IJE Advance Access originally published online on December 14, 2006
International Journal of Epidemiology 2007 36(1):93-101; doi:10.1093/ije/dyl252
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The relationship between parity and overweight varies with household wealth and national development
1 Nutrition and Health Sciences Program, Graduate Division of Biological and Biomedical Sciences, Emory University, Atlanta, GA 30322, USA.
2 Hubert Department of Global Health, Rollins School of Public Health, Emory University, Atlanta, GA 30322, USA.
3 Department of Sociology, Emory University, Atlanta, GA 30322, USA.
* Corresponding author. Hubert Department of Global Health, Rollins School of Public Health of Emory University, 1518 Clifton Road, N.E. Atlanta, GA 30322, USA. E-mail: rmart77{at}sph.emory.edu
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Background Recent studies support a positive relationship between parity and overweight among women of developing countries; however, it is unclear whether these effects vary by household wealth and national development. Our objective was to determine whether the association between parity and overweight [body mass index (BMI)
25 kg/m2] in women living in developing countries varies with levels of national human development and/or household wealth. Methods We used data from 28 nationally representative, cross-sectional surveys conducted between 1996 and 2003 (n = 275 704 women, 15–49 years). The relationship between parity and overweight was modelled using logistic regression, controlling for several biological and sociodemographic factors and national development, as reflected by the United Nations’ Human Development Index. We also modelled the interaction between parity and national development, and the three-way interaction between parity, household wealth and national development.
Results Parity had a weak, positive association with overweight, which varied by household wealth and national development. Among the poorest women and women in the second tertile of household wealth, parity was positively related to overweight only in the most developed countries. Among the wealthiest women, parity was positively related to overweight regardless of the level of national development.
Conclusions As development increases, the burden of parity-related overweight shifts to include poor as well as wealthy women. In the least-developed countries, programmes to prevent parity-related overweight should target wealthy women, whereas such programmes should be provided to all women in more developed countries.
Keywords Parity, reproduction, overweight, obesity, developing countries
Accepted 21 October 2006
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Once thought to be a condition of only wealthy countries, overweight [body mass index (BMI)
25 kg/m2] is now reaching epidemic levels in high, middle and low income countries. In the United States, 62% of women aged 20–74 years are overweight.1 In some low- and middle-income countries, rates of overweight in women are similar to or even higher than rates in the US. In Turkey, for example, the prevalence of overweight in mothers of reproductive age (15–49 years), was 52% in 1998,2 and in Egypt it was 71% in 2000.2 Reproduction is thought to play a role in overweight in developed countries,3–9 but scientific interest in the role of reproduction in developing countries has focused mainly on underweight and undernutrition. Because of the increasingly obesogenic environment of developing countries, which has resulted from the nutrition transition,10–14 researchers have begun to investigate whether parity now plays a role in overweight in these countries, as it does in developed ones. Recent studies in Mexico, countries in North Africa/West Asia and China have found a weak to moderate, positive association between parity and overweight.15–18 In our own recent research, we analysed data for women of reproductive age (15–49 years) from cross-sectional, nationally representative surveys conducted in 50 low- and middle-income countries in five regions around the world.19 In the majority of countries, we found a positive association between parity and overweight, after controlling for age and other confounding factors. We also found that this parity–overweight association was stronger in more developed countries. We defined country development as the overall level of well-being of the people in each country, represented by a summary score constructed from 12 established measures of national income, urbanization, health and education.
A limitation of the country level analysis19 was that we assumed that the effects of country development on the parity–overweight relationship were uniform throughout the population, which may not be the case. For example, the current literature suggests that the associations between the prevalence of overweight and a country's level of development vary by individual socioeconomic status (SES).20 Therefore, the present study examines the effect of national development on the parity–overweight relationship, by different levels of individual wealth (approximated by a measure of household wealth).
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Sources of data
Data for this analysis come from the Demographic Health Surveys (DHS) (www.measuredhs.com) and the Reproductive Health Surveys (RHS) (www.cdc.gov/reproductivehealth/surveys) which local organizations conduct in collaboration with ORC Macro, Inc. and the Division of Reproductive Health at the Centers for Disease Control and Prevention, respectively.
