Development and preliminary testing of a technology-enhanced intervention to improve energy intake regulation in children

Abstract

Interventions designed to improve children’s self-regulation of energy intake have yielded mixed results. We tested the efficacy of a technology-enhanced intervention designed to teach children to eat in response to internal hunger and fullness cues. Thirty-two children (mean age 4.9±0.8 y) completed this within-subjects, pre-post design study that took place across 10 laboratory sessions, each scheduled approximately 1 week apart. The intervention was conducted across weeks 4–7 in small groups focused on teaching children how food travels through the body and how to respond to hunger and fullness signals. Children’s short-term energy compensation, a measure of intake regulation, was collected at baseline and follow-up using a preloading protocol. Twenty-five minutes prior to receiving a standardized test meal, children consumed a low-energy (3 kcal) or high-energy (150 kcal) preload beverage, presented in random order at baseline and follow-up. Knowledge of intervention concepts was also assessed at baseline and follow-up. Linear mixed models were used to examine changes in short-term energy compensation and knowledge from baseline to follow-up. Knowledge related to the intervention improved from baseline to follow-up (3.5±0.3 to 7.0±0.3 correct responses out of a possible 10; P<0.001). Children’s energy compensation also improved from baseline to follow-up, as evidenced by a time-by-preload condition interaction (P=0.02). However, this improvement was driven by boys who increased the adjustment for beverage energy content from baseline to follow-up (P=0.04). Girls showed no change in energy compensation with the intervention (P=0.58). The overall increase in knowledge, paired with the improvement in energy compensation in boys, suggests that this technology-enhanced intervention may be efficacious for some children. Further research is needed to determine whether boys and girls will benefit from different, personalized intervention strategies for obesity prevention.

1. INTRODUCTION

The ability to eat in response to homeostatic needs, rather than environmental food cues, is critical to preventing excess weight gain. However, there remains a lack of effective, evidence-based approaches to teach children to regulate their energy consumption. The purpose of this study was to test the efficacy of a laboratory-based, technology-enhanced intervention to improve energy intake regulation in 4–6 year-old children.

Previous interventions aimed at improving energy intake regulation in children have produced various levels of success (1–5). For example, Lumeng and colleagues conducted an intervention in Head Start aimed at improving general and food-specific self-regulation, but no changes in child weight status were observed (4). Several other interventions that have focused specifically on improving children’s food-intake regulation have been based on the premise that obesity is the result of overeating in response to external and situational cues (6). As a result, these interventions have sought to strengthen appetite awareness, a concept which has been broadly defined as eating in response to internal hunger and fullness cues and minimizing food intake elicited by external cues (7). Using this concept, Bloom et al. found that appetite awareness training reduced weight status among 6–12 year-old children with overweight, but these results were not sustained at 6 months (3). Other interventions have observed short-term improvements in weight-related outcomes. For example, Boutelle and colleagues (2) examined a combination of appetite awareness training, which focused on teaching children to regulate their food intake by responding to hunger and fullness, and food cue exposure training, which aimed to reduce the strength of the association between food cues (e.g. the smell and sight of food) and physiological experiences (e.g. salivation, insulin release). Results indicated that these strategies were effective at reducing binge eating among 8–12 year-old children. Most closely related to the current intervention, Johnson et al. (1) trained preschoolers in childcare settings to improve appetite regulation, i.e. eating in response to physiological hunger and fullness cues, through the use of dolls, videos, and skits to teach concepts of digestion, hunger and fullness, improving their short-term energy compensation. However, these findings have not been replicated.

A key dimension of homeostatic food intake regulation is accurately responding to internal satiety signals. Homeostatic regulation can be assessed behaviorally by measuring short-term energy compensation, which is operationally defined as adjustments in energy intake in response to the previous ingestion of a first course, or ‘preload’ (8). Participants are served preloads that vary by some attribute (e.g., energy, macronutrients) to determine their effects on intake at subsequent ad libitum test-meals (9). Compensation ability can then be quantitatively assessed by comparing intake across the various preloading conditions (i.e., standard approach) or by calculating the compensation index (COMPx), a linear transformation of the difference in intake across two preload conditions that has often been used with children (1,10–14). Regardless of how the data are treated, studies in children have shown that energy compensation measured via preloading paradigms begins to decrease after the preschool years (15–17) and is inversely associated with weight status (11,14,18). These individual differences in short-term energy compensation demonstrate that this measure exhibits variability that is related to weight status, and therefore may be an ideal target for intervention.

In addition to internal satiety cues, preschoolers (age 3–5 years) are also influenced by environmental food cues (e.g. 19), and therefore the ability to successfully regulate energy intake may be related to how responsive children are to a food’s hedonic properties. A behavioral measure of hedonic food cue responsiveness is eating in the absence of hunger (EAH; 20–21). To quantify EAH, children are provided with a meal to consume to satiety, followed by the opportunity to eat palatable snacks and/or to engage in alternative activities (i.e., often coloring or free play). The amount of energy children consume from the palatable snacks is used to determine EAH. Eating in the absence of hunger is stable across development (22–23) and prospectively associated with excess weight gain (22). However, it is also malleable in response to interventions that target food responsiveness (2), thus supporting its selection as an additional outcome measure in the current study.

