Neural correlates of inhibitory control in youth with symptoms of food addiction

Abstract

Prior research has found that food addiction is associated with reward-related neural differences, but research has yet to examine whether there are also neural differences in inhibitory control. This may be particularly relevant during adolescence as it is a key developmental period where difficulties in inhibitory control are more prevalent. The Yale Food Addiction Scale is a self-report questionnaire that applies substance use disorder diagnostic criteria to certain foods that has also been adapted for children. Here we investigate the association between addictive-like eating and brain functioning during inhibitory control in youth. Seventy-six right-handed participants 8.2–17.8 years (44 male) were recruited. Participants performed a go/no-go task during functional magnetic resonance imaging and completed the Yale Food Addiction Scale for Children, after which they were categorized into two groups according to their scores (No Symptom Group = 0; YFAS-C Group: score ≥ 1). Inhibitory control was probed with a contrast of correct no-go versus go trials. An independent-samples t-test comparing groups revealed a significant difference in three primary clusters, all exclusively in the left hemisphere (No Symptoms Group > YFAS-C Group): middle temporal gyrus/occipital gyrus, precuneus/calcarine sulcus, and inferior frontal gyrus. Specifically, the YFAS-C Group showed deactivation in all three clusters. Adolescents who endorse food addiction appear to show hypo-activation in response to the inhibitory control portion of a go/no-go task, which suggests possible inhibitory control difficulties.

1. Introduction

The term food addiction refers to a process involving physical and psychological dependence on certain foods, particularly those high in fat and refined carbohydrates (Davis & Cater, 2009; Davis et al., 2011; Gearhardt, Corbin, & Brownell, 2009; Gearhardt, Davis, Kuschner, & Brownell, 2011a; Schulte, Grilo, & Gearhardt, 2016). Similar to drug and alcohol addiction, addiction to these highly palatable foods can involve tolerance, withdrawal, and loss of control (Gearhardt et al., 2009). Food addiction has been shown to be strongly associated with obesity, body mass index (BMI)¹, and body fat percentage (Davis & Carter, 2009; Gearhardt et al., 2012; Pedram et al., 2013). Thus, food addiction may be a contributing factor in problem eating behaviors, including binge eating disorders and obesity (Schulte et al., 2013, Flint et al., 2014).

Just as inhibitory control problems can contribute to problematic substance use and substance use disorders (for review, see Mitchell & Potenza, [2014]), poor inhibitory control is thought to be one contributing factor to the “loss of control” aspect of food addiction, leading to excessive food intake (Chen et al., 2018). Indeed, there is growing evidence suggesting that decreased inhibitory control is associated with increases in unhealthy eating (Jasinka et al., 2012) and overeating (Guerrieri et al., 2007), including overeating in response to negative emotional states (Bekker, van de Meerekdonk, & Mollerus, 2004; Racine et al., 2009). Previous studies have also shown that individuals who are more impulsive and have poor inhibitory control are more likely to be overweight or obese than those who have greater inhibitory control (Guerrieri, Nederkoorn, & Jansen, 2008) or are of a healthy weight (Nederkoorn, Braet, Van Eijs, Tanghe, & Jansen, 2006). Furthermore, individuals with weaker versus those with stronger inhibitory control have been shown to have a higher BMI (Batterink, Yokum, & Stice, 2010; Cohen, Yates, Duong, & Convit, 2011).