These cross-sectional surveys were administered to a nationally representative sample of households, generally selected using a multistage, stratified cluster sample design. In most cases, households were sampled either directly from primary sampling units (PSUs) or from one or two clusters that were sampled from larger PSUs. The PSUs were usually census enumeration units (e.g. villages or urban areas), as defined by each country's government, and the clusters were usually areas of equal population size within each PSU. Some surveys also relied on a stratified sample design (e.g. sampled within urban/rural areas) and may have over-sampled some sub-groups in the population for specific analytic purposes.21 For these reasons, national sample weights and robust variance estimates were used to account for the sample designs.
The data sets contain information about the standard of living of sample households, as well as the health status of women of reproductive age (15–49 years) and their children (with ages variously defined as 
35, 59, 72 months). A household listing and characteristics questionnaire generally was administered first, in which information about household members, as well as household assets owned, characteristics of the dwelling and sanitation conditions was collected. The household listing also was used to identify women who were eligible for an individual interview, in which questions were asked about the health status and sociodemographic attributes of women and their children. A majority of the questions that were asked were standardized across the national surveys, permitting cross-national comparisons.17 Fieldworkers participating in the DHS and RHS were trained to take height and weight measurements according to the methods described in the United Nations manual, How to Weigh and Measure Children.17 Women were weighed with digital scales (weight read to the nearest 0.1 kg), and their height was measured using an ‘adjustable wooden measuring board,’ with a scale read to the nearest 0.1 cm,17 [and personal communication, Rafael Flores, February 13, 2006].
The present analysis included data for the 28 countries in which anthropometric data were collected for parous and nulliparous women. These 28 data sets represent countries from five regions: North Africa/West Asia, Central Asia, South and Southeast Asia, Latin America and the Caribbean and Sub-Saharan Africa. The number of eligible households ranged from 2168 to 94 845, and the household response rates were universally high (93.8%–99.6%).22 The number of women who were eligible for the individual interview ranged from 3155 to 94 641, and the individual response rates were also high (90.6%–99.5%).22
Analytic sample
To be included in the analysis, women had to be non-pregnant and at least 3 months post partum. All women who had any non-singleton births or had missing data for covariates were excluded. For each national sample, we examined the distribution of height and weight and excluded women who had extreme outlying measurements, defined as being above the 75th percentile plus three times the interquartile range, or below the 25th percentile minus three times the interquartile range.23 These exclusions were performed separately for each country because the range and distribution of measurements for height and weight varied greatly across included countries, and so applying aggregate cut points for outliers would be inappropriate.
We calculated the percentage of women who were excluded from our analytic sample as a result of missing data (after the a priori exclusion criteria were applied). Urban/rural status, age and previous birth interval were never missing in the data sets; analysts imputed missing values for the latter two variables before the data sets were made publicly available.24 Women who were missing information on breastfeeding duration and/or years of schooling were excluded. Less than 1% of the women in all national analytic samples were missing years of schooling, and 0.1–1.6% of the women in all national samples were missing breastfeeding duration. Thus, neither interview non-response nor item non-response were likely to be sources of bias in the analyses.
These criteria resulted in national data set sample sizes ranging from 2162 to 72 769, with a median sample size of 6778 women. One multicountry data set was constructed from these data sets, with a total sample size of 275 704 women.
Individual and household-level variables
Individual and household-level variables were created for each country separately. The outcome analysed was overweight, defined as BMI 25 kg/m2,25 calculated from measured height and weight. The main independent variable of interest was parity, defined as the total number of pregnancies resulting in a live or still birth. However, this information was not available for all data sets, so we approximated parity with the number of live births. There were four categories for this variable: 0, 1 (here considered the reference group), 2–3 and 4 or more live births.
Additional covariates that we controlled for were age, short birth interval, months of breastfeeding, urban/rural status, schooling, and household wealth. Age was grouped into the following categories: 15–19, 20–24, 25–29, 30–34, 35–39, and 40–49 years. A woman was classified as having a short birth interval if <24 months had elapsed between her two most recent live births. Women with no live births or one live birth were considered not to have experienced a short birth interval. Information on months of breastfeeding was available for the majority of countries (24) for all surviving or dead children who would have been 59 months at the time of the interview. In three countries, it was collected for children who would have been 35 months at the time of the interview; in one of these three countries, data were limited to the last two live births. In one country, information on months of breastfeeding was collected for all children who would have been 72 months at the time of the interview. We categorized breastfeeding as 0, 1–12, 13–24, 25 months; it was defined as total months of breastfeeding during the country-specific recall period (35, 59 or 72 months). We explored the effects of the inconsistencies in the breastfeeding duration information collected on the regression estimates for the main effect, parity. Our main concern was for the effect of including countries whose period of recall was 35 months (Kazakhstan, Uzbekistan and India). When the breastfeeding variable was omitted from the model, the parity estimates did not change meaningfully. When we excluded the 35-month countries from the data set, and ran the model with and without the breastfeeding variable, the estimates again did not change appreciably. Therefore, we felt that mixing the 35, 59 and 72-month recall periods did not meaningfully affect our results.
