Snack food consumption among Bangladeshi children, supplementary data from a large RCT

Abstract Childhood obesity has been associated with consumption of energy‐dense foods such as caloric beverages and fast foods. Many low‐ and middle‐income countries like Bangladesh are now experiencing a rising problem of noncommunicable diseases along with the long‐standing problem of stunting and undernutrition. WASH Benefits Bangladesh was a large community‐based cluster randomized controlled trial conducted in rural Bangladesh. Study clusters were randomized into seven arms: single nutrition (N); water (W); sanitation (S); hygiene (H); combined water, sanitation, and hygiene (WSH); WSH and nutrition (N + WSH); and a double sized control (C). Nutrition intervention messages included four promotional components: maternal nutrition, breastfeeding, complementary feeding, and lipid‐based nutrient supplements. The World Health Organization infant food frequency questionnaire (24‐hr recall and 7‐day recall) was administered at Year 1 and Year 2 of intervention. The likelihood of any snack food consumption was significantly lower (odds ratio 0.37: 95% confidence interval [0.28, 0.49]) in the nutrition intervention arms compared to the control arm in Year 2 follow‐up. In addition, in the water intervention arm, fewer children (about 50% less) consumed soft drinks, but not the other sugar‐sweetened beverages, compared with control in Year 2. There were no other differences between groups. Simple messages about balanced diet and feeding family foods were effective in lowering commercially produced snack food consumption of the young children in low‐income rural communities of Bangladesh. Provision of safe water apparently encouraged mothers to reduce offering unhealthy beverages to the young children.


| INTRODUCTION
A substantial worldwide increase in childhood obesity has posed a major global health challenge. The Global Burden of Disease Study 2013 identified that overweight and obesity among children have increased markedly in high-, middle-, and low-income countries (Ng et al., 2014). Although the level of undernutrition continues to be high, some low-and middle-income countries have experienced similar or greater increases in childhood obesity compared with highincome countries (Lobstein et al., 2015). All dimensions of health, physical, emotional, and social, are profoundly influenced by childhood obesity (Sahoo et al., 2015). The World Health Assembly set a global nutrition target to halt the rise in childhood overweight by 2025(WHO, 2013. Childhood obesity has been associated with consumption of energy-dense foods like caloric beverages and fast foods (Drewnowski & Specter, 2004). A study with more than 4,000 USchildren aged 5 and 11 years showed that there was an increased likelihood of developing overweight/obesity with increased frequency of fast food and sugar-sweetened drink consumption (Berry, Burton, & Howlett, 2017). A longitudinal study on children of low socioeconomic status from Brazil also found that early ultraprocessed food consumption as a proportion of daily energy was associated with altered lipoprotein profile (Rauber, Campagnolo, Hoffman, & Vitolo, 2015); ultraprocessed foods are not made from whole foods but from their derivatives and food additives (Monteiro et al., 2012).
Snacking habits such as consumption of sweet snacks, candy, and chips were associated with other adverse health outcomes including dental caries in young children (Johansson, Holgerson, Kressin, Nunn, & Tanner, 2010). Energy-dense, nutrient-poor foods have become more accessible to people at a lower cost due to developments in agriculture and food technologies (Drewnowski & Specter, 2004).
South Asian countries, including Bangladesh, have shown remarkable economic advancement in recent years. Improvement in household economic status, local production of packaged food, improved transportation, and extensive food marketing have bought changes in dietary patterns even to rural populations (Bishwajit, 2015). Up to 75% of 2-year-old Asian children from countries like Philippines and Nepal were found to consume sugary snack foods (Huffman, Piwoz, Vosti, & Dewey, 2014). A study in rural Indonesia showed that snack food that included fast food, soft drinks, traditional snack foods, candies, and desserts, and modern snack foods consumption, in respect to total energy and nutrient intake, was associated with lower height-for-age z-score among schoolchildren (Sekiyama, Roosita, & Ohtsuka, 2012). Countries like Bangladesh and India are now detecting increasing rates of noncommunicable diseases along with the long-standing problem of stunting and undernutrition (Bishwajit, 2015). With a national stunting prevalence of 36% (BDHS, 2014), Bangladesh is simultaneously experiencing an increase in the prevalence of overweight (3.6 to 7.9% from 1998 to 2015) and obesity (2 to 9% from 2004 to 2015) over time (Biswas, Islam, Islam, Pervin, & Rawal, 2017).
A parenting program in Bangladesh to address early childhood health, growth, and development found the intervention was effective in changing mother's practices related to dietary diversity (Aboud, Singla, Nahil, & Borisova, 2013). A recent systematic review on the factors determining eating behaviour among preschool children in low-and middle-income countries found that better nutritional knowledge of caregivers was associated with an increased healthy eating practices of the children (Sirasa, Mitchell, Rigby, & Harris, 2019). Therefore, an intervention focused on improved caregiver knowledge of appropriate complementary feeding practice may result in improved dietary quality and less unhealthy snack food consumption in their children.
The primary objective of this study was to measure commercially available snack food and packaged food consumption patterns among children less than 3 years of age participating since birth in a randomized controlled trial of nutrition, water, sanitation, and hygiene interventions. We previously reported a significant increase in dietary diversity among the children in the nutrition intervention group compared with control . Thus, we were interested to explore if there were any relationships between snack food consumption and complementary food diversity among the children who took part in the intervention trial. We hypothesized that a nutrition inter- These sites were selected according to the quality of ground water