There is also neural evidence suggesting that poor inhibitory control may contribute to risk for being overweight or obese. Several neuroimaging studies report decreased activation in prefrontal inhibitory control regions, such as the ventrolateral prefrontal and orbitofrontal cortices, for obese compared to lean participants in response to high-calorie food images (Silvers et al., 2014) or when trying to inhibit responses to high-calorie foods (Batterink et al., 2010). Additionally, decreased recruitment of the inferior, middle, and superior frontal gyri during an executive control task was found to predict a greater rate of weight gain in subsequent years (Kishinevsky et al., 2012). Thus, hypo-activation of the prefrontal cortex in obese individuals has been proposed as an explanation for poor control over food intake (Alonso-Alonso & Pascual-Leone, 2007). Reduced activation has been found in inhibitory control and performance monitoring regions in women with food addiction (Gearhardt et al., 2011b; Franken, Nijs, Toes, & van der Veen, 2016), where higher food addiction scores were associated with less activation in lateral orbitofrontal cortex when receiving highly palatable food (i.e. chocolate milkshake) (Gearhardt et al., 2011b). The authors measured food addiction with the Yale Food Addiction Scale (YFAS), a self-report questionnaire that applies the diagnostic criteria of substance use disorders to food (Gearhardt et al., 2009). The patterns of neural responses reviewed above have been implicated in other consumptive addictions, such as alcohol or tobacco use (Sutherland & Stein, 2018; Zahr, Pfefferbaum, & Sullivan, 2017), further extending the similarities between substance use and addictive-like eating. However, as two of these studies did not use fMRI tasks that directly measure inhibitory control (Gearhardt et al., 2011b; Silvers et al., 2014), it is possible that other neural processes are at work; thus, further research is needed to directly link food addiction with inhibitory control problems.

The YFAS is the most commonly used measure to assess food addiction (Meule & Gearhardt, 2014). It was initially developed for the assessment of food addiction in adults and has since been adapted for children (YFAS-C; Gearhardt, Roberto, Seamans, Corbin, & Brownell, 2013). The downward extension of the scale was important because addictive eating in children has been shown to be related to elevated BMI and reduced satiety. Although research on how addictive processes contribute to problematic food consumption in adults is growing, research in children and adolescents is still limited. Childhood-onset obesity typically persists into adulthood (Epstein, Wing, & Valoski, 1985), similar to the way that substance use disorders that emerge during early adolescence often last throughout an individual’s life (Chen, Storr, & Anthony, 2009). This risk is thought to stem from variations in an already vulnerable neural system (i.e. reward dysfunction, increased impulsivity) (Tapert, Caldwell, & Burke, 2004), in addition to an increased likelihood of using substances to cope (Clark, Thatcher, & Tapert, 2008). Highly palatable foods, if they have an addictive potential, may have more of an impact on children as opposed to adults due to neural and psychological vulnerabilities. To date, however, no studies have investigated the link between food addiction symptoms and patterns of neural activation in children and adolescents.

Here we investigated inhibitory control neural circuitry in adolescents with symptoms of food addiction versus those who exhibit no symptoms. Previous fMRI and electrophysiology studies examining addictive-like eating have used reward and/or a salience-based tasks (Imperatori et al., 2015; Kishinevsky et al., 2012; Gearhardt et al., 2011b; Schulte, Yokum, Jahn, & Gearhardt, 2019), thus the investigation of inhibitory control is innovative. Inhibitory control was assessed using functional magnetic resonance imaging (fMRI) and the go/no-go paradigm, and food addiction symptomology was measured with the YFAS-C. Due to the novelty of this study, we used a whole-brain analysis approach in order to avoid making biased assumptions about the data. We predicted that adolescents scoring higher on the YFAS-C would show less activation in key prefrontal inhibitory control regions in the brain compared to adolescents who show no food addiction symptomology.

2. Methods

2.1. Participants

Seventy-six right-handed individuals (44 males) aged 8.2–17.8 years participated in this study. Participants were recruited from the Michigan Longitudinal Study, a prospective, multi-wave study of families with high levels of parental alcohol use disorder (AUD) and a contrast sample of nonalcoholic families from mid-Michigan that began in 1987 (Zucker, 1996). Since 2005, neuroimaging data has been collected every 1–2 years on a subset of these participants, starting between the ages of 7–12 years. The YFAS-C (Gearhardt et al., 2013), described in Section 2.2, was incorporated into the neuroimaging protocol in 2014 and is collected after each fMRI scan. All participants with useable fMRI scans and complete YFAS-C data were eligible for the current study. Participants were split into two groups based on their YFAS-C scores (see Section 2.2). The study was conducted according to the guidelines provided by the Declaration of Helsinki and the ethical requirements of the University of Michigan (HUM00084126).

