcome measures at the 10 % level of significance from univariate analysis are included in the tables. A test for collinearity among the variables included in the multivariate analysis was done; there was no significant collinearity. Additionally, interaction was tested between various variables significant at the univariate level; no significant interactions were found. Significant predictors of adolescent weight status and central obesity at the multivariate level were at child, maternal and household levels.

Table 2 Determinants of BMI-for-age Z-score and combined overweight and obesityFootnote * in adolescents aged 10–20 years (n 1848), Agincourt sub-district, rural South Africa, 2007

HHH, household head; ref., reference category.

* International Obesity Taskforce criteria(Reference Cole, Bellizzi and Flegal23) for those <18 years and BMI ≥ 25 kg/m2 for those ≥18 years.

Linear regression.

Logistic regression.

Table 3 Determinants of waist-for-height ratio Z-score for adolescents aged 10–20 years (n 1848) and central obesityFootnote * for adolescents in Tanner stages 3–5 (n 1110), Agincourt sub-district, rural South Africa, 2007

HHH, household head; ref., reference category.

* Waist-to-height ratio of >0·5(Reference Ashwell24).

Linear regression.

Logistic regression.

§ Pre-pubertal used as reference in the model for central obesity and waist-for-height Z-score for adolescents aged 10–20 years.

|| Pubertal used as reference in the model for central obesity for adolescents in Tanner stages 3–5.

Child factors

All of the child-level factors examined (age, sex and pubertal development status) emerged as significant predictors of adolescent weight status and central obesity. After controlling for other covariates, for every increase in year of age, the odds of central obesity increased by 10 % (P = 0·041). Compared with boys, girls had higher BMI-for-age and waist-to-height ratio Z-scores, had fourfold higher odds of being overweight or obese, and sevenfold higher odds of having central obesity (all P < 0·001). Compared with pre-pubertal adolescents, pubertal and post-pubertal adolescents had higher BMI-for-age Z-scores and post-pubertal adolescents had higher waist-to-height ratio Z-scores; while post-pubertal adolescents had fourfold higher odds of overweight/obesity compared with pre-pubertal adolescents and threefold higher odds of having central obesity compared with pubertal adolescents (all P < 0·001; Tables 2 and 3).

Maternal factors

Among maternal factors included, only the mother's age was a significant determinant. Adolescents of older mothers aged 50+ years had close to twofold higher odds of central obesity compared with adolescents whose mothers were aged 35–49 years (P = 0·020; Table 3).

Household factors

After controlling for other covariates, significant household-level predictors included household head's highest education level, food security and SES. Compared with adolescents in households headed by someone with no formal education, adolescents in households where the household head had education below secondary level certificate had lower BMI-for-age Z-scores and had 40 % lower odds of being overweight/obese (all P < 0·05). Adolescents in food-secure households had higher waist-to-height ratio Z-scores (P = 0·046). Compared with adolescents in the lowest SES households, those in the medium and the highest SES households had higher BMI-for-age Z-scores (both P ≤ 0·001), while those from the highest SES households had higher waist-to-height ratio Z-scores (P = 0·031). Similarly, those from the highest SES households had about twofold higher odds of overweight/obesity and central obesity (both P < 0·05; Tables 2 and 3).

Discussion

The present study has described key predictors of weight status and central obesity among adolescents aged 10–20 years living in rural South Africa. The study has demonstrated that significant predictors are at child level including age, sex and pubertal development; at maternal level including mother's age; and at household level including highest education of the household head, food security status and SES. Identifying context-specific predictors of the increasingly important problem of child obesity is an important step in its prompt containment given its link to the risk for paediatric metabolic diseases in LMIC(Reference Kelishadi26, Reference Reilly, Methven and McDowell27).

