Skip Navigation


American Journal of Epidemiology Advance Access originally published online on December 24, 2007
American Journal of Epidemiology 2008 167(5):615-623; doi:10.1093/aje/kwm340
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
167/5/615    most recent
kwm340v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Arantxa Colchero, M.
Right arrow Articles by Bishai, D.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Arantxa Colchero, M.
Right arrow Articles by Bishai, D.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

American Journal of Epidemiology © The Author 2007. Published by the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org.

ORIGINAL CONTRIBUTIONS

Effect of Neighborhood Exposures on Changes in Weight among Women in Cebu, Philippines (1983–2002)

M. Arantxa Colchero1,2 and David Bishai2

1 Department of Health Economics and Evaluation, National Institute of Public Health, Cuernavaca, Mexico
2 Department of Population, Family and Reproductive Health, Bloomberg School of Public Health, Johns Hopkins University, Baltimore, MD

Correspondence to Dr. Arantxa Colchero, Room 217, National Institute of Public Health, Av. Universidad No. 655, Col. Sta. Maria Ahuacatitlán, CP 62508 Cuernavaca, Morelos, Mexico (e-mail: acolcher{at}jhsph.edu).

Received for publication May 23, 2007. Accepted for publication October 24, 2007.


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 
The authors aimed to identify the contributions of community factors to weight change in a cohort of women from Metropolitan Cebu, Philippines, between 1983 and 2002. The authors created a three-level random-intercept model to see whether mean body mass index (BMI; weight (kg)/height (m)2) varied by individual- and cluster-level variables and identified community characteristics associated with changes in BMI among 2,952 nonpregnant women. The average BMI among women living in places with four public amenities (telephones, electricity, mail delivery, and newspapers) was 0.16 kg/m2 (95% confidence interval: 0.07, 0.26) higher than that of women living in places with fewer than three amenities. An increase in population density of 10,000 persons per km2 was associated with a BMI increase of 0.09 kg/m2 (95% confidence interval: 0.05, 0.13). A model with interactions revealed that the effect of population density increased significantly over time. These findings confirm earlier observations that in low-income countries, obesity starts among the wealthiest communities. Secondary and tertiary prevention policies designed to reduce obesity should be implemented in the most economically developed areas first. Primary prevention would be most needed in less developed areas, where the obesity epidemic is just beginning.

body mass index; body weight changes; developing countries; multilevel model; obesity; occupations; overweight; residence characteristics

Abbreviations: BMI, body mass index; CI, confidence interval


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 
More than one billion adults in the world are overweight, and at least 300 million of them are clinically obese (1). The prevalences of overweight and obesity are rapidly increasing in the developing world, even in regions suffering from malnutrition and high rates of infant mortality (2–4). Genetic factors may explain individual susceptibility to obesity, but the rapid increase in population weight in the last three decades has been attributed to profound social and economic changes worldwide (5, 6). Urbanization and economic development are driving major transformations in dietary habits and physical activity that may be linked to this rapid increase in overweight and obesity in low- and middle-income countries (7, 8). People today are facing a new environment that is modifying their options regarding eating and physical activity (9).

The effect of environmental exposures or community factors on weight gain has been consistently emphasized (5, 6, 10–13). In developed countries, dietary patterns and obesity have been associated with proximity to urban areas, per capita number of fast-food and full-service restaurants, and food prices (14–16). In most high- and middle-income countries, income is negatively associated with obesity, with higher rates of obesity being observed among the poorest people (15, 17, 18). Wealthier neighborhoods have more access to supermarkets, a greater variety of foods, and more places for physical activity (9).

Studies on the effect of environmental exposures on obesity in developing countries are limited because of the unavailability of data at the neighborhood or community level and the scarcity of longitudinal studies on trends in obesity. The obesity epidemic in the developing world may have unique characteristics, with patterns varying by economic and urbanization level (8, 19–21). In low-income countries, the prevalences of overweight and obesity increase first in wealthier areas and may spread to the poor later. The evidence suggests that obesity is less prevalent in rural areas, where the food supply is limited and work-related activities are more physically demanding.