Computer-based interventions using characters can increase children’s motivation and learning (24–31). To our knowledge, technology-enhanced interventions have not yet been applied to improving children’s ability to regulate energy intake. Therefore, we developed and conducted preliminary testing of a technology-enhanced intervention based on the curriculum originally designed by Johnson (1) for improving children’s energy intake regulation. In this new iteration of the intervention, an interactive character-based technology platform was added to enhance observational learning (32), and parents were provided with educational materials on child feeding. We hypothesized that the updated intervention would improve children’s 1) knowledge of concepts related to digestion, hunger, and fullness; 2) short-term energy regulation, assessed via both the standard approach (i.e., comparing energy intake across the various preload conditions) and COMPx; and 3) intake of palatable snacks when not hungry, assessed behaviorally with the EAH paradigm.

2. METHODS

2.1. Participants

We recruited families with children ages 4–6 years old from a university-sponsored database of potential families and by posting flyers at local preschools and community events. Interested families were screened for eligibility and children were excluded from the study if they had a medical condition that would affect their ability to participate in the study (e.g. dysphagia, learning disorder, food allergies), were taking prescription medication that could affect appetite, or were not between 4–6 years old. Our entire sample included 33 children (n=33, 16 male) ages 4–6 years old (4.9±0.8 years) who ranged in weight status (mean ± SD BMI percentile 70.2±17.9). One child (female) was unable to complete the protocol on visit 1 and was excluded from further testing, for a final sample of 32 children who completed all 10 visits (97% retention rate, see Figure 1 for enrollment diagram). We powered the study based off of effect sizes for the change in COMPx (pre- vs. post-intervention) reported in Johnson (1). Based on these effect sizes (Cohen’s d=0.50), we calculated that 30 children would be needed to have 80% power to detect a significant change in COMPx across the intervention.

Figure 1.

Flow diagram of the progress through screening, enrollment, and data analysis of the final sample.

The parent or guardian primarily in charge of feeding was required to accompany the child to each visit; in all but one instance, this was the mother. Descriptive statistics of participant demographics are included in Table 1. The study was approved by the Institutional Review Board of The Pennsylvania State University, and all parents provided written informed consent before the child’s participation.

Table 1.

Children’s characteristics who participated in a technology-based intervention to improve knowledge and energy compensation.

	BOYS		GIRLS
Continuous Variables	Mean±SD	Range	Mean±SD	Range
Age (years, n=32)	5.1±0.8	4.0–6.8	4.8±0.8	4.2–6.9
BMI z-score (baseline, n=32)	0.7±0.6	−0.3–2.0	0.6±0.8	−0.5–2.8
BMI z-score (follow-up, n=31a)	0.6±0.8	−0.5–1.9	0.6±0.8	−0.9–2.8
Categorical Variables	% (n)		% (n)
Household income (annual, n=32)
≤$50,000	25 (4)		13 (2)
>$50,000	75 (12)		87 (14)
Race (n=32)
White	81 (13)		81 (13)
Non-Whiteb	19 (3)		19 (3)

^aOne child (female) was excluded from anthropometric measures at follow-up because precise height and weight could not be collected due to this child having a cast for a broken leg.

^bChildren who were in the following racial categories were condensed into Non-White category: American Indian/Alaska Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, Mixed-Race, or Other.

2.2. Study Design

A within-subjects, pre-post design was used to test intervention efficacy on child knowledge and regulation of food intake. Participants completed 10, ~1.5 hour visits, scheduled one week apart, in the Children’s Eating Behavior Laboratory. A summary of the timeline for the 10 visits is provided in supplementary material. On visit 1, anthropometric data were collected, EAH was assessed, and a 10-question interview was completed with the child to collect information on his/her knowledge related to digestion, hunger, and fullness. On visits 2 and 3, children’s energy compensation was assessed using a standard preloading protocol (10). The intervention was delivered at visits 4–7, and post-intervention assessments (i.e., EAH, compensation, and knowledge) were completed again on visits 8–10. All study visits were conducted during lunchtime (between 11:00 A.M. and 1:00 P.M.) or during dinnertime (between 4:00 P.M. and 5:30 P.M.), based on the family’s availability. These visit times were selected to be convenient for the family (e.g. families could come directly to the laboratory after picking the child up from school), and children did not begin eating their meal until ~30–45 minutes after their visit start time. Meal days and times were kept consistent within an individual for each of the 10 visits. Children were instructed not to eat or drink anything, with the exception of plain water, for a minimum of three hours prior to their visit to achieve a level of hunger similar to what they would experience when approaching a meal.

2.3. Anthropometrics

At both baseline (visit 1) and follow-up (visit 8), child height and weight were measured in duplicate. Children were weighed and measured without shoes and in light clothing. Weight was measured to the nearest tenth of a pound using a digital scale (Tanita^®, Arlington Heights, IL). Height was measured to the nearest tenth of a centimeter using a portable stadiometer (Seca^®, Chino, CA). The mean of the two measurements at each time point was used for analyses. Body mass index (BMI), BMI percentile, and BMI z-scores were calculated using the Center for Disease Control’s BMI Calculator for Children and Teens (33). Children at or above the 85^th percentile BMI-for-age and sex were classified as having overweight; those at or above the 95^th percentile were classified as having obesity (34).

