Role of Child Weight Status and the Relative Reinforcing Value of Food in Children’s Response to Portion Size Increases

Abstract

Objective

Food portion size is an important determinant of intake in children. It remains unknown if children’s weight status and relative reinforcing value of food (RRVF) interact with portion size to affect intake.

Design and Methods

In a randomized crossover design, 25 normal-weight and 25 obese children, ages 8 to 10 years, consumed dinner once a week for 3 weeks. At each dinner, the same meal was served, but the portion size of all foods (chicken nuggets, hash browns, ketchup, green beans, brownies) and a sugar-sweetened beverage (fruit punch) varied across conditions (100%, 150%, 200%). Children’s RRVF was assessed using a behavioral choice task.

Results

There was a significant main effect of portion size condition (P=0.003) and weight status (P=0.0005) and a non-significant trend for a portion size-by-weight status interaction (P=0.108) on intake. Mean intakes across conditions (100%, 150%, 200%) were 801±57, 964±58, and 873±57 kcal for normal-weight children and 1041±57, 1129±57, and 1210±57 kcal for obese children, respectively. Neither the main effect of RRVF status nor the condition-by-RRVF status interaction was significant (P>0.48).

Conclusions

Environments that offer large portions of palatable foods affect all children’s intake irrespective of their weight status or how reinforcing they find food to be.

Introduction

One in 5 children between ages 6 and 11 years are obese (1, 2). There is a pressing need to better understand the controls of eating in children and their interactions with the obesogenic food environment and to identify children who are most susceptible to overeating.

The high prevalence of childhood obesity has been attributed, in part, to environmental influences on children’s eating behavior. Obesity-promoting environments are thought to encourage excess intake by offering access to large food portions. Laboratory studies showed that children as young as two increased their intakes when served larger portions (3-8). Little is known, however, if obese children show more pronounced intakes when food portions increase.

Despite the omnipresence of large food portions, not all children are equally susceptible to overeating. Individual differences in the relative reinforcing value of food (RRVF) may account for some of these differences. The RRVF examines how hard individuals are willing to work to gain access to a preferred food rather than an appealing nonfood alternative (9). The construct can also be used to determine the RRV of one food or type of food (e.g., high energy-dense food) compared to another food or type of food (e.g., low energy-dense food) (10). Studies have shown that overweight/obese children, on average, find food more reinforcing than non-obese children (11). No study has examined the extent to which children’s RRVF may interact with environmental cues, such as food portion size, to affect intake.

The primary aim of this study was to compare energy intake at a meal in normal-weight and obese children when the portion size of energy-dense foods and a sugar-sweetened beverage was systematically increased. We hypothesized that increasing the portion size of all foods and beverages would lead to a significant increase in energy intake in both normal-weight and obese children. Obese children, however, would show a significantly greater increase in intake than would normal-weight children. A second aim was to test if children’s response to increases in portion size was affected by how reinforcing they find food to be. We hypothesized that, when controlling for BMI, children who find food very reinforcing relative to a nonfood (activity) alternative would show a significantly greater increase in intake than would children who find food less reinforcing.

Methods

Design

This study used a within-subjects mixed design with weight status as a between-subjects factor and portion size as a within-subjects factor. Children ate dinner at the Center for Weight and Eating Disorders (CWED) at the University of Pennsylvania once a week for three weeks. Test visits were scheduled one week apart. On each test day, the same foods were served for dinner, but all foods and the beverage varied in portion size (100%, 150%, 200%). The order of presenting the conditions was randomized across groups of children eating together. Children’s RRVF was assessed during a screening visit.

Participants and Recruitment

Subjects included 50 racially/ethnically diverse boys and girls, ages 8 to 10 years, and their mothers. Half of the sample was normal-weight (BMI-for-age 5–84^th percentile); the other half was obese (BMI-for-age ≥95^th percentile). Overweight children (BMI-for-age 85-94^th percentile) were excluded to maximize the weight discrepancy between the weight groups.