Studies conducted in developing countries emphasize the association between urban residence and obesity (7, 22–24). However, urbanization is a complex process that occurs at different rates and in different periods (25). Conventional distinctions between urban and rural areas may not capture more subtle differences in the effect of economic development on health between regions. Detecting areas with higher risks of overweight and obesity requires identifying the level at which urbanization affects health and the regional factors driving the obesity epidemic.

Policy interventions for obesity and overweight can focus on persons or communities. Individual-level interventions could include diet and exercise, while community interventions might include constructing more walkable neighborhoods and providing greater access to fresh fruits and vegetables. Knowing what approach to emphasize requires determining how much of the obesity epidemic is due to factors operating at the individual level versus the community level. Our objective in this study was to identify the respective contributions of community and individual factors to obesity and overweight in a cohort of women in Cebu, Philippines.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 References
 
Data
The Cebu Longitudinal Health and Nutrition Survey follows a cohort of women who gave birth in 1983 in Cebu, the second-largest metropolis in the Philippines. Metropolitan Cebu is divided into six municipalities and three cities, both subdivided into neighborhoods called barangays, the smallest local government unit (average size, 4 km2). A single-stage cluster sampling design was used to locate all pregnant women from 33 randomly selected barangays (26). The surveys obtained information on the index children, their mothers, their households, and the communities in which they lived.

We selected seven surveys for which information on household income and mother's occupation was available. Community surveys were conducted in the same years, except for 1985, because in the first phase of the study only two community surveys were conducted: one at baseline in 1983 and the other at the end of this first stage in 1986. We interpolated values for all community variables for which information was missing in 1985, using information about the community obtained in 1983.

From 3,327 women surveyed in 1983 at baseline, 345 had dropped out of the study by 2002, mainly because of migration outside Cebu. Compared with women who completed at least one survey, dropouts were slightly younger, more educated, and of lower parity, and they had higher household incomes and were less likely to be working (p < 0.001). The significant differences between these two groups may have led to selection bias. A small number of women disappeared from the study due to mortality, but we ignored other reasons associated with attrition. We controlled for all observable factors associated with attrition, but we must acknowledge the possibility that women with multiple waves of participation differed in unobservable ways from the baseline sample.

In the analysis, we included all women who participated in the surveys except those who were pregnant. Pregnant women were excluded only from the survey round in which they reported being pregnant. Table 1 presents the number of nonpregnant women included in our analysis and the number of pregnant and missing women in each survey. Some women may have missed a survey because of temporal migration but were present in subsequent waves. Our analytical sample was composed of 2,952 nonpregnant women.


View this table:
[in this window]
[in a new window]

  		TABLE 1. Numbers of pregnant and nonpregnant women included in the Cebu Longitudinal Health and Nutrition Survey, by round, Cebu, Philippines, 1983–2002

 
Variables
Body mass index (BMI; weight (kg)/height (m)2) was used as a standard measure of weight independent of height. The dependent variable was change in weight, calculated as the difference between current BMI and baseline BMI.

Public amenities and population density measured at the community level constituted the principal explanatory variables. The number of amenities available in a barangay reflects the existing infrastructure and economic development. Based on four dichotomous variables revealing the availability of four public amenities (electricity, mail delivery, telephones, and newspapers, coded as a sum from 0 to 4), we created a new variable that was coded 0 for barangays with up to three amenities and 1 for barangays with all four amenities. Population density, defined as the number of persons per square kilometer, was used as an additional measure of urbanization. This measure captured more subtle differences in economic development than the urban/rural codes assigned by the Cebu Longitudinal Health and Nutrition Survey at baseline. In the models, population density was expressed in terms of 10,000 persons per km2.