2.4. Knowledge Questionnaire

To assess children’s knowledge of digestion, hunger, and fullness, we developed a 10-question assessment that was administered at baseline and follow-up using a one-on-one interview format. Correct responses (i.e., 1 point each) were summed across the assessment for a range of 0–10 possible for the total score. Assessments were scored by the same researcher to ensure consistency, and a second research assistant scored all questionnaires independently to determine inter-scorer reliability. Intraclass correlations indicated excellent reliability for both baseline (ICC=0.983, P<0.001) and follow-up (ICC=0.963, P<0.001) knowledge scores. Questionnaire items are included in Table 2.

Table 2.

Knowledge questionnaire items.

Item	Question
1	Can you point to your mouth?
2	Can you name 2 things your mouth does to help you eat food?
3	Can you name 3 parts of your body that help you digest food?
4	Can you point to your esophagus?
5	Can you point to your stomach?
6	Can you name 1 thing your stomach does with food?
7	Where does your food go after it leaves your stomach?
8	Can you name 2 things your body feels when you are hungry?
9	Can you name 2 things your body feels when you are full?
10	What happens to extra food your body doesn’t need to use for energy or to grow?

2.5. Energy Compensation: Pre- and Post-Intervention

As a measure of homeostatic eating regulation, a standard preloading protocol was used to assess energy compensation before (visits 2 and 3) and after (visits 9 and 10) the intervention. Twenty-five minutes prior to receiving the test meal, children drank a high-energy (~150.0 kcal; 0.87 kcal/g) or low-energy (~3.0 kcal; 0.02 kcal/g) preload beverage and were encouraged to finish in ≤ 5 min. For children who had difficulty finishing, research assistants provided additional encouragement by engaging in a race to see who could finish their drink first. At baseline, 100% (n=32) of children consumed the entire high-energy preload, and 91% (n=29) of children consumed the entire low-energy preload. At follow-up, 91% (n=29) of children consumed the entire high-energy preload, and 94% (n=30) of children consumed the entire low-energy preload. The order in which each child received the preloads was counter-balanced and randomly assigned. Beverage preloads were prepared according to recipes developed by Johnson & Birch (11) and were matched for taste, volume (173.0 g), and appearance. The high-energy preload contained one 3.6 gram packet of Cherry Kool Aid^® drink mix, 227.0 g of sucrose, and 250.0 grams of maltodextrin (a relatively tasteless starch used to increase energy density without altering sweetness). The low-energy preload contained one 3.6 gram packet of Cherry Kool Aid^® drink mix and twenty-four, 1-gram packets of Equal^® sweetener. Following consumption of each preload, children were served a standardized test meal from which they could eat ad libitum. The same procedure was conducted on visits 9 and 10 to assess energy compensation.

The foods presented at each test-meal included: macaroni and cheese (Kraft^®, Kraft Heinz Co., Chicago, IL), frozen broccoli florets (Green Giant^®, B&G Foods, Parsippany-Troy Hills, NJ), red grapes (Wegmans^®, Wegmans Food Markets, Rochester, NY), baby carrots (Wegmans^®, Wegmans Food Markets, Rochester, NY), graham crackers (Nabisco Original^®, Nabisco, East Hanover, NJ), reduced fat mozzarella cheese stick (Kraft^®, Kraft Heinz Co., Chicago, IL), and 2% fat content milk (Wegmans^®, Wegmans Food Markets, Rochester, NY). For each food, standard, age-appropriate serving sizes were presented, as informed by data from The Continuing Survey of Food Intakes by Individuals (35). Total energy offered was 561.0 kcal for the meal, assuming the child did not have multiple servings.

Prior to the meals, children tasted small samples (< 5.0 grams) and rated liking for all foods using a 5-point hedonic smiley face scale ranging from “super bad” to “super good” (36). After rating liking, children reported level of fullness using a 4-point silhouette scale adapted from Fisher and Birch (37).

Items were arranged identically on a tray for all test meals. Before placing the meal in front of the child, a research assistant informed them that they had 30 minutes to eat as much or as little as they would like. The research assistants also told the child that they could have additional servings of any item throughout the meal, and that they could tell the research assistant if they were finished eating prior to the end of their time limit. While the child was eating, a research assistant read a book of the child’s choosing to serve as a neutral and consistent distraction. All books that the children could choose from were pre-screened and did not contain food references. Children’s fullness ratings were collected again following the completion of each test meal.

