Food commercials do not affect energy intake in a laboratory meal but do alter brain responses to visual food cues in children

Abstract

Food commercials promote snack intake and alter food decision-making, yet the influence of exposure to food commercials on subsequent neural processing of food cues and intake at a meal is unclear. This study tested whether exposing children to food or toy commercials altered subsequent brain response to high- and low-energy dense food cues and influenced laboratory intake at a multi-item, ad libitum meal. Forty-one 7–9-year-old children (25 healthy weight; 16 with overweight/obesity) completed five visits as part of a within-subjects design where they consumed multi-item test-meals under three conditions: no exposure, food commercial exposure, and toy commercial exposure. On the fourth and fifth visits, functional magnetic resonance imaging (fMRI) was performed while children viewed low- and high-energy dense food images following exposure to either food or toy commercials. Linear mixed models tested for differences in meal energy intake by commercial condition. A whole-brain analysis was conducted to compare differences in response by commercial condition and child weight status. Meal intake did not differ by commercial condition (p = 0.40). Relative to toy commercials, food commercials reduced brain response to high-energy food stimuli in cognitive control regions, including bilateral superior temporal gyri, middle temporal gyrus, and inferior frontal gyrus. Commercial condition * weight status interactions were observed in orbitofrontal cortex, fusiform gyrus, and supramarginal gyrus. Children with overweight/obesity showed increased response in these regions to high-energy stimuli following food commercials. Food commercial exposure affected children’s subsequent processing of food cues by reducing engagement of the prefrontal cortex, a region implicated in cognitive control. Even though food commercial exposure did not increase intake at a meal, the effect of reduced prefrontal cortical engagement on a broader range of consumption patterns warrants investigation.

1. Background

The high prevalence of childhood obesity both in the United States (Van Grouw & Volpe, 2013) and globally (Karnik & Kanekar, 2012) continues to be a significant problem. Currently, 33% of children are overweight, and over 13% are obese (Ogden, Carroll, Kit, & Flegal, 2012). Childhood obesity is a public health concern as children with obesity often present with co-morbidities that negatively impact overall quality of life (Schwimmer, Burwinkle, & Varni, 2003) and increase total healthcare costs (Wang, McPherson, Marsh, Gortmaker, & Brown, 2011). A significant contributor to the rise in childhood obesity is increased intake of high-fat/high-sugar foods, drinks, and snacks which are aggressively marketed to children (Andreyeva, Kelly, & Harris, 2011; Cullen, Baranowski, Rittenberry, & Olvera, 2000; Halford, Boyland, Hughes, Oliveira, & Dovey, 2007; Story & French, 2004). Food marketing has explicitly been implicated as a contributor to the rise in childhood obesity through a variety of mechanisms (Harris, Bargh, & Brownell, 2009), but primarily as a promoter of unhealthy food intake (Brownell & Horgen, 2004). While many environmental factors influence children’s eating behaviors, food advertising has been proposed as one of the most powerful (Story & French, 2004).

Food commercials, in particular, are associated with an increase in children’s preferences (Borzekowski & Robinson, 2001; Goldberg, Gorn, & Gibson, 1978), requests for (Buijzen & Valkenburg, 2002; McDermott, O’Sullivan, Stead, & Hastings, 2006), and consumption of (Halford et al., 2004, 2007) the advertised products. On average, children see more than 15 food commercials per day (Commission FT, 2007). A recent evaluation of television food commercial content reported that 98% of the advertised foods are high in sugar, fat, and sodium (Powell, Schermbeck, & Chaloupka, 2013). Furthermore, food commercials often contain powerful food consumption cues including images of peers eating and snacking, associations between positive emotions and energy dense snacks, and salient images of energy dense foods that lack critical micronutrients (Folta, Goldberg, Economos, Bell, & Meltzer, 2006). It has been suggested that such messaging may increase snacking behaviors (Harris et al., 2009). Boyland and colleagues recently analyzed data from 18 studies to determine the impact of food commercials on children’s consumption. They found a statistically and clinically meaningful association between food commercial exposure with increased food intake in children (Boyland et al., 2016). Several studies also suggest that children with obesity may be particularly vulnerable to these effects (Halford et al., 2004, 2007; Rapuano, Huckins, Sargent, Heatherton, & Kelley, 2015). However, most current studies evaluating the effects of food commercials on intake have been primarily focused on measuring consumption of snack foods while children watch television (Boyland et al., 2016). It is currently unclear if these effects carry over into homeostatic eating events, such as meals.

Neurobiological differences in brain regions implicated in cognitive control and reward processing may underlie individual differences in response to food commercials (Bruce et al., 2011, 2013a; Rapuano et al., 2015). Two brain imaging studies have demonstrated that food commercials are associated with engagement of limbic neurocircuitry implicated in food motivation and reward (Gearhardt, Yokum, Stice, Harris, & Brownell, 2014; Yokum, Gearhardt, Harris, Brownell, & Stice, 2014). Additionally, Bruce and colleagues found that following food commercial exposure there was both increased brain activation in the ventral medial prefrontal cortex and an increased speed in food and beverage decision-making. They suggested that this shift in selection may be because food commercials bias children to place more emphasis on hedonic as opposed to health attributes of food (Bruce et al., 2016). While these studies have evaluated the effects of commercials on the brain, none to date have measured how exposing children to commercials may impact subsequent brain response to high- and low-energy food cues and related these responses to objectively assessed eating behaviors. This study design is ecologically relevant because often children are exposed to food advertising content when they watch television before a meal.