Families were recruited through newspaper/online advertisements and flyers distributed in local pediatrician offices and grocery stores/pharmacies (5/2011–3/2012). To be included, children had to be: 8 to 10 years; normal-weight or obese; and like most study foods (see Taste Preference Assessment). Children were excluded if they had medical conditions or were taking medications known to affect food intake or body weight; were diagnosed with a learning disability or sight/hearing impairment; had food allergies/intolerances; or were on a special diet. Figure 1 shows the CONSORT flow diagram for subject enrollment.

Figure 1.

CONSORT flow diagram for subject enrollment.

The study was approved by the Institutional Review Board of the University of Pennsylvania. Subjects were asked to provide voluntary consent (mothers) and assent (children) to participate in the study by signing the consent and assent forms.

Assessment of Child Height and Weight

At the screening visit, a trained staff member measured children’s weight on a digital scale (Tanita BWB-800, Arlington Heights, IL; accurate to 0.1kg) and their standing height on a stadiometer (Veder-Root, Elizabethtown, NC; accurate to 0.1cm) with children wearing light clothing and having their shoes removed. Child age- and sex-specific BMI percentiles and z-scores were calculated using the CDC Growth Charts (12). Computation of children’s daily estimated energy requirement (EER) was based on age-, sex-, weight-, and height-specific equations (13), using a physical activity coefficient of 1.00 (i.e., sedentary) for all subjects.

Taste Preference Assessment

At the screening, children’s liking for all study foods was assessed using a taste preference assessment (14, 15). Children were asked to taste a small amount (<1 teaspoon) of each of the dinner foods and rate each food using a hedonic Likert-type scale with 5 cartoons ranging from a frowning to a smiling face (1=‘dislike very much’, 2=‘dislike a little’, 3=‘just ok’, 4=‘like a little, 5=‘like very much’). Only children who rated the chicken nuggets and hash browns as “Just OK” or above (with at least one of these foods being rated as “Like a Little” or “Like a Lot”) and who rated the remaining foods as “Just OK” or above were included in the study.

Children repeated the assessment for the snacks and activity alternatives used in the BCT. Children sampled each of four snacks (potato chips, crackers, cookies, candy) and spent 1-2 minutes sampling each of the activity alternatives (magazine, coloring book, puzzle, video game). They then rated and rank-ordered all snacks and activity alternatives using the 5-cartoon scale. The most preferred snack and activity alternative were used for the RRVF assessment.

RRVF Assessment

At the screening, children’s RRVF was assessed using a validated, computerized Behavioral Choice Task (BCT) (9-11, 17). When scheduling the screening, we instructed mothers to have her child fast for at least 2 hours before the visit. We verified the time of last food consumption with the child before the taste preference assessment. The RRVF was assessed after ingestion of ~160 kcal from food samples used in the taste preference assessment to reduce hunger and thereby optimize the conditions for RRVF assessment (11, 18-20). Children worked on either one of two computer screen sections. One section was associated with earning points for their preferred snack; the other was associated with their preferred activity alternative (10-11). Children could choose, by mouse click, which reinforcer they wished to work for. Each screen contained three objects of different shapes and colors, which rotated every time the child made a mouse click. Each time the objects matched in shape and color, the child earned one point for the reinforcer they worked for. Children were asked to make choices about earning points for either of the two reinforcers over a total of 5 trials. Each trial was terminated when children had earned 20 total points between the two reinforcers. The variable ratio (VR) concurrent reinforcement schedules varied the number of responses that were required to earn a point across trials. For the first trial, the reinforcement schedule for the snack and the activity alternative were set to reinforce every second response (VR2). In the subsequent trials, the reinforcement schedule for the snack was doubled (VR4, VR8, VR16, and VR32). The schedule for the activity alternative was kept constant at VR2 across all 5 trials. Children were instructed that they would be given access to the reinforcer that earned the most points over the 5 trials after they finished the RRVF task. They were also informed that 10 points earned for the snack could be exchanged for one 3.25 oz soufflé cup of that snack; 10 points earned toward the activity alternative could be exchanged for 2 minutes of time spent in that activity. The maximum amount that could be earned was 10 cups of the snack (for a maximum of 100 points [20 points times 5 trials]) or 20 minutes to be spent in the activity alternative (for a maximum of 100 points [20 points times 5 trials]). Snack portions were similar in volume, but not weight or calories. If children earned more points for the activity alternative, they were given access to that activity in the lab (they were not allowed to take the game home); if children earned more points for the snack, they were handed the snacks at the end of the task, which they could take home. The %RRV of the snack and activity alternative were computed by dividing the total number of points earned for the snack by 20 and multiplying by 100. This yielded a percentage of the total number of points earned for snack in each of the 5 trials.

