The Association of Addictive-like Eating with Food Intake in Children

Abstract

Objectives:

The potential role of an addictive process in problematic eating is a growing area of interest and debate. Children are more vulnerable to the negative effects of addictive substances than adults and may be at increased risk for the development of addictive-like eating behavior. No prior study has evaluated the association of addictive-like eating with objectively measured eating behavior. We examined the relationship between “food addiction” and observed food consumption among children while controlling for body mass index (BMI) percentile, age, and race/ethnicity.

Method:

Seventy children participated in an observed dinner meal, completed a dietary recall interview, and answered a questionnaire about addictive-like eating behaviors. Children’s total calories ordered, calories consumed at dinner, calories consumed post-dinner, and a total of calories consumed at dinner and post-dinner were calculated along with their BMI percentile. We used generalized estimated equation models to investigate the relationship between addictive-like eating and food consumption.

Results:

Elevated “food addiction” symptoms, but not BMI percentile, significantly predicted an increased amount of calories consumed at dinner. Addictive-like eating was not associated with the amount of calories ordered, but addictive-like eating significantly predicted the total amount of combined calories consumed at dinner and post-dinner.

Conclusions:

As the first study of objectively measured eating behavior, we found addictive-like eating in children significantly predicted the total amount of calories consumed. “Food addiction” was a better predictor than BMI percentile of the total amount of calories that children consumed, highlighting the importance of assessing behavioral phenotypes when evaluating caloric intake in children.

Introduction

The potential role of an addictive process in problematic eating is an area of growing scientific interest and ongoing debate. Animal studies demonstrate that rats given highly processed foods or intermittent access to sugar demonstrate neurobiological and behavioral indicators of addiction (i.e. tolerance, withdrawal, binging, dopaminergic downgrading).^1.2 Neuroimaging studies in adult humans suggest overlapping neurobiological systems are activated by both drugs of abuse and hyperpalatable food (e.g. pizza, ice cream).³ Individuals with problematic eating-related (e.g. binge eating disorder, obesity) and addictive behaviors exhibit similar patterns of neural reactivity to food or drug cues, respectively.^4–6 However, there is not a scientific consensus on whether “food addiction” is a valid concept^7,8 and additional research is needed.

To further investigate the contribution of addictive processes to problematic eating behavior, the Yale Food Addiction Scale (YFAS)^* was developed to operationalize “food addiction”.⁹ The YFAS applies the diagnostic criteria of substance dependence based upon the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV)¹⁰ to the consumption of highly palatable foods⁹ (see Table 1 in the Supplement materials). Elevated scores on the YFAS have been associated with more frequent episodes of binge eating, increased impulsivity, higher body mass index (BMI), and stronger cravings for high-fat foods in adults (for a review see Meule & Gearhardt).¹¹ Additionally, in adults, elevated YFAS scores have been linked to genetic profiles and neural response patterns implicated in addiction.^12,13

Although research on whether an addictive process may contribute to problematic food consumption in adults is growing, research in children is limited. Based on the addiction literature, children are more vulnerable to the negative effects of addictive substances than adults.¹⁴ Addictive substances are more deleterious to children, in part, by disrupting the normative development of neural and psychological processes.^15,16 Thus, if certain foods are addictive, children relative to adults might be at increased risk for addictive-like eating behavior. Consistent with this hypothesis, a qualitative study of messages written by overweight/obese 8-to-21 year olds on an intervention website observed that responders frequently described their relationship with food in a manner that was consistent with the DSM-IV criteria for substance dependence and 66% of participants reported feeling addicted to food.¹⁷ Merlo and colleagues¹⁸ reported that 33% of children receiving treatment at a pediatric lipids clinic reported that they were “sometimes” or “often” addicted to food. Further, children with higher scores on the children’s version of the YFAS (YFAS-C) have increased BMIs, greater levels of emotional eating, and lower satiety responsiveness.¹⁹ In adolescents with overweight and obesity, YFAS “food addiction” is associated with more binge eating, elevated food cravings, and higher attentional and motor impulsivity.²⁰

This initial evidence provides preliminary support for the concept of addictive-like eating in children and adolescents. However, no prior study has examined whether “food addiction” is related to objectively measured eating behavior. As addiction is associated with an increased desire for and consumption of the addictive substance,²¹ we would predict that elevated “food addiction” in children would be associated with greater caloric intake. To address this gap in the literature, the goal of the current study is to examine the association of the YFAS-C with observed food consumption among children while controlling for BMI percentile, age, and race/ethnicity.

