Household food insecurity is particularly prevalent in developing countries(1). Even in developed countries, household food insecurity has been strongly associated with poor health status of adults(Reference Stuff, Casey, Szeto, Gossett, Robbins, Simpson, Connell and Bogle2–Reference Hadley and Patil9) and children(Reference Cook, Frank and Berkowitz10–Reference Casey, Szeto, Lensing, Bogle and Weber13). In the USA and Canada, adults in food-insecure households were significantly more likely to rate their health as poor than adults in food-secure households(Reference Stuff, Casey, Szeto, Gossett, Robbins, Simpson, Connell and Bogle2–Reference Vozoris and Tarasuk4). Household food insecurity was associated with poorer health status(Reference Cook, Frank and Berkowitz10), higher rate of Fe-deficiency anaemia(Reference Skalicky, Meyers, Adams, Yang, Cook and Frank11), more frequent stomach aches and headaches(Reference Alaimo, Olson and Frongillo12) and higher rate of hospitalization(Reference Cook, Frank and Berkowitz10) among low-income infants and young children in food-insecure households compared with their counterparts in food-secure households.
Although the association between household food insecurity and adverse health outcomes of adults is established, its relationship to nutritional status is not. In most(Reference Olson5, Reference Townsend, Peerson, Love, Achterberg and Murphy7, Reference Townsend14) but not all(Reference Lee and Frongillo3) studies in developed countries, household food insecurity has been associated with overweight and obesity of adults. Household food insecurity was associated with obesity among rural women in Malaysia(Reference Shariff and Khor15) but instead with underweight among adults in Trinidad(Reference Gulliford, Mahabir and Rocke8). These results suggest that the direction of the association between household food insecurity and nutritional status of adults may vary depending on the context.
Among children, household food insecurity was associated with underweight, wasting and stunting in South Africa(Reference Kruger, Lemke, Phometsi, van’t Riet, Pienaar and Kotze16) and Pakistan(Reference Baig-Ansari, Rahbar, Bhutta and Badruddin17). On the other hand, household food insecurity was associated with obesity among urban children in Korea(Reference Oh and Hong18). The association between household food insecurity and the growth of infants and young children has not been studied well even in developed countries, much less in developing countries.
The purpose of the present research was to investigate the association between household food security and subsequent growth of infants and young children in rural Bangladesh. Use of longitudinal data and appropriate analytical techniques for longitudinal data permitted us to establish a plausible temporal relationship between household food security and growth by preventing a potential problem of reverse causality that may be associated with cross-sectional studies. The results from the study provide an understanding of the consequences associated with household food insecurity in relation to the growth of infants and young children in rural Bangladesh and other low-income countries.
Methods
Data
We used the data from the Maternal and Infant Nutrition Intervention in Matlab (MINIMat) study, a large intervention trial in Bangladesh in which infants were followed from birth until 24 months of age and beyond. The enrolment of pregnant women in the MINIMat study took place between November 2001 and October 2003. They were followed during pregnancy, when data on household food security and socio-economic status were collected. Weight and length measurements of the children were collected on seventeen occasions: at birth, monthly during the first 12 months of life and quarterly thereafter until the children were 24 months old. Weight was measured by electronic or beam scales, which were precise to 10 g (UNICEF Uniscale; SECA GmbH & Co., Hamburg, Germany). Locally manufactured, collapsible length boards, which were precise to 1 mm, were used to measure recumbent length.
MINIMat study
In the MINIMat study, pregnant women were randomly assigned to either ‘usual start’ or ‘early start’ of food supplementation just after identifying and confirming their pregnancy at 6–13 weeks of gestation. The mothers of the children in the present study had all been randomized to the ‘usual start’ food supplementation group. There was an ongoing, government-supported, Bangladesh Integrated Nutrition Project/National Nutrition Project (BINP/NNP) that provided a food (energy–protein) supplement to pregnant and lactating women. The supplement provided 2512 kJ/d (600 kcal/d) for six days per week. It contained roasted rice powder (80 g), roasted lentil powder (40 g), molasses (20 g) and soyabean oil (12 ml). The BINP/NNP provided supplementation to pregnant women starting at about 17 weeks of gestation (usual start). All women were also further randomly assigned to receive a pill that contained one of three different combinations of micronutrients. The micronutrient combinations were: (i) 60 mg Fe plus 400 μg folic acid; (ii) 30 mg Fe plus 400 μg folic acid; and (iii) 30 mg Fe plus 400 μg folic acid with thirteen other micronutrients.
Informed consent was obtained from the women about their participation in the study. The study was approved by the Research Review Committee and Ethical Review Committee of the International Centre for Diarrhoeal Disease Research, Bangladesh and the Cornell University Committee on Human Subjects.