2.6. EAH: Pre- and Post- Intervention

To assess children’s tendency to eat when not hungry (i.e., hedonic eating), a standard EAH protocol (22) was conducted before (visit 1) and after (visit 8) the intervention. Twenty minutes following the completion of the ad libitum meal (the same meal served on the compensation visits), children were presented with a tray of highly-palatable snack foods including: potato chips (Lay’s^® original, PepsiCo, Harrison, NY), cookies (Chips Ahoy!^® chocolate chip cookies, Nabisco, East Hanover, NJ), fruit candy (Starbursts^® original chews, Mars Inc. ^®, McLean, VA), M&M’s(Mars Inc.^®, McLean, VA), cheese crackers (Ritz^® cheese cracker sandwiches, Nabisco, East Hanover, NJ), and brownies (Entenmann’s^® Little Bites Fudge Brownies, Bimbo Bakeries, Horsham). In addition to the snacks, toys and games were made available and children were told that they could play or eat any of the foods that they would like. Children were left alone in the room with the snacks and toys for 10 minutes before a research assistant came back to collect the tray. Prior to the EAH procedure, children reported liking for each of the 6 foods following tasting of a small (<5.0 grams) sample (36). Children reported fullness before and after the meal using the 4-point silhouette scale.

2.7. Parent Questionnaires

On visits 1 and 8, parents completed various questionnaires to assess family demographics, parent feeding practices, and children’s appetitive traits. These questionnaires included the Children’s Eating Behavior Questionnaire (CEBQ), the Child Feeding Questionnaire (CFQ), the Caregiver Feeding Style Questionnaire (CFSQ) and the Division of Responsibility in Feeding Questionnaire (sDOR). The CEBQ assesses eight subscales of child appetitive traits: satiety responsiveness, food fussiness, emotional undereating, slowness in eating, enjoyment of food, desire to drink, food responsiveness, and emotional overeating (38). The CFQ assesses parents’ perceptions, attitudes, and practices toward child feeding, with a specific focus on children’s disposition towards obesity (39). The CFSQ assesses parent feeding styles, which fall into two subscales (responsiveness and demandingness) to categorize parents as having an authoritative, authoritarian, indulgent, or uninvolved feeding style (40). The sDOR measures parents’ adherence to the Division of Responsibility in feeding where parents provide the what, when, and where of feeding while children determine whether and how much they eat (41). Measurements of family racial and ethnic composition, household income and education, and parent weight status were also collected by parental self-report.

2.8. Intervention

The intervention was conducted in small group settings (ranging from 2–3 children per group) over visits 4–7 by trained research assistants. Descriptions of each lesson are included in Table 3 and a conceptual figure of how they are linked with the outcomes in Figure 2. The intervention was modeled after the lessons developed by Johnson (1), but was updated to include a virtual module to teach children about digestion and provide feedback on optimal portion size selection, parent education materials to reinforce the Division of Responsibility in feeding (42), and interactive meals to allow for experiential learning about hunger and satiety. The standardized test meal described above was not served on these visits in order to prevent children from growing fatigued with the foods; instead, children took part in preparing familiar, age-appropriate foods (e.g., sandwiches, tacos, and pizza bagels) at each of the four lessons and the consumption of these meals was used to reinforce concepts from the lessons. For example, in lesson 2, children were queried to find out how their body feels when they are hungry (before meal) and full (after meal).

Table 3.

Intervention lesson descriptions

Lesson	Visit	Description	Associated Learning Theory
1. “What parts of my body help me eat?”	4	Children used a virtual anatomy software (Food Munchers Free Play Module http://146.186.106.208/) to help visualize the digestive system. As children watched food travel through the digestive tract, age-appropriate explanations of the role of the mouth, esophagus, stomach, small intestine, and colon in digestion were provided.	Observational learning- children observed food travel through an avatar’s body and saw that different foods and amounts filled the avatar’s stomach to different level
2. “How do our bodies feel when we are hungry or full?” and “Why do we eat? Why do we stop eating?”	5	Images were shown depicting children eating for various reasons, including “because an adult told me to,” “because the food looks yummy,” and “because the food tastes yummy.” Children then consumed a meal to experience these scenarios. After the meal, children were shown images of reasons children stop eating, including “because my friends are done,” “because there is no food left on my plate,” and “because I want to go play.” The research assistant then discussed the importance of eating in response to internal hunger and fullness signals rather than environmental cues.	Observational learning- children observed images of children eating/stop eating for various reasons Experiential learning- children were guided during the meal to discuss their feelings of hunger and fullness with the researcher, and were assisted in creating a meal based on their hunger levels
3. “What happens if I eat too much?”	6	Children were read “Peter-Peter Pumpkin Eater,” a book to teach them what it can feel like to overeat. After the story, children were asked to discuss why they thought Peter got a stomach ache and if they had ever experienced discomfort associated with overeating. Children were then asked to discuss ways Peter could avoid overeating in the future.	Observational learning- children observed what happens to Peter- Peter Pumpkin Eater when he eats too much
4. “How much does your stomach fill up when you eat?”	7	Children played the “Make a Meal” module of the interactive virtual eating simulation game. In this module, children must select a meal for their avatar that contains the “just right” amount. If they select too little, their avatar tells them they don’t have enough energy to play. If they select too much, their avatar appears sick and sluggish and gets a stomach ache. Children must continue through the simulation until they select an amount that is “just right.”	Experiential learning- children had to use trial and error to select a meal that would be appropriate for themselves Observational learning- children received feedback from the avatar about whether they had chosen too much, too little, or just the right amount of food

Figure 2.

A conceptual figure linking each of the lessons (light yellow boxes) to the concepts explored in these lessons (light blue boxes) and the associated intervention outcomes (light orange boxes).

