Association between consumption of nonessential energy-dense food and body mass index among Mexican school-aged children: A prospective cohort study

BACKGROUND/OBJECTIVES: Obesity prevalence in Mexican children has increased rapidly and is among the highest in the world. We aimed to estimate the longitudinal association between nonessential energy-dense food (NEDF) consumption and body mass index (BMI) in school-aged children 5 to 11 years, using a cohort study with 6 years of follow-up. SUBJECTS/METHODS: We studied the offspring of women in the Prenatal omega-3 fatty acid supplementation, child growth, and development (POSGRAD) cohort study. NEDF were classified into four main groups: chips and popcorn, sweet bakery products, non-cereal based sweets, and ready-to-eat cereals. We fitted fixed effects models to assess the association between change in 418.6 kJ (100 kcal) of NEDF consumption and changes in BMI. RESULTS: Between 5 and 11 years, children increased their consumption of NEDF by 225 kJ/day (53.9 kcal/day). In fully adjusted models, we found that change in total NEDF was not associated with change in children’s BMI (0.033 kg/m2, [p=0.246]). However, BMI increased 0.078 kg/m2 for every 418.6 kJ/day (100 kcal/day) of sweet bakery products (p=0.035) in fully adjusted models. For chips and popcorn, BMI increased 0.208 kg/m2 (p=0.035), yet, the association was attenuated after adjustment (p=0.303). CONCLUSIONS: Changes in total NEDF consumption were not associated with changes in BMI in children. However, increases in the consumption of sweet bakery products were associated with BMI gain. NEDF are widely recognized as providing poor nutrition yet, their impact in Mexican children BMI seems to be heterogeneous.


Introduction
In the last four decades, the prevalence of obesity in children has increased in developing countries from 8.3-13.2% [1]. In Mexico, the prevalence of obesity in school-aged children has increased rapidly from 9.0% in 1999 to 18.6% in 2021 [1][2][3][4]. Childhood obesity has been linked to lower quality of life, higher risk of non-communicable diseases in adulthood, such as hypertension, dyslipidemia, type-2 diabetes, and premature death [5][6][7][8]. The hypothesis is that a low-quality diet, rich in nonessential energy-dense food (NEDF), often termed "junk food", "discretionary food" or "processed food", is a key factor for weight gain and obesity [9][10][11]. NEDF could be a critical risk factor for obesity in children and adolescents in Mexico as they displace healthy foods [12], are high in sugar, re ned grains, unhealthy fats [10,13], and have a high glycemic index [12]. NEDF consumption, de ned in previous studies as energy-dense food without a cut-off point or > 13% of total energy from added sugars and/or saturated fat is high in Mexico, particularly among school-aged children living in big cities [10,14,15], who on average consume 21% of their total daily energy requirement from NEDF, with 50% of this consumed at school [14,15]. School-aged children are particularly vulnerable to NEDF consumption due to targeted marketing and advertisement to young children as well as misleading nutritional information [16,17]. Increased access to NEDF in this age group was large due to high availability in elementary schools and school-aged children have not fully developed their cognitive capacity of resistance towards these foods, and are less able to avoid or reduce their consumption [18].
Although NEDF might play a critical role as a risk factor for childhood obesity, due to their low diet quality and high intake among children, evidence of the effect of NEDF is limited. One of the main limitations is the lack of a consistent de nition or classi cation of NEDF. Moreover, longitudinal studies of the effect of several NEDF classi cations to weight gain in children are scarce and with mixed results [19][20][21][22][23]. Some of the limitations identi ed in those studies may be small samples sizes, short follow-ups, and limited information on changes in NEDF consumption over time, which could explain the lack of consistent ndings. Further evidence on the potential link between NEDF consumption and weight gain in children is needed, to understand the impact of these food group in child wellbeing and health. Our aim was to estimate the longitudinal association of the change in NEDF consumption over time and changes in body mass index of school-aged children 5 to 11 years of age, using a cohort study with 6 years of follow-up.