Selection of sample
We selected all infants (n 1343) whose mothers had been in the ‘usual start’ food supplementation group, regardless of micronutrient group (448 per group), and who were born by December 2003 and completed their 24-month follow-up in December 2005.
Household food security
We created a household food security scale based on the eleven items in the food security component questionnaire of the MINIMat study(Reference Frongillo, Chowdhury, Ekström and Naved19). These data covered frequency of food purchased (rice and perishable foods, e.g. vegetables, fish, meat); frequency of cooking, borrowing or lending (food and money); and whether there was ready access to adequate meals and snacks. This information was collected during household visits at 8 and 30 weeks of pregnancy. Scores were assigned to each item of the food security measure according to the classification of households and their experiences(Reference Frongillo, Chowdhury, Ekström and Naved19). Higher scores were assigned to those experiences that indicated better food security status.
Score on the food security scale ranged from 10 to 44. The mean food security score at 8 and 30 weeks was 38·6 (sd 3·6). A higher score on the scale indicated a better situation of household food security. We used this scale as a continuous as well as a categorical variable. For the categorical variable, households were divided into quartiles: 1 (extremely food-insecure), 2 (moderately food-insecure), 3 (occasionally food-insecure) and 4 (food-secure). The reliability (i.e. Cronbach’s alpha) of the sum of the scales at 8 weeks and 30 weeks was 0·78.
Anthropometric indices
Weight and length measurements were converted to Z-scores – weight-for-age Z-score (WAZ), length-for-age Z-score (LAZ), weight-for-length Z-score (WLZ) and BMI-for-age Z-score (BMIZ) – according to the 2006 WHO child growth standards(20). A Z-score cut-off for these indices of −2 was used for classifying children as undernourished. Underweight and stunting were defined as WAZ < −2 and LAZ < −2, respectively. Wasting was defined as WLZ < −2 or BMIZ < −2.
Statistical analysis
We used multilevel models for change to model weight and length trajectories of the infants and young children(Reference Singer and Willet21, Reference Peugh and Enders22). These models are appropriate for longitudinal data with repeated measures in which several individual measurements (follow-up measurements on weight and length) are nested within individuals (infants and children). The models included sub-models at two levels: (i) a level-1 sub-model that described how individual children changed over time (within child); and (ii) a level-2 sub-model that described how these changes varied across children (among children). These two components were combined to form the composite multilevel model for change with random intercepts and random coefficients for household food security. The model included interaction terms of food security with child age and age-squared to evaluate the effect of food security on attained growth, growth velocity and growth deceleration. Our dependent variable was growth status (weight, length and anthropometric indices) of infants and children from 1 to 24 months of age. Models were run with and without adjustment for child sex and maternal age. We also used χ 2 tests to examine whether the attained weight and length as well as anthropometric indices of the infants and children in different food security quartiles were statistically different from one another.
All statistical analyses were done using the mixed model procedure(Reference Singer and Willet21, Reference Peugh and Enders22) in the SPSS for Windows statistical software package version 14·0 (SPSS Inc., Chicago, IL, USA).
Results
General and baseline characteristics
Mean birth weight and birth length of the newborns was 2697 (sd 401) g and 47·8 (sd 2·1) cm, respectively. Mean birth weight of boys (2741 (sd 411) g) was greater (P < 0·001) than that of girls (2650 (sd 384) g). Mean birth length of boys and girls was 48·0 (sd 2·1) cm and 47·5 (sd 2·1 cm), respectively (P < 0·001). Overall, 30 % of the babies were born with low birth weight (<2500 g). The rate of low birth weight was higher (P < 0·01) for girls (34·5 %) than for boys (26·0 %). Mean age and weight of the mothers was 26·3 (sd 5·8) years and 46·5 (sd 6·6) kg, respectively. Median parity was 1 (range 0–7). The mean years of education completed by these women and their husbands was 5·0 (sd 4·1) years and 5·4 (sd 4·5) years, respectively.
Growth patterns
The children were small-for-age from birth to 24 months. Attained body weight at 6, 12 and 24 months was 6·7 (sd 0·9) kg, 7·9 (sd 1·1) kg and 9·7 (sd 1·3) kg, respectively. Growth in weight from 1 to 6, 12 and 24 months was 3·1, 4·3 and 6·1 kg, respectively. Compared with the 2006 WHO standards, mean weight deficit at 6, 12 and 24 months was 0·9, 1·4 and 2·1 kg, respectively. Attained length at 6, 12 and 24 months was 63·5 (sd 2·6) cm, 70·7 (sd 2·8) cm and 80·4 (sd 3·5) cm, respectively. Growth in length from 1 to 6, 12 and 24 months was 12·0, 19·2 and 28·9 cm, respectively. Growth faltering of these children occurred at 6 months of age or later and no catch-up growth was achieved before 24 months.