Occupational activity was ranked in three intensity categories as low, medium, or heavy. Managers, clerical workers, and professionals were classified as engaging in low levels of activity; salespersons, service workers, and homemakers were classified as engaging in medium activity; and persons involved in farming, fishing, and production processes were classified as engaging in heavy activity. Household income was estimated from wages on a per-time basis, from piece-rate wages, or as farming, fishing, and livestock activities and self-employment income. Other variables measured at the individual level were breastfeeding, education (reported as the highest grade completed), daily energy intake (in kilocalories), age, age squared, parity, and number of children under 5 years of age in the household. Dummy variables for each round were included to control for period effects on changes in BMI.

Energy intake was assessed through 24-hour food recall questionnaires, except for 1991, in which a quantitative food frequency questionnaire was used. To improve validity in all surveys, food recalls were recorded on 2 days of the week, and an average was estimated. The University of North Carolina converted food consumption values into kilocalories per day using food composition tables from the Philippines (L. Adair, University of North Carolina at Chapel Hill, personal communication, 2006).

Analysis
In longitudinal studies with cluster designs, the assumption that the errors are independent and that the variance is constant is violated. Traditional methods that assume independence and constant variance may result in imprecise standard errors and incorrect inferences (27–29). Hierarchical models can be used to account for variation at different levels of analysis, providing more accurate standard errors and inferences.

If the processes contributing to obesity were spatially correlated, then estimation using ordinary least squares would produce incorrect standard errors. Furthermore, if the observable features of barangays were correlated with unobservable features that affected obesity, the effects of the observable features on BMI would be biased. Ordinary least squares estimation requires that random errors be independent and normally distributed and have constant variance. The assumptions in cluster design are violated, given that random errors are dependent within each cluster and the errors have unequal variances. Another source of bias is the correlation between repeated measures taken in the same individual over time. We used hierarchical models to estimate change in BMI, taking into account variation at the individual level over time and at the cluster level.

We specify a three-level hierarchical model, where level 1 units are occasions, defined as repeated measurements in the same individual over time, level 2 units are individuals (women), and level 3 corresponds to barangay units. In this three-level model, occasions for individuals are clustered in barangays.

The level 1 equation defines change in BMI for the jth woman in the kth barangay at time t, as a function of repeated measurements taken over time, such as income, occupation, and period dummies for each year of the survey:

Formula (1)

At level 2, the intercept β0jk is decomposed into a linear combination of a population-level grand mean {gamma}01 and a woman-specific random effect u0jk.:

Formula (2)

At level 3, the intercept {gamma}01 is decomposed into a general mean {gamma}000, contributions from barangay variables such as population density and amenities available, and a barangay-specific random effect u0k:

Formula (3)

Combining equations 1–3 yields the following random-intercept specification:

Formula (4)

The random parameters of the model u0jk, u0k, and eijk compose the disturbance term Var[yijk]. The model assumes that random parameters and error terms are normally distributed and have a mean of zero and constant variance:

Formula

Two intraclass correlation coefficients can be obtained from a three-level model. For the same barangay but a different woman, {rho}(barangay) = {tau}3/({tau}3 + {tau}2 + o2). For the same barangay, woman j, the coefficient is {rho}(woman/barangay) = ({tau}2 + {tau}3)/({tau}3 + {tau}2 + o2). We expect to see {rho}(woman/barangay) > {rho}(barangay), given that occasions within a woman are more similar than occasions within a barangay.

All continuous variables were scaled to facilitate the interpretation of the intercept in the models. We centered variables by subtracting from each observation its mean in each respondent's barangay in the corresponding time period: Xjkt – Formula.jk. Interactions between population density and period dummies were included in the model.