This study addressed three primary aims, 1) to compare the impact of exposing children to food versus toy commercials on subsequent ad libitum intake of high- and low-energy dense foods at a laboratory meal;2) to compare the impact of exposing children to food versus toy commercials on subsequent brain response to high- and low-energy dense food stimuli; 3) to determine whether children’s brain responses to high- and low-energy dense food images following food commercial exposure are associated with individual differences in energy intake following food commercial exposure. We hypothesized that children would consume more high-energy dense food and less low-energy-dense food after exposure to food vs. toy commercials. We also hypothesized that exposure to food commercials would result in decreased food cue reactivity in areas related to cognitive control (e.g., dorsolateral prefrontal cortex (dlPFC), etc.) and increased response in areas related to reward (e.g., amygdala, ventral striatum, etc.) to high-relative to low-energy dense food stimuli. Additionally, we hypothesized that differences in children’s brain response after exposure to food vs. toy commercials would be related to differences in their laboratory meal intake after exposure to food vs. toy commercials. Finally, because previous studies have shown that children with overweight/obesity may be more susceptible to food advertising (Bruce et al., 2013b; Gearhardt et al., 2014; Halford et al., 2004, 2007; Stice, Spoor, Bohon, Veldhuizen, & Small, 2008), we also examined whether child weight status moderated the relationship between advertising exposure and brain and behavioral responses.

2. Methods

2.1. Participants

Forty-four (N = 44) children (53.70% male) between the ages of 7–9 years (7.90 ± 0.70) were recruited from Central Pennsylvania for testing. Children were recruited primarily through flyers distributed throughout the community (e.g., schools, stores, and afterschool care facilities) and at local family-friendly events. Three children did not finish the study and did not have complete data for either the meals or the MRI and were therefore excluded from the final analysis, resulting in a sample of 41 (25 healthy weight, 10 with overweight, and 6 with obesity). Parents of the children were primarily college-educated with relatively high household incomes ($76,000 – $100,000/year). Detailed anthropometric characteristics of the 41 participants are summarized in Table 1. All children were screened via phone calls with parents to ensure they were in overall good health, had no metabolic diseases, eating disorders, or learning disabilities (e.g., ADHD). Children were also required to watch at least 1 h of television per week, according to parental report, to exclude those with minimal exposure to television commercials. Children were also excluded if they were left-handed or had other characteristics that would impact fMRI performance (e.g., braces, claustrophobia, etc.). This study was approved by the Institutional Review Board of The Pennsylvania State University. All parents provided written informed consent and children provided written assent before participating.

Table 1.

Participant characteristics (n = 41).

	Healthy Weight (n = 25)	Overweight/Obese (n = 16)
Sex	12 female (48%); 13 male (52%)	10 female (62.5%); 6 male (37.5%)
Age	7.84 ± 0.68	8.00 ± 0.73
Height (cm)	129.48 ± 7.29	135.35 ± 7.93
Weight (kg)	26.92 ± 3.64	37.77 ± 8.35
BMI (kg/m2)	15.99 ± 0.98	20.38 ± 2.38
BMI-Age-Sex-Percentile	0.48 ± 0.18	0.91 ± 0.05
Body Fat (%)	15.99 ± 4.98	27.41 ± 7.36

Values represented are mean ± SD.

Healthy Weight = < 85% BMI-Age-Sex Percentile.

Overweight/Obese = ≥85% BMI-Age-Sex Percentile.

BMI = Body Mass Index.

2.2. Study design and overview

Using a within-subjects, cross-over design with repeated measures, participants completed five visits, each scheduled one week apart to determine the impact of viewing food versus toy commercials on food intake and brain response to food cues. The first three visits were completed at the Metabolic Kitchen and Children’s Eating Behavior Laboratory. In these visits, children were presented with test-meals under three conditions: baseline and two commercial conditions (food or toy commercials). During the final two visits, participants completed two fMRI scans at the Social Life and Engineering Imaging Center (see Table 2). In these visits, children viewed pictures of high- and low-energy dense foods in the MRI scanner following two commercial conditions (food or toy commercials).

Table 2.

Study visits timeline.

Visit 1	Visit 2	Visit 3	Visit 4	Visit 5
	Food or Toy Commercial Exposure		Food or Toy Commercial Exposure
Anthropometrics	Ad Libitum	Ad Libitum	fMRI	fMRI
Questionnaires	Laboratory	Laboratory	scan	scan
Ad Libitum	Meal	Meal
Laboratory Meal	Mock MRI	Mock MRI

2.3. Overview of procedures

Participants who qualified for the study selected a preferred day of the week and then either a lunch (11:30 a.m. to 2:00 p.m.) or dinner time (4:30 p.m. to 7:30 p.m.) to attend for five consecutive weeks. Children were instructed not to eat 3 h before all appointments. Compliance was verified by parental report at the start of each visit. Before the start of the study, all children were assigned a randomized order for viewing the food and toy commercials before visits two through five, and this order was counter-balanced across participants.

On the first visit, parents and children were given an overview of the study and completed informed consent/assent to participate. Parents and children also completed questionnaires (described in detail below) administered through Qualtrics (Qualtrics, Provo, UT). Anthropometric measurements (i.e., height, weight, and body fat percentage) were collected on the children. Before the meal, children rated their liking of all test foods using 5-point pictorial Likert scales anchored from “Hate it” to “Love it.” Children then consumed an ad libitum test-meal to serve as a baseline control condition for food intake in the laboratory setting (Birch, McPhee, Bryant, & Johnson, 1993). At the end of the visit, children completed a mock MRI scan (described below).

On visits two and three, children viewed a cartoon embedded with commercials from their assigned condition (i.e., food or toy) followed by an ad libitum test-meal. On visits four and five, children again viewed a cartoon embedded with commercials from their assigned condition(i.e., food or toy) immediately followed by a 30-min fMRI scan where they passively viewed pictures of low- and high-energy dense foods.