The mean WAZ was about −1·5 from 1 to 24 months of age. Both boys and girls had mean WAZ greater than −1·5 during the first year, but less than −1·5 during the second year of life. The mean LAZ was between −1 and −2 during the first year of life, but less than −2 during the second year of life. Girls had greater LAZ, particularly before 15 months of age. After this age, both groups showed a decline towards the value of −2. Mean WLZ for both boys and girls was greater than zero at 1 month and decreased gradually to −0·2 by 6 months of age. After this age, both groups declined towards a value of −1 that plateaued from 18 months onwards. All mean anthropometric indices of the infants and children decreased with age.
Food security and growth
Household food security was significantly associated (P < 0·001) with gain in weight, length and BMI from 1 to 24 months of age, as indicated by the interaction terms of food security with age and age-squared (Table 1). The interaction terms mean that higher food security was associated with greater gain and greater degree of curvature. These associations remained in the models not adjusted for child sex and maternal age (data not shown). Similarly, household food security was also significantly associated with higher gain in all anthropometric indices (WAZ, LAZ, WLZ and BMIZ) from 1 to 24 months of age (data not shown).
Multilevel model effects of household food security on gains in weight, length and BMI of infants and young children (n 1343) from 1 to 24 months of age: Maternal and Infant Nutrition Intervention in Matlab (MINIMat) study in rural Bangladesh
Association was statistically significant: ***P < 0·001.
Food security status and anthropometric indices
There was a gradient in WAZ of children from 1 to 24 months of age in the four quartiles of household food security status (Fig. 1(a)). The WAZ values of the children differed significantly (P < 0·001) among the four categories of household food security status – the food-secure children having the highest and the extremely food-insecure children having the lowest WAZ. Similar patterns were found for LAZ of children in the four categories of households (Fig. 1(b)). The WLZ values of children in the food-secure households were higher (P < 0·001) than those of the other three groups of children, but the WLZ values of the children in the three groups of food-insecure households were similar (data not shown). Similar patterns were also observed for BMIZ of children in the four groups of food security status. Similar gradients with the four categories of household food security were found with attained weight, length and BMI (data not shown).
(a) Mean weight-for-age Z-score (WAZ) and (b) mean length-for-age Z-score (LAZ) of infants and children (n 1343) from 1 to 24 months of age in four categories of household food security (—●—, extremely food-insecure; — –○—, occasionally food-insecure; – –▾– –, moderately food-insecure; - -▵- -, food-secure): Maternal and Infant Nutrition Intervention in Matlab (MINIMat) study in rural Bangladesh
Food security and prevalence of nutritional indicators
Proportions of underweight children from 1 to 24 months of age differed significantly (P < 0·05) among the four groups of household food security status, particularly after 6 months of age (Fig. 2(a)). There was a gradient in the proportion of underweight among children from 1 to 24 months of age – food-secure households had the lowest and the extremely food-insecure households the highest proportion. Similar patterns were found for the rates of stunting among children from 1 to 24 months of age (Fig. 2(b)). Although the proportions of wasting based on WLZ < −2 differed among the four groups of households, these differences were statistically significant (P < 0·05) only during the second year of life (data not shown). In contrast, the proportions of wasting according to BMIZ < −2 differed significantly among the four groups of households during the second half of infancy.
Proportions of (a) underweight (weight-for-age Z-score <−2) and (b) stunting (length-for-age Z-score <−2) among infants and children (n 1343) from 1 to 24 months of age in four categories of household food security (—●—, extremely food-insecure; — –○—, occasionally food-insecure; – –▾– –, moderately food-insecure; - -▵- -, food-secure): Maternal and Infant Nutrition Intervention in Matlab (MINIMat) study in rural Bangladesh
Discussion
We examined the association between household food security and subsequent growth of infants and children from 1 to 24 months of age in rural Bangladesh. Use of longitudinal data and multilevel models for change permitted us to examine growth over time rather than at just a single point. The nature of the data and the analytical techniques increased the plausibility of the association between household food security and infant and young child growth by reducing the potential for reverse causality(Reference Singer and Willet21, Reference Peugh and Enders22). Our results suggested that there is a strong association between household food security and growth of these rural Bangladeshi infants and young children.
More than 80 % of the total population in Bangladesh lives in rural areas and directly or indirectly depends on agriculture(23). Poverty is deep-rooted all over Bangladesh, particularly in rural areas, resulting in greater vulnerability to adverse health and nutritional status among infants and children. Poor households in Bangladesh are food-insecure because they lack sufficient food from own production and cash income and other resources to acquire enough food. Therefore, the context was an impetus for conducting the present research on the relationship between household food security and growth of infants and young children in a rural area of Bangladesh.