Urban or rural residence was based on the value available in each country's data set. Schooling and household wealth were country-specific variables, and each woman was ranked relative to other women in her own country. Tertiles for total completed grades of schooling were used to represent schooling attainment. For countries in which most of the women had zero grades of schooling, ‘no schooling’ was considered the lowest group, and the remaining women were ranked and divided at their median. Because data on income were not available, individual wealth was approximated by an index for household wealth. We created this index based on household assets owned, attributes of the dwelling and sanitary conditions of the dwelling, using a methodology that researchers at ORC Macro, Inc. and the World Bank developed jointly.26,27 Individuals were assigned the value of the wealth index for the household in which they lived, and scores were grouped into tertiles. Notably, this measure for wealth was developed for each country separately; therefore, each woman was ranked relative to other women in her own country, and this variable should be interpreted accordingly. We elected this approach because the items that were used to measure household standard of living varied somewhat across countries in the analysis. Using all of the items for each country enabled us to maximize the variation in household standard of living among women in the same country and to use contextually appropriate indicators for standard of living.
Country development
Level of country development is often defined in terms of economic output, such as gross domestic product (GDP). However, such measures do not capture the distribution of resources or quality of life, and this limitation has led to the development of more comprehensive indices of ‘human development.’28 Therefore, we used the United Nations’ human development index (HDI) to represent country development. The HDI is a composite measure that captures three aspects of human development in a country: lifespan, measured by life expectancy at birth; knowledge, measured by the literacy rate of men and women aged 15 years and older and the combined gross enrolment ratios for primary, secondary and tertiary schools; and standard of living, measured by GDP per capita adjusted for price purchasing parity.29 The creation of the HDI from these measures has been described in detail elsewhere.29
HDI values generally are calculated every 5 years. The values for the countries in this study were obtained from the 2004 Human Development Report. Because values were available for only 1990, 1995, 2000 and 2002, for each country we used the HDI value for the year closest to the survey year. If the HDI was available for two different years that were the same temporal distance from the survey year, the value for the earlier year was used.
We grouped HDI values into tertiles. Because the HDI was created so that an increase in value indicates an increase in national human development, an increase in tertile also indicates an increase in overall human development.
Statistical analysis
All statistical tests and regression analyses were done in SUDAAN (SAS9-callable, Release 9.0.0, Research Triangle Institute) to account for the multistage, stratified cluster sample design of the surveys. Country-specific sample weights were applied in the calculations of sample sizes, means and ratios.
For the full sample, and within each HDI tertile, we tested whether there was a linear trend in the mean BMI across the levels of parity and each covariate, unadjusted for the other covariates. We also tested whether there was a trend in the percentage of women who were overweight across the levels of parity and each covariate, unadjusted for the other covariates.
Individual and household-level variables and HDI tertiles were included in a full main effects logistic regression model. The interaction between parity and national human development was modelled by adding a product term for parity and HDI tertiles to this model. The three-way interaction between parity, household wealth and national human development was qualitatively examined by stratifying the sample by HDI tertile, adding a product term for parity and household wealth tertiles to the main effects model, and estimating this model within each HDI tertile. Thus, three models were run to assess the three-way interaction.
In additional analyses, we ran all models with regional fixed effects, and results for the main coefficients of interest were very similar (results not shown; available upon request). Because of the strong correlation between region and HDI tertile (Table 1), the models that are presented exclude the regional fixed effects.
|
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
The countries included in this analysis, year of data collection for each survey, HDI value and placement of countries in HDI tertiles are listed in Table 1. The first tertile represents countries with the lowest level of human development. Region and level of human development were clearly related. The countries in the first and second HDI tertiles were found mainly in Sub-Saharan Africa and South and Southeast Asia. The countries with the highest levels of human development were found mainly in Latin America and the Caribbean, Central Asia and North Africa/West Asia.