2.4. Questionnaires

Parents completed a standard demographic assessment to determine socioeconomic status, ethnicity, and other variables. Parents reported child pubertal development using the Puberty Development scale (Cole, Bellizzi, Flegal, & Dietz, 2000). Parents also completed the Children’s Eating Behavior Questionnaire (CEBQ), a validated 35- item assessment of child appetitive traits (Carnell & Wardle, 2007; Wardle, Guthrie, Sanderson, & Rapoport, 2001). Furthermore, parents reported child advertisement exposure using The Parent Brand Inventory (available on request), a questionnaire developed by our laboratory to measure parents’ perception of their child’s familiarity with food brands. This questionnaire asked parents to indicate whether or not their child would be familiar with particular food and non-food brands. Information from these questionnaires was used to assess for potential covariates that might influence children’s food intake and/or brain response to food cues, but will not be presented in detail in the current paper.

Children completed the Children’s TV survey, a questionnaire developed by our laboratory (available on request) to assess children’s exposure to television. In brief, this survey asks children to self-report the amount of time spent watching television and using the internet. Furthermore, it asks children to indicate how familiar they are with television channels and children’s programming on a scale anchored from “never heard of it” to “watch it a lot.” Children also reported fullness before and after all test-meals and before the fMRI scan using an age-appropriate, validated analog scale (Keller et al., 2006).

2.5. Anthropometric measurements

Children were assessed for height, weight, and body fat percentage in light clothing (e.g., no jackets, socks, or shoes). 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 stadiometer (Seca®, Chino, CA). Body fat was measured to the nearest tenth percent using a standing bioelectrical impedance body composition analyzer (Tanita®, Arlington Heights, IL). Child BMI (BMI kg/m²) was used to categorize children by weight status. Children were included in the overweight and obese group if their BMI for age and sex was ≥85th percentile (Cole et al., 2000).

2.6. Commercial exposure

At the beginning of visits 2–5, participants watched a 12-min cartoon embedded with ten commercials presented in two commercial breaks lasting 90 s each for a total run time of 15-min; this is similar to established protocols used in other studies (Boyland, Kavanagh-Safran, & Halford, 2015; Halford et al., 2004, 2007). Commercials were selected using Nielsen data (Nielsen Ad Intel, 2015) for the most advertised brands seen by children. A list of commercials used is available upon request. Four different episodes of a popular cartoon show (Phineas and Ferb ^©, Disney Channel) were used to ensure children did not tire of watching the same episode over the four visits. The episodes were screened to not contain any overt conversations or actions related to food.

2.7. Ad libitum test-meals

In order to test exploratory hypotheses related to the impact of food commercials on meal energy density, children were presented with three high-energy dense foods: Four Cheese Pizza (DiGiorno®, NestleS.A., Vevey, Switzerland), Chips Deluxe Chocolate Chip Cookies (Keebler®, Battle Creek, MI), French Fries (Ore Ida®, Kraft Heinz, Philadelphia, PA) and three low-energy dense foods:Grilled Chicken (Tyson®, Springdale, AR), Frozen Broccoli Florets (Wegmans®, Rochester, NY), and Red Grapes. Serving sizes were based on prior studies (Fearnbach et al., 2015, 2016) and were intended to provide children with enough food so as not to limit intake of any item. Details of serving sizes and energy densities of the foods can be found in Table 3. All test-meal foods were weighed to the nearest hundredth of a gram before serving using a digital food scale. Food items were arranged identically on two trays for all meals.

Table 3.

Test-meal foods.

Meal Item	Serving Size	Weight (g)	Energy (kcal)	Energy Density (kcal/g)
High-Energy Options
Cheese Pizza	½ pizza	130	360	2.8
Chocolate Chip	3 Cookies	33	160	4.8
Cookies
French Fries	27 pieces	84	190	2.3
Low-Energy Options
Grilled Chicken	3 oz	84	100	1.2
Red Grapes	1 cup	127	90	0.8
Frozen Broccoli	1 cup	93	20	0.2
Florets
Ketchup	1.2 oz	35	30	0.9

Children were informed that they had 30-min to eat as much or as little as they wanted and that they could end the meal early if they were finished. A research assistant read an age-appropriate story (e.g., Matilda) that did not contain food references while the children ate to serve as a consistent, neutral distraction that also provided a reason for the researcher to be in the room to monitor the child. Children were also allowed to request additional servings of any item throughout the meal. If the child finished any food item, they were asked if they would like seconds, and if requested, the researcher brought out another serving of the same size. We have used similar protocols in past studies of ad libitum test-meal assessment because we have found that some children may be too shy to request seconds even if they would like them (Lee & Keller, 2012). Immediately following the meal, leftovers were weighed to the nearest hundredth of a gram.

2.8. Mock MRI procedure

Following each of the three test-meals, children completed a training session conducted at the MRI imaging facility mock scanner. The mock-training sessions were based on a previous protocol developed to facilitate the quality of fMRI data collection in children (English, Lasschuijt, & Keller, 2015). The mock scanner allows children to experience the MRI environment and practice the viewing task before testing. A trained research assistant explained the various components of the MRI machine such as the table, head coil, and mirror. Because excessive motion in the scanner can cause motion artifacts and loss of data (Byars et al., 2002), children were given examples of movements to avoid, and research staff coached them on how to remain still. Children then practiced the task for 5-min with images not presented during the fMRI scan. Recorded sounds similar to those heard during MRI were also played over speakers to prepare children for the upcoming scans.