Study design and population
We studied the offspring of women in the ongoing POSGRAD (Prenatal omega-3 fatty acid supplementation and child growth and development) study a double-blind, randomized, controlled trial in which women were supplemented with DHA or placebo from mid-pregnancy to parturition. The original study included 978 live births between June 2005 and June 2007 to 973 women who remained in the study as previously described [24]. Mother-child pairs have been followed prospectively. All children in the cohort had access to the Mexican Social Security Institute in Cuernavaca, Mexico, which provides services to formal employees and their families.

Outcome variables
Weight and height were measured twice by trained personnel using standardized procedures [25]; the average of both measurements was used. Weight with light clothes or a hospital gown was measured to the nearest of 100 g using a digital step-up scale (SECA 803). Height was measured without shoes, hat, hairclips, headbands or other items that could obstruct the procedure using a portable stadiometer (SECA 213

Exposure variables
Dietary intake was evaluated at the three waves. At baseline a 24-hour recall (24HR) questionnaire was applied by a trained interviewer to the person in the household who prepared the meals (frequently the mother); for waves 2 and 3 the same 24HR standardized method were applied but through an automated software previously used in a National Survey [28]. This is a method which capture more accurate information of the interviewees through 5 iterative steps that complement each other for memory improvement in food intake and thus, reducing under-reporting [29]. During the interview, children participated in the report of their diet, complementing, and validating their mothers' report and in some cases adding missing food, correcting the size of a portion or removing the food reported by the adult.
24HR were performed from Sunday to Friday. To increase the accuracy of portions size, we used standardized food replicas, images of products, spoons, and cups of different sizes. Dietary information was collected as follows: 1) individual foods, 2) custom recipes (recipe reported and described in detail by the participant), and 3) standard recipes (set of ingredients in a documented and standard recipe when unknown to the subject). For our analysis all the recipes were disaggregated into their ingredients (with exception of beverages) to facilitate identifying all NEDF in recipes, (e.g. chips, puffed wheat snacks, candies, chocolate, sweets, others).
To address outliers in food items, we identify those when the reported amount was > 4 SD from the mean for the same food and age group to minimize their in uence in total diet and analysis. We identify less than 0.1% as outliers and were truncated at the highest value (median + four SD) to minimize their in uence in total diet and analysis. We did not identify implausible reporters by using the ratio for total energy intake to estimated energy requirement out of the interval between − 3 and + 3 SD in each wave.
After the rst stage of data cleaning and processing, energy, nutrients and added sugar from food were obtained with the Mexican Food Database in its 18.1.1version that include 1978 different foods including standardized recipes, and labeling information from some processed products [30].
Total NEDF was classi ed using the de nition by the Ministry of Finance and Public Credit and Ministry of Health of Mexico in 2014 [31], which considers two criteria: an energy density of > 1151 kJ/100g (275 kcal/100g) and to be classi ed as "nonessential foods". NEDF were classi ed in four groups according to nutrition composition and consumption patterns: 1) Chips and popcorn, 2) Sweet bakery products, 3) Non-cereal based sweets and, 4) Ready-to-eat cereals. Subgroups of these principal groups were constructed to obtain more homogenous groups as presented in Table 1. We excluded salty seeds or other seed products from the rst group because there is solid evidence that seeds are associated with weight lost and are considered a healthy food [32][33][34]; they were included in the tax because of their high sodium content but our outcome of interest was weight gain and not sodium-related outcomes. Also, we classi ed beverages into the following food groups: plain water, sugar-sweetened beverages (regular soda, homemade fruit water with added sugar, sweetened milk, coffee or tea with sugar, fruit drinks and sport beverages), 100%-fruit juice, and milk without sugar. Fried potato, our and corn chips, packaged fried pork skin, ready-to-eat popcorn, microwave popcorn.

Sweet bakery products
Whole grain with added sugar Bars, enriched bread and cookies made with whole grain our but all with added sugar.

Sweet bread
Sweet cookies, sweet bread, energy bars and cereal bars.

Pie and cakes
All kinds of pies and cakes.

Noncereal based sweets
Cocoa and other sweet products Cocoa, raisin, plum and legumes covered with chocolate and gums with or without sugar.

Readyto-eat cereals
All the pre-prepared and ready-to-eat cereal with added sugar. 1 NEDF, Nonessential energy-dense food.