Table 2 shows the distribution of the sample with respect to the modelling variables, for the full sample and stratified by HDI tertile. In the full sample, the distribution of the sample was even across parity groups, with the exception of the one-child group: only 13% of the sample was in this group. The sample of women was roughly evenly distributed across age groups and levels of household wealth. A large percentage (43.9%) of the women was in the lowest schooling group. The majority of the sample (62.1%) lived in rural areas. Most women did not have a previous short birth interval (85.3%) and did not breastfeed during the recall period for their country (63.2%). About 51% of the sample lived in countries that had a medium level of human development. Notably, this distribution of women by level of national human development was driven by the India data set, which had the largest sample of women (n = 72 769, or 26% of the entire sample). The mean BMI for the total sample was 22.5, which falls within the normal classification (18.5–24.9 kg/m2),25 but the prevalence of overweight was 23.9%.
|
Within the full sample and each HDI tertile, we found significant, positive trends for mean BMI and percent overweight across categories of parity and the covariates, except for schooling and breastfeeding. In the full sample and in the first two HDI tertiles, we found significant, positive trends for mean BMI and percent overweight across categories of schooling, and significant, negative trends across categories of breastfeeding. However, in the third tertile, we found significant, negative trends for mean BMI and percent overweight across categories of schooling, and significant, positive trends across categories of breastfeeding.
In the full main effects model (which included the variables parity, age, household wealth, schooling, urban residence, short birth interval, duration of breastfeeding during the recall period and HDI tertiles), the associations of overweight with parity and the covariates did not change appreciably from their unadjusted associations; each variable had a positive relationship with the odds of overweight, except for breastfeeding, which had a negative relationship. In this model, the relationship between parity and overweight was significant and positive; the biggest difference in the odds ratios (ORs) was between the parity 0 vs parity 1 groups. The odds of being overweight were 35% lower among women of parity 0 compared with those with parity 1 (OR = 0.65, 95% CI: 0.62–0.69). The odds of being overweight were 12% higher among women with 2–3 live births compared with those with parity 1, and 5% higher for women with four or more live births, compared with women with parity 1, OR = 1.12 (1.08–1.17) and OR = 1.05 (1.00, 1.10), respectively. The odds of being overweight also was positively associated with HDI tertile; there was over a 3-fold difference in the ORs for overweight between the 2nd and 3rd HDI tertiles, 2.07 (1.95–2.20) and 7.35 (6.95–7.78), respectively.
Table 3 shows the relative odds of overweight by parity level and HDI tertile. The results that are presented were generated by adding interaction terms for parity and HDI tertiles to the main effects model. All of the ORs that are presented are with reference to one group: women with one live birth, living in countries in the lowest HDI tertile. Most of the ORs presented were significant as a result of the large sample sizes within each group. Thus, we were interested in examining meaningful differences between ORs as a function of parity and national human development. There was virtually no relationship between parity and overweight in the first two HDI tertiles. In the third tertile, there was a strong, positive relationship between parity and overweight, with a 2.4-fold difference between the highest (4+) and lowest (0) parity categories.
|
Table 4 shows the results of the three-way interaction between parity, household wealth and national human development. The relative odds of overweight by parity and tertile of household wealth are shown by tertile of HDI. The reference category within each HDI tertile was the group of women with one live birth, in the lowest tertile of household wealth. Overall, there was a negative relationship between parity and overweight among the poorest women living in the countries in the lowest HDI tertile, and a positive relationship between parity and overweight among the wealthiest women in the same countries. Women in the middle of the wealth distribution showed no association between parity and overweight. In the 2nd HDI tertile, the same pattern was evident: a negative association between parity and overweight among the poorest women, a positive association among the wealthiest women, and no association among women in the middle. In the 3rd HDI tertile, a different pattern was observed. Among the poorest women there was an increase in the OR of overweight from parity 0 to parity 1, and no differences at the higher parity categories. In both the 2nd and 3rd household wealth tertiles, higher parity was associated with higher odds of overweight.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Our goal was to examine whether the parity–overweight relationship varied by the level of household wealth and national human development. We found that the parity–overweight relationship was dependent on the relative wealth of a woman within her own country, as well as the aggregate level of human development of the country in which she lived. In the first two HDI tertiles, the negative relationship between parity and overweight among the poorest women was consistent with reproductive behaviour having a depleting or wasting effect. In the highest HDI tertile, parity was positively related to overweight among women in the 1st and 2nd household wealth tertiles. Among the poorest women, this positive relationship was seen only for women with no live births compared with women with at least one live birth. Among the wealthiest women, parity was positively related to overweight, regardless of the country's level of development.
