Household food insecurity and excess weight/ obesity among Brazilian women and children: a life-course approach

Household food insecurity (HFI) may increase obesity risk, but results are not consistent across the life course or between developed/underdeveloped settings. The objective of this paper is to review findings from previous analyses in Brazil among adult women, female adolescents, and children up to five. Data were derived from the 2006 Brazilian Demographic and Health Survey. Associations between HFI (measured with the Brazilian Food Insecurity Scale) and excess weight/obesity were investigated through Poisson regression models. While severe HFI was associated with obesity risk among adult women (PR: 1.49; 95%CI: 1.17-1.90), moderate HFI was associated with excess weight among female adolescents (PR: 1.96; 95%CI: 1.18-3.27). There was no association between HFI and obesity among children (either boys or girls). The nutrition transition in Brazil may be shaping the differential deleterious effect of HFI on body fat accumulation across the life course; the association is already evident among female adolescents and adult women but still not among children. Nutritional Transition; Food Security; Nutritional Status; Life Cycle Stages DEBATE DEBATE


Household food insecurity and excess weight/ obesity among Brazilian women and children: a life-course approach
Insegurança alimentar e excesso de peso/ obesidade entre mulheres e crianças brasileiras: uma abordagem para o curso da vida Inseguridad alimentaria y exceso de peso/ obesidad entre mujeres y niños brasileños: un enfoque para el transcurso de la vida Abstract Household food insecurity (HFI) may increase obesity risk, but results are not consistent across the life course or between developed/underdeveloped settings. The objective of this paper is to review findings from previous analyses in Brazil among adult women, female adolescents, and children up to five. Data

Introduction
Household food insecurity (HFI) is related to poverty, which in turn has traditionally been linked to undernutrition. However, the nutrition paradox proposed over fifteen years ago 1 considered the hypothesis of an association between HFI and excess weight, a possibility that since then has been at least partially supported by empirical evidence predominantly generated in the USA 2 .
Research shows that women living with HFI are more likely to present excess weight when compared to their food-secure counterparts. There is also growing evidence of a positive association between HFI and obesity among adolescents. For children, support of an association between HFI and weight status has been mixed, and there is sparse evidence for the adult male population. Furthermore, there is little evidence that HFI promotes increased weight gain over time because most data were derived from crosssectional studies 3,4,5,6 .
With obesity rates rising worldwide and the eradication of HFI far beyond reach, a complete understanding of the relationship between these two apparently paradoxical conditions is essential for the development, implementation, and expansion of effective policies and programs 7 .
The goal of this study is to present a summary of results on the association between HFI and excess weight/obesity, using data from the 2006 Brazilian Demographic and Health Survey (DHS), following a life-course approach, examining data from adult women, female adolescents, and children under five 8,9,10 .

Methods
Three different sub-samples derived from the 2006 Brazilian DHS were analyzed: adult women (18-49 years); female adolescents (15-19 years); and children of both sexes (0-60 months). The Brazilian DHS is a nationally representative cross-sectional household survey with a complex sampling design. The DHS collected data from childbearing-age women, including mothers and their children aged 0-5 years 11,12 . Detailed information on study design and data collection can be found in the publications used as the basis for the summary results presented here 8,9,10 .
The study outcomes were excess weight/ obesity, defined in different ways for each subset. For children, body mass index/age z-scores (BMI-Z) were calculated according to the World Health Organization (WHO) growth reference standards 13 . BMI-Z can measure either excess or insufficient weight for a certain length or height and is expressed as sex-and age-(in months) specific z-scores. Obesity was defined as BMI-Z > 2 standard deviations (SD) 8 . For female adolescents aged 15-18 years, the WHO age-adjusted BMI growth charts were used to classify nutritional status. The 85 th percentile was defined as the cut-off for excess weight. For adolescents aged 19 years we defined excess weight as BMI ≥ 25kg/m 2 9 . For adult women, obesity was defined as BMI ≥ 30kg/m 2 10 .
HFI level (security, mild insecurity, moderate insecurity, severe insecurity) was the key independent variable and was measured with the previously used and extensively validated Brazilian Food Insecurity Scale (EBIA). EBIA is an adaptation of the US Household Food Security Survey Module (US-HFSSM), developed during the early 1990s and first fielded in the 1995 US Current Population Survey. The detailed description of the adaptation and validation of the EBIA scale can be found elsewhere 14, 15 . The EBIA is composed of 15 dichotomous (yes/no) questions, which evaluate food insecurity experiences representing different levels of food insecurity. Each household is assigned a summative food insecurity score based on the number of affirmative answers to the scale items. Households were classified either as food secure (HFI score = 0), mildly food insecure (score = 1-5), moderately food insecure (score = 6-10), or severely food insecure (score = 11-15).
Statistical analyses comprised the prevalence distribution of HFI and excess weight/obesity for each of the age groups. Subsequently, associations between HFI and the outcomes were investigated through adjusted prevalence ratios (PR) with 95% confidence intervals (CI) using multivariate Poisson regression models. The following confounders were considered in the multivariate analysis (i) children: geographic region of the household (North, Northeast, Southeast, South, Midwest), area of residence (urban, rural), family income quartiles, mother's skin color (white, black, brown, yellow, indigenous), maternal schooling (0-4, 5-8, ≥ 9 years), maternal smoking (yes, no), maternal marital status (single/ widowed/divorced or married/cohabiting), sex, age, birth order, and number of children aged 0-60 months living in the household; (ii) female adolescents: geographic region of the household, area of residence, per capita family income (log-transformed), skin color, years of schooling, smoking, marital status, number of people living in the household, and age; and (iii) adult women: geographic region of the household, family income quartiles, skin color, years of schooling, marital status, and age. These covariates were selected based on theoretical and empirical con-siderations and allow for direct comparison between studies.
Analyses on obesity prevalence and the Poisson regression model for children are new. All other estimates presented in the paper have been reported previously 8,9,10 . All estimates were weighted, and standard errors were corrected to account for the complex sampling design by "svy" commands in Stata version 10.1 (Stata Corp., College Station, USA).
The relationship between HFI and obesity was not significant for children ( Figure 2). Among female adolescents, the likelihood of excess weight was 1.96 times higher among those with severe HFI when compared to their food-secure counterparts (Figure 3). Among adult women, the likelihood of obesity was 1.49 times higher among those with moderate HFI when compared to their food-secure counterparts (Figure 4).