Key messages
• The promotion of optimal infant and young child feeding practices was associated with a passive reduction of commercially produced snack food consumption.
• Provision of safe water appeared to encourage mothers not to offer unhealthy beverages to young children.
• Snack food consumption was not associated with displacement of nutritious complementary foods; however, more research is needed using quantitative dietary intake assessment methods. and presence or absence of ongoing or upcoming water, sanitation, and nutrition interventions. At the initial stage of the study, research assistants conducted a community census to identify pregnant women. Women at their first or second trimester of pregnancy were eligible to enrol in the study. Eight nearest pregnant women, identified using a global positioning system, formed a study cluster. These study clusters were randomized into seven arms: single nutrition (N); water (W); sanitation (S); hygiene (H); combined water, sanitation, and hygiene (WSH); combined WSH and nutrition (N + WSH); and a double sized control (C). Each intervention arm contained 90 clusters, and the control arm contained 180 clusters. A total of 5,551 women and their offspring were enrolled in the study. The randomized group assignment could not be blinded due to the nature of the interventions. Detailed study design and rationale have been published previously (Arnold et al., 2013;Jannat et al., 2018;Luby et al., 2018).  (Menon et al., 2016) and delivered small quantity lipid-based nutrient supplements to children from 6 to 24 months of age. To deliver the intervention to the households, especially to the mothers, women from the local community were recruited as community health promoters (CHPs). CHPs were under regular supervision of the research assistants. Supervision of the research staffs were arranged in a ranked order so that the implementation remained even and precise. Supervisors were trained in direct intervention delivery; subsequently, they trained the CHPs. First round of training was more intensive including basic project description, principals of behaviour change, communication skills, intervention specific training, and role play. This was followed by regular monthly meeting between supervisors and CHPs. Refresher training was also arranged about a year after initiation of intervention .

| Intervention delivery
Each cluster of 6-8 months was monitored by one CHP. On average, a cluster was about 1 km in diameter that could easily be covered by a CHP. We maintained a 1 km of buffer zone in between the study clusters to prevent possible spillover effects. to the households. Lipid-based nutrient supplements were distributed to the mothers on a monthly basis. Two 10-g sachets were recommended for daily consumption starting from 6 months until 2 years of age.