The intervention was developed based on principles of social cognitive and experiential learning (43), both of which are validated learning techniques for this age group. The use of diagrams and dolls by Johnson (1) relied on children’s ability to learn about concepts of digestion through observation (32); to allow more flexibility and personalization in depictions of digestion, hunger, and fullness, we developed a pilot version of a virtual reality game (“Food Munchers,” http://146.186.106.208/). At the start of the game, children chose either a male or female avatar and then selected from a range of food items varying in portion size and energy density to feed their avatar (Figure 3). Once the avatar ate the selected foods, the child followed the food through the digestive system (i.e., esophagus, stomach, small intestines, and large intestines).

Figure 3.

Image from a pilot version of a virtual reality game (“Food Munchers,” http://146.186.106.208/) used to teach children how to identify and respond to internal cues of hunger and fullness. Children could choose various types and amounts of food to feed their avatar, and received real-time feedback from their avatar about whether they fed the avatar too much, too little, or just the right amount of food based on that child’s own age- and sex-specific energy requirements.

The game included two modules. The goal of the first module (i.e., “Free Play”) was for the child to practice selecting different amounts of foods to observe their impact on the digestive system. Larger portion sizes and higher energy-dense foods took longer to travel through the digestive system and filled the stomach to a greater amount. This allowed children to learn about differences in a food’s energy content without specifically introducing the concept of calories. During the second module (“Make a Meal”), children were instructed to make a meal that was “just right” with respect to amount. For this pilot version of the game, we did not focus on nutritional quality of the items, although this will be an important direction for future implementations. A research assistant prompted children to feed their avatar types and amounts of foods that would leave them with the perfect amount of energy to run and play. Once children made their selections, the avatars provided tailored feedback on the amount of food the child selected; if the child selected too much food, their avatar acted sluggish and sick because it had eaten too much. If the child selected too little food, the avatar did not have enough energy to play and provided feedback to the child that it was still hungry. Children played this game until they selected a meal that was “just right” for their meal energy requirements. In designing the game, we defined “just right” as 25% (± 5%) of the child’s daily age- and sex-specific energy requirements. We based this amount on the assumption that for a typical child who consumes 3 meals and 1 snack per day (i.e., 4 eating episodes), 1 meal would be equivalent to approximately 25% of the child’s total daily energy needs determined by the Dietary Reference Intakes (44).

2.9. Statistical Analyses

The main outcomes of this study were changes in knowledge, short-term energy compensation, and EAH. Normality of each main outcome at baseline and follow-up was assessed using the Shapiro-Wilk test (45) and any outliers were excluded from further analyses. One child (male) was considered an outlier for EAH kilocalories consumed at both baseline and follow-up; no other outliers were detected. Descriptive statistics (i.e., means, standard deviations, and frequencies) were used to analyze demographic data. Independent samples t-tests were used to determine differences in continuous outcome variables by child sex, and Chi-square analyses tested for sex-differences in categorical variables.

The primary outcomes were analyzed using linear mixed models with time-point (baseline vs. follow-up) as the independent variable. The primary outcome of short-term energy composition was analyzed using two approaches for treating the outcome variable. In the standard approach, total energy consumed during each visit was calculated by adding the energy consumed from the preload to the energy consumed from the meal that followed. Linear mixed models were then performed to examine differences in total intake (dependent variable) as a function of preload condition (low kcal vs. high kcal) and timepoint (baseline vs. follow-up). For the second approach, the dependent variable was transformed to calculate COMPx using the following formula: [(meal kcal following low-energy preload - meal kilocalories following high-energy preload)/(high-energy preload kilocalories – low-energy preload kilocalories)] x 100 (11). The amount of energy consumed, rather than served, during each preload was used in this equation to more accurately capture each child’s COMPx. A COMPx below 100% indicates that children undercompensated, or overate at the meal following the high-energy preload compared to the low-energy preload. COMPx above 100% indicates that children overcompensated, or underate at the meal following the high-energy preload compared to the low-energy preload, and a COMPx of 100% indicates perfect compensation or adjustment following the preloads.

To determine which covariates to examine in the mixed models, we tested for correlations between variables of interest (e.g., age, sex, BMI z-score, socioeconomic factors, parental feeding practices, etc.) and the outcomes. Variables that were correlated with an outcome at α≤0.10 were included in the corresponding model. Covariates that did not have a significant main effect or interactive effect with time on the respective outcome at α≤0.05 were removed from the final models. Significance for the mixed models was set at α≤0.05. Analyses were conducted with Statistical Package for the Social Sciences version 25 (SPSS Inc., Chicago, IL).

3. RESULTS

3.1. Child characteristics:

Participant characteristics are presented by sex in Table 1. No significant differences between boys and girls were found for any of the variables. Mean ± SD age was 5.1±0.8 years for boys and 4.8±0.8 years for girls. The majority of boys (75%) and girls (87%) came from households with an annual income above $50,000, and were Caucasian (81% for both). One girl was excluded from anthropometric measures (BMI z-score and weight classification) at follow-up because she had a cast for a broken leg, and height and weight were not collected.