Discussion
Following a life-course approach, this paper shows that Brazilian adult women living in moderately food-insecure households had a 49% higher risk of being obese, and female adolescents living in severely food-insecure households were almost twice as likely to show excess weight when compared to their food-secure counterparts. No association was shown between HFI and obesity among Brazilian children (either boys or girls) aged 0-60 months.
The study has some limitations, namely the inability to ascertain the temporal direction of Note: fi gure constructed with data from Kac et al. 8,9 and Velásquez-Meléndez et al. 10 . Table 1 Prevalence of body mass index/age in z-score (BMI-Z) > 2 standard deviations (SD), for children (0-60 months), excess weight in female adolescents (15-19 years), and obesity in adult women (18- 9 and Velásquez-Meléndez et al. 10 .
* Prevalence ratios were adjusted for geographic region of the household, area of residence (urban, rural), income quartiles, mother's self-reported skin color, maternal schooling, maternal smoking, maternal marital status, sex, age, birth order, and number of children 0-5 years living in the household; ** Estimates were weighted and standard errors were corrected to take the complex sampling design into account. *** Estimates were weighted and standard errors were corrected to take the complex sampling design into account.
Note: fi gure constructed with data from Kac et al. 9 . * Prevalence ratios were adjusted for skin color, schooling, geographic region, income quartiles, age, and marital status; ** Obesity was defi ned as BMI ≥ 30kg/m 2 ; *** Estimates were weighted and standard errors were corrected to take the complex sampling design into account.
associations between HFI and excess weight/ obesity (a characteristic of cross-sectional design). In addition, since population-based surveys with complex samples have broad and general goals, important potential mediators or confounders were not measured (e.g. self-rated health, dietary intake, alcohol consumption, and physical activity). However, estimates for adult women and female adolescents remained statistically significant after adjusting for several known potential confounders, including geographic region of the household, family income, skin color, schooling, marital status, and age. The results agree with those of previous studies using U.S. population samples, in which mild HFI was associated with obesity in adult women 16,17,18 . However, in other low and middle-income countries experiencing early stages of the nutrition transition, such as Trinidad and Tobago 19 and Colombia 20 , HFI was associated with undernutrition rather than obesity. This could be due to the fact that Brazil is at a more advanced stage in the nutrition transition, with higher prevalence rates for obesity than for underweight in poor women 21 .
A possible explanation for the difference in the level of HFI at which adolescent and adult women are affected (i.e. severe HFI for adolescents and moderate HFI for adults) lies in the physiological changes related to puberty which make adolescents more resistant to body fat accumulation. By contrast, adult women's obesity risk increased with moderate HFI. In addition parents may also protect children and adolescents from some of the harmful effects of HFI. It is also possible that strong concerns with body image, common among young Brazilian women 22,23 , may protect them from the obesogenic influence of HFI until the problem becomes so severe that they can no longer cope without changing lifestyle behaviors, which lead to body fat accumulation.
To explain the paradoxical association between HFI and excess weight/obesity in this study, we hypothesize that poorer families are purchasing high-fat and energy-dense foods, following classic economic consumer theories 24,25 . Furthermore, due to their limited access to food, when any foods are available these individuals may overeat or even eat foods they normally would not have access to in an attempt to compensate for periods of scarcity 3 , despite their ability to identify healthy foods 26 . In this sense, the paradox would be directly related to the way different countries experience the nutrition transition; only when this process reaches a stage at which energy-dense foods become available at affordable prices, HFI becomes a risk factor for excess weight/obesity. Moreover, Brazil's current stage of nutrition transition may also help explain the lack of association between HFI and obesity among young children. It has been suggested that the dynamics regarding the occurrence of a shift from undernutrition to obesity tend to follow a path that starts among adults, followed by adolescents, and only later affects children 27 . This nutrition transition pattern has been documented previously in the Brazilian population by studies using data from the most important nationally representative surveys in the last two decades 21,27 . The possible effect of HFI on obesity in children may constitute a feature of a very advanced stage of the nutrition transition that Brazil has still not reached. However, this hypothesis has not been confirmed in the U.S. population. Alternatively, the relatively high-energy requirements per unit of body weight may protect children against the negative influence of HFI on obesity.
In conclusion, the nutrition transition in Brazil may be shaping the differentially deleterious effect of HFI on body fat accumulation over the life course. This may explain why the association between HFI and excess weight/obesity is still not evident among Brazilian children but is found among female adolescents and adult women. Cad