Food Cue Reactivity in Food Addiction: a Functional Magnetic Resonance Imaging Study

Abstract

Background:

While neuroimaging studies have revealed that reward dysfunction may similarly contribute to obesity and addiction, no prior studies have examined neural responses in individuals who meet the “clinical” food addiction phenotype.

Methods:

Women (n=44) with overweight and obesity, nearly half of whom (n=20) met criteria for moderate-to-severe Yale Food Addiction Scale 2.0 (YFAS 2.0) food addiction, participated in a functional magnetic resonance imaging cue reactivity task. Participants viewed images of highly processed foods, minimally processed foods, and household objects while thinking about how much they wanted each item. Differences in neural responses by YFAS 2.0 food addiction to highly processed and minimally processed food cues were investigated.

Results:

There was a significant interaction between participant group and neural response in the right superior frontal gyrus to highly versus minimally processed food cues (r=.57). Individuals with YFAS 2.0 food addiction exhibited modest, elevated responses in the superior frontal gyrus for highly processed food images and more robust, decreased activations for minimally processed food cues, whereas participants in the control group showed the opposite responses in this region. Across all participants, the household items elicited greater activation than the food cues in regions associated with interoceptive awareness and visuospatial attention (e.g., insula, inferior frontal gyrus, inferior parietal lobe).

Conclusions:

Women with overweight or obesity and YFAS 2.0 food addiction, compared to those with only overweight or obesity, exhibited differential responses to highly and minimally processed food cues in a region previously associated with cue-induced craving in persons with a substance-use disorder. Overall, the present work provides further support for the utility of the food addiction phenotype within overweight and obesity.

1. Introduction

Obesity remains one of the most significant risk factors for numerous diseases (e.g., hypertension, diabetes) and a shortened lifespan comparable to the effects of smoking (1, 2). The causes of obesity are multifactorial (3, 4) and there has been a call within the field of obesity and, more broadly, psychiatry, to move towards personalized, precision medicine by identifying specific phenotypes that may inform the development of mechanism-focused interventions (5). One possible phenotype in obesity may be food addiction, which posits that addictive-like responses to certain foods may contribute to compulsive overeating (6–8). Food addiction is most commonly operationalized using the Yale Food Addiction Scale (YFAS), which adapts the Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria for substance-use disorders (e.g., loss of control, craving, withdrawal) to the addictive-like consumption of certain foods (9, 10). Behavioral data have revealed that individuals with obesity who meet for YFAS food addiction may have different risk factors for overconsuming food, relative to those with only obesity, such as greater food cravings and emotionally triggered eating (11). However, research into biological mechanisms that may be particularly relevant for persons with obesity and food addiction remains limited and warrants further empirical attention.

One of the key tenets of food addiction is that certain foods may exhibit an addictive potential, akin to drugs of abuse (12), and the limited research to date in this area remains a significant point of controversy (13, 14). Neuroimaging research has demonstrated that highly processed foods, defined as foods with added amounts of fat, refined carbohydrates, and/or salt, (e.g., pizza, chocolate, potato chips), and cues signaling their availability, engage reward-related neural circuitry in a similar manner as drugs of abuse (15), whereas minimally processed foods (e.g., fruits, vegetables) have less reward potential (15, 16). In healthy subjects, highly processed foods with both fat and carbohydrate have demonstrated supra-additive responses in the dorsal striatum, a region associated with reward motivation, compared to minimally processed foods, or foods high in only fat or carbohydrate. (17).

The rewarding nature of highly processed foods may be particularly relevant within obesity. Functional magnetic resonance imaging (fMRI) studies have observed that individuals with obesity, relative to those of healthy weight, have exhibited greater blood-oxygen-level dependent (BOLD) response in brain regions implicated in motivation, craving, and reward-based learning (e.g., dorsal striatum) for highly processed, compared to minimally processed, food cues or neutral images (18–20). Notably, the observed responses to highly processed foods and food cues in persons with obesity parallel neural reactivity to drug cues in persons with a substance-use disorder (21–25). Collectively, these findings have bolstered support for the theory that highly processed foods may be most likely to exhibit an addictive potential, and persons with obesity may be particularly susceptible to these foods’ rewarding properties. Differences in food-related reward response may be particularly relevant to individuals with obesity who exhibit food addiction and may highlight intervention targets for this phenotype. Thus, future research is needed to examine whether neurobiological responses to highly processed and minimally processed foods differ in persons with elevated body weight who meet for YFAS food addiction, relative to those with only elevated body weight.