Test Meal

On all study days, participants were served chicken nuggets, hash browns, green beans (w/small amount of butter), brownie, and fruit punch (Table 1). All foods served in this study are commonly consumed by children of this age (22). Portion sizes served in the 100% condition were based on intakes from a pilot group of 8 children and on data from the Continuing Survey of the Food Intakes by Individuals (23). None of the children in the actual study finished all of the food that was provided to them in any of the conditions. The meals were prepared in a research kitchen according to a standardized protocol. All foods were pre-weighed prior to being served and reweighed after the children had finished eating to determine the weight consumed by each child to the nearest 0.1g. Information from nutrition facts labels were used to convert the amounts consumed to energy intake.

Table 1.

Types and amounts of foods served during dinner across portion size conditions

Foods and Beverage	Energy density (kcal/g)	100% Portion		150% Portion		200% Portion
		Weight (g) / # of pieces	Energy (kcal)	Weight (g) / # of pieces	Energy (kcal)	Weight (g) / # of pieces	Energy (kcal)
Chicken nuggets1	3.00	10 pieces	540	15 pieces	810	20 pieces	1,080
Hash browns2	1.98	20 pieces	378	30 pieces	567	40 pieces	756
Ketchup3	1.18	80	94	120	141	160	188
Green beans (w/ butter)4	0.31	100	31	150	47	200	62
Brownies (bite-size)5	4.38	6 pieces	420	9 pieces	630	12 pieces	840
Fruit punch6	0.33	300 (10 fl oz)	100	450 (15 fl oz)	150	600 (20 fl oz)	200
Total kcal served:		1,563		2,345		3,126

¹Tyson, Chicken Nuggets (prepared from frozen), Tyson Foods, Springdale, AR

²Ore-Ida Tater Tots (prepared from frozen), Heinz North America, Pittsburgh, PA

³Heinz, Tomato Ketchup, Heinz North America, Pittsburgh, PA

⁴Del Monte, Cut Green Beans (canned, salted), Del Monte Foods, San Francisco, CA

⁵Entenmann’s, Little Bites Fudge Brownies, Orograin Bakeries Products Inc., Horsham, PA

⁶Hawaiian Punch, Fruit Juicy Red/Lemon Berry Squeeze, Manufactured under license by Mott’s LLP, Rye Brook, NY; The nutritional information of the fruit punch changed over the course of the study. Twenty-two participants were served a fruit juice which provided 70 kcal per 8 fl oz instead of 80 kcal per 8 fl oz (original formulation).

Procedures

On the day of their visits, mothers were instructed to have their child consume lunch and an afternoon snack (if desired) and to not consume any foods/beverages (except water) after 3:00PM. Upon arrival to the Center at 5:00PM, mothers were asked to complete a meal report to ensure that they had complied with these instructions. 75% of mothers complied with the instructions (others indicated that their child had a small snack after 3:00PM). Children’s hunger was assessed using cartoons depicting an empty, a half-empty, and a full stomach (24, 25). 94% of children perceived their stomach to be empty or half-empty before dinner across conditions. Research staff engaged children in games for 30 minutes before dinner was served. Children ate in groups of 3-4 children for 25 minutes in the presence of staff. Groups consisted of either all normal-weight or all obese children to minimize peer influences on eating with respect to weight status (26). Children ate in groups, rather than alone, to simulate mealtime at school/home. Children were asked not to share foods and were told that they could eat/drink as much or as little as they desired during dinner. Mothers waited in a room nearby for the duration of the visit.