Material and methods

Participants

One hundred seventeen children were recruited from the New Haven community with their parents to participate in a study on family dining preferences and eating habits. The YFAS-C was later added to the study’s battery of assessments after the initial data collection started; therefore, 70 children completed the YFAS-C. As shown in Table 1, the sample was racially and ethnically diverse. The average age of participants was 8.34 years (SD = 2.7; range 4 – 16 years of age) and 42.9% were female. Of the 70 children, the majority (n = 41) had at least one sibling who also completed the YFAS-C in the study. The Yale University Institutional Review Board approved this study. Parental guardians provided informed consent, and the children provided verbal assent.

Table 1.

Participant Demographics

		N(%)	M	SD
Age (years old)			8.24	2.70
Sex
Male		40 (57.1%)
Female		30 (42.9%)
BMI			20.39	6.18
BMI Percentile			71.41	31.8
Weight Category
Underweight		3(4.84%)
Normal Weight		27 (43.55%)
Overweight		9 (14.52%)
Obese		23 (37.09%)
Ethnicity
White		39 (55.7%)
African American		20 (28.6%)
Hispanic		4 (5.7%)
Other		7 (10%)

Study Procedures

Families were recruited for a consumer market research study on dining preferences and eating habits. Participants arrived at 5:30pm and participated in a focus group-like activity answering questions about their restaurant preferences. Half way through, they were asked to order dinner from a restaurant menu and eat a meal provided at no cost. They were told they could not take home leftovers and were not allowed to share food. Participants returned the next day to complete a dietary recall interview. Parents and children were asked to abstain from eating after 3pm on the first day of the study to standardize hunger levels.

Although not the focus of this paper, one aim of the data collection was to examine how calorie information on restaurant menus impacted parents’ and children’s eating behavior. Families were randomized to one of three calorie labeling conditions: 1) a menu without calorie labels; 2) a menu with calorie labels and a label stating that the recommended daily caloric requirement for adults; or 3) a menu with calorie labels and labels stating the recommended daily caloric requirements for adults as well as children of different age ranges.

All menus displayed items from Au Bon Pain (including salads, sandwiches, beverages, and desserts) and a non-chain restaurant (including appetizers, entrees, and desserts such as mozzarella sticks, pizza, hamburgers, and cheesecake). The menus had a kid’s menu section with items such as chicken fingers, sandwiches, salads, and vegetables. Calorie information was obtained from Au Bon Pain’s website and the caloric content of items from the other restaurant was estimated by weighing the food with an Ohaus digital scale accurate up to + 0.1 grams and using the Food Processor SQL (esha Salem, OR) calorie content database. Unbeknownst to participants, their food was weighed before serving and again once they were done eating. To calculate total calories consumed, the weight of each plate collected after the meal was subtracted from the weight of each plate prior to being served. The weight of the food in conjunction with the caloric density of each item was used to calculate the caloric intake for each parent and child at the study meal.

The next evening, families returned to the laboratory to complete a dietary recall interview using the Multiple Pass Method²² assisted by pictures and measuring props to illustrate portion sizes. Several studies have found that the multiple-pass method is a valid measure of intake among adults and produces a valid and unbiased estimate of intake in children when parents report the child’s intake.^22–24 Parents were instructed to report everything they and their child/children ate the previous night after the study meal. Older children independently completed the dietary recall. Participants were then debriefed and the family was paid $15 (See the Supplement materials for additional study procedure information).

Study Measures

Total Dinner Calories Ordered was computed by summing the calories of each menu item selected for the study meal (M = 1869.50, SD = 739.49; range 630 – 4520).

Total Dinner Calories Consumed was estimated by weighing all of the food before and after the meal (M = 1066.50, SD = 484.38; range 205.42 – 2370.42).