Association between household food security and growth
Household food security was strongly associated with higher gain in weight and length from 1 to 24 months of life in this sample. In contrast, household food insecurity was associated with overweight and obesity among children in the USA(Reference Casey, Szeto, Lensing, Bogle and Weber13, Reference Casey, Simpson and Gossett24–Reference Matheson, Varady, Varady and Killen26) and Korea(Reference Oh and Hong18). It is not surprising that household food insecurity was associated with poorer growth of these Bangladeshi infants and young children in such a resource-poor environment.
There are multiple pathways by which household food security could be associated with growth. Several studies suggested that household food insecurity, as a stressor, was associated with maternal anxiety and depressive illnesses(Reference Hadley and Patil9, Reference Whitaker, Phillips and Orzol27) that contributed to unfavourable parenting practices(Reference McLean, Minkovitz, Strobino, Marks and Hou28) including a negative effect on breast-feeding(Reference Henderson, Evans, Straton, Priest and Hagan29). High maternal depression was associated with poor child nutritional status in India and Vietnam(Reference Harpham, Huttly, De Silva and Abramsky30). Maternal–infant interaction during feeding was associated with household food insecurity in the present population(Reference Frith31). In other MINIMat analyses, we found that household food security was associated with better feeding practices during the second half of infancy(Reference Saha, Frongillo, Alam, Arifeen, Persson and Rasmussen32). Each of these factors could potentially explain the association between household food security and growth.
Food security scale
We used the food security scale that was based on in-depth understanding of the experience of food insecurity developed through qualitative research in rural areas of Bangladesh(Reference Frongillo, Chowdhury, Ekström and Naved19). Such locally developed food insecurity measures can validly assess the status of household food security in developed, as well as in developing countries(Reference Frongillo, Rauschenbach, Olson, Kendall and Colmenares33, Reference Coates34). In Burkina Faso, an experience-based measure of household food insecurity was valid within and across seasons(Reference Frongillo and Nanama35). The nine themes of the food insecurity scale of the present study were similar to those suggested by a study done in Bangladesh by Coates et al.(Reference Coates, Wilde, Webb, Rogers and Houser36). They found a high degree of agreement between the items of food insecurity measures identified by qualitative and quantitative methods. In the food insecurity scales developed from both studies in Bangladesh(Reference Frongillo, Chowdhury, Ekström and Naved19, Reference Coates, Wilde, Webb, Rogers and Houser36), all of the items were related to one of the four domains – uncertainty and worry, inadequate quantity, insufficient quality and social unacceptability – which have been the basis for the food insecurity measures in most countries of the world(Reference Coates, Swindale and Bilinsky37).
One limitation of the present study was that food security was measured when the mothers of these children were pregnant. We assumed that food security in this population was relatively static and that food security during pregnancy was a valid proxy of food security during the postpartum period.
There are two seasonal dimensions of food security in Bangladesh. One is the high exposure to flood and drought at certain times of the year; the other is related to the cycle of food production and seasonal variation in food availability and prices. Floods and droughts are common in Bangladesh, and are the main causes of fluctuations in food availability, employment and food prices. March–April and October–November are the two lean seasons; the latter is particularly severe for the rural people as this is the pre-harvest period with low employment opportunities in agriculture. Further research is required to investigate if there is any seasonal variation of household food security and its effect on growth of these children. It will also be important to measure household food security during the postpartum period to investigate its association with growth during infancy and early childhood.
Household food insecurity has both short- and long-term nutritional consequences for caregivers. In the short term, caregivers may change their dietary patterns, limit portion sizes and skip meals(Reference Oldewage-Theron, Dicks and Napier38). These changes have been observed among the households with chronic food insecurity in South Africa(Reference Oldewage-Theron, Dicks and Napier38). As a result of these management strategies, there could be compromised maternal nutrition that would ultimately influence infant care behaviour, including infant feeding(Reference Henderson, Evans, Straton, Priest and Hagan29).
Conclusion
Children in food-secure households grew better in the first two years of life than their counterparts in food-insecure households. Our analytical approach strengthened the plausibility of this association between household food security and better growth during infancy and early childhood. Nevertheless, the growth of the children who were in the food-secure households still was not adequate compared with the new WHO standards. This indicates that the households that were food-secure were not food-secure in absolute terms, but rather were food-secure only relative to the other three food insecurity categories.
Acknowledgements
The MINIMat research study was funded by several organizations. For the present analysis, K.K.S. received support from the Fogarty-NIH Training Grant (5 D 43 TW 001271) to K.M.R. None of the authors had any financial or personal conflict of interest to disclose. The authors’ responsibilities were as follows: K.K.S. designed the study, analysed and interpreted the data and drafted the manuscript; S.E.A., L.A.P., K.M.R. and E.A.F. designed the MINIMat study; K.K.S., D.S.A., S.E.A. and L.A.P. supervised data collection; E.A.F. and K.M.R. supervised data analysis and interpretation. All authors contributed to the revision of the manuscript. We gratefully acknowledge the participation of all pregnant women and families in Matlab.