The number of barangays increased from 33 at baseline to 171 in 2002, because of internal migration within Cebu. As a result, new barangays had very small numbers of women. The variance explained at the barangay level increased with the number of persons per cluster (see Results section). The small number of women in some barangays may have caused us to overestimate the variability of changes in BMI. In contrast, models excluding barangays may have given rise to selection bias. For comparison, we fitted four models: ordinary least squares and random-intercept models including all women and ordinary least squares and random-intercept models excluding barangays with less than 10 women. We applied multilevel mixed-effects linear regression (xtmixed) using Stata, version 9.0 (Stata Corporation, College Station, Texas).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 References
 
Important changes took place in the 33 baseline barangays between 1983 and 2002 (table 2). The proportion of overweight women (BMI > 25) (30) increased from 7 percent to 43 percent, and the number of barangays with more than 15 percent of women overweight increased from 1 to 28. Mean population density increased from 12,000 persons per km2 to 14,000 persons per km2. The prevalence of amenities provided in the community significantly increased.


View this table:
[in this window]
[in a new window]

  		TABLE 2. Characteristics of women from 33 sampled barangays in the Cebu Longitudinal Health and Nutrition Survey (n = 2,952), Cebu, Philippines, 1983 and 2002

 
Mean BMI increased from 20.5 in 1983 to 24.3 in 2002. Figure 1 shows the mean BMI and 95 percent confidence interval for each barangay in 1983 and 2002. Whereas in 1983 barangays had similar BMI distributions, 20 years later barangays appeared more differentiated. For some barangays, the distribution of BMI moved significantly to the right and the dispersion increased, whereas other barangays continued to have a low dispersion and a low mean BMI.


Figure 1
View larger version (16K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		FIGURE 1. Mean body mass index (weight (kg)/height (m)2) for each of 33 sampled barangays in the Cebu Longitudinal Health and Nutrition Survey, Cebu, Philippines, 1983 and 2002. Each point in the graph represents the mean body mass index of all women living in a specific barangay. Bars, 95% confidence interval.

 
We compared the basic characteristics of women living in barangays with fewer than 10 women with those of the sample that included barangays with 10 or more women in round 1. We found differences in parity and household size (significant at 5 percent) and age and education (significant at 10 percent). Household income did not differ between the two samples. Barangays with less than 10 women had a lower population density and fewer amenities available than barangays with 10 or more women. These results suggest that exclusion of barangays may not have been random, leading to potential selection bias. However, coefficients between the two models were similar, with only slight differences in low occupational activity and household income (tables 3 and 4).


View this table:
[in this window]
[in a new window]

  		TABLE 3. Results from ordinary least squares and random-intercept models of change in body mass index* for all barangays and for barangays with 10 or more women, Cebu Longitudinal Health and Nutrition Survey, Cebu, Philippines, 1983–2002{dagger}

 

View this table:
[in this window]
[in a new window]

  		TABLE 4. Results from ordinary least squares and random-intercept models of change in body mass index* for all barangays and for barangays with 10 or more women, with interactions between population density and period dummies, Cebu Longitudinal Health and Nutrition Survey, Cebu, Philippines, 1983–2002{dagger}

 
Table 3 presents the results from ordinary least squares and random-intercept models without interactions, comparing models with all barangays and models excluding barangays with less than 10 women. As expected, the ordinary least squares estimates and 95 percent confidence intervals were different from those of the random-intercept models. Ordinary least squares underestimated the effect of population density at the barangay level and overestimated the effects of occupational activity, household income, and public amenities on weight.

An increase in population density of 10,000 persons per km2 at the barangay level was associated with an average increase in BMI of 0.09 kg/m2 (95 percent confidence interval (CI): 0.05, 0.13). On average, BMI among women living in places with all four amenities (telephones, electricity, mail delivery, and newspapers) was 0.16 kg/m2 (95 percent CI: 0.07, 0.26) higher than that among women living in places with fewer than three amenities.