Parental AUD diagnosis for the larger Michigan Longitudinal Study was based on DSM-IV criteria, and assessed by way of the Diagnostic Interview Schedule–Version 4 (Robins, Cottler, Bucholz, Compton, North, & Rourke, 2000; Robins, Helzer, Croughan, & Ratcliff, 1981), supplemented with the Drinking and Drug History (Zucker, Ellis, Fitzgerald, Bingham, & Sanford, 1996; Zucker et al., 2000). For details on parental AUD diagnosis, see Supplementary Material. Participants in the current neuroimaging study came from families with and without a family history of AUD (there were no significant differences between groups for family history; see Table 1). Exclusionary criteria for this study included: neurological, acute, uncorrected, or chronic medical illness; current or recent (within 6 months) treatment with centrally active medications; and history of psychosis or schizophrenia in first-degree relatives. The presence of Axis I psychiatric or developmental disorders that would interfere with the interpretation of the data was also exclusionary; this did not include past history of mood disorder or current unmedicated mood disorder, or current or past history of conduct or attention deficit hyperactivity disorders. Diagnosis was determined using the Diagnostic Interview Schedule–Child (Costello, Edelbrock, Kalas, Kessler, & Klaric, 1982). Families in which the target child displayed evidence of fetal alcohol effects were excluded from the original ascertainment.

Table 1.

Participant Characteristics

	All Participants (n = 76)	NFAS Group (n = 41)	YFAS-C Group (n = 35)	Statistic	Significance
Males/Female (n)	44/32	26/15	18/17	—	p = .35a
Age at fMRI scan	14.3 (2.8)	14.6 (2.9)	14.0 (2.8)	t(74) = 0.89	p = .38b
Full-scale IQ*	101.8 (12.6)	102.5 (12.1)	100.9 (13.3)	t(72) = 0.78	p = .57b*
Family history of AUD (n)	13 FH−/63 FH+	7 FH−/34 FH+	6 FH−/29 FH+	—	p = 1.00a
BMI	21.8 (4.1)	22.1 (4.7)	21.5 (3.4)	t(72) = 0.62	p = .54c
YFAS-C	0.7 (1.0)	0.0 (0.0)	1.5 (1.0)	t(74) = −9.0	p = 0.001c
Substance Use Prior to fMRI
Scan (n; present)
Used alcohol	17	9	8	—	p > .99a
Used cannabis	12	7	5	—	p > .99a
Used nicotine	6	4	2	—	p = .68a
Used other drugs	4	3	1	—	p = .62a
Used multiple substances	10	6	4	—	p > .99a
Lifetime Diagnosis (n; present)
Conduct disorder	0	—	—	—	—
ADHD	4	1	3	—	p = .31a
Generalized anxiety disorder	0	—	—	—	—
Depressive disorder	0	—	—	—	—

Note. NFAS: No Food Addiction Symptom; YFAS-C: Yale Food Addiction Scale for Children; fMRI, functional magnetic resonance imaging; AUD, alcohol use disorder; FH−/+, family history negative/positive for alcohol use disorder; BMI, body mass index; ADHD, attention deficit hyperactivity disorder. Numbers represent means, with standard deviations in parentheses, unless otherwise noted.

^aTwo-tailed Fisher’s exact test.

^bTwo-tailed independent samples t-test, equal variances assumed

^cTwo-tailed independent samples t-test, equal variances not assumed

^*One participant in each group was missing an IQ score

As part of the Michigan Longitudinal Study, all offspring were assessed annually on substance use and related problems (Zucker et al., 1996). For details on annual substance use assessments, please see Supplementary Material. All participants were told to abstain from alcohol and illicit substances for 48 hours prior to the fMRI scan. For participants age 15 years and older, urine drug screens were conducted immediately prior to the fMRI scan; positive results were exclusionary. In participants age 14 years and younger, we relied on verbal confirmation of drug and alcohol abstinence on the day of the scan. No participants had to be excluded from the current study due to a positive drug screen or affirmative self-report of alcohol or drug use. All participants gave written consent/assent after explanation of the experimental protocol and at least one parent gave written informed consent, as approved by the University of Michigan Institutional Review Board. Height (in) and weight (lbs) for each participant were collected at the time of the scan. Table 1 contains participant information.