2.9. Functional MRI paradigm

Standardized food images used in the fMRI task were provided by the Image Sciences Institute, UMC Utrecht, and created as part of the Full4Health project (www.full4health.eu), funded by the European Union Seventh Framework Program (FP7/2007–2013) under grant agreement nr. 266408, and the I.Family project (http://www.ifamilystudy.eu. ifamilystudy.eu), grant agreement nr. 266044 (Charbonnier, van Meer, van der Laan, Viergever, & Smeets, 2016). Pictures of food shown during the fMRI scan included: low-energy dense foods (e.g., vegetables and fruit) (n = 120); high-energy dense foods (e.g., candy and ice-cream) (n = 120); and blurred image controls (n = 240). Blurred pictures were used as an implicit baseline to control for visual stimulation. Images were presented to subjects using E-Prime (Psychology Software Tools Inc., Sharpsburg, PA) by projecting images onto a screen at the head of the MRI scanner for viewing using a mirror mounted to the head coil. Pictures were each presented for 2.5 s with a 0.5 s (blank screen) inter-stimulus interval.

The MRI task procedure was based off a previously used paradigm (Masterson, Kirwan, Davidson, & LeCheminant, 2016). Each MRI visit consisted of 2 functional runs each 6-min in length (120 pictures in each run) with a 1-min rest period between blocks to provide children time to ask any questions or express any concerns about the procedure. A total of 48 blocks of 10 images each were presented to each participant. To ensure presentation of the stimuli was random while still being balanced for food types, the 48 blocks of images were randomized into two master blocks. Master blocks consisted of 6 sub-blocks of high-energy foods and six low-energy sub-blocks. A master block of stimuli (6 low- and six high-energy blocks) was randomly assigned to be shown on the first visit with the other block being used in the subsequent visit. Blocks of stimuli were always followed by a sub-block of identical corresponding blurred images. The presentation order of the sub-blocks was randomized. Furthermore, the presentation order of the images within a sub-block was randomized. There was no inter-block rest period and no jittering of images.

To improve engagement in the task, children were asked to respond to the question, “Would you eat this in the morning or the evening?” while viewing each picture. Children indicated their response by pressing one of two buttons on a fiber optic response pad with his/her right hand. For blurred stimuli, children were asked to press a button each time the picture changed.

2.10. MRI data acquisition

MRI data were acquired using a Siemens MAGNETOM Trio 3T MRI scanner (Siemens Medical Solutions, Erlangen, Germany) with a standard 12-channel head coil. The use of padding around the participants head, arms, and body was used to reduce excessive movement during the scan. Scans were identical for both sessions. The following parameters were used to obtain a T1-weighted MPRAGE structural scan for each subject: TE = 2.26 ms; T = 1900 ms; flip angle = 9°; matrix size = 256 × 215 mm; field of view = 250 × 218 mm; 176 slices; slice thickness = 1 mm; voxel size = 0.977 × 0.977 × 1 mm; 1 total acquisition. T2*-weighted echo-planar images were obtained using the following parameters: TE = 28 ms; TR = 1800 ms; flip angle = 90°; matrix size = 64 × 64; field of view = 220 × 220 mm; 36 slices; slice thickness = 3 mm; voxel size = 3.4 × 3.4 × 3 mm; 270 total acquisitions.

2.11. Data analysis

Participants’ descriptive data, food intake, fullness scores, and BOLD responses extracted from fMRI were analyzed in SPSS 22.0 (IBM Corp., Chicago, IL, USA). Energy intake from test-meals was calculated by converting gram amounts of foods consumed to energy (kcal). Linear mixed models were used to test for differences in pre-meal fullness and energy intake between the three meal conditions. Models were tested to assess if sex, BMI-z scores, parental education, time of the meals, social economic status, or pre-meal fullness impacted the relationship between commercial condition and test-meal intake. A paired-samples t-test was used to test for differences in pre-scan fullness for the two MRI days.

Statistical analyses for whole-brain response from the fMRI data were conducted using BrainVoyager 20.6 (Brain Innovation, Maastricht, The Netherlands). Functional data were 3D motion corrected using six vectors (3 translations, 3 rotations), temporal high-pass filtered using a GLM-Fourier basis set with three cycles per time course, and 3D spatial smoothed using a 6-mm³ FWHM Gaussian filter. Anatomical data were normalized to MNI space using the MNI-152 template. Functional data were co-registered to each participant’s anatomical data.

Functional runs were analyzed using BrainVoyager’s Motion Correction Processor plugin. Regressors of no interest were created for any volume with over 1 mm of translation or 1° of rotation in any direction. Runs were excluded if more than 25% of the data was censored using this process. Only 15% of runs contained motion over the desired threshold with only 1% being unusable. Therefore 2 of 164 total runs were excluded from the final analysis.

Functional data were analyzed using a random-effects general linear model (RFX-GLM). Regressors were created coding for low- and high-energy food blocks. Control blocks were not explicitly modeled and served as the implicit baseline. Blocks were modeled by convolving the standard hemodynamic response function with a 30-s boxcar function.