3.2. Outcome 1- Knowledge:

An effect of time was found on children’s knowledge related to the intervention (F_1,62=57.32, P<0.001), such that knowledge increased from baseline (3.5±0.3) to follow-up (7.0±0.3). Further analyses revealed a significant time-by-age interaction such that the increase in knowledge from baseline to follow-up was greater for older than younger children (F_2,63=37.75, P<0.001) (Figure 4).

Figure 4.

Knowledge scores for the entire sample increased from baseline (3.50±0.32) to follow-up (7.00±0.33) (P<0.001). However, a significant time-by-age interaction showed that, as age increased, the increase in knowledge across the intervention was greater (F_2,63=37.75, P<0.001).

3.3. Outcome 2 –Short-term energy compensation

3.3.1. Energy compensation quantified with the standard approach (i.e., total energy intake):

A mixed linear model showed no main effect of preload condition on total energy intake (preload energy + meal energy) (F_1,124=2.03, P=0.16). There was a main effect of time such that total energy intake was higher at follow-up than at baseline (F_1,124=5.62, P=0.02), however, this was attributable to a time-by-preload condition interaction demonstrating that children’s meal energy intake following the low-energy preload was higher at follow-up than at baseline. There were no differences between baseline and follow-up for meal energy intake after the high-energy preload conditions (F_1,126=5.57, P=0.02) (Figure 5; Table 4). Covariates including child age, BMI-z-score, body weight, and estimated energy requirements did not affect the final model (P > 0.05 for all). An exploratory analysis revealed a three-way interaction between preload condition-by-time-by-child sex interaction on total energy intake (F_5,126=2.61, P=0.03). This interaction revealed that following the intervention, boys showed improved compensation in that they consumed similar amounts in the two preload conditions, while girls consumed more at the high-energy than the low-energy condition (a similar response to what was observed at baseline). Figure 6 depicts all pairwise comparisons for this three-way interaction.

Figure 5.

A time-by-preload condition interaction revealed that total energy intake (preload energy + meal energy) at baseline was higher following the high-energy preload compared to the low-energy preload. Additionally, total energy intake was higher at follow-up than at baseline following the low-energy preload. No differences were found for total energy intake during the high-energy preload conditions (F1,126=5.57, P=0.02). Energy intakes during the no-preload conditions are shown in the dotted lines for baseline (406.05±184.84 kcal) and follow-up (433.12±169.01 kcal), but were not included in analyses since these visits were not randomized.

Table 4.

Mean ± SD energy and weight intakes in kilocalories and grams, respectively, across the various meal conditions at baseline and follow-up

	Baseline		Follow-Up
	Energy(kcal)	Weight(g)	Energy(kcal)	Weight(g)
Low-energy preload	3.1±0.3	166.5±18.0	3.1±0.3	169.0±16.9
Ad libitum meal following low-energy preload	373.8±163.9	314.9±146.2	457.1±242.3	353.7±200.0
Total intake (meal + preload) following low-energy preload	376.8±163.8	481.4±144.4	460.2±242.3	522.7±202.3
High-energy preload	157.4±1.1	171.1±1.2	153.9±14.8	167.3±16.1
Ad libitum meal following high-energy preload	334.6±163.0	286.9±144.0	350.8±184.8	286.6±161.9
Total intake (meal+ preload) following high-energy preload	235.7±163.2	458.1±144.5	504.7±181.8	453.9±158.3
Meal preceding EAH	406.1±184.8	363.4±189.7	433.1±169.0	374.4±167.2
EAH	203.9±113.2	56.3±41.8	256.1±186.1	44.8±26.0

Figure 6.

A mixed linear model revealed a three-way interaction between time, preload condition, and child sex on total energy consumed (preload + meal) (F_5,126=2.61, P=0.03). This interaction revealed that following the intervention, boys consumed similar amounts in the two preload conditions, while girls consumed more at the high-energy than the low-energy condition (a similar response to what was observed at baseline). Different letters denote significant differences at α<0.05.

The same analyses were repeated after excluding milk from total intake, as milk intake could reflect thirst rather than hunger. In agreement with the results that included milk, a time-by-preload interaction was found showing that the greater intake from foods at the meal at follow-up was attributable to increases after the low-energy preload (F_1,126=6.75, P=0.01). Additionally, the time-by-sex-by-preload condition interaction remained significant (F_5,126=2.60, P=0.03). These results demonstrate that overall, boys improved short-term energy compensation with the intervention while girls showed no change.

In addition to the main outcomes of intake, exploratory analyses were performed to test for potential sex differences in meal energy density (ED) and to determine whether change in knowledge was related to change in compensation. There were no main effects or interactions between time and preload condition on meal ED; however, a time-by-sex interaction on meal ED was found such that boys consumed a higher meal ED following both preloads at follow-up than girls, but no sex differences were found in meal ED at baseline (F_1,126=4.47, P=0.04). These results were not explained by differences in fluid milk intake between boys and girls (P>0.05.) However, when excluding one child (male, age 5 years) who was an outlier, the time-by-sex interaction was no longer significant (F_2,119=2.12, P=0.13). Finally, a mixed linear model indicated that change in knowledge across the intervention did not predict follow-up energy compensation when controlling for baseline energy compensation and child sex (F_7,56=1.11, P=0.37).