Covariates
A socioeconomic index was calculated at baseline using a questionnaire administered to the head of household that included sanitation and household characteristics and assets. Using this information, we generated an index using principal components analysis. Maternal BMI was calculated using measured weight and height; maternal formal education was categorized as less than secondary school, secondary school, and high school or higher. Marital status was categorized as single (single, separated, divorced or widower participants) and marriage or free union (participants in marriage or living together with a couple). Maternal age was obtain using birth date, while children's sex was obtained from the birth certi cate.
Physical activity and sedentary activities. Physical activity and sedentary activities were assessed at 7 and 11 years old, using a validated semi-quantitative questionary based on the Youth Activity Questionnaire developed and validated by Hernández et al [35]. Physical activity included: playing soccer, volleyball, cycling, skating or skateboarding, basketball, dancing, swimming, walking, taking care of pets, cleaning the house, playing games at home or in the school, among others. Sedentary activities included: time spend watching television, playing videogames, reading ,and doing homework and was de ned using the time doing this activity during weekdays and weekends. Available responses included: "0 h", "<0.5 h", "0.5-2 h", "2-4 h", "4-6 h", "> 6 h" and responses were scored "0 h", "0.25 h", "1.25 h", "3 h", "5 h", and "6 h", respectively. Items from physical activity and sedentary activities were added to obtain the total minutes in a week and then per day.

Statistical methods
Children's age and sex and mother's age, education, marital status, and BMI were described at baseline.
We evaluated trends across the three cohort waves using grams, joules (calories), percent total energy of NEDF and its subgroups and for BMI. As physical activity and sedentary behavior was just measure at 7 and 11 years, we assumed no change from 5 to 7 years old, imputing for age 5 the same values of age 7. Then we test for trend for dietetic, anthropometric and physical activity variables across 5, 7 and 11 years using an extension of the Wilcoxon rank-sum test created for this purpose. Fixed effect models were used to assess the association between within-individual change in NEDF consumption and the change in BMI. Fixed effect models considered the nesting structure of data waves nested within children, using age as time of observation [36]. Energy from NEDF and its subgroups was rescaled to produce coe cients relative to 418.6 kJ (100 kcal) change. We ran three different models for each main exposure variable. The rst model was unadjusted, the second model was controlled by change in joules different from the main exposure (joules/day) [37] and the third was adjusted by model 2, plus change in physical activity (min/day) and change in tv watching (min/day).
Some sensitivity analyses were performed to assess the robustness of the results. First, in the xed effect model we changed the adjustment of change in energy different from NEDF for those dietary groups that are closely related with change of BMI, like sugar-sweeten beverages, 100% fruit juice, milk, fruits, dairy food, meat and eggs and processed meet. Second, we left the raw value of the detected dietary outliers to understand their in uence in our results. Finally, we changed our model from xed to a random effects model to estimate the association between BMI and NEDF consumption using the confounders that change in time and without changing in time. All statistical analyses were performed in STATA® Version 13 [38].
Children contributed with 1885 data points, which represents and average of 2.4 visits per child over a mean of 6.1 years of follow-up. Table 3 presents the longitudinal change in body weight, BMI, other anthropometric and physical indicators. Mean BMI increased in the rst period 1.0 ± 1.46 kg/m 2 and 4.07 ± 2.9 kg/m 2 in the second; this represents an increase of 0.56 ± 0.96 in BMI z-score and 24.2% increase in the proportion of overweight and obesity from baseline to the third wave.  Table 4 presents baseline and change levels of NEDFL consumption in joules (calories), grams and percentage of total energy consumption to 7 and 11 years old. All sweet bakery products, especially sweet bread, were the main contributors to NEDF with 644 kJ/day (154 kcal/day) (10.5% of total caloric intake). Non-cereal based sweets, ready-to-eat cereals, and chips and popcorn represented a total caloric intake of 3.8%, 3.1% and 2.1%, respectively. On average, the amount of NEDF consumption increased 113 ± 1289 kJ/day (27 ± 308 kcal/day) from baseline to the rst wave and 226 ± 1536 kJ (54 ± 367 kcal/day) from baseline to the second wave (p = 0.034); nevertheless, total caloric intake decreased by 2.3 and 3.1 percent points, respectively. In all NEDF groups and subgroups, except for ready-to-eat cereals, the amount in grams and caloric intake increased from 5 to 11 years, yet, as a percent of total energy the change was either negative or null. Consumption of chips and popcorn increased in joules [∆ 98 kj/day (∆ 23.4 kcal/day)] and percent total energy (∆ 0.7 pp/wave) from 5 to 11 years old. Sweet bakery consumption increased on average 54 ± 1050 kJ (13 ± 251 kcal) in rst period and 129 ± 1213 kJ (31 ± 290 kcal) in the second, mainly due to sweet bread.  Table 5 includes the unadjusted and adjusted association between BMI change and NEDF change. In the fully adjusted model, the increase in 418.6 kJ/day (100 kcal/day) in total NEDF was associated with a     In further sensitivity analyses, adjustment for those dietary groups that are closely related with change of BMI (Supplemental table 2) and not truncating outlier truncation, did not change results of the models. In the random effect analysis, we found a similar pattern of association, with weaker coe cients for most of the outcomes. Chips and popcorn, sweet bakery products and sweet bread were associated for each  (Supplemental table 3).