Random effects at the individual and barangay levels made a statistically significant contribution to the estimates. The intraclass correlation coefficient {rho}(barangay) in a model of BMI change without covariates that included all barangays was 18 percent and decreased to 4 percent after exclusion of barangays with less than 10 women. After covariates were included, the variability explained at the barangay level fell to 7 percent in the model with all barangays and to 2.1 percent in the model excluding barangays with fewer women.

Table 4 presents results from the models with interactions between population density and period dummies. In this specification, the effect of public amenities is magnified, showing that barangays with all four amenities had a mean BMI change that was 0.23 kg/m2 (95 percent CI: 0.13, 0.32) higher than that in barangays with up to three amenities. The model shows that the effect of a one-unit increase in population density became significantly larger over time. Women living in areas with higher population density had larger increases in BMI over time, which is shown by positive coefficients for period dummy interactions. The effect of population density in each period as compared with the reference year (1984) was positive and significant, starting in 1991. To graphically illustrate the interactions, we plotted the predicted change in BMI for each year of the survey against population density using the coefficients estimated in the model (figure 2). The graph illustrates that the effect of population density became positive after 1985 and significantly larger with time.


Figure 2
View larger version (13K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		FIGURE 2. Effect of population density on predicted change in body mass index (weight (kg)/height (m)2), by survey year, in the Cebu Longitudinal Health and Nutrition Survey, Cebu, Philippines, 1983–2002.

 
Random parameters were significant and slightly decreased in the model with interactions. The intraclass correlation coefficient at the barangay level decreased from 4.0 percent in the model without interactions to 3.4 percent in the model including all barangays and from 2.1 percent to 1.7 percent in the model excluding barangays with less than 10 women. The likelihood ratio test favored the random-intercept model with interactions compared with the model without interactions for specification using all barangays (likelihood ratio test: {chi}2 = 123.9, p < 0.0001) and barangays with 10 or more women (likelihood ratio test: {chi}2 = 131.7, p < 0.0001).

Overall variation in BMI at the barangay level was higher when we included all barangays and much lower when we excluded barangays with fewer than 10 women. Despite some differences in the samples that may reflect selection bias, coefficients between the two models were similar. However, conclusions on how much variation is explained at the barangay level differ. Given that new barangays incorporated into the surveys due to internal migration had small numbers of women, variability in BMI may have been overestimated. Moreover, a small number of women may not be representative of all women living in a specific barangay. In addition, changes in weight among women migrating to a new barangay can hardly be related to exposure to this new environment. Therefore, the result from the model excluding barangays with less than 10 women is our preferred estimate.

The models showed a positive and significant association between daily energy intake and changes in weight, but the coefficient was relatively small: An increase of 100 kilocalories consumed per day was associated with a mean increase in BMI of 0.02 kg/m2 (95 percent CI: 0.01, 0.03). Although mean energy intake or beverage consumption did not increase dramatically in this population, there has been an important rise in the percentage of energy derived from fat, from 10.5 percent in 1983 to 15.5 percent in 2002.

Sedentary leisure-time activities, such as time spent watching television, may be associated with weight gain (31–34). Information on leisure activities was not collected in the Cebu Longitudinal Health and Nutrition Survey. However, the proportion of households owning a television rose from 22 percent in 1983 to 62 percent in 2002. BMI among women who owned a television was 0.44 kg/m2 (95 percent CI: 0.36, 0.54) higher than that among women who did not possess a television. Owning a car was also a significant predictor of greater changes in weight (table 3). Although information on time spent watching television or driving was not recorded, having a television or a car may be an indicator of an increased amount of time spent in more sedentary leisure activities.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
References
 
Our results suggest that women gained weight at different rates depending on the barangay in which they lived. We identified two variables at the barangay level that were associated with change in BMI between 1983 and 2002: number of public amenities and population density. We found that BMI among women living in barangays with four public amenities (telephones, mail delivery, newspapers, and electricity) was greater than that among women living in barangays with up to three amenities. An increase in population density was positively associated with BMI.