3.3.2. Energy compensation calculated via COMPx:

Overall, COMPx did not change from baseline to follow-up (F_1,60=2.60, P=0.11). However, a time-by-sex interaction was found (F_1,60=4.68, P=0.03) such that in boys, COMPx improved from baseline (Mean ±SEM= 6.0±32.6%) to follow-up (107.4±33.4%), but in girls, COMPx did not change from baseline (44.3±20.6%) to follow-up (29.6±16.6%) (Figure 7). Covariates including child age, BMI-z-score, body weight, and estimated energy requirements did not affect the final model (P > 0.05 for all). Additionally, there was no effect of preload order on COMPx (F_1,62=0.00, P=0.95). In addition, a multiple linear regression with follow-up COMPx as the dependent variable showed that change in knowledge with the intervention did not predict children’s final COMPx when controlling for baseline COMPx and sex (F_3,31=1.41, P=0.26). In line with results for energy compensation calculated using the standard approach, boys showed improvements in COMPx score from baseline to follow-up, but overall, change in knowledge did not predict change in COMPx.

Figure 7.

COMPx scores for the entire sample did not differ significantly from baseline (25.18±19.27) to follow-up (68.47±19.64) (P=0.112). However, a significant sex-by-time interaction found that boys’ COMPx score significantly improved from baseline to follow-up, but girls’ COMPx did not significantly change across the intervention. No significant difference in COMPx was found between boys and girls at baseline (P>0.05), but boys’ COMPx was significantly higher than girls’ COMPx at follow-up (P=0.048).

Exploratory analyses were conducted to determine if baseline parental feeding practices and child appetitive traits (as measured by the parent questionnaires) interacted with time to affect children’s change in energy compensation, computed by both the standard approach and COMPx; however, no significant interactions were found (P > 0.05 for all models tested). Additionally, there were no systematic differences in parental survey responses between boys and girls, and the sex differences in COMPx were not explained by any of the feeding practices or appetitive behaviors (P > 0.05 for all models tested).

3.4. Eating in the absence of hunger:

There was no main effect of time (F_1,58=2.15, P=0.15), sex (F_1,58=1.18, P=0.28), or a time-by-sex interaction (F_1,58=0.00, P=0.98) on the amount children consumed in the absence of hunger (EAH) when all 31 children were included in the analysis. A model that did not include sex as a covariate confirmed this finding (F_1,60=2.18, P=0.15), indicating that EAH did not change from baseline to follow-up in boys (206.0±25.9 kcal at baseline; 246.7±36.3 kcal at follow-up) or girls (176.9±18.7 kcal at baseline; 216.5±26.8 kcal at follow-up). Additional analyses were performed after excluding data from children who did not report sufficient levels of fullness prior to EAH, and results remained the same after this exclusion.

4. DISCUSSION

This technology-enhanced intervention was associated with increases in 4–6 year-old children’s knowledge of digestion, hunger and fullness, and improvements in boys’ short-term energy compensation following a 4-week intervention delivered in the laboratory. Our findings support prior work from Johnson (1) by demonstrating that short-term energy regulation can be improved in some children, potentially by using interventions that target appetite awareness. The overall increase in knowledge among all children, paired with the improvement in energy compensation in boys, suggests that this technology-enhanced intervention may be beneficial for improving children’s appetite awareness.

One of our primary outcome variables, knowledge related to hunger and fullness, showed robust improvements following the intervention. Overall, children increased knowledge related to digestion, anatomy, and the bodily sensations that accompany feelings of hunger and fullness. However, change in knowledge did not explain change in short-term energy compensation. Interventions that target nutrition knowledge alone are unlikely to produce changes in behavior (46). However, given that boys improved both knowledge and energy compensation from baseline to follow-up, it is possible that the knowledge assessment was not comprehensive enough to capture aspects of the intervention that lead to improvements in short-term energy regulation. The majority of the knowledge questions focused on concepts related to the digestive system. Although digestion was an integral portion of the intervention, the lessons also focused on teaching children to identify and respond to internal hunger and fullness cues through experiential and observational learning. The fact that the knowledge survey was not tested for initial validity and only captured a limited portion of the concepts taught in the intervention may help explain the lack of congruency between knowledge and energy compensation behavior. In addition, the fact that age was positively related to change in knowledge across the intervention suggests that the lessons may have been better targeted at school-aged children (age 5–6 years), or that older children were simply better at knowledge recall. This illustrates that, even within a narrow age range, comprehension, internalization, and retention of intervention messages vary.

In addition to increases in knowledge, we also showed overall improvements in children’s ability to adjust for the energy content of liquid preloads. This was evidenced by children consuming similar amounts of total energy (preload + meal) following the intervention compared to baseline consumption. This was driven by the boys, who accomplished this by consuming more at the meal following the low-calorie beverage after the intervention than they did before the intervention. Potentially, they were more focused on internal hunger and fullness cues during the follow-up assessment period, and as a result, showed better short-term adjustments. While several studies have shown that young children tend to show accurate adjustment for calories delivered in a first course or preload, even without intervention (10,12,47), other studies dispute these findings (e.g. 48), and it is likely that compensation ability varies widely across children. The current pilot study suggests that short-term energy adjustment can improve among some children and supports the use of technology-enhanced platforms to deliver this curriculum.