The study reports a substantial prevalence of overweight/obesity particularly among girls, which is consistent with other findings in South Africa(Reference Reddy, Resnicow and James28). This is also consistent with findings in other LMIC undergoing a nutrition transition(Reference Kelishadi26, Reference Neutzling, Taddei and Rodrigues29). Substantial levels of child overweight/obesity have been reported in Africa particularly in North African countries such as Egypt, in other southern African countries, and even in countries as poor as Malawi(Reference de Onis and Blossner30).

Consistent with other studies(Reference Davison and Birch10), child-level factors emerged as key predictors of adolescent weight status and central obesity. Obesity increased with increase in age and was associated with pubertal development. This may reflect the effect of factors such as increased sedentary behaviour and decreased physical activity with age and pubertal onset(Reference Davison and Birch10, Reference Hardy, Bass and Booth31). These changes are explained by physical, social and emotional changes(Reference Lindquist, Reynolds and Goran32). However, sedentary behaviour and physical activity, which could have influenced the outcomes, were not measured in the present study. The increased risk of obesity among girls in our study is in keeping with studies in other LMIC(Reference Neutzling, Taddei and Rodrigues29) and in South Africa in particular(Reference Jinabhai, Reddy and Taylor6). Several factors may explain the sex differences in obesity: (i) biologically, energy needs differ for boys and girls and also in relation to rate of growth(Reference Wisniewski and Chernausek33); and (ii) behaviourally, boys are generally more physically active than girls especially during adolescence(Reference Kruger, Kruger and Macintyre11, Reference Lindquist, Reynolds and Goran32, Reference Reddy, Panday and Swart34). Concerns about body image, particularly among adolescent girls, can lead to problematic eating behaviours such as irregular meal patterns which may result in increased weight gain(Reference Neumark-Sztainer, Paxton and Hannan35). Differential problematic eating behaviours by sex have been reported among youth in South Africa(Reference le Grange, Telch and Tibbs36).

While many studies have documented the effect of mother's age on child undernutrition, little has been documented on the effect of mother's age on adolescent obesity. The present study indicates that adolescents with older mothers are significantly more likely to be centrally obese. Although this needs further study, we postulate that this may be associated with less knowledge of the adverse health effects of obesity in the older age group, particularly given their lower literacy levels in the study community(Reference Collinson37), and a stronger adherence to the cultural value ascribed to larger women. This may lead to practices that promote obesity since studies indicate that parents serve as role models for their children; for example, children often prefer foods eaten by their parents. Additionally, older mothers may monitor their child's behaviour less, which may affect the child's diet and physical activity patterns(Reference Davison and Birch10).

The household-level predictors of obesity observed in the present study include household head's highest education level, food security and SES. Education may affect nutritional status through knowledge of a healthy diet and of the harmful effects of overnutrition. It may also affect income levels leading to sedentary lifestyles, availability of food and dietary changes. In our study, education less than secondary certificate was protective, while secondary and tertiary education – levels that could lead to higher SES – were not significantly associated with obesity. There was, however, a tendency towards a positive association. There seems, therefore, to be a mixed effect: while on the one hand education is protective, it also leads to higher BMI and central obesity. However, the household head's education level was observed to be significantly associated with SES (although there was no significant collinearity); this may indicate attenuation of each of these factors in the overall effect.

Available literature on the association between food (in)security and childhood/adolescent overweight/obesity is generally conflicting(Reference Kursmark and Weitzman38, Reference Amuna and Zotor39). In high-income countries, food insecurity is often associated with child and adolescent overweight/obesity(Reference Kursmark and Weitzman38) because the poor may opt for cheaper, energy-dense processed foods(Reference Drewnowski40). However, there are also studies that indicate otherwise(Reference Amuna and Zotor39). Similarly, in the limited literature from LMIC, some studies have found a positive association between food insecurity and overweight/obesity, and others the opposite(Reference Kursmark and Weitzman38). The Agincourt study area, a former homeland, has low local food production; hence food security may relate strongly to the ability to afford purchased (processed) food.