Statistical Analysis

Data were analyzed using SAS (Version 9.2; SAS Institute Inc, Cary, NC). The Shapiro-Wilk test in conjunction with distribution plots and summary statistics confirmed normal distribution of continuous variables. Chi-Square and Fisher’s Exact tests compared normal-weight and obese children in preference ratings. To test Aim 1, a mixed-effects linear model with repeated measures assessed total energy intake (primary outcome) by portion size condition between normal-weight and obese children. Secondary outcomes were calories consumed from only the beverage and total calories consumed as %EER. The fixed factor effects used in all models were condition and weight status. For all models, the condition-by-weight status interaction and effects of week and sex were tested as covariates.

To test Aim 2, a 2 weight group × 5 trial (trials 1-5) mixed-effects model tested the main effects of weight status and trial on %RRV. The weight group-by-trial status interaction was tested for significance. Children were classified as ‘high RRVF’ when the total points earned for the snack reinforcer exceeded the total points earned for the activity alternative or as ‘low RRVF’ when the total points earned for the activity alternative exceeded the total points earned for the snack. A 2 RRVF status (low RRVF/high RRVF) × condition mixed-effects model tested the main effects of RRVF status and condition on total energy intake. Children’s BMI z-score was added as a covariate to the model to control for differences in weight status. The condition-by-RRVF status interaction was tested for significance. We repeated this analysis with number of mouse clicks, rather than points earned, in the model.

Descriptive statistics are reported as mean (±SD) for continuous variables or as frequency (percentage) for categorical variables. Results from the mixed linear model analysis are presented as model-based means (±SEM). Reported P values are two-sided and P<0.050 was considered significant for all tests.

Results

Child characteristics

Table 2 depicts the demographic and anthropometric characteristics for children by weight group. Approximately half the children in each group were male and the majority of children were African American. Obese children significantly differed from normal-weight children in all weight measures (P<0.001).

Table 2.

Demographic and anthropometric characteristics (N (%) or mean ± SD) of normal-weight (N = 25) and obese (N = 25) child participants

Child Characteristic	Normal-weight (N (%) or mean ± SD)	Obese (N (%) or mean ± SD)
Age (years)	9.7 ± 0.8	9.5 ± 0.8
Sex (male / female)	14 (56%) / 11 (44%)	10 (40%) / 15 (60%)
Race
Black or African American	15 (60%)	19 (76%)
White	6 (24%)	5 (20%)
More than one race	4 (16%)	1 (4%)
Ethnicity
Hispanic	3 (12%)	1 (4%)
Not Hispanic	20 (80%)	21 (84%)
Unknown	2 (8%)	3 (12%)
Height (cm)	137.0 ± 8.3	144 7 ± 77 ***
Weight (kg)	31.0 ± 4.8	52.5 ± 7.7 ***
BMI (kg/m2)a	16.5 ± 1.5	25.0 ± 2.0 ***
BMI z-score	−0.2 ± 0.7	2.0 ± 0.2 ***
BMI-for-age percentile	45.8 ± 25.0	97 7 ± 1 1 ***

^***P ≤ 0.001;

^aBody Mass Index (BMI)

Taste Preference Ratings

The majority of children indicated ‘like very much’ and ‘like a little’ for the chicken nuggets (96%), brownies (88%), and punch (92%). The hash browns and beans were less well liked, with 62% of children indicating ‘like very much’ or ‘like a little’ for the hash browns and 72% of children giving these ratings for the beans. Normal-weight and obese children did not differ in their preference ratings for chicken nuggets, green beans, brownies, and punch (P>0.24), but there was a significant difference in their liking ratings for hash browns (P=0.04). The percentage of children who rated the hash browns as ‘Just ok’, Like a little’, or ‘Like very much’ were 40%, 8%, and 52% among normal-weight children and 36%, 36%, and 28% among obese children, respectively.