An exploratory whole-brain analysis was first conducted to assess the effects of food commercial exposure on brain response to high- and low-energy food images. Prior to the group-level analysis contrast maps for energy density of food images (high > low) were created on the single subject level for both the food and toy commercial conditions. These contrast maps were then subjected to further analysis using ANOVA with commercial condition as a within-subject factor, weight status as a between-subject factor, and participants as a random factor. To correct for multiple testing, results were initially run at a voxel-wise threshold of P < 0.001 (Woo, Krishnan, & Wager, 2014), and a spatial extent threshold of five, determined by using the ClusterThresh plugin of the BrainVoyager software (Forman et al., 1995; Goebel, Esposito, & Formisano, 2006). However, for the main effect of commercial condition, only two regions survived at a threshold of P < 0.001 (i.e., left parahippocampal gyrus and left inferior frontal gyrus). Because this was a preliminary study, we also included regions that survived at a threshold of P < 0.005 to inform the direction of future investigations. As recommended by Woo and colleagues, 10,000 Monte Carlo simulations were conducted to provide estimations for an overall P-value of P < 0.05 (Woo et al., 2014). Contrast maps for the main effects (condition and weight status) and the interaction between factors were then evaluated. Average beta coefficients were extracted from all cluster-based regions of interest that survived multiple comparison corrections and values were subjected to further analysis in SPSS. A multivariate ANOVA was conducted with all extracted beta weights from the significant clusters. Beta weights were entered in as dependent variables and all other variables were entered in as fixed factors (i.e., commercial condition and weight status) or covariates (i.e., socioeconomic status, condition order, parental education, daily television exposure, child age and sex, and commercial exposure). None of these factors influenced the significance of the models.

To test the third hypothesis regarding the relationship between brain response to commercial condition and eating behavior, we conducted a targeted region of interest (ROI) analysis according to suggestions from Berkman & Falk (Berkman & Falk, 2013). Independently from the whole-brain fMRI group-based analyses which identified regions where food cue processing differed as a function of prior exposure to food or toy commercials, we identified separate regions implicated in reward processing and inhibitory control that were expected to be related to laboratory eating behavior. Ten brain regions that are commonly implicated in the “appetitive network” (Dagher, 2012) were selected from 3 recent meta-analyses and included regions implicated in reward processing (i.e. insula) and inhibitory control [i.e., dorsolateral prefrontal cortex (dlPFC)] (Brooks, Cedernaes, & Schiöth, 2013; van der Laan, De Ridder, Viergever, & Smeets, 2011; van Meer, van der Laan, Adan, Viergever, & Smeets, 2015). A list of regions and coordinates used are listed in Table 5. Five-millimeter spheres were drawn in Brain Voyager around the bilateral peak coordinates reported for each region. Average beta weights were extracted for the contrast between high- and low-energy food cue response after both the food and toy commercial exposure conditions. For example, higher values would be indicative of greater BOLD response in a region for high-relative to low-energy dense food cues. For each ROI, linear mixed models were conducted to determine whether brain response to high relative to low-energy dense food cues in response to commercial exposure was related to children’s laboratory intake following commercial exposure. In these models, intake at both conditions (food vs. toy) was the dependent variable, brain response at both conditions (food vs. toy) was a covariate, and commercial condition (food vs. toy) was included as a fixed factor. For all relationships tested, a P-value ≤ 0.05 was used to determine significance.

Table 5.

Linear mixed model results for targeted ROI analysis with energy intake as the dependent variable.

Region of Interest	L/R	X	y	z	References
vlPFC	R	40	36	12	van Meer et al., 2015
	L	−40	36	12
Ventral Striatum	R	6	0	−12	van der Laan et al., 2011
	L	−6	0	−12
Dorsal Striatum	R	8	10	−8	van Meer et al., 2015
	L	−8	10	−8
Insula	R	40	6	−12	van Meer et al., 2015
	L	−38	−10	4
dlPFC	R	30	28	39	Brooks et al., 2013
	L	−30	28	39
Amygdala	R	24	−4	−16	van Meer et al., 2015
	L	−22	−2	−20
Anterior cingulate cortex	R	0	48	0	van Meer et al., 2015
	L	−2	34	22
Orbitofrontal cortex	R	32	34	−12	van Meer et al., 2015
	L	−32	34	−12
Fusiform	R	39	−62	−14	van Meer et al., 2015
	L	−40	−58	−14
IFG	R	42	38	10	van der Laan et al., 2011
	L	−42	38	10

R = Right; L = Left.

3. Results

Participant characteristics are summarized in Table 1. The majority of participants were Caucasian (n = 34); the others were Black (n = 4), Asian (n = 2), and Hispanic (n = 1). The sample included 22 males(53.7%) and 19 females (46.3%). Participants reported no difference in pre-meal fullness between meal sessions (F(1,39) = 0.62; P = 0.54). Similarly, there was no difference in pre-scan fullness by commercial condition (t = −0.461; P = 0.65). There was no difference in energy intake between conditions (baseline, food commercials, toy commercials conditions) (F(1,39) = 0.927; P = 0.40). Additionally, energy intake was highly correlated across meals (ranging from α = 0.81–0.90), demonstrating that children ate similarly from one meal to the next. There was a main effect of weight status (normal vs. overweight/obese) (F(1,39) = 18.905; P ≤ 0.001) indicating that those with overweight/obesity consumed 212.77 (± 48.94) kcal more on average than their healthy weight counterparts, regardless of commercial condition. When separating out high- and low-energy dense foods in the analysis, there were no differences in high-energy food intake by commercial condition (F(1,39) = 0.407; P = 0.66). However, children with overweight/obesity consumed more [231.87 (± 47.33)] high-energy dense foods than those who were healthy weight (F(1,39) = 24.001; P ≤ 0.0001). There was no effect of commercial condition (P = 0.14) or weight status (P = 0.21) for intake of low-energy dense foods.

3.1. Exploratory whole-brain analysis results

Brain Regions for the Main Effect of Commercial Condition:

There was a main effect of commercial condition in right superior temporal gyrus (F(1,39) = 12.62; P = 0.001), two regions of the left superior temporal gyrus (F(1,39) = 15.47; P = 0.0003 and F(1,39) = 11.29; P = 0.001), bilateral superior frontal gyrus (right - F(1,39) = 19.18; P = 0.00008); left - F(1,39) = 10.25; P = 0.002), left parahippocampal gyrus (F (1,39) = 13.99; P = 0.0005), two regions of left inferior frontal gyrus (F(1,39) = 20.97; P = 0.00004; F(1,39) = 13.28; P = 0.0007), and two regions of left middle temporal gyrus (F(1,39) = 11.85; P = 0.001; F (1,39) = 11.34; P = 0.001). In all regions exposure to food commercials before viewing food images resulted in a decreased response to high-relative to low-energy density stimuli (Table 4; Fig. 1).