The relationship between SES and obesity varies across different countries depending on economic development(Reference Wang41). The pathways through which SES is associated with overnutrition include income, education and occupation, resulting in behaviours which influence the balance between energy intake, expenditure and metabolism(Reference Sobal42). The positive relationship found in the present study is in keeping with several other studies in LMIC(Reference Griffiths, Rousham and Norris9, Reference Neutzling, Taddei and Rodrigues29, Reference Wang41), but contradicts findings in high-income countries(Reference Wang41). However, even in high-income countries such as the USA the reverse association between SES and overnutrition is said to be weakening over time and is population specific, particularly in children(Reference Wang and Zhang43, Reference Matijasevich, Victora and Golding44). Although South Africa is classified as a middle-income country, there is high intra-country inequality with regard to income distribution and high levels of poverty in some regions. The study area is low-income, located in one of the three provinces with the highest poverty rates in South Africa(Reference Gelb14). The relationship observed may be associated with the sedentary lifestyle and ability to afford processed foods of those in the highest SES category relative to those in the lowest SES category. Results not shown on the assets that constitute the socio-economic score indicate that while almost all of the highest SES households owned a television (96 %), slightly below 50 % of the lowest SES did; while 35 % of the highest SES households owned a car, only 7 % of the lowest SES households did. SES and food security variables were significantly associated, although the association was not significantly collinear. This may indicate attenuation of the overall effect of each of these variables.

Extensive labour migration to larger towns outside the study area(Reference Collinson, Tollman and Kahn15, Reference Collinson, Tollman and Wolff16) facilitates the transfer and introduction of urban practices to rural settings with consequent change in diet, resulting in substitution of traditional foods with energy-dense processed foods as seen in some settings in South Africa(Reference Vorster, Venter and Wissing45). While those in higher SES households may afford to purchase such foods, the poorest may not and would therefore rely on limited home-grown produce and wild foods. Consumption of animal products and energy-dense foods has been associated with overweight/obesity in countries undergoing nutrition transition(Reference Popkin and Gordon-Larsen1).

The present study did not address the association between weight status and central obesity and their more proximate determinants; i.e. imbalance between energy intake and expenditure such as dietary patterns and physical activity. However, there is little controversy concerning these factors. The study therefore focused on the more distal determinants which give rise to the imbalance and for which current evidence is more limited. A further limitation relates to the measurement of food security, which was designed primarily as a simple tool to measure trends in household food security in the Agincourt health and sociodemographic surveillance area over time rather than to detail multiple dimensions of food security.

Our study has identified key predictors of adolescent weight status and central obesity in rural South Africa. The findings show that female sex, pubertal development, being in a household that is food secure and of higher SES (measured using household assets) are the most important predictors of weight status and central obesity. These findings indicate a need for gender-sensitive intervention strategies that take into consideration pubertal development, relative wealth and related behaviours in curbing the rising problem of child and adolescent obesity. Further research into dietary patterns and physical activity among adolescents in the study area and related settings, and their associations with child- and household-level characteristics, is necessary for effective interventions.

Acknowledgements

Sources of funding: Funding from the National Research Foundation (NRF) and the Medical Research Council (MRC), South Africa, is acknowledged. The Agincourt health and sociodemographic surveillance system is funded by the Wellcome Trust, UK (069683/Z/02/Z) and is a member of the INDEPTH Network. E.W.K.-M. had a PhD fellowship funded by the Flora and William Hewlett Foundation, USA, while S.A.N. was on a Wellcome Trust-funded fellowship. Conflict of interest: The authors have no conflict of interest. Authors’ contributions: E.W.K.-M. contributed to study design, project implementation and management, data analysis and writing of the manuscript. K.K. contributed to study design, overall project co-ordination and review of the manuscript. J.M.P. contributed to study design and review of the manuscript. S.M.T. contributed to study design and review of the manuscript. K.K.-G. contributed to analytic guidance and review of the manuscript. S.A.N. contributed to study design, overall project management, analytic guidance and review of the manuscript. All authors read and approved the final manuscript for submission. Acknowledgments: We acknowledge logistical support from the MRC/Wits Rural Public Health and Health Transitions Research Unit (Agincourt) during data collection and from the African Population and Health Research Center, Kenya, during preparation of the manuscript. We thank Dr Mark Collinson, Dr Xavier Gomez-Olivé and Professor David Dunger for their technical contribution during design of the study. We are grateful to the data collection team and the LINC office team at the MRC/Wits-Agincourt Unit, specifically Rhian Twine, Jeffrey Tibane and Audrey Khosa for their role in community mobilisation. We also acknowledge the training team and the data entry team from the Birth-to-Twenty Programme, University of the Witwatersrand, South Africa, funded by the Wellcome Trust, UK (077210/Z/05/Z).