The percentage of children who indicated “like very much” or “like a little” were above 80% for potato chips, cookies, M&Ms, and the video game. The highest ranked snack reinforcer was potato chips (30% of children); the highest ranked activity alternative was the video game (58% of children).

Energy Intake by Weight Status

There was a trend towards a significant interaction between condition and weight status (P=0.108). Both the main effects of condition (P=0.003) and weight status (P=0.0005) were statistically significant. Mean intakes across the 100%, 150%, and 200% conditions, with groups combined, were 921±40, 1046±41, and 1041±40kcal, respectively, which included 83±5, 99±5, and 112±5kcal consumed from the beverage. The results did not change when excluding the calories consumed from the beverage from the analysis or intakes from children who had not fasted for 2 hours and reported a half-empty stomach before the meal or when controlling for sex. When presented as %EER, mean intakes across the 100%, 150%, and 200% conditions were 53%±4, 63%±4, and 58%±4 for normal-weight children and 56%±4, 61%±4, and 64%±4 for obese children (P=0.16).

Planned comparisons showed that obese children consumed significantly more calories during the meal compared to normal-weight children in all conditions (P<0.046). Across the 100%, 150%, and 200% conditions, obese children consumed 30% (240kcal), 17% (166kcal), and 39% (337kcal) more calories than normal-weight children (Figure 2).

Figure 2.

Total energy intake (model-based means ± SEM) across portion size conditions for normal-weight (N = 25) and obese (N = 25) children.

RRVF

There was a significant main effect of trial for the %RRV of food and the activity (P<0.001), but no significant main effect of weight status (P=0.59) or trial-by-weight status interaction (P=0.69) for either reinforcer (Figure 3). The results did not change when basing the %RRV on the number of mouse clicks instead of points earned. These findings indicate that in this study normal-weight and obese children did not differ in the way they allocated their choices between a snack reinforcer and an activity alternative.

Figure 3.

Relative reinforcing value (%RRV; model-based means ± SEM) by food reinforcer and activity alternative across 5 trials for normal-weight (N = 25) and obese (N = 25) children

Normal-weight and obese children also did not differ in the percentage of children classified as high RRVF (36% normal-weight, 44% obese) versus low RRVF (64% normal-weight, 56% obese; chi square: 0.33; P=0.56).

Energy Intake by RRVF Status

There was a significant main effect of condition on energy intake (P=0.003), but no significant main effect of children’s RRVF status (P=0.48) or condition-by-RRVF status interaction (P=0.55; Figure 4). The results did not change when children’s BMI z-score was added as a covariate to the model or when excluding intakes from children who had not fasted for 2 hours and reported a half-empty stomach before the meal or when RRVF status was based on the number of mouse clicks instead of points earned. These findings indicate that changes in food portion size affected all children in the study, irrespective of their RRVF.

Figure 4.

Total energy intake (model-based means ± SEM) across portion size conditions for children with a low RRVF (N = 30) and a high RRVF (N = 20).

Discussion

This study showed that obese children, independent of portions served, consumed, on average, 29% more calories when served a palatable meal than normal-weight children. With respect to the RRVF, this study showed that increasing food/beverage portions affected all children’s intake irrespective of their RRVF. Together, these findings suggest that obesogenic environments that offer large amounts of energy-dense foods promote overeating both in normal-weight and obese children and in those who find food more or less reinforcing relative to an activity alternative.

The primary aim of this study was to compare energy intake at a meal in normal-weight and obese children when the portion size of all foods and a sugar-sweetened beverage was doubled. Findings from prior portion size experiments indicate that obese children were more susceptible to overeating when served large portions than normal-weight children (27, 28). In our study, the findings pointed in the same direction, however, we only found a non-significant trend for a condition-by-weight status interaction. It is possible that larger sample sizes may be required to show significant differences in children’s response to portion changes based on their weight status. Both normal-weight and obese children consumed a large number of calories during a single meal, which corresponded to >50–60% of their EER across conditions. We also showed that doubling the size of a sugar-sweetened beverage from 10 fl oz to 20 fl oz led to a 33% increase in calorie intake from that beverage. If sustained, these intakes can lead to excess weight gain in children over time. Important clinical implications of these findings are that obesity prevention efforts that target portion size modifications should focus on children with a range in BMI. Given that normal-weight children appear to be equally susceptible to overeating when served large portions, they, too, should be considered targets for intervention, especially children with a family history of obesity.