Table 4.

Significant clusters of activation for the contrasts of the main effect of commercial exposure, the main effect of the stimulus, and the commercial exposure by stimulus interaction.

ANOVA Contrast	Brain Region	No. Voxels	X	Y	Z	F-value	P-value
ME Commercial Condition	R. superior temporal gyrus	16	36	−49	18	12.62	0.001
	L. superior temporal gyrus	14	−66	−35	15	15.47	0.0003
	L. superior temporal gyrus	24	−34	−58	26	11.29	0.001
	R. superior frontal gyrus	18	26	55	36	19.18	0.00008
	L. superior frontal gyrus	9	−16	52	42	10.25	0.002
	L. inferior frontal gyrus	6	−31	36	−3	13.28	0.0007
	L. parahippocampal gyrus	50	−45	−41	−2	13.99	0.0005
	L. inferior frontal gyrus	60	−56	31	1	20.97	0.00004
	L. middle temporal gyrus	12	−55	−49	1	11.85	0.001
	L. middle temporal gyrus	12	−56	−56	21	11.34	0.001
ME Weight Status	R. supramarginal gyrus	7	62	−48	33	12.10	0.001
	R. transverse temporal gyrus	10	57	−15	10	13.26	0.0007
	R. fusiform gyrus	16	48	−57	−24	15.88	0.0002
	R. fusiform gyrus	18	27	−63	−18	13.55	0.0007
	R. middle temporal gyrus	14	46	−77	16	12.81	0.0009
	R. cuneus	8	21	−98	9	16.00	0.0002
Commercial Condition X Weight Status	L. orbitofrontal cortex	12	−14	30	−21	11.75	0.001
	L. orbitofrontal cortex	14	−22	21	−25	15.24	0.0003
	L. fusiform gyrus	8	−34	−59	−17	12.30	0.001
	L. supramarginal gyrus	19	−57	−38	25	13.61	0.0006

ME = Main Effect; ME Commercials = food commercials vs. toy commercials.

R = Right; L = Left.

Fig. 1.

Results for the main effect of commercial condition from the whole brain fMRI analysis (n = 44) corrected for multiple comparisons. Contrasts maps for high-energy > low-energy food images were created at the individual subject level. These maps were then subjected to group level analysis using 2 × 2 ANOVA with condition (food commercial or toy commercial) and weight stats (healthy weight or overweight/obese) as fixed factors and participants as a random factor. Results were thresholded with a voxel-wise P-value of P < 0.005 and spatial extent threshold of six (3 × 3×3mm) voxels. Regions that remained significant after correction include: a) right superior temporal gyrus, b) right superior frontal gyrus, c) left middle temporal gyrus, d) left superior temporal gyrus, e) left superior temporal gyrus,f) left superior frontal gyrus, g) left parahippocampal gyrus, h) left inferior frontal gyrus, i) left middle temporal gyrus, j) left inferior frontal gyrus.

Brain Regions for the Interaction between Commercial Condition and Weight Status:

Interactions between commercial condition and child weight status were observed in two regions of left orbitofrontal cortex (F(1,39) = 11.75; P = 0.001; F(1,39) = 15.24; P = 0.0003), left fusiform gyrus (F(1,39) = 12.30; P = 0.001) and left supramarginal gyrus (F(1,39) = 13.61; P = 0.0006) (Table 4; Fig. 2). Post-hoc analyses revealed significant differences in response between children with healthy weight and those with overweight/obesity in left fusiform gyrus, left supramarginal gyrus, and one region of left orbitofrontal cortex. In these regions, following exposure to toy commercials, children with healthy weight showed greater response to higher energy dense food cues when compared to children with overweight/obesity (P = 0.039, P = 0.04, P = 0.017 respectively). However, following food commercials the opposite pattern was observed, as children with healthy weight showed reduced response to higher energy-dense food cues compared to children with overweight/obesity (P = 0.014, P = 0.008, P = 0.025, respectively). In one region of left orbitofrontal cortex there was no difference in brain response between children with healthy weight and those with overweight/obesity following exposure to toy commercials (P = 0.52). However, following food commercial exposure, those with overweight/obesity again showed a significantly greater response to higher energy dense foods relative to children of healthy weight (P = 0.003).

Fig. 2.

Results for the interaction between commercial condition and weight status from the whole brain fMRI analysis (n = 44) corrected for multiple comparisons. Contrasts maps for high-energy > low-energy food images were created at the individual subject level. These maps were then subjected to group level analysis using 2 × 2 ANOVA with condition (food commercial or toy commercial) and weight stats (healthy weight or overweight/obese) as fixed factors and participants as a random factor. Results were thresholded with a voxel-wise P-value of P < 0.005 and spatial extent threshold of six (3 × 3×3mm) voxels. Regions that remained significant after correction include: a) left fusiform gyrus, b) left supramarginal gyrus, c) left orbital frontal cortex, d) left orbital frontal cortex.

Brain Regions for the Main Effect of Weight Status:

There was a main effect of weight status in right supramarginal gyrus (F(1,39) = 12.10; P = 0.001), right transverse temporal gyrus (F(1,39) = 13.26; P = 0.0007), right middle temporal gyrus (F(1,39) = 12.81; P = 0.0009), right cuneus (F(1,39) = 16.00; P = 0.0002), and two regions of right fusiform gyrus (F(1,39) = 15.88; P = 0.0002 and F (1,39) = 13.55; P = 0.0007). All regions showed that compared to healthy weight children, those with overweight/obesity had decreased BOLD response to higher energy dense food cues, regardless of commercial condition (Table 4; Fig. 3).