Total Calories Consumed Post-Dinner was based on reported intake during the dietary recall interview (M = 68.30, SD = 135.95; range 0.0 – 661.73).

Total Dinner Calories and Post-Dinner Calories Consumed was calculated by adding total dinner calories consumed to total calories consumed post-study meal (M = 1124.18, SD = 444.87; range 379.80 – 2364.62).

Demographic and Eating Habits Questionnaire was a self-report questionnaire constructed for this study that asked parents to provide age, race/ethnicity, and education level. If both parents were present at the study meal, the highest education level achieved by either parent was used to create a variable reflecting family education as an ordinal variable. Study investigators classified children’s race/ethnicity based on their parents’ self-report.

Parents were also asked to provide self-reported height and weight for themselves and their children, which was used to calculate BMI. Children’s calculated BMI was then assigned a BMI percentile based on the Centers for Disease Control age- and sex-specific growth curves, which allows the child’s BMI to be compared to other children of the same age and sex.²⁶

Yale Food Addiction Scale for Children (YFAS-C).

Adapted from the YFAS⁹, the YFAS-C¹⁹ uses a lower reading level and more age appropriate content to assess signs of addictive-like eating in children. The YFAS-C includes two scoring options: a continuous count of the number of symptoms endorsed and a dichotomous “diagnosis”, which reflects the DSM-IV¹⁰ substance dependence diagnostic criteria (i.e., three or more symptoms and clinically significant impairment/distress). The symptom count version of the YFAS-C was used in the current study to capture the full spectrum of addictive-like eating. The YFAS-C appears to have adequate internal consistency, as well as convergent and incremental validity.¹⁹ The YFAS-C in the current study had good internal consistency (Cronbach’s α=.84). Parental guardians were given the opportunity to assist their child in completing the YFAS-C if they felt it was beyond their reading comprehension level.

Data analytic plan

All variables were examined for missing data and normality. The measure of total dinner calories consumed during the meal was positively skewed. One outlier (SD > 3) was identified and removed to ensure normal distribution. The measure of total dinner plus post-dinner calories consumed was also positively skewed, and one outlier (SD > 4) was removed. To determine which covariates to include in a final model, we examined the relationship between the potential covariates (menu type, age, sex, and race/ethnicity) and our primary predictor variable (the YFAS-C symptom count) as well as our primary outcome variable (food intake) using t-tests, ANOVAs, and Pearson’s correlations. We used generalized estimating equations (GEE) to control for the interdependence related to the inclusion of siblings in the sample SPSS 17.0 and ran two models. The first GEE model investigated the relationship between the YFAS-C and food consumption. The second multivariate GEE model examined this relationship controlling for BMI percentile, race/ethnicity, and age. Due to the small sample size of some race/ethnicities, we compared White participants to all non-White participants in the analyses. Although this does not allow for the comparison of effects between different non-White race/ethnicities, we thought it was worthy of exploration because of the higher obesity prevalence among African-American and Hispanic populations relative to White populations.^27,28

Results

YFAS-C

The average YFAS-C symptom count score in the current study was 2.2 (SD = 1.81, range = 0 – 7). Of the 70 participants, 7.2% (n = 5) met criteria for “food addiction.”

Examining covariates

Table 2 summarizes the correlational relationships between the study measures of eating behaviors and food consumption with continuous demographic variables. Participants’ BMI percentile was positively correlated with YFAS-C scores. Age was significantly correlated with total dinner calories ordered, total dinner calories consumed, and total dinner plus-post dinner calories consumed. Race/ethnicity was significantly associated with total dinner calories consumed with non-White participants consuming more calories relative to White participants (t(68) = −2.09, p = .04) Neither menu condition nor sex were associated with the total dinner calories ordered (all p-values > .26), total dinner calories consumed (all p-values > .27), total calories consumed post-dinner (all p-values > .41), total dinner plus post-dinner calories consumed (all p-values > .09), and the YFAS-C (all p-values > .46). Race/ethnicity was also not associated with the YFAS-C, total dinner calories ordered, total post-dinner calories consumed, and total dinner plus post-dinner calories consumed (all p-values > .12); Based on these analyses, BMI percentile, race/ethnicity, and age were included as covariates in the multivariate GEE models.