We applied hierarchical models to account for variation at the individual and barangay levels. The increased variability of BMI over time and its concentration by women and barangay violate the assumption of independence and homoskedasticity using ordinary least squares. Therefore, we conclude that hierarchical models provided more accurate estimations of the geographic effects on body size. The model excluding barangays with fewer than 10 women was our preferred specification because of the possibility that effects would be unstable for barangays with only a few respondents.

Changes in weight over time were associated with population density and public amenities. Differences in weight between places with higher and lower population densities became larger with time. Public amenities and population density may not have a direct clinical effect on obesity but may be an indicator of economic development and wealth. Our results suggest that women living in places with less economic development had smaller changes in weight over time. Areas with lower population density may have developed more slowly. Fast-food options and transportation in these areas might be limited, so that people walk more and eat healthier foods than people in more developed areas.

Changes in the food environment may have played an important role in the obesity epidemic (35). In contrast with the situation in some developed countries, fast food is not yet a cheaper option than home-cooked food in the Philippines. Street vendors sell a variety of cheap home-cooked food, although in urban areas deep-fried food is becoming increasingly available. This may partly explain why in countries and regions such as Metropolitan Cebu, obesity starts among the wealthiest groups—persons who have more access to and can afford more energy-dense and fast foods and who have more sedentary lifestyles.

The analysis presented in this paper had some limitations. Although we found that hierarchical models provided more accurate estimates, differences in coefficients and standard errors between ordinary least squares and hierarchical models were moderate. A larger part of the variation in BMI was explained at the individual level rather than at the barangay level. The proportion of the variance explained at the cluster level was negligible at baseline and increased with time. The small degree of cluster-level variation may reflect a relatively homogenous population within Metropolitan Cebu. In the Cebu Longitudinal Health and Nutrition Survey, approximately 75 percent of the sampled barangays were urban. In addition, the close proximity of rural areas to cities in the sample might have attenuated differences between urban and rural areas within Metropolitan Cebu.

The strong period effects show that women in this cohort experienced weight gains over time but at different rates, depending on the barangay or city where they lived. There is some evidence in developed countries regarding the influence of the built environment on obesity (36). The built environment involves factors related to urban design, land use, and availability of public transportation, such that it can either promote or discourage physical activity and healthy eating behaviors. In our analysis, despite the richness of the community surveys in the Cebu Longitudinal Health and Nutrition Survey, few variables at the barangay level were associated with changes in weight. This might reflect the reality that with a sample of 2,952 women and 37 barangays, the effect size was so small that patterns escaped detection. Measurement error could have contributed to our inability to detect effects of barangay-level variables. Food prices were collected in the community surveys from two different stores in each barangay. However, prices were not included in our analysis because a large number of price measures were missing, partly due to unavailability of certain items in the stores. Limiting our analysis to barangays with complete information on prices would have created selection bias.

We found that places with higher economic development, as reflected by public amenities and population density, had the greatest gains in weight. The associations found between cluster-level variables and weight changes do not necessarily reflect causal effects. If unobservable factors were associated with changes in weight or with any of the covariates included in the models, our estimates would be biased. It is possible that the association we found between cluster variables and weight represents an example of the ecological fallacy. The similarity of coefficients for cluster-level amenities and population density obtained by the ordinary least squares and random-intercept models is reassuring, but it is not proof that unobservable factors are not playing a large role.

Identifying community- and individual-level factors associated with obesity helps in targeting high-risk groups for intervention. Our findings confirm earlier observations that in low-income countries, obesity starts among the wealthiest groups and communities (20). Secondary and tertiary prevention policies designed to reduce obesity should be implemented in the most economically developed areas first. However, primary prevention of obesity in adult women would be most needed in less developed areas, where the obesity epidemic is just beginning. Interventions should be carefully designed in low-income countries where overweight and underweight coexist within the same region, city, and household (4).


   ACKNOWLEDGMENTS
 
This research was supported by a grant awarded to Dr. Arantxa Colchero from the Center for a Livable Future, Johns Hopkins Bloomberg School of Public Health (Innovation Grant August 2006–2007).

The authors are grateful to Drs. Francesca Dominici, Kevin Frick, and Benjamin Caballero from the Johns Hopkins School of Public Health for their interesting and useful comments on the manuscript. They also thank Drs. Constance Gultiano and Abet Bas from the Office of Population Studies at the University of San Carlos in Cebu for their help in the use of data from the Cebu Longitudinal Health and Nutrition Survey.

Conflict of interest: none declared.


   References
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 

  1. Health World. Organization. Obesity and overweight. (Fact sheet). (2003) Geneva, Switzerland: World Health Organization.
  2. Raymond SU, Leeder S, Greenberg HM. Obesity and cardiovascular disease in developing countries: a growing problem and an economic threat. Curr Opin Clin Nutr Metab Care (2006) 9:111–16.[Web of Science][Medline]
  3. Popkin BM, Gordon-Larsen P. The nutrition transition: worldwide obesity dynamics and their determinants. Int J Obes Relat Metab Disord (2004) 28((suppl 3)):S2–9.[CrossRef]
  4. Caballero B. A nutrition paradox—underweight and obesity in developing countries. N Engl J Med (2005) 352:1514–16.[Free Full Text]
  5. Hill JO, Wyatt HR, Reed GW, et al. Obesity and the environment: where do we go from here? Science (2003) 299:853–5.[Abstract/Free Full Text]
  6. Hill JO, Peters JC. Environmental contributions to the obesity epidemic. Science (1998) 280:1371–4.[Abstract/Free Full Text]
  7. Popkin B. Urbanization, lifestyle changes and the nutrition transition. World Dev (1999) 27:1905–16.[CrossRef]
  8. Caballero B. Introduction. Symposium: obesity in developing countries: biological and ecological factors. J Nutr (2001) 131((suppl)):866S–70S.[Abstract/Free Full Text]
  9. Popkin BM, Duffey K, Gordon-Larsen P. Environmental influences on food choice, physical activity and energy balance. Physiol Behav (2005) 86:603–13.[CrossRef][Medline]
  10. Finkelstein EA, Ruhm CJ, Kosa KM. Economic causes and consequences of obesity. Annu Rev Public Health (2005) 26:239–57.[CrossRef][Web of Science][Medline]
  11. French SA, Story M, Jeffery RW. Environmental influences on eating and physical activity. Annu Rev Public Health (2001) 22:309–35.[CrossRef][Web of Science][Medline]
  12. Hill JO, Sallis JF, Peters JC. Economic analysis of eating and physical activity: a next step for research and policy change. Am J Prev Med (2004) 27:111–16.[CrossRef][Web of Science][Medline]
  13. Walker P, Rhubart-Berg P, McKenzie S, et al. Public health implications of meat production and consumption. Public Health Nutr (2005) 8:348–56.[CrossRef][Web of Science][Medline]
  14. Chou S-Y, Grossman M, Saffer H. An economic analysis of adult obesity: results from the Behavioral Risk Factor Surveillance System. (NBER Working Paper no. W9247). (2002) Cambridge, MA: National Bureau of Economic Research.
  15. Drewnowski A, Specter SE. Poverty and obesity: the role of energy density and energy costs. Am J Clin Nutr (2004) 79:6–16.[Abstract/Free Full Text]
  16. Lopez R. Urban sprawl and risk for being overweight or obese. Am J Public Health (2004) 94:1574–9.[Abstract/Free Full Text]
  17. Drewnowski A, Darmon N. The economics of obesity: dietary energy density and energy cost. Am J Clin Nutr (2005) 82((suppl)):265S–73S.[Abstract/Free Full Text]
  18. Ball K, Crawford D. Socioeconomic status and weight change in adults: a review. Soc Sci Med (2005) 60:1987–2010.[CrossRef][Web of Science][Medline]
  19. Monteiro CA, Conde WL, Popkin BM. Independent effects of income and education on the risk of obesity in the Brazilian adult population. J Nutr (2001) 131((suppl)):881S–6S.[Abstract/Free Full Text]
  20. Monteiro CA, Moura EC, Conde WL, et al. Socioeconomic status and obesity in adult populations of developing countries: a review. Bull World Health Organ (2004) 82:940–6.[Web of Science][Medline]
  21. Popkin BM. An overview on the nutrition transition and its health implications: the Bellagio meeting. Public Health Nutr (2002) 5:93–103.[CrossRef][Web of Science][Medline]
  22. Adair LS. Dramatic rise in overweight and obesity in adult Filipino women and risk of hypertension. Obes Res (2004) 12:1335–41.[Web of Science][Medline]
  23. Paeratakul S, Popkin BM, Keyou G, et al. Changes in diet and physical activity affect the body mass index of Chinese adults. Int J Obes Relat Metab Disord (1998) 22:424–31.[CrossRef][Web of Science][Medline]
  24. Popkin BM. The nutrition transition and its health implications in lower-income countries. Public Health Nutr (1998) 1:5–21.[CrossRef][Medline]
  25. Kasarda JD, Crenshaw EM. Third world urbanization: dimensions, theories, and determinants. Annu Rev Sociol (1991) 17:467–501.[CrossRef][Web of Science][Medline]
  26. Adair L. The Cebu. Longitudinal Health and Nutrition Survey: history and major contributions of the project. Philipp Q Cult Soc (2001) 29:5–37.
  27. Bryck AS, Raudenbush SW. Hierarchical linear models: applications and data analysis methods. 1st ed. (1992) Newbury Park, CA: Sage Publications, Inc.
  28. Snijders TAB, Bokser RJ. Multilevel analysis: an introduction to basic and advanced multilevel modeling. (1999) Thousand Oaks, CA: Sage Publications, Inc.
  29. Steenbergen MR, Jones BS. Modeling multilevel data structures. Am J Pol Sci (2002) 46:218–37.[CrossRef][Web of Science]
  30. Gibson RS. Principles of nutritional assessment. (2005) New York, NY: Oxford University Press.
  31. Martinez-Gonzalez MA, Martinez JA, Hu FB, et al. Physical inactivity, sedentary lifestyle and obesity in the European Union. Int J Obes Relat Metab Disord (1999) 23:1192–201.[CrossRef][Web of Science][Medline]
  32. Gutierrez-Fisac JL, Guallar-Castillon P, Diez-Ganan L, et al. Work-related physical activity is not associated with body mass index and obesity. Obes Res (2002) 10:270–6.[Web of Science][Medline]
  33. Hu FB, Li TY, Colditz GA, et al. Television watching and other sedentary behaviors in relation to risk of obesity and type 2 diabetes mellitus in women. JAMA (2003) 289:1785–91.[Abstract/Free Full Text]
  34. Fung TT, Hu FB, Yu J, et al. Leisure-time physical activity, television watching, and plasma biomarkers of obesity and cardiovascular disease risk. Am J Epidemiol (2000) 152:1171–8.[Abstract/Free Full Text]
  35. Jeffery RW, Baxter J, McGuire M, et al. Are fast food restaurants an environmental risk factor for obesity? Int J Behav Nutr Phys Act (2006) 3:2.[CrossRef][Medline]
  36. Booth KM, Pinkston MM, Poston WS. Obesity and the built environment. J Am Diet Assoc (2005) 105((suppl)):S110–17.[Web of Science][Medline]

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?



This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
167/5/615    most recent
kwm340v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Arantxa Colchero, M.
Right arrow Articles by Bishai, D.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Arantxa Colchero, M.
Right arrow Articles by Bishai, D.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?