References

1. Popkin, BM & Gordon-Larsen, P (2004) The nutrition transition: worldwide obesity dynamics and their determinants. Int J Obes Relat Metab Disord 28, Suppl. 3, S2S9.CrossRefGoogle ScholarPubMed
2. World Health Organization (2002) The World Health Report 2002: Reducing Risks, Promoting Healthy Life. Geneva: WHO; available at http://www.who.int/whr/2002/en/whr02_en.pdfGoogle Scholar
3. Bibbins-Domingo, K, Coxson, P, Pletcher, MJ et al. (2007) Adolescent overweight and future adult coronary heart disease. N Engl J Med 357, 23712379.CrossRefGoogle Scholar
4. Dietz, WH (1998) Childhood weight affects adult morbidity and mortality. J Nutr 128, Suppl 2, S411S414.CrossRefGoogle ScholarPubMed
5. Department of Health, Medical Research Council & ORC Macro (2007) South African Demographic and Health Survey 2003. Pretoria: Department of Health; available at http://www.measuredhs.com/pubs/pdf/FR206/FR206.pdfGoogle Scholar
6. Jinabhai, CC, Reddy, P, Taylor, M et al. (2007) Sex differences in under and over nutrition among school-going Black teenagers in South Africa: an uneven nutrition trajectory. Trop Med Int Health 12, 944952.CrossRefGoogle ScholarPubMed
7. Labadarios, D, Swart, R, Maunder, EMW et al. (2008) National Food Consumption Survey–Fortification Baseline (NFCS-FB-I): South Africa, 2005. S Afr J Clin Nutr 21, Suppl. 2, 245300.Google Scholar
8. Kimani-Murage, E, Kahn, K, Pettifor, J et al. (2010) The prevalence of stunting, overweight and obesity, and metabolic disease risk in rural South African children. BMC Public Health 10, 158.CrossRefGoogle ScholarPubMed
9. Griffiths, PL, Rousham, EK, Norris, SA et al. (2008) Socio-economic status and body composition outcomes in urban South African children. Arch Dis Child 93, 862867.CrossRefGoogle ScholarPubMed
10. Davison, KK & Birch, LL (2001) Childhood overweight: a contextual model and recommendations for future research. Obes Rev 2, 159171.CrossRefGoogle ScholarPubMed
11. Kruger, R, Kruger, HS & Macintyre, UE (2006) The determinants of overweight and obesity among 10- to 15-year-old schoolchildren in the North West Province, South Africa – the THUSA BANA (Transition and Health during Urbanisation of South Africans; BANA, children) study. Public Health Nutr 9, 351358.CrossRefGoogle ScholarPubMed
12. Hargreaves, JR, Collinson, MA, Kahn, K et al. (2004) Childhood mortality among former Mozambican refugees and their hosts in rural South Africa. Int J Epidemiol 33, 12711278.CrossRefGoogle ScholarPubMed
13. Dolan, CG, Tollman, SM, Nkuna, VG et al. (1997) The links between legal status and environmental health: a case study of Mozambican refugees and their hosts in the Mpumalanga (Eastern Transvaal) Lowveld, South Africa. Health Hum Rights 2, 6284.CrossRefGoogle ScholarPubMed
14. Gelb, S (2003) Inequality in South Africa: nature, causes and responses. http://www.commerce.uct.ac.za/Research_Units/DPRU/DPRU-Conference2004/Papers/Gelb_Inequality_in_SouthAfrica.pdf (accessed February 2010).Google Scholar
15. Collinson, MA, Tollman, SM & Kahn, K (2007) Migration, settlement change and health in post-apartheid South Africa: triangulating health and demographic surveillance with national census data. Scand J Public Health 35, Suppl. 69, S77S84.CrossRefGoogle Scholar