Fig. 3.

Results for the main effect of weight status from the whole brain fMRI analysis (n = 44) corrected for multiple comparisons. Contrasts maps for high-energy > low-energy food images were created at the individual subject level. These maps were then subjected to group level analysis using 2 × 2 ANOVA with condition (food commercial or toy commercial) and weight stats (healthy weight or overweight/obese) as fixed factors and participants as a random factor. Results were thresholded with a voxel-wise P-value of P < 0.005 and spatial extent threshold of six (3 × 3×3mm) voxels. Regions that remained significant after correction include: a) right cuneus, b) right fusiform gyrus, c) right fusiform gyrus, d) right supramarginal gyrus, e) right transverese temporal gyrus.

3.2. Targeted ROI and test-meal intake results

None of the additional ROIs tested showed a main effect of commercial exposure condition. Three regions including left orbitofrontal cortex (F(1,78) = 4.26; p = 0.04), right (F(1,78) = 4.18; p = 0.04), and left (F(1,78) = 5.34; p = 0.02) dorsal striatum showed significant main effects reflective of associations between brain response and intake. All three regions showed that increased response to higher energy dense food images was associated with decreased energy intake in the laboratory meals regardless of commercial condition. Statistical outliers were noted in the bilateral striatum, and removal of the outliers improved the relationship in both cases (F(1,78) = 6.81; p = 0.01; F (1,78) = 4.69; p = 0.03 respectively). One region, right insula, (F (1,78) = 4.02; p = 0.04) showed a significant interaction between commercial condition and brain response. Increased response to high-energy food images in the toy commercial condition was associated with decreased energy intake whereas increased response to high-energy food images following food commercials was associated with increased energy intake. When controlling for child BMI-z score in the above relationships, only results in left orbitofrontal cortex and left dorsal striatum remained significant. When controlling for pre-meal hunger levels no results remained significant.

4. Discussion

This study makes two significant contributions to the literature. First, while previous studies have shown that food commercials have a small to moderate effect on snack food intake (Boyland et al., 2016), this is the first study to test the impact of food commercials on subsequent intake at a meal in children. Our results suggest that exposing children to food commercials before a laboratory meal does not increase short-term intake (compared to viewing toy commercials). Furthermore, the effect of food commercials on intake did not differ as a function of child weight status, which contradicts prior studies that have found children with obesity to be more responsive to the intake-promoting effects of food advertising compared to healthy weight counterparts (Anderson et al., 2014; Halford et al., 2008). Second, although previous research has observed the effects of food commercials on the brain (Bruce et al., 2016; Gearhardt et al., 2014; Yokum et al., 2014), we are the first to show that food commercials impact processing of subsequent visual food cues in regions related to sensory processing, reward, control, and emotion. These results demonstrate that exposure to food commercials alters the subsequent processing of visual food cues, and this may have implications for children’s eating behaviors beyond the laboratory.

Regarding the primary study aim, we observed no effect of food commercial exposure on children’s intake at a multi-item test-meal. There are several possible explanations for this outcome. First, we note that not all previous studies have found an influence of food commercials on consumption (Anschutz et al., 2009, 2010; Boyland et al., 2013; Harris, Speers, Schwartz, & Brownell, 2012). The lack of results in some studies suggests that observed effects may vary depending on differences in participant characteristics, commercial exposure paradigms, and food items tested. For example, Dovey et al. found that following food commercial exposure, children with high levels of food neophobia did not increase consumption of snack foods to the same extent as those with lower levels of neophobia (Dovey, Taylor, Stow, Boyland, & Halford, 2011). Second, the effect sizes reported in a recent meta-analysis conducted by Boyland and colleagues (Boyland et al., 2016) suggest that commercials have only a small-to-moderate influence on the consumption of snack foods in children. Meals and snacks vary substantially concerning food variety (Rolls et al., 1981), portion sizes (Rolls, Morris, & Roe, 2002), and palatability (Rolls, 1979). Therefore, it is possible that food commercials have a more considerable influence on snack food intake than they do on intake at a meal, where competing influences such as variety, portion size, and palatability may override any exposure effect of food adverts. Another possible reason for the lack of effect may be due to the laboratory eating environment. Previous studies have measured eating behavior in a more naturalistic setting (Boyland et al., 2016), and children may eat differently in the laboratory than they do at home or school. It is worth noting that the estimated effect size for energy intake differences between the two commercial conditions in this study was minimal (d = 0.003) and therefore the lack of an effect is not likely due to inadequate sample size. Therefore, this study provides preliminary evidence that the intake promoting effects of food commercials may not extend into multi-item meals, at least not under the conditions tested in the current study.

A primary aim of this study was to evaluate response to visual food stimuli following exposure to food and toy commercials. Our whole-brain analysis revealed 10 regions in which exposure to food commercials affected the subsequent processing of food cues. These regions have previously been shown to be impacted by food commercial exposure (Bruce et al., 2016; Gearhardt et al., 2014; Yokum et al., 2014) and are associated with a variety of functions that can be broadly described as inhibitory control, sensory processing, memory, and attention (Aron & Poldrack, 2005; Batterink, Yokum, & Stice, 2010; Berkman, Falk, & Lieberman, 2011; Critchley & Rolls, 1996; Gearhardt et al., 2014; Rolls, 1996; Tabibnia et al., 2011; Yokum et al., 2014). The direction of these effects suggested that food commercials tended to reduce children’s response to high-relative to low-energy dense foods, while an opposite pattern was observed following toy commercials. Of particular interest is the decreased response to higher energy food cues in regions related to cognitive control following food commercials (i.e., superior frontal gyrus and inferior frontal gyrus). These observed responses provide support for our hypothesis that food commercials impact food cue processing in part by reducing engagement of brain regions implicated in cognitive control. Furthermore, this finding compliments behavioral data suggesting that food marketing can override the effects of impulse control leading to increases in snack food intake (Folkvord, Anschütz, Nederkoorn, Westerik, & Buijzen, 2014).