| Data collection
During the baseline assessment, before the index children were born, research assistants visited participants in their home to collect household socio-economic and demographic data. The research assistants conducted two follow-up visits at about 1 year and 2 years of commencement of intervention delivery. The infant food frequency questionnaire (24-hr recall and 7-days recall) was completed at both follow-up visits. The infant and young child feeding module was adapted from indicators recommended by the World Health Organization (WHO, 2010a) with an added module to assess consumption of commercially produced snack and packaged foods. Commercially produced snacks were given special consideration due to their high salt, sugar, and trans-fat content, and also taking into account the presence of food substances that are not commonly used in culinary preparations such as hydrogenated oil and food additives like colouring agents, flavourings, and nonsugar sweeteners (Araújo, Ribeiro, Padrão, & Moreira, 2019). We grouped different commonly available snack foods and packaged foods in the rural market according to their taste and pattern. The questionnaire module focused on store-bought sweet and savoury snacks rather than those that were home prepared. Store-bought foods may be ultraprocessed, contain high amounts of sugar, high sodium, or are deep fried with frying oil used and reheated several times, which increases the trans-fat content of the food (Bhardwaj et al., 2016). We grouped them into six categories: If any item from the mentioned food categories was consumed, mothers were asked for how many days the particular food item was consumed in past 7 days.

| Data analysis
The outcomes presented in this paper were added during the course of the study and were not considered primary or secondary outcomes of the trial. Those results have been published elsewhere Tofail et al., 2018). Nevertheless, the statistical analyses were planned and pre-specified prior to beginning the analysis. The WHO has defined snack foods for young children as foods that are easy to prepare, mostly self-fed, and eaten in between main meals (WHO, 2005). In this paper, we defined snack food as unhealthy when they contained excess sugar (these included as soft drinks, artificial fruit juice, sweetened milk, and sweet snacks), high salt ( groups among seven recommended food groups all days within past 7 days. We used a multivariate logistic regression model to measure if there was any association between snack food consumption and minimum dietary diversity in the past 7 days. We added the covariates: child age and sex, parent's education, father's occupation, study districts, household socio-economic status, and food insecurity in the model. This model was used to assess the association among children from the nutrition or nonnutrition arms and controls separately. We examined that the association between snack food consumption and dietary diversity in the nutrition arm and other arms as dietary diversity was similar among children from control and nonnutrition arms . Food insecurity was measured using WHO Household Food Insecurity Access Scale (Coates, Swindale, & Bilinsky, 2007). Because participating children were randomized at cluster level, it was expected that responses were correlated within cluster. In order to obtain cluster-robust standard errors, all analyses were adjusted for clustering using the sandwich estimator (Carroll, Wang, Simpson, Stromberg, & Ruppert, 1998). Baseline characteristics, child age, and sex were balanced across study arms (Table 1). About half of the mothers had an educational level of more than grade five. Electricity connection was available in around 60% of the households. The household size was around five persons (2 SD) per family. Food insecurity was reported by only around a quarter (30%) of the households.

| Ethical considerations
The prevalence of snack food consumption in past 7 days in different arms was 0-7% at the Year 1 follow-up. This prevalence increased to 3-76% in Year 2 follow-up (Table 2) Table 2). The consumption of any snack food at least once in the past 7 days was significantly lower (N: 73%; N + WSH: 69% vs. C: 87%) in the nutrition intervention arms compared with the control arm at Year 2 follow-up (Table 3).
Some of the snack foods consumption was associated with increased dietary diversity (Table 4). Specifically, among the control and nonnutrition arms, canned fruit juices consumption was associated with a higher odds of dietary diversity at Year 2 (Table 4). In the nutrition arms at Year 2 follow-up, consumption of soft drinks, sweet snacks, and pickles was associated with an increased consumption of diverse complementary foods (Table 4). In addition, when measuring association between snack food consumption and dietary diversity, higher educational attainment of fathers was consistently associated with an increased food diversity.