Although these results are promising, there are several caveats to consider. First, although boys compensated more accurately following the intervention, they did so by increasing intake following the low-calorie preload instead of decreasing intake after the high-calorie preload. Therefore, it is possible that some boys were overeating after the intervention. Second, by both COMPx and the standard approach to assessing energy compensation, we found that boys were more responsive to preload energy following the intervention than girls. However, it is critical to point out that the sample size was small, and we were underpowered to test two- and three-way interactions (49), and results should be interpreted cautiously. Despite this, the pattern of results was similar across the two approaches, and previous literature supports the notion that boys can demonstrate more accurate energy compensation than girls (1,11,50–51). Though this finding has not been consistently reported across other studies (12–13,16,52–53), and the reasons for these sex differences are presently unclear, they suggest that different intervention strategies for boys and girls may be needed to improve energy intake regulation. Additionally, although evidence has shown that character-based gaming apps may be effective for delivering nutrition education to preschoolers (54), it remains possible that boys and girls respond differently depending on mode of delivery. Regardless of these caveats, the current intervention offers a promising new approach to improve children’s energy intake regulation.

Despite observing improvements in children’s knowledge and short-term energy compensation, EAH did not change from baseline to follow-up. Because the intervention specifically targeted children’s ability to respond to internal cues of hunger and fullness, it is possible that additional lessons aimed at reducing children’s tendency to eat in response to external cues would be needed to reduce EAH. For example, the appetite awareness intervention conducted by Boutelle and colleagues (2) was only effective at improving EAH among 8–12 year-olds when it was paired with additional training to teach children how to moderate the effects of food cues. This suggests that EAH is malleable, but must be targeted differently than other self-regulatory behaviors such as energy compensation.

To our knowledge, this study is the first to use a technology component to improve children’s ability to regulate energy intake. Prior interventions that included interactive technology (e.g. video games, mobile apps) have demonstrated modest effects on children’s short-term health behavior changes such as physical activity (55), fruit and vegetable intake (56–57), and weight status (58), although results vary across studies. Multi-component interventions (e.g. web-based app in addition to school-based programs) tend to demonstrate greater efficacy than standalone technology (58). However, given the ubiquity of mobile phones, tablets, etc., technology-enhanced interventions are a promising approach for targeting young children for prevention and/or treatment of pediatric obesity (59). Our findings contribute to the growing body of literature suggesting that interactive, web-based platforms can enhance the delivery of nutrition education interventions for children.

The current study had several strengths worth noting. We had excellent retention (97%) across a 10-visit study, and the majority of children finished the preloads, allowing for objective measurement of children’s short-term energy compensation. The multi-modal nature of the intervention, which integrated technology with experiential learning, reinforced concepts in multiple ways to keep children engaged. Parent-directed materials were also provided, which is critical given that caregivers are primarily responsible for their children’s health at this age (59). However, we did not measure parent engagement with these materials, a critical question for future studies. Additionally, several other limitations should be mentioned. Most notably, because this was a pilot study, it was not feasible to include a control group. Because of this, we cannot attribute changes in knowledge and energy compensation to intervention material specifically, as compared to time or contact hours with the intervention staff. Second, the questions that were used to assess children’s knowledge related to appetite regulation were developed in our laboratory and were not tested for internal validity. Conclusions made regarding this outcome should be interpreted in the context of this limitation with our knowledge assessment. Additionally, it is also important to note that EAH was not randomized, and always took place on visits 1 and 8. Children may have learned to expect palatable snacks after their meals since this took place on their first visit, and this may have influenced intake at subsequent test sessions. Because this intervention was conducted in a laboratory setting, it is unknown how these findings translate to eating behaviors in childcare, home, and other settings. To make the laboratory setting more comfortable for children, we read books to the children during their meals. In most cases, the book that was read to the child during their meal varied across visits and although these books were screened and did not contain food cues, it is unknown how different stories may affected food intake. Finally, the intervention was time- and resource-intensive and may not be feasible to deliver on a larger scale in other settings (e.g., schools, childcare centers). Adapting the lesson materials for more efficient delivery outside the laboratory, with a more culturally diverse group of children, will be critical for broad uptake of this intervention in the future.

To conclude, this technology-enhanced intervention increased children’s knowledge about digestion and satiety and, among boys, was associated with improved short-term energy compensation. Our findings suggest that boys and girls respond differently to an intervention designed to improve self-regulation of food intake, which supports the notion that more personalized and tailored approaches are needed to prevent overeating. Future work should focus on understanding the nature of these sex differences and expanding the intervention so it can be tested in a more culturally diverse cohort.

1. LIST OF ABBREVIATIONS

COMPx

energy compensation index

EAH

eating in the absence of hunger

BMI

body mass index

CEBQ

Children’s Eating Behavior Questionnaire

CFQ

Child Feeding Questionnaire

CFSQ

Caregiver Feeding Style Questionnaire

sDOR

Division of Responsibility in Feeding Questionnaire

ED

energy density