2.2. Yale Food Addiction Scale for Children

Participants were categorized into one of two groups based on their symptom count scores on the YFAS-C; the No Food Addiction Symptom Group (NFAS Group) contained participants who scored 0 on the YFAS-C (n = 41), and the YFAS-C Group contained participants who endorsed ≥1 food addiction symptoms on the YFAS-C (n = 35). Groups were matched on age at fMRI scan. See Table 1.

The Yale Food Addiction Scale is the most commonly used measure to assess the construct of food addiction (Meule & Gearhardt, 2014). The full YFAS-C is a 25-item self-report questionnaire that applies the diagnostic criteria for substance use disorders to consumption of certain foods, namely highly palatable foods such as ice cream, pizza, and chocolate. Because of participant burden concerns, we used a modified YFAS-C with nine questions; one question representing each symptom relating to the seven diagnostic criteria for substance dependence, as well as two questions for clinically significant impairment/distress, was used. This same approach has been used to abbreviate the adult version of the YFAS; tests for internal consistency reveal that the YFAS remains psychometrically sound in this form (Flint et al., 2014; Schulte & Gearhardt, 2017). The question representative of each symptom was based on the highest factor loading from the original validation (Gearhardt et al., 2013). Each question is scored 0 (symptom absent) or 1 (symptom present). For example, answering “always” to Question 1 (“I eat foods even when I am not hungry”) gives a score of 1, but answering “never”, “rarely”, “sometimes”, and “very often” leads to a score of 0. Scores for all nine questions are totaled based on these answers. Table 2 outlines each question and scoring information. Table S2 details how each participant in the YFAS-C Group answered each of the nine questions.

Table 2.

Modified YFAS-C Scoring Dichotomy and Scale

In the last year (past 12 months):	Never	Rarely	Sometimes	Very Often	Always
Q1. I eat foods even when I am not hungry	0	0	0	0	1
Q2. I feel tired a lot because I eat too much	0	0	0	1	1
Q3. I avoid places where I cannot eat the food I want	0	0	1	1	1
Q4. I eat certain foods to stop from feeling upset or sick	0	0	0	1	1
Q5. The way I eat makes me really unhappy	0	0	0	1	1
Q6. The way I eat causes me problems (for example, problems at school, with my parents, with my friends)	0	0	0	1	1
In the last years (past 12 months):				No	Yes
Q7. I eat in the same way even though it is causing problems				0	1
Q8. I need to eat more to get the food feelings I want (for example, for happy, calm, relaxed)				0	1
Q9. I am unable to cut down on certain foods				1	0

Symptom count scoring dichotomy is represented by 0 or 1. Number in bold is number that contributes to symptom count for each particular question.

2.3. fMRI task

A go/no-go task (Durston, Thomas, Yang, Ulug, Zimmerman, & Casey, 2002) was used to probe response inhibition. Participants were instructed to respond to target stimuli (letters other than X) by pressing a button (go trials) but make no response to infrequent non-target stimuli (letter X; no-go trials). Letters were white and presented on a black screen. Stimulus duration was 500 ms, followed by 3500 ms of fixation. Each no-go trial could be preceded by 1, 3, or 5 go trials; order of the trials was pseudorandomized. There were 5 runs of 49 trials, each run lasting 3 minutes and 2 sec and containing 11, 12, or 13 no-go trials for a total of 60 no-go trials out of 245 trials. Reaction times for correct go responses (Hit RT), accuracy for correct go trials (Hit), accuracy for false alarms (False Alarms), and reaction times for false alarms (False Alarm RT) were recorded. Before the fMRI scan, all participants had a practice session of 49 trials on a desktop computer.

2.4. MRI data acquisition

Whole-brain blood oxygenated level-dependent images were acquired on a 3.0 Tesla GE Signa scanner (Milwaukee, WI) using a T2*-weighted single-shot combined spiral in-out sequence (Glover & Law, 2001) with the following parameters: TR = 2000 ms; TE = 30 ms; flip angle = 90°; FOV = 200 mm; 64 × 64 matrix; in-plane resolution = 3.12 × 3.12 mm; slice thickness = 4 mm; 29 slices. A high-resolution anatomical T1 scan was obtained for spatial normalization (three-dimensional spoiled gradient-recalled echo; TR = 25 ms; min TE; FOV= 25 cm; 256 × 256 matrix; slice thickness = 1.4 mm). Participant head motion was minimized using foam pads placed around the head along with a forehead strap. In addition, the importance of keeping still was emphasized.