| DISCUSSION
In this randomized controlled trial of nutrition, water, sanitation, and hygiene, children participating in the nutrition intervention were less likely to consume sweets or packaged snack food items. We previously reported that the nutrition intervention, which focused on the timely introduction of diverse complementary foods, resulted in a higher dietary diversity score at both time points  and high adherence to lipid-based nutrient supplementation recommendations . The present data suggest that parents prioritized feeding the promoted nutritious foods or lipid-based nutrient supplements over commercially available sweet, salty, or packaged snack foods to their children. Parents have more control over children's food environment and F I G U R E 1 Summary of participant enrolment, randomization, retention, and analysis for snack food consumption experiences at an early age (Anzman, Rollins, & Birch, 2010). Similar to the WASH Benefits intervention, studies that delivered interventions targeting the parents were found to be effective in improving children's dietary behaviour (Kader, Sundblom, & Elinder, 2015). The The significant reduction in soft drinks in the water intervention arm is also notable. It is possible that when households felt that their own water was safer, they were less likely to feed their children with packaged drinks. Studies in the United States have also found that perceived water safety was associated with lower intake of sugarsweetened beverages (Onufrak, Park, Sharkey, Merlo, et al., 2014; Abbreviations: CI, confidence interval; OR, odds ratio; SD, standard deviation; WSH, water, sanitation and hygiene; WSHN, WSH and nutrition. a None of the indicators were significantly different compared to the control arm. b Food insecurity was measured using World Health Organization Household Food Insecurity Access Scale.

T A B L E 2
Effect of the nutrition intervention on snack and packaged food consumption (at least once in last 7 days)  (Laitala, Vehkalahti, & Virtanen, 2018). Consumption of energy dense foods and sugar-sweetened drinks may result in an increased body mass index in children and adults (Collison et al., 2010;Pérez-Escamilla et al., 2012). To regulate this practice of increased consumption, positive food parenting has an important role (Larsen et al., 2015). It is thus important to educate caregivers to support building healthy food habits among children.
We were interested to know how snack food consumption was associated with dietary diversity. We presumed that children who consumed more snack foods would have less diverse complementary The WASH Benefits Bangladesh trial was a large randomized control trial, which was designed to measure small differences in outcome variables, maintained high levels of implementation fidelity, and achieved high follow-up rates throughout the 2-year trial . Promoted behaviours showed a high uptake , including an increase in dietary diversity in the nutrition intervention arm . There were some limitations to this study that should be noted, however. Both complementary feeding and snack food consumption information were reported by the mothers. All food consumption was measured by frequency of intake not by quantity; thus, precise estimation of calories consumed was not possible. Therefore, we could not calculate snack food consumption as a percent of energy or directly estimate the possibility of displacement of healthy foods. Moreover, there might be inconsistencies in food categories with respect to their nutrient density. In the future, a nutrient profiling method would allow for a more standardized method for categorization of food types (Drewnowski, Dwyer, King, & Weaver, 2019). Because the intervention was not blinded to either the participants or the data collectors, the chance of information bias cannot be ruled out. However, we used standardized dietary indicators to assess food diversity (WHO, 2010a). We adapted a new module for assessing commercially produced snack food consumption that covered most of the common snack items available in the rural community of Bangladesh. Rigorous training was conducted for the data collectors. The survey team worked independent of the intervention team to minimize surveyor's bias. Loss to follow-up was reasonable (~15%) over a 2-year period.
Simple messages about balanced diet and feeding family food were effective in lowering commercially produced snack food consumption of the young children in low-income rural communities of Bangladesh. Provision of safe water appeared to encourage mothers to reduce offering unhealthy soft drinks to the children. Nevertheless, a sharp increase in snack food consumption in the second year of child age suggests that greater attention is needed to prevent the establishment of unhealthy eating behaviours in children.

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.

AUTHOR CONTRIBUTIONS
KJ drafted the manuscript under the guidance of CPS and input from all listed co-authors. SPL drafted the research protocol; he coordinated input from the study team throughout the project. Peter J. Winch https://orcid.org/0000-0001-8569-5507 Christine P. Stewart https://orcid.org/0000-0003-4575-8571