Table 2.

Correlations among food consumption outcome variables, BMI, age, and food addiction measured by the YFAS-C

	YFAS-C	Total Calories Ordered	Total Dinner Calories Consumed	Total Calories Consumed Post Dinner	Total Dinner Plus Post-Dinner Calories Consumed	Age	BMI Percentile
YFAS-C	-
Total Calories Ordered	.19	-
Total Dinner Calories Consumed	.27*	.67**	-
Total Calories Consumed Post Dinner	.03	−.18	−.21	-
Total Dinner Plus Post-Dinner Calories Consumed	.25*	.65**	.95**	.53**	-
Age	−.01	.29*	.48**	.02	.53*	-
BMI Percentile	.29*	.10	.16	.09	.20	−.04	-

^*p < .05

^**p < .01

Outcomes

Table 3 summarizes the relationship between “food addiction” with the amount of food ordered and food consumption while controlling for covariates (i.e. BMI percentile, race/ethnicity, and age)

Table 3.

GEE models of the amount of food ordered and food consumption

Model/Independent Variables & Covariates	Dependent Variables
	Total Dinner Calories Ordered		Total Dinner Calories Consumed		Total Calories Consumed Post Dinner		Total Dinner Plus Post-Dinner Calories Consumed
	χ2(1)	β	χ2(1)	β	χ2(1)	β	χ2(1)	β
1st Univariate Model
YFAS- C Total Score	1.42	.17	7.65	.30**	.03	.02	5.35	.27*
2nd Multivariate Model
YFAS-C Total Score	2.91	.24	7.36	.26**	.10	.04	8.32	.28**
Age	12.66	.45**	34.72	.72**	.59	.09	38.03	.77**
BMI Percentile	.42	.08	1.76	.15	1.59	.09	2.43	.17
Race/Ethnicity	.63	−.20	6.55	−.55*	1.19	.25	6.07	−.50**

^*p < .05

^**p < .01

Total Dinner Calories Ordered

In the first GEE model, YFAS-C was not found to be significantly associated with total calories ordered. In the multivariate GEE model, age was significantly related to ordering behavior such that older children ordered more food. The YFAS-C was a trend-level, but was not significantly associated with total calories ordered. Neither BMI percentile nor race/ethnicity were significantly related to total calories ordered.

Total Dinner Calories Consumed

The YFAS-C was significantly associated with total dinner calories in the first GEE model, such that those children who consumed more calories at dinner had scores indicating a greater level of “food addiction”. After controlling for covariates, the YFAS-C continued to be significantly related to total dinner calories consumed. Age and race/ethnicity were significantly associated dinner calories eaten such that older children ate more calories as did non-White children (M = 1198.90, SD = 513.85) relative to White children (M = 961.26, SD = 438.08). BMI was not significantly related to dinner calories consumed.

Total Calories Consumed Post-Dinner

The YFAS-C was not significantly associated with total calories consumed post-dinner in the first GEE model or the multivariate model. Age, BMI percentile, and race/ethnicity also were not significantly related to food consumed after dinner.

Total Dinner plus Post-Dinner Calories Consumed

In the first of the two GEE models, the YFAS-C was significantly associated with the total number of calories consumed at dinner plus any calories consumed after dinner. The YFAS-C continued to be significantly related to dinner plus post-dinner calories after controlling for covariates. In the multivariate GEE model, age and race/ethnicity were both significantly associated with dinner plus post dinner calories consumed with older children having greater food intake as did non-White participants (M = 1214.88, SD = 478.14) relative to White (M = 1050.66, SD = 407.55). BMI percentile was not significantly related to dinner plus post dinner calories consumed.