A recent British study found a relationship between parity and measures of fatness (BMI and waist–hip ratio) in both men and women aged 60–79 years.30 Each additional child was associated with 0.12 and 0.36 BMI units in men and women, respectively. Adjustment for confounding attenuated the findings for waist–hip ratio to the null in men only but did not alter those for BMI. The authors suggested that lifestyle factors associated with having large families may lead to obesity in both genders. However, other factors must also be operating because of the larger coefficients associated per additional child in women as opposed to men. These results prompted us to conduct similar analyses for men as we did for women. However, only the Guatemala data set had anthropometric data for men. We did not find any evidence of a relationship between parity and overweight in men (results not shown; available upon request). The differences in setting (e.g. developing country vs the UK), study design (e.g. longitudinal vs cross-sectional) and control variables included in the analyses may explain the different findings. Because we were able to examine the parity–overweight relationship for men only in one of the countries included in this analysis, we are limited in the conclusions we can draw about the parity–overweight relationship in men vs women. The finding of a relationship between overweight and parity among women but not men suggests the influence of biological factors, although the role of gender-specific lifestyle factors cannot be discounted.
In a separate analysis, we ran the same models using linear regression, to determine the magnitude in average weight change corresponding to the ORs for overweight in Table 4. In HDI tertile 1, among the poorest women, for a woman of height 1.56 m (the median value in the sample), a change in parity from 0 to 4+ corresponded to an average decrease of 0.4 kg. In HDI tertile 2, for the same group, the same comparison corresponded to an average decrease of 2.2 kg, whereas in HDI tertile 3 there was an increase of 0.9 kg. In HDI tertile 1, among the wealthiest women, a change in parity from 0 to 4+ corresponded to an average increase in weight of 2.2 kg. For the same comparisons, in HDI tertile 2 there was an increase of 3.7 kg, and in tertile 3 there was an increase of 6.1 kg. Thus, there was no clear pattern among the poorest women, but among the wealthiest women these are important effects that increase with level of human development.
Existing research on the parity–overweight relationship has mainly come from developed countries. Evidence points to cumulative cycles of post partum weight retention as the mechanism by which parity leads to overweight.7,8,31,32 Excess gestational weight gain has been shown to lead to post partum weight retention and has been associated with increased weight gain in the long-term.8,9,31,33,34 Lifestyle factors, including physical activity and dietary intake in the post partum period have also been associated with post partum weight retention.31,35 Because of the nutrition transition, these same mechanisms may be now operating in some developing countries. For example, a recent Brazilian study found that gestational weight gain above recommendations was associated with significantly more weight retention at 9 months post partum than gaining within or below the guidelines, independent of pre-pregnant BMI. The same study also found that weight gain above recommendations was common in this sample of 208 women (28.8%), especially among overweight women (50%).36 In a longitudinal study of 169 pregnant and non-pregnant non-lactating (NPNL) urban women in Colombia, researchers found that by the end of the subjects’ pregnancies, the pregnant women had spent significantly less time in two energy expending activities and significantly more time in two energy-saving activities than the NPNL women.37 They concluded that decreasing activity was the main way the pregnant women met the energetic needs of pregnancy. If these women continued the pattern of decreased activity after giving birth, post partum weight retention could occur.
We found that among wealthier women, parity was positively related to overweight, regardless of the level of national human development. This finding suggests that wealthier women, even in the least-developed countries, were in positive energy balance, i.e. they had ample energy to meet their needs, which may explain the positive relationship between parity and overweight. In her review of the physiology of pregnancy, Janet King stated that ‘[e]nergy in excess of fetal needs has a high potential to be stored as maternal fat because of the overall anabolic milieu of pregnancy’.38
At the highest level of national human development, among the low- and middle-income countries we studied, the parity–overweight relationship included not only the wealthiest women, but also women in the 2nd tertile of household wealth. The poorest women also showed a higher OR for overweight at parity 1 compared with no live births. In studies done in the United States, high-gestational weight gain, post partum weight retention and factors that affect those outcomes have been associated with low SES.31,39,40 To our knowledge, the present study is the first to examine the influence of SES on risk factors for parity-related overweight in low- and middle-income countries. However, literature is available regarding the relationship between SES and obesity in general. In a recent review that examined the relationship between SES and obesity within developing countries as a function of national income, using data from several single and multicountry studies published between 1989 and 2003, Monteiro et al.41 concluded that in the poorest countries, high SES was positively associated with obesity and low SES was protective. However, in countries at the highest end of national income represented, obesity was more common in the low SES groups.