16. Collinson, MA, Tollman, SM, Wolff, B et al. (2006) Trends in internal labour migration from rural Limpopo Province, male risk behaviour, and implications for the spread of HIV/AIDS in rural South Africa. J Ethn Migr Stud 32, 633648.CrossRefGoogle ScholarPubMed
17. Kahn, K, Tollman, SM, Collinson, MA et al. (2007) Research into health, population and social transitions in rural South Africa: data and methods of the Agincourt Health and Demographic Surveillance System. Scand J Public Health 35, Suppl. 69, S8S20.CrossRefGoogle Scholar
18. Lohman, TG, Roche, AF & Martorell, R (1991) Anthropometric Standardization Reference Manual. Champaign, IL: Human Kinetics Books.Google Scholar
19. Tanner, JM (1962) Growth at Adolescence. Oxford: Blackwell.Google Scholar
20. Norris, SA & Richter, LM (2005) Usefulness and reliability of Tanner pubertal self-rating to urban black adolescents in South Africa. J Res Adolesc 15, 609624.CrossRefGoogle Scholar
21. Filmer, D & Pritchett, LH (2001) Estimating wealth effects without expenditure data – or tears: an application to educational enrollments in states of India. Demography 38, 115132.Google ScholarPubMed
22. World Health Organization (2009) WHO AnthroPlus for Personal Computers Manual: Software for Assessing Growth of the World's Children and Adolescents. Geneva: WHO.Google Scholar
23. Cole, TJ, Bellizzi, MC, Flegal, KM et al. (2000) Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ 320, 12401243.CrossRefGoogle ScholarPubMed
24. Ashwell, M (2009) Obesity risk: importance of the waist-to-height ratio. Nurs Stand 23, 4954.CrossRefGoogle ScholarPubMed
25. World Health Organization (1995) Physical Status: The Use and Interpretation of Anthropometry. Report of a WHO Expert Committee. WHO Technical Report Series no. 854. Geneva: WHO.Google Scholar
26. Kelishadi, R (2007) Childhood overweight, obesity, and the metabolic syndrome in developing countries. Epidemiol Rev 29, 6276.CrossRefGoogle ScholarPubMed
27. Reilly, JJ, Methven, E, McDowell, ZC et al. (2003) Health consequences of obesity. Arch Dis Child 88, 748752.CrossRefGoogle ScholarPubMed
28. Reddy, S, Resnicow, K, James, S et al. (2009) Underweight, overweight and obesity among South African adolescents: results of the 2002 National Youth Risk Behaviour Survey. Public Health Nutr 12, 203207.Google ScholarPubMed
29. Neutzling, MB, Taddei, JA, Rodrigues, EM et al. (2000) Overweight and obesity in Brazilian adolescents. Int J Obes Relat Metab Disord 24, 869874.CrossRefGoogle ScholarPubMed
30. de Onis, M & Blossner, M (2000) Prevalence and trends of overweight among preschool children in developing countries. Am J Clin Nutr 72, 10321039.CrossRefGoogle ScholarPubMed
31. Hardy, LL, Bass, SL & Booth, ML (2007) Changes in sedentary behavior among adolescent girls: a 2.5-year prospective cohort study. J Adolesc Health 40, 158165.CrossRefGoogle Scholar
32. Lindquist, CH, Reynolds, KD & Goran, MI (1999) Sociocultural determinants of physical activity among children. Prev Med 29, 305312.CrossRefGoogle ScholarPubMed
33. Wisniewski, AB & Chernausek, SD (2009) Gender in childhood obesity: family environment, hormones, and genes. Gend Med 6, Suppl. 1, S76S85.CrossRefGoogle ScholarPubMed