2.5. Data Analysis

Hypotheses were specified before data was collected. Additionally, the analytic plan was pre-specified before any analyses were performed.

2.5.1. Demographics, performance, substance use variables:

Independent-samples t-tests were used to look for group differences in age, IQ (as assessed by the Wechsler Intelligence Scale for Children, or WISC-III), body mass index (BMI; measured at the time of the fMRI scan via height and weight measurements), and each of the performance variables: Hit, Hit RT, False Alarms, and False Alarm RT. Fisher’s Exact Test was used to look for group differences in sex and family history of AUD. The number of participants for each group who reported substance use prior to their fMRI scan (having used alcohol, nicotine, cannabis, other drugs, or any combination of these) is reported in Table 1, and Fisher’s Exact Test was used to look for group differences between these. For diagnoses, Fisher’s Exact Test was used for attention deficit hyperactivity disorder as no other diagnoses (conduct disorder, generalized anxiety disorder, or depressive disorder) were present in this sample.

2.5.2. fMRI data:

Functional images were reconstructed using an iterative algorithm (Noll, Fessler, & Sutton, 2005; Sutton, Noll, & Fessler, 2003). Subject head motion was corrected using FSL 5.0.2.2. (Analysis Group, FMRIB, Oxford, United Kingdom) (Jenkinson, Bannister, Brady, & Smith, 2002). Analysis of estimation of motion parameters confirmed that overall head motion within each run did not exceed 3 mm translation or 3° rotation in any direction. All remaining image processing (including slice timing correction) and statistical analysis were completed using statistical parametric mapping (SPM8; Wellcome Institute of Cognitive Neurology, London, United Kingdom). Functional images were spatially normalized to a standard stereotaxic space as defined by the Montreal Neurological Institute. A 6 mm full-width half-maximum Gaussian spatial smoothing kernel was applied to improve signal-to-noise ratio and to account for differences in anatomy.

Individual-level analysis was completed using a general linear model. Three regressors of interest (correct no-go, failed no-go, and correct go) were convolved with the canonical hemodynamic response function, with event durations of 4 seconds from stimulus presentation. Motion parameters were modeled as nuisance regressors to remove residual motion artifacts. The main contrast of interest was correct no-go (correct reject) > go trials (CR > GO). This was calculated for group-level analysis by linearly combining parameter estimates over all five runs of the task. The pre-specified significance level was set at a primary threshold of p < 0.001, cluster-level of p < .05 (family-wise error rate [FWE]-corrected for multiple comparisons). To confirm that the GNG task elicited the expected activation, a one-sample t-test was run using the entire sample (n = 76) for the contrast of interest (CR > GO).

The main hypothesis of interest (group differences for the CR > GO contrast) was tested on a whole-brain basis using an independent-samples t-test. Values from significant clusters were extracted using MarsBaR Region of Interest toolbox (Brett, Anton, Valabregue, & Poline, 2002). Extracted values were then imported into SPSS (Version 24, IBM Corp, Armonk, NY) for graphical purposes and for post-hoc analyses. Associations among scan age, BMI, YFAS-C group, and mean activation in significant clusters were investigated using independent-samples t-tests or Pearson’s correlations.

3. Results

3.1. Participant Characteristics

There were no significant differences (i.e., all ps > .05) between groups on sex, age at fMRI scan, IQ, family history of AUD, BMI, or ADHD diagnosis. There were also no significant differences between the groups for substance use initiation prior to fMRI scan. There was a significant difference between groups for YFAS-C scores. Age was significantly positively correlated with BMI across the whole group, r = .34, p = .002. See Table 1 for statistics.

3.2. Task Performance

There were no significant differences between groups on task performance measures—Hits, Hit RT, False Alarms, and False Alarm RT. See Table 3 for statistics.

Table 3.

Go/No-Go Task Performance by Group

	NFAS Group	YFAS-C Group	Statistic	Significance
Hits (%)	96.8 (3.2)	96.3 (4.0)	t(74) = 0.52	p = .60
Hit RT (ms)	429.7 (59.2)	438.2 (58.4)	t(72) = −0.63	p = .53a
False Alarms (%)	41.1 (20.7)	44.5 (19.8)	t(74) = −0.72	p = .47
False Alarm RT (ms)	388.2 (58.3)	392.3 (49.8)	t(74) = −0.34	p = .75

Note. ms, milliseconds. For NFAS and YFAS-C columns, the numbers given are means, with standard devotion in parentheses. Statistic and significance columns refer to two-tailed independent-samples t-tests.

^aDegrees of freedom reduction reflects equal variances not assumed.

3.3. fMRI Data

3.3.1. Task effect:

Activation for the task effect passed correction for multiple comparisons in six regions: right middle frontal gyrus, right inferior temporal gyrus, right posterior cingulate gyrus, left angular gyrus, left middle frontal gyrus, and left inferior frontal gyrus. See Table S2 (Supplementary Material) for task effect coordinates and statistics. These regions are agreement with previous studies that have investigated the neural correlates of response inhibition (see Zhang, Geng, & Lee, 2017 for meta-analysis) and confirm that our task elicited the expected activation.

3.3.2. Group differences:

An independent-samples t-test revealed a significant group difference for CR > GO at primary threshold of p < .001, cluster-level FWE-corrected p < .05, when comparing the NFAS to the YFAS-C Group, in three clusters: 1) left middle temporal gyrus, extending into the middle occipital gyrus; 2) left precuneus/left calcarine sulcus, extending into the posterior cingulate, and 3) left inferior frontal triangularis/opercularis region (Figure 1). See Table 4 for coordinates and statistics. In all three clusters, the YFAS-C Group showed less activation during successful inhibitory control compared to the NFAS Group (Figure 2). Post-hoc correlations were run to demonstrate that results were not due to associations with age or BMI; activation in the three clusters was not significantly correlated with age (all ps > .437) or BMI (all ps > .329) across all subjects. The YFAS Group did not demonstrate significantly greater activation in any brain regions compared to the NFAS Group for CR > GO.

Figure 1.

Regions displaying a significant difference for the CR > GO contrast at initial threshold of p < .001 uncorrected with a cluster-wise threshold of p < .05 FWE. Three clusters passed criteria: Cluster 1 encompassed the left middle temporal and occipital gyrus (red arrow); Cluster 2 encompassed the left precuneus and calcarine sulcus (green arrow); Cluster 3 encompassed the left inferior frontal gyrus – pars triangularis and pars opercularis (white arrow). See Table 4 for full coordinates.

Table 4.

CR > GO contrast from independent-samples t-test of NFAS vs YFAS-C Group

	% ROI	x	y	z	k	t-value	FWE p-value
Cluster 1		−40	−66	16	257	4.42	.034
Left Middle Temporal Gyrus	53.7%
Left Occipital Gyrus	26.1%
Outside (not identified)	20.2%
Cluster 2		−14	−52	12	236	4.10	.046
Left Precuneus	40.3%
Left Calcarine Sulcus	30.5%
Outside (not identified)	13.6%
Lingual Gyrus	9.8%
Vermis 4/5	5.9%
Cluster 3		−56	24	20	312	3.86	.016
Left Inferior Frontal Gyrus - Pars Triangularis	68.3%
Left Inferior Frontal Gyrus – Pars Opercularis	31.7%

Note. ROI = region of interest; x, y, z are peak voxel coordinates for clusters of activation in Montreal Neurological Institute coordinates; k = cluster size; threshold p < .05, FWE cluster corrected.

Figure 2.

Mean activations extracted from each cluster from the CR > GO contrast in Figure 1 for the NFAS and YFAS-C groups. See Table 4 for cluster labels and coordinates. Error bars represent one standard error of the mean.

As one participant had a substantially higher YFAS-C score compared to the rest of the YFAS-C group (participant #35, see Table S1), the group difference analyses for CR > GO were rerun with this participant removed. All three clusters were still significant for at p < .001, cluster-level FWE-corrected p < .05, when comparing the NFAS to the YFAS-C Group. No other significant clusters were found.

4. Discussion

In the current study, adolescents who had experienced food addiction symptoms demonstrated less neural activation during successful inhibitory control compared to individuals who had not experienced any symptoms. Although studies have explored the association between brain activity and BMI/obesity (Bauer & Houston, 2017; Gearhardt, Yokum, Stice, Harris, & Brownell, 2014; Val-Laillet et al., 2015; Yokum, Gearhardt, Harris, Brownell, & Stice, 2014) as well as disordered eating in adults and adolescents (for review, see Donnelly, Touyz, Hay, Burton, Russell, & Caterson, [2018]), to our knowledge this is the first study to examine the neural circuitry of food addiction symptomology in adolescents using the YFAS-C. Here we found that three clusters, encompassing the left middle temporal/occipital gyrus, the left precuneus/occipital gyrus, and left inferior frontal gyrus, showed less activation in the YFAS-C group compared to the NFAS group.

Activation in both the left and right inferior frontal gyrus have been implicated in inhibitory control during the response inhibition portion of a go/no-go task (for left inferior frontal gyrus specifically, see Swick, Ashley, & Turk [2008]; for review, see Zhang et al. [2017]), suggesting that the inferior frontal gyrus is linked to suppressing prepotent responses. The inferior frontal gyrus is part of a larger network, including subcortical and cortical regions known to be involved in executive and motor control, that supports response inhibition during the go/no-go task. While we do interpret decreased activation here as less efficient processing, it is important to note that this could also be explained as more efficient processing—as task activation differences without performance differences can often be difficult to interpret. However, prior studies have linked a weakness, interpreted as less efficient processing, in this network to risk for substance use problems (for review, see Heitzeg, Cope, Martz, & Hardee [2015]); similarly, weaknesses in inhibitory control may be a relevant mechanism underlying risk for food addiction. In the only other neuroimaging study to use the YFAS, young adults with elevated symptomology exhibited decreased activation in inhibitory control systems when consuming a highly processed food reward, compared to young adults with one or fewer YFAS symptoms (Gearhardt et al., 2011b; Hsu, Wang, Ko, Hsieh, Chen, & Yen, 2017). Here, we used non-food cues to examine inhibitory control in a group of children and adolescents; this enabled us to examine the impact of non-food related inhibitory control at an earlier point in the developmental process.

Despite the fact that adolescence is a high-risk period for the emergence of addictive behaviors, little is known regarding how addictive-like eating develops during this time. In the dual systems model of adolescent brain development, the earlier maturation of subcortical reward networks relative to prefrontal cognitive (inhibitory) control networks creates a maturational mismatch between the two systems (Bourgeois & Goldman-Rakic, 1994; Casey, Jones & Hare, 2008; Steinberg, 2008). Prefrontal control matures linearly across development from late childhood through early adulthood (Galvan et al., 2006; Gogtay et al., 2004; Hare, Tottenham, Galvan, Voss, Glover, & Casey, 2008; Huttenlocher, 1979; Sowell, Peterson, Thompson, Welcome, Henkenius, & Toga, 2003), while the reward network follows a curvilinear trajectory that increases during late childhood, peaks during adolescence, and then declines during the late teens and early twenties (for review, see Shulman et al. [2016]). This imbalance is thought to contribute to higher rates of problematic substance use, such as binge drinking, as well as the increased risk for obesity (Dietz, 1994; Eaton et al., 2010). The YFAS has been associated with factors related to addiction, such as craving and impulsivity, and the negative consequences of overeating in both clinical (e.g. obese and overweight) and non-clinical adolescents (Gearhardt et al., 2013; Chen, Tang, Guo, Liu, & Xiao, 2015; Meule, Muller, Gearhardt, & Blechert, 2017; Richmond, Roberto, & Gearhardt, 2017; Schulte, Jacques-Tiura, Gearhardt, & Naar, 2018). Inhibition of a prepotent response contributes to self-control over excessive behaviors, such as continuing to take drugs or eat unhealthy foods. Less activation in the YFAS-C group during inhibitory control may signify a general weakness in the mechanism supporting response inhibition during adolescence. Weak inhibitory control circuitry in adolescents paired with the experience of addictive eating behaviors may be useful indicators for investigating the later emergence of clinical levels of food addiction.

The role of the other two activation clusters—the middle temporal/occipital gyrus and the precuneus/occipital gyrus—is more speculative. It has been argued that, in addition to motor response inhibition, the go/no-go task is a measure of sustained attention (Robertson, Manly, Andrade, Baddeley, & Yiend, 1997). Thus, lower activation in both the middle temporal/occipital gyrus and the precuneus/occipital gyrus in the YFAS-C group could reflect disengagement of attention away from the task. Research has shown that brain regions known to mediate attention selectively to external stimuli (i.e. task stimuli) are also important for actively directing attention internally (i.e. to focus on the visceral or body state), otherwise known as interoception (Critchley, Wiens, Rotshtein, Ohman, & Dolan, 2004; Nebel, Wiese, Stude, de Greiff, Diener, & Keidel, 2005; Pollatos, Kirsch, & Schandry, 2005; Tracy et al., 2007). Volkow and colleagues (Volkow, Wang, Tomasi, & Baler, 2013) suggest that impairments in self-control and interoceptive circuitry is a functional model of addiction that can be used to understand why some obese individuals have difficulty in regulating caloric intake and maintaining energy homeostasis. The ability to inhibit a prepotent response, such as in the go/no-go task, and exert self-control is a likely contributor to one’s ability to avoid engaging in excessive behaviors (e.g. taking drugs, overeating); thus, this would increase vulnerability to substance-related addictive behaviors or obesity (Volkow & Fowler, 2000; Volkow et al., 2008).

These activation differences were found despite the fact that there were no significant differences between the groups with respect to go/no-go task performance. This is similar to results found in adolescents at risk for AUD, whose performance is analogous to controls for simple measures of response inhibition, such as the go/no-go task (Hardee et al., 2014; Heitzeg, Nigg, Yau, Zucker, & Zubieta, 2010; Norman, Pulido, Squeglia, Spadoni, Paulus, & Tapert, 2011). It is possible that performance differences may accompany aberrant neural responses with a more challenging task in the YFAS-C group.

This study has limitations. First, the study was not powered to investigate sex differences. Given that males and females may show different risk factors for substance use (Foster, Hicks, Iacono, & McGue, 2015; Heitzeg, Hardee, & Beltz, 2018; Kendler, Edwards, & Gardner, 2015; Kuhn, 2015), it is possible their risk profiles for problem eating also differ. Future studies should investigate sex differences in the association between neural mechanisms involved in inhibitory control and vulnerability to food addiction. Second, this study was cross-sectional, which did not allow for the evaluation of whether deactivation in inhibitory control networks precede or follow elevated food addiction symptomology. Future work using a longitudinal design will allow for a better understanding of the developmental antecedents to problem eating. Additionally, nearly 75% of participants are considered at-risk as they have a family history of AUD, which precludes these results from generalizing to individuals without a family history of substance use. However, when controlling for a family history of AUD in supplementary analyses, we found it did not affect our results. Third, we did not have data on eating disorder psychopathology (e.g. binge eating severity), which is known to be strongly correlated with food addiction symptoms; thus, we could not look for differences between the groups. Additionally, general psychopathologies, such as depression and anxiety, are known to be significantly correlated with food addiction symptoms (Burrows, Kay-Lambkin, Pursey, Skinner, & Dayas, 2018). While we did have measures of general psychopathology for this sample, it was missing for 13 out of the 76 participants; therefore, we did not include this data.

In summary, the current study is the first to investigate neural differences based on food addiction symptoms in adolescents. Adolescents who endorse food addiction appear to show hypo-activation in response to the inhibitory control portion of a go/no-go task, which suggests possible inhibitory control difficulties. These findings converge with research in adults where food addiction is associated with behavioral inhibition difficulties; they also indicate that at least some part of the process is in place earlier. However, it is important to note that the implications of adolescents endorsing food addiction symptoms are currently unknown, as this research is still relatively nascent. It is essential to investigate whether this predicts the development of clinical levels of food addiction and related problems (e.g. obesity, binge eating). This work suggests that interventions looking to increase inhibitory control in adolescents who endorse food addiction, such as neurofeedback and/or transcranial direct current stimulation, may be a promising future direction.