Certain limitations of this analysis merit remark. First, because this analysis relied on cross-sectional data, causality cannot be inferred as we cannot establish the temporal ordering between some covariates and the outcome of interest (e.g. household wealth and BMI were both measured at the time of the survey). Second, measured weight before reproduction began was not available, and this variable may play an important role in parity-related overweight and obesity.3,16 Third, the total number of pregnancies or the total number of live and still births may more fully represent the reproductive experiences of women better than the number of live births, but these variables were not available in every data set and thus could not be used in the analysis. Fourth, a disproportionate number of women in the sample came from selected countries (e.g. India, n = 72 769 women, representing 26% of the total sample). To explore the effects of the high representation of Indian women, we ran the same analyses without India and found similar results, indicating that the large sample of Indian women did not bias the results. Fifth, because three regions included only a small number of countries, we were unable to separate the effects of region and national human development on the parity–overweight relationship. As data become available from a greater number of developing countries, it may be possible to explore regional differences, independent of national human development, that affect the parity–overweight relationship. Sixth, there may be important, sub-national differences by ethnicity in the parity–overweight relationship, and this would be an interesting topic for future research that did not have as an explicit objective broad, cross-national comparisons. Finally, neither data on income nor consumption were available in the data sets analysed; therefore, individual wealth was approximated by a household wealth index. This methodology was developed by the World Bank and ORC Macro, Inc. based on research in developing countries that showed that the wealth index was a good proxy for consumption.26 Each woman was assigned the wealth score of the household in which she lived. Implicit in this procedure is the assumption that women have full and equal access to the wealth and resources of the household. This assumption may not be true because women live in households of varying size. Also, in some cultures, women may have less power in the household to negotiate access to these and other household resources that are associated with their BMI. Additionally, the degree to which household sizes and power structures affect the relative wealth of a woman may vary by level of household wealth, region and level of country development.42 Future analyses might use a measure for relative wealth or resources that accounts for household size and/or women's status in the household. Further work that examines whether the burden of parity-related overweight shifts away from women of high SES, or if it simply broadens to include women of all SES groups as national human development increases also will be useful for policy makers in low- and middle-income countries.
This analysis also had many strengths. It included a wide range of countries from different regions and levels of national human development in the world, as represented by the HDI. The HDI values ranged from 0.302 to 0.754. When the analysis was done using a more comprehensive ‘development index’ that was used in our previous research,19 the results did not change appreciably. Multiple geographic regions were represented, and the data set analysed had a very large sample size.
Our results support family planning initiatives in developing countries. The United Nations has estimated that about 20% of women in developing countries have an ‘unmet need’ for family planning,43 and prevention of parity-related overweight provides another argument for increased efforts to make family planning available when desired by women. Although the nature of family planning programmes may differ in countries where fertility rates are near replacement levels, our findings show that parity-related overweight may affect even women who choose to have only one child. Our results show that the odds of overweight were 61% higher for women with one live birth compared with women with no live births, among women in the highest level of national human development (Table 3). When women in the highest HDI tertile were examined by household wealth, the increase was even greater, 67 and 80% for women in the 2nd and 3rd tertiles of household wealth, respectively (Table 4). These findings suggest an urgent need for the development and implementation of programmes to prevent and treat parity-related overweight in some developing countries. Such programmes could include efforts to prevent excessive gestational weight gain, and/or to prevent post partum weight retention by promoting exclusive breastfeeding concomitant with moderate levels of physical activity. In countries at the lowest level of human development, programmes to prevent parity-related overweight should target the wealthiest women, whereas in more developed countries such programmes should be made available to all women.
KEY MESSAGES
|
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
This study is supported by NIH R01 TW005598.