Discussion
We aimed to estimate the association between changes in NEDF consumption and BMI change in schoolage children. Over an average of 6.1 years of follow-up children increased their NEDF consumption by 225 kJ (53.9 kcal), yet, we did not observe an association between NEDF increases and BMI increase (0.033 kg/m2, [p = 0.246]). However, in fully adjusted models BMI increased 0.078 kg/m 2 for every 418.6 kJ/day (100 kcal/day) of sweet bakery products intake (p = 0.035). Increases in the consumption of chips and popcorn were associated with a 0.208 kg/m 2 increase in BMI in unadjusted models; however, this association was attenuated in fully-adjusted models. Changes in the consumption of other food groups within the NEDF classi cation did not show an association with changes in BMI.
Few efforts have been made to estimate the association between NEDF consumption and weight in school-aged children. Some cross-sectional studies have described a positive association between NEDF and weight gain [12,39], however, the few longitudinal studies available have shown mixed results. In two studies, the rst under ve years of age and the second with a mean age of 16 years, children's weight tended to decrease with higher levels of consumption of NEDF [19,20]. A positive association was observed in a study of 961 children 5 to 12 years of age, that de ned dietary patterns rich in high-energy and low-nutrient-density foods as exposure [21] with 2.5 years follow-up. Similarly, Phillips, et al, found a non-signi cative result that children's weight tend to increase with more joules (calories) from NEDF consumption in a cohort of 166 non-overweight school-aged girls followed for seven years [22], both of these cohort studies used mixed-effect models which can not control for time-invariant confounders as effectively as xed effects models. Our study is unique in that we are assessing the impact of changes in NEDF consumption and changes in BMI; under this approach we did not detect an association between NEDF and BMI overall, but we did identify a signi cant association with the consumption of sweet bakery products.
The association between sweet bakery product consumption and weight gain has been reported in two other prospective studies. Phillips et al., in a girls' cohort study with an average follow-up of seven years, found an increment of z-score in BMI (0.003; p = 0.11), with more intake of cookies, pies, cakes, brownies, chocolate candy, nonchocolate candy, ice cream, milkshakes sherbet, potato chips and corn chips [22]. The non-signi cant result for potato chips and popcorn un our study could be explained by a lack of power to detect the effect due to a small sample size and the small number of energy-dense foods included in the food frequency questionary. In the preset study we identi ed more than 90 different types of NEDF in children's diet. Also, in a prospective cohort study with more than 120 thousand adults followed for 20 years the consumption of one serving of potato chips per day or re ned grains (including sweet bakery products) was associated with increment in body weight of 0.77 kg and 0.25 kg in a 4-year period, to each food product, respectively [34]. Negative associations between chips and bakery and BMI have also been reported. Field et al., in a cohort of children and adolescents with three years of follow-up found a reduction of 0.006 in BMI z-score (p < 0.05) among those eating energy-dense foods; however, this association became non-signi cant after adjusting for dieting status and maternal overweight [23].
In our study we found that eating ready-to-eat cereals was marginally associated with a BMI reduction of -0.098 kg/m 2 for every 418.6 kJ/day (100 kcal/day). Consumption of ready-to-eat cereals has been related to a healthy dietary pattern in children [40] and this include more consumption of vitamins and minerals, less of saturated fat and cholesterol, but also, with a higher intake of added sugar [40,41].
Negative associations between ready-to-eat cereals and BMI, have also been reported in longitudinal analyses [42,43]. However, ready-to-eat cereals comprises many different products, and their nutritional impact will depend on the composition of the cereal, and the food consumed with them (such as fruit or milk); our study, as well as all previous studies available could be confounded by these characteristics.