No studies to date have examined neural responses to food cues in persons with YFAS food addiction, although one fMRI study using participants with subclinical YFAS food addiction provides preliminary insight into the biological mechanisms that may contribute to addictive-like eating behavior. Gearhardt and colleagues (26) found that individuals with versus without subclinical YFAS food addiction exhibited elevated responses in mesocorticolimbic reward regions (e.g., dorsolateral prefrontal cortex (dlPFC), caudate) while anticipating a palatable, highly processed food (chocolate milkshake) and less response in an inhibitory control region (lateral orbitofrontal cortex (OFC)) during the receipt of this food (26). These patterns have also been observed in individuals with substance-use disorders with respect to the anticipation and receipt of drug-specific rewards (22, 27) and are in line with the incentive sensitization theory (28, 29), which posits that persons with addiction experience increased “wanting” in response to drug-relevant cues. Thus, individuals with indicators of addictive-like eating behavior may similarly exhibit greater anticipatory “wanting” triggered by highly processed food cues. However, this study consisted of participants of all weight classes, only two of whom met the YFAS food addiction criteria. Further, reward responses were only investigated for one food, which did not allow for conclusions to be drawn about whether highly processed, relative to minimally processed, foods are more implicated in addictive-like reward responses in the context of food addiction.

The present study advances the knowledge of neurobiological mechanisms that may contribute to addictive-like eating behavior in two ways. First, this work explored between-group differences in neural responses to food cues in individuals with elevated body weight who met the moderate or severe food addiction thresholds on the current version of the YFAS (YFAS 2.0) versus those with only elevated body weight but not YFAS 2.0 food addiction. This approach allowed for the investigation for whether food addiction may be a meaningful obesity phenotype. Second, the current study examined neural food cue reactivity to nutritionally diverse foods (highly and minimally processed) in order to assess whether foods vary in their ability to engage reward-related neural circuitry, particularly for those with YFAS 2.0 food addiction. Akin to prior studies of food and drug cue reactivity (15), neural responses to both types of food cues were contrasted with activity to a household item cue control condition, which was selected because household cues may have more similar colors, complexities, and valences as the food cues compared to other control conditions (e.g., fixation).

Thus, the current study was the first to examine neural responses to highly and minimally processed foods in a sample of individuals with elevated body weight, half of whom exhibited the food addiction phenotype of interest. Broadly, given prior work suggesting that highly processed foods may be especially associated with compulsive consumption (30, 31), it was expected that the highly processed food cues would engage mesocorticolimbic regions to a greater degree than minimally processed foods or the household item control condition. Further, it was hypothesized that persons with, relative to without, YFAS 2.0 food addiction would exhibit greater neural responses for highly processed food cues in regions implicated in prior studies of drug cue reactivity in individuals with substance-use disorders (e.g., dorsal striatum, dlPFC, lateral OFC).

2. Methods and Materials

2.1. Participants

Given that overweight is a risk factor for the development of obesity and is associated with similar negative health consequences as obesity (32), both weight classes were included in the present study. Women with overweight or obesity were recruited through the University of Michigan’s clinical research listserv (umhealthresearch.org) and flyers posted in the community and online. Exclusion criteria were: a lifetime history of eating or substance-use disorders or history of psychiatric disorders (e.g., depression) in the past six months, which were assessed through clinical interviews using the Mini International Neuropsychiatric Interview (33), fMRI contraindications (e.g., metal in body), or current enrollment in a weight loss program (e.g., Weight Watchers). Women with overweight or obesity and without eating disorders were recruited in order to isolate food addiction as the key between-group difference and allow for the findings to demonstrate unique associations with food addiction in the absence of commonly comorbid eating-related conditions (34). Among individuals who participated in the study (N=49), two were excluded for having fMRI data with movement >3mm and three were excluded due to technical difficulties during the scan. Participants included in analyses (n=44) were 25–40 years old (M=30.55, SD=4.04) and racially diverse: 52.3% White/Caucasian, 27.3% Black/African-American, 9.1% Hispanic, 6.8% Multiracial, 2.3% Asian, and 2.3% Arab. BMI ranged from 24.70–51.00 (M=33.80, SD=5.48), with 20.45% of women having overweight and 79.55% having obesity. The number of individuals with overweight and obesity was not statistically significant by food addiction status (p=.18). For women with food addiction, two met criteria for overweight and 17 for obesity, and one participant met for a BMI in the normal weight range despite meeting screening criteria for overweight. Among participants without