The authors would like to thank Bridgette James and Kiersten Johnson at ORC Macro, Inc. for their help in working with DHS. We thank Paul Stupp and Alicia Ruiz at CDC and Rafael Flores at Emory University for their help in working with RHS. We also thank Meng Wang for support with data analysis.
Conflict of interest: None declared.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
1 CDC. Table 69, Health, United States, 2004 With Chartbook on Trends in the Health of Americans. (2004) Hyattsville, Maryland: National Center for Health Statistics. 2004.
2 Nutrition for Improved Development Outcomes. In: Standing Committee on Nutrition, 2004. 5th Report on the World Nutrition Situation, March: Geneva: United Nations.
3 Linné Y, Dye L, Barkeling B, Rössner S. Weight development over time in parous women-The SPAWN Study-15 years follow-up. Int J Obes (2003) 27:1516–22.[CrossRef][ISI][Medline]
4 Bastian LA, West NA, Corcoran C, Munger RG. Number of children and the risk of obesity in older women. Prev Med (2005) 40:99–104.[CrossRef][ISI][Medline]
5 Linné Y, Barkeling B, Rössner S. Long-term weight development after pregnancy. Obes Rev (2002) 3:75–83.[CrossRef][Medline]
6 Szklarska A, Jankowska EA. Independent effects of social position and parity on body mass index among Polish adult women. J Biosoc Sci (2003) 35:575–83.[CrossRef][ISI][Medline]
7 Gunderson EP, Abrams B. Epidemiology of gestational weight gain and body weight changes after pregnancy. Epidemiol Rev (1999) 21:261–75.
8 Rooney BL, Schauberger CW. Excess pregnancy weight gain and long-term obesity: one decade later. Obstet Gynecol (2002) 100:245–52.
9 Winkvist A, Rasmussen KM, Lissner L. Associations between reproduction and maternal body weight: examining the component parts of a full reproductive cycle. Eur J Clin Nutr (2003) 57:114–27.[CrossRef][ISI][Medline]
10 Popkin BM. Urbanization, lifestyle changes, and the nutrition transition. World Development (1999) 27:1905–16.[CrossRef][ISI]
11 Popkin BM. The nutrition transition and its relationship to demographic change. In: Nutrition and Health in Developing Countries—Semba RD, Bloem MW, eds. (2001;) Totowa: Humana Press Inc. 427–45.
12 Popkin BM. An overview on the nutrition transition and its health implications: the Bellagio meeting. Public Health Nutr (2002) 5:93–103.[CrossRef][ISI][Medline]
13 Uauy R, Albala C, Kain J. Obesity trends in Latin America: transiting from under- to overweight. (2001) Paper presented at: Experimental Biology, April 15–April 19: San Diego, CA.
14 Drewnowski A, Popkin BM. The nutrition transition: new trends in the global diet. Nutr Rev (1997) 55:31–43.[ISI][Medline]
15 Arroyo P, Avila-Rosas H, Fernández V, Casanueva E, Galván D. Parity and the prevalence of overweight. Int J Gynaecol Obstet (1995) 48:269–72.[CrossRef][Medline]
16 Coitinho DC, Sichieri R, D’Aquino Benicio MH. Obesity and weight change related to parity and breast-feeding among parous women in Brazil. Public Health Nutr (2001) 4:865–70.[ISI][Medline]
17 Loaiza E. Maternal nutritional status. In: Demographic and Health Surveys Comparative Studies (1997) Calverton, Maryland: ORC Macro, Inc. No. 24. December.
18 Wen W, Gao YT, Shu XO, et al. Sociodemographic, behavioral, and reproductive factors associated with weight gain in Chinese women. Int J Obes (2003) 27:933–40.[CrossRef][ISI][Medline]
19 Kim SA, Stein AD, Martorell R. Country development and the association between parity and overweight. Int J Obes. October 24, 2006 [Epub ahead of print].
20 Monteiro CA, Conde WL, Popkin BM. The burden of disease from undernutrition and overnutrition in countries undergoing rapid nutrition transition: a view from Brazil. Am J Public Health (2004) 94:433–34.
21 Sampling Manual. (1996) Vol. 6. Calverton, Maryland: Macro International Inc.
22 ORC Macro, 2006. In: MEASURE DHS STATcompiler. Available at: www.measuredhs.com/statcompiler. Accessed February 13, 2006.