Even though ready-to-eat cereals may be associated with weight lost in children, children consuming a non-high ber cereal, had worse type 2 diabetes risk pro le than children consuming a high ber cereal in a longitudinal study [44].
There are three different explanations for the association between sweet bakery products consumption and weight gain. First, sweet bakery products tend to be high in added sugar, saturated fat and of course high quantities of energy in small portions of food [10] and may promote excess energy intake without control over the joules (calories) consumed. In a recent crossover trial, adults were randomized to receive an ultra-processed (generally energy-dense food) or unprocessed diet for a period of 2 weeks. Participants in the ultra-processed diet increased 0.8 kg and those in the unprocessed diet reduced 1.1 kg.
Those in the ultra-processed diet consumed 2126 kJ/day (508 kcal/day) more than the other group, mainly by fats and carbohydrates [45]. Second, many sweet bakery products are high in re ned carbohydrates and starches and may induce stronger insulin secretion. This promotes less satiating signals, increasing subsequent hunger feelings [46] and suppresses the release of fatty acids from adipose tissue into circulation, while keeping glucose and fatty acids away from the oxidation process to store them in the adipose tissue [47]. Third, a diet rich in energy-dense food is associated with less protein consumption in children and according with the "protein leverage hypothesis" this may be related to the disturbance of the appetite system through increased postprandial hunger and reduced postprandial satiety [48].
Strengths of this study include the prospective cohort design, including three measurements of anthropometric and dietary information, large sample size and the approach analysis with the possibility to assess the change in change effect. Our study also has some important limitations that must be taken into account to interpret our results. First, at each wave we had one dietary evaluation, instead of two or more 24-hour recalls, which may not re ect usual NEDF consumption. Second, we lost 14.1% of the sample in the rst period and 29.3% in the second; however, baseline socioeconomic index, children sex at baseline, age and maternal education, and overweight status were no different between children lost and those who stayed in the cohort.
In summary, our results showed that consumption of NEDF overall was not associated with BMI in children. However, NEDF subgroups such as sweet bakery products and, possibly, chips and popcorn, showed an association with BMI. The longer-term effect of NEDF consumption in school-aged children on BMI requires further studies with bigger samples, better follow-up, and the use of an objective measure of adipose tissue in order to obtain more reliable results. Decreasing the consumption of NEDF is a key step to improve dietary quality and prevent obesity, aligned with the WHO 25x25 goals [49] and the Sustainable Development Goals [50]. In Mexico, a strategy to reduce NEDF consumption is in place, limiting access to these foods in elementary schools [51], restricting food marketing to children on television and public areas [52], and implementing an 8% tax to all NEDF [31]. However, further public health efforts need to be directed to reduce the consumption of NEDF, particularly early on in life.
Declarations AUTHOR CONTRIBUTIONS DI-Z conceived the design research, performed the computations and was the major contributor in writing the manuscript. BS, SG, MD and TB-G, assist the statistical analysis and contributed to writing the manuscript. BV-A, RS-I and RI contributed to the follow-up of the cohort and supervised the cleaning and processing data. RS-I derived the nonessential energy-dense food database. All authors read and approved the nal manuscript.

FUNDING
Primary funding for the study came from the National Institute of Public Health and from National Institutes of Health Grant Number R01DK108148.

COMPETING INTERESTS
The authors declare that they have no competing interests.