34. Reddy, SP, Panday, S, Swart, D et al. (2003) Umthenthe Uhlaba Usamila – The South African Youth Risk Behaviour Survey 2002. Cape Town: South African Medical Research Council; available at http://www.mrc.ac.za/healthpromotion/reports.htmGoogle Scholar
35. Neumark-Sztainer, D, Paxton, SJ, Hannan, PJ et al. (2006) Does body satisfaction matter? Five-year longitudinal associations between body satisfaction and health behaviors in adolescent females and males. J Adolesc Health 39, 244251.CrossRefGoogle ScholarPubMed
36. le Grange, D, Telch, CF & Tibbs, J (1998) Eating attitudes and behaviors in 1,435 South African Caucasian and non-Caucasian college students. Am J Psychiatry 155, 250254.CrossRefGoogle Scholar
37. Collinson, MA (2009) Striving against adversity: the dynamics of migration, health and poverty in rural South Africa. PhD Thesis, Umeå University.CrossRefGoogle Scholar
38. Kursmark, M & Weitzman, M (2009) Recent findings concerning childhood food insecurity. Curr Opin Clin Nutr Metab Care 12, 310316.CrossRefGoogle ScholarPubMed
39. Amuna, P & Zotor, FB (2008) Epidemiological and nutrition transition in developing countries: impact on human health and development. Proc Nutr Soc 67, 8290.CrossRefGoogle ScholarPubMed
40. Drewnowski, A (2003) Fat and sugar: an economic analysis. J Nutr 133, Suppl. 3, S838S840.CrossRefGoogle ScholarPubMed
41. Wang, Y (2001) Cross-national comparison of childhood obesity: the epidemic and the relationship between obesity and socioeconomic status. Int J Epidemiol 30, 11291136.CrossRefGoogle ScholarPubMed
42. Sobal, J (1991) Obesity and socioeconomic status: a framework for examining relationships between physical and social variables. Med Anthropol 13, 231247.CrossRefGoogle ScholarPubMed
43. Wang, Y & Zhang, Q (2006) Are American children and adolescents of low socioeconomic status at increased risk of obesity? Changes in the association between overweight and family income between 1971 and 2002. Am J Clin Nutr 84, 707716.CrossRefGoogle ScholarPubMed
44. Matijasevich, A, Victora, CG, Golding, J et al. (2009) Socioeconomic position and overweight among adolescents: data from birth cohort studies in Brazil and the UK. BMC Public Health 9, 105.CrossRefGoogle ScholarPubMed
45. Vorster, HH, Venter, CS, Wissing, MP et al. (2005) The nutrition and health transition in the North West Province of South Africa: a review of the THUSA (Transition and Health during Urbanisation of South Africans) study. Public Health Nutr 8, 480490.CrossRefGoogle ScholarPubMed
Figure 0

Table 1 Distribution of combined overweight and obesity* for adolescents aged 10–20 years (n 1848) and central obesity† for adolescents in Tanner stages 3–5 (n 1110) by various covariates, Agincourt sub-district, South Africa, 2007

Figure 1

Table 2 Determinants of BMI-for-age Z-score and combined overweight and obesity* in adolescents aged 10–20 years (n 1848), Agincourt sub-district, rural South Africa, 2007

Figure 2

Table 3 Determinants of waist-for-height ratio Z-score for adolescents aged 10–20 years (n 1848) and central obesity* for adolescents in Tanner stages 3–5 (n 1110), Agincourt sub-district, rural South Africa, 2007