Discussion

The current study is the first to explore the association of “food addiction” and objectively measured food consumption and to investigate this relationship among children. Although children with higher YFAS scores did not order more food, they did consume significantly more calories at a dinner meal, while controlling for BMI percentile, race/ethnicity and age. Prior research observed that elevated YFAS scores in children were associated with greater emotional eating and decreased satiety responsiveness.¹⁹ In adults, higher YFAS “food addiction” scores are associated with decreased activation of a neural region implicated in inhibitory control (i.e., lateral orbitofrontal cortex) during food consumption and greater activation in reward-related regions (e.g., striatum) in response to food cues.¹² Thus, when food is present, children with more addictive-like eating may be prone to consume more calories due to increased reward responsivity to the food, coupled with lower sensitivity to satiety cues and decreased inhibitory control. However, more research is needed to understand the mechanisms underlying the association between addictive-like eating in children and greater caloric intake.

Higher YFAS scores were associated with food consumption, but BMI percentile was not. Elevated BMI can be the result of many heterogeneous contributors that can include factors other than high caloric intake (e.g. genetic conditions²⁹, medication side effects³⁰, physical inactivity³¹, increased muscle mass³²). To more fully understand excess food consumption, it may be important to assess different phenotypes of eating behavior in addition to BMI. For example, Eisenstein and colleagues³³ found that an emotional eating phenotype, but not BMI, was associated with dysfunction in reward-related neural functioning implicated in problematic eating (i.e., central dopamine D2 receptor binding). Similarly, addictive-like eating in children, but not BMI, was related to the quantity of calories consumed in the current study. Thus, the assessment of behavioral phenotypes (e.g., “food addiction,” emotional eating) in children may be important for understanding contributors to eating behavior.

The current study also adds to the limited literature on “food addiction” in children. The foods that are most associated with addictive-like eating are those high in added fats and refined carbohydrates (e.g. ice cream, pizza, chocolate).³⁴ Unlike drugs of abuse, which are more commonly used for the first time in adolescence and early adulthood³⁵, exposure to potentially addictive foods occurs earlier in development.^36,37 For example, Pan and colleagues³⁸ found that 25.9% of infants had been fed a sugar-sweetened beverage in their first year of life. Therefore, even if foods high in added fats and refined carbohydrates are less addictive than drugs of abuse, children are being exposed to them earlier in development when their neural and psychological systems are more vulnerable.^39,40 This earlier age of exposure may increase the likelihood that children at-risk for addictive-like eating will develop problematic patterns of food consumption.

There are several limitations to consider for the current study. First, the study was cross-sectional, which limits our ability to make causal inferences. Longitudinal studies are needed to investigate whether addictive-like eating in children predicts future excessive weight gain and health-related illness. Additionally, our study relied on parent-reported height and weight to calculate participants’ BMI. Error in parent-reported height and weight vary in direction and magnitude across studies with an estimated average discrepancy between mean measured and mean parent-reported height and weight of ± 1 cm and ±1 kg, respectively.⁴¹ The largest error of parent-report height and weight occurs in young, preschool-aged children^42,43 and decreases with older children.^43–45 Studies using objectively measured height and weight to calculate BMI are needed to further investigate the relationship between BMI, “food addiction” and calorie consumption. The clinical significance of these findings may be limited, as study participants were drawn from a community sample. To understand the clinical utility of the YFAS, it will be important to examine the association of “food addiction” with caloric intake in children receiving treatment for eating-related problems (e.g., obesity). In addition, although the sample was relatively diverse, we were underpowered to directly compare different racial/ethnic minority groups. Further evaluation of the relationship between race/ethnicity and additive-like eating will be an important next step. Another study limitation is that parental control over eating behavior was not assessed. Parental behavior may impact the likelihood that children will exhibit addictive-like eating and the amount of food they consume. Understanding the impact of parenting behavior on addictive-like eating behaviors in children will be an important future direction.

Conclusion

The current study is the first to examine whether elevated YFAS “food addiction” scores are related to objectively measured eating behavior; moreover, it is the first study to do so in children. We found that addictive-like eating in children was more strongly associated with increased caloric intake than BMI percentile, highlighting the importance of assessing behavioral phenotypes of eating behavior. If an addictive process is contributing to higher caloric intake for some children, it will be important to develop prevention and intervention approaches that target these mechanisms.

Figure 1.

Relationship between YFAS-C symptoms and Total Dinner Calories

Figure 2.

Relationship between YFAS-C symptoms and Total Dinner Plus Post-Dinner Calories