A second aim of this study was to test if children’s response to increases in portion size was affected by their RRVF. Contrary to our hypothesis, this study showed that children’s RRVF did not impact their intake at dinner or their response to portion changes even when controlling for BMI z-score. This suggests that all children, irrespective of how reinforcing they find a snack reinforcer to be, may be susceptible to overeating in environments that offer large portions of energy-dense foods. These results, however, must be viewed in the context of the study set-up. For example, we used a snack, rather than a meal, as the food reinforcer to assess children’s RRVF. It is possible that interactive effects on intake between children’s RRVF and food portions may be reinforcer-specific. Future studies should determine the impact of changing portions of the specific food reinforcer that was used to assess children’s RRVF on intake. Second, the BCT provided children with an activity alternative, but the dinner did not. It is possible that children’s intake during dinner may have differed had they been allowed to engage in activities when they were done eating (i.e., they may have stopped eating earlier). Third, the dinner was consumed in a group, while the BCT was completed individually, which may have also affected the outcomes. Lastly, for the BCT, we chose the highest ranked activity alternative and kept the VR constant. It is possible that selecting a different type (e.g., least liked) and/or different reinforcer schedule for the activity may have differentially affected how children allocated their choices.

In our study, normal-weight and obese children did not differ in their RRVF. This finding differs from previous studies, which showed that overweight/obese children found food more reinforcing than normal-weight children (11). It is possible that the type of snacks and games used for the BCT may have affected how children allocated choices towards the two reinforcers. For example, the most popular activity alternative was the video game. Gaining access to this novel game may have trumped some children’s motivation to work for the more familiar snack reinforcer.

The strengths of this study include the unique sample and the large number of minority children. To our knowledge, this also is the first study which assessed multiple eating phenotypes concurrently under controlled laboratory conditions to test the extent to which environmental factors (portion size) interact with individual predispositions (RRVF) to affect intake in normal-weight and obese children. The study also had limitations. One, while unique, the small sample size may have limited the statistical power for this study. Cohen’s D effects sizes for the 100%, 150%, and 200% conditions were 0.60, 0.41, and 0.85, respectively. Second, while methodologically rigorous, having children eat with children of the same weight status may have created an unusual eating environment and it is unknown to what extent this may have affected intakes. Third, differences in liking ratings for hash browns among normal-weight and obese children may have differentially affected their intake of hash browns. Fourth, differences in some children’s hunger or physical activity levels (not assessed) may have impacted their intake at meals. Lastly, given that the study used an estimated, rather than observed, measure of physical activity for the computation of EER, children’s true daily energy requirement may differ from the one used in this analysis.

Together, the findings from this study suggest that the portion size of foods/beverages exert a robust effect on child energy intake with individual differences with respect to children’s weight status and the RRVF appearing to play a lesser role. Therefore, modifying the food environment to offer smaller food and beverage portion sizes and implementing obesity prevention and treatment programs that target obesogenic eating behaviors are expected to benefit both normal-weight and obese children.

What is already known about this subject?

• The portion size of food is an important determinant of energy intake in children.

• Overweight and obese children, on average, find food more reinforcing than non-obese children.

What does this study add?

• This study, for the first time, assessed multiple child eating phenotypes concurrently under controlled laboratory conditions to test the extent to which environmental factors (portion size) interact with individual predispositions (relative reinforcing value of food) to affect eating behavior in normal-weight and obese children.

Abbreviations

BMI

body mass index

EER

estimated energy requirement

RRVF

relative reinforcing value of food

VR

variable ratio