The whole-brain analysis also revealed interactions between child weight status and commercial condition in four regions including fusiform gyrus, supramarginal gyrus, and two regions of the orbitofrontal cortex. Compared to children of overweight/obese status, children in the healthy weight group showed greater engagement of these regions to higher energy-dense foods following toy commercial exposure. On the other hand, after food commercial exposure, children with overweight/obesity had increased engagement of these regions to higher energy dense foods. These findings highlight the differential effects of food marketing on children with overweight/obesity that have been previously observed (Gearhardt et al., 2014), despite the fact that our ad libitum test-meal was unable to capture these differences behaviorally. The orbitofrontal cortex is of particular interest as it is a vital region of the “appetitive network” that is involved in flavor processing (De Araujo, Rolls, Kringelbach, McGlone, & Phillips, 2003), sensory-specific satiety (Kringelbach, O’Doherty, Rolls, & Andrews, 2003), and evaluating the pleasantness and reward value of a food-related stimulus (Kringelbach, 2005; van der Laan et al., 2011). The orbitofrontal cortex is consistently seen to activate in response to food stimuli. Specifically, children with obesity have shown greater response in the orbitofrontal cortex compared to healthy weight counterparts when viewing food images food (Bruce et al., 2010). Furthermore, previous studies have shown increased activation of the orbitofrontal cortex following food commercials (Gearhardt et al., 2014; Rapuano et al., 2015). In prior studies from our lab, we have found increased BOLD response in the orbitofrontal cortex to large portion or rewarding food cues is associated with excess consumption in the laboratory (Adise, Geier, Roberts, White, & Keller, 2018; Keller et al., 2018). Our findings help to bridge the gap in the previous literature by revealing that food commercials modify response to higher-energy food stimuli in the orbitofrontal cortex in a manner that is dependent on child weight status.

While not explicitly part of our primary aims, we identified several brain regions that showed a main effect of weight status including right cuneus, right fusiform gyrus, right middle temporal gyrus, right supramarginal gyrus, and right transverse temporal gyrus. In all six cases there was reduced response to higher energy stimuli in children with overweight/obesity compared to those with healthy weight. These regions are broadly implicated in cognitive control but may have other functions as well. We also note that similarly reduced response to food cues has been observed in children with overweight/obesity when viewing food and non-food brands compared to healthy weight counterparts (Bruce et al., 2013b). Therefore, our results compliment these findings by showing that children of higher weight status generally have reduced engagement in control-related neurocircuitry in response to palatable, higher energy food cues.

The targeted ROI analyses identified three additional regions that were associated with food intake regardless of commercial exposure condition (left orbitofrontal cortex and bilateral dorsal striatum). Additionally, an interaction between commercial condition and brain response was observed in the right insula. However, caution is recommended in interpreting these results as they were not adjusted for multiple testing. In addition, these relationships were likely driven by children’s hunger level prior to the prior to the meal, as adjusting for reported fullness eliminated any statistical significance. One possible explanation for the lack of relationship between brain response to commercials and laboratory eating behavior is the overall limited variation in test-meal intake observed across conditions. On the group level there was virtually no difference by condition (d = 0.003) and on the subject level there was a strong correlation between conditions (α = 0.89). Future studies should attempt to replicate this design using a non-homeostatic eating occasion (e.g., snack or eating in the absence of hunger), as this protocol may be better suited to detect variations in behavior that are influenced by food marketing.

This study has several strengths and limitations. One strength is the use of toy commercials for comparison as opposed to using commercials for other less salient items (e.g., cars, clothes) that are not explicitly targeted towards children. This allowed us to test food commercials against a salient and age-appropriate contrast. Also, we experienced high success rates for the fMRI due to a comprehensive mock training protocol (de Bie et al., 2010; English et al., 2015). One limitation is the relatively modest sample size which may have reduced the sensitivity of both brain and behavioral responses, and especially may have limited our ability to find correlations between the two measures. Using a larger sample should be considered for future studies. We also note that repeated exposure to commercial stimuli, meals, and food images may have contributed to habituation to the stimuli, and therefore may have reduced the effect sizes. Future studies should consider also conducting the MRI on the same day as the meal. This would allow for a temporally closer comparison between brain response and changes in eating behavior. In addition, future studies using similar designs should be done to determine whether food commercial exposure influences non-homeostatic eating events, like snacks and/or eating in the absence of hunger. A final limitation is the overall homogeneity of the population sampled. Future studies should examine these relationships in cohorts that vary by demographic characteristics that have been implicated in response to food commercials, such as ethnicity (Harris, Schwartz, & Brownell, 2010), and socioeconomic status (Gorn & Goldberg, 1977).

In conclusion, the current study demonstrates that food commercials impact food cue processing in brain regions implicated in cognitive control, sensory processing, and attention. Further, child weight status moderated the influence of food commercials on food cue processing in the orbitofrontal cortex, a key region implicated in reward valuation as part of the appetitive network. These results suggest that food commercials interfere with the neural processing of food cues, and these responses differ depending on child weight status. Upon replication in a larger cohort, these results may help to explain why some children are more vulnerable to advertising content than others and may ultimately inform the development of more targeted interventions.