23 Rosner B. Fundamentals of Biostatistics (2000) 5th edn. Boston: Duxbury Press.
24 ORC Macro, Inc. In: Description of the Demographic and Health Surveys Individual Recode Data File (1996).
25 WHO. Obesity - Preventing and Managing the Global Epidemic. (1998) Report of a WHO Consultation on Obesity, June 3–5, 1998. Geneva: World Health Organization.
26 ORC Macro, Inc. In: DHS and World Bank use wealth index to measure socioeconomic status (2002) Vol. 4. 1–2, DHS+ Dimensions.
27 Gwatkin D, Rustein S, Johnson K, Pande R, Wagstaff A. Socioeconomic differences in health, nutrition, and population in Bangladesh (2000) Washington, DC: The World Bank.
28 Soubbotina T. Beyond Economic Growth, Second Edition: An Introduction to Sustainable Development (2004) 2nd edn. The World Bank.
29 Fukuda-Parr S. Human Development Report 2004 (2004) New York: United Nations Development Programme.
30 Lawlor DA, Emberson JR, Ebrahim S, et al. Is the association between parity and coronary heart disease due to biological effects of pregnancy or adverse lifestyle risk factors associated with child-rearing?: findings from the British Women's Heart and Health Study and the British Regional Heart Study. Circulation (2003) 107:1260–64.
31 Olson CM, Strawderman MS, Hinton PS, Pearson TA. Gestational weight gain and postpartum behaviors associated with weight change from early pregnancy to 1 y postpartum. Int J Obes (2003) 27:117–27.[CrossRef][ISI][Medline]
32 Harris HE, Ellison GT, Holliday M. Is there an independent association between parity and maternal weight gain? Ann Hum Biol (1997) 24:507–19.[CrossRef][ISI][Medline]
33 Rössner S. Weight gain in pregnancy. Hum Reprod (1997) 1:110–15.
34 Butte NF, Ellis KJ, Wong WW, Hopkinson JM, Smith EO. Composition of gestational weight gain impacts maternal fat retention and infant birth weight. Am J Obstet Gynecol (2003) 189:1423–32.[CrossRef][ISI][Medline]
35 Olson CM. The Importance of Behaviors to Postpartum Weight Change. Cornell University—Cornell Cooperative Extension [Internet]. July 15, 2002, Accessed 2004.
36 Kac G, Benicio M, Velasquez-Melendez G, Valente J, Struchiner C. Gestational weight gain and prepregnancy weight influence postpartum weight retention in a cohort of Brazilian women. J Nutr (2004) 134:661–66.
37 Dufour DL, Reina JC, Spurr G. Energy intake and expenditure of free-living, pregnant Colombian women in an urban setting. Am J Clin Nutr (1999) 70:269–76.
38 King JC. Physiology of pregnancy and nutrient metabolism. Am J Clin Nutr (2000) 71(Suppl. 5):1218S–25S.
39 Wolfe WS, Sobal J, Olson CM, Frongillo EA Jr, Williamson DF. Parity-associated weight gain and its modification by sociodemographic and behavioral factors: a prospective analysis in US women. Int J Obes Relat Metab Disord (1997) 21:802–10.[CrossRef][ISI][Medline]
40 Murphy SP, Abrams BF. Changes in energy intakes during pregnancy and lactation in a national sample of US women. Am J Public Health (1993) 83:1161–63.
41 Monteiro CA, Moura EC, Conde WL, Popkin BM. Socioeconomic status and obesity in adult populations of developing countries: a review. Bull World Health Organ (2004) 82:940–46.[ISI][Medline]
42 Smith L, Ramakrishnan U, Ndiaye A, Haddad L, Martorell R. The Importance of Women's Status for Child Nutrition in Developing Countries (2003.) Washington DC: International Food Policy Research Institute. 131.
43 World population monitoring 2002. In: Reproductive Rights and Reproductive Health (2004) New York: United Nations.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  E Koch, M Bogado, F Araya, T Romero, C Diaz, L Manriquez, M Paredes, C Roman, A Taylor, and A Kirschbaum Impact of parity on anthropometric measures of obesity controlling by multiple confounders: a cross-sectional study in Chilean women J. Epidemiol. Community Health, May 1, 2008; 62(5): 461 - 470. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. Mishra and D. Kuh Commentary: The relationship between parity and overweight--a life course perspective Int. J. Epidemiol., February 1, 2007; 36(1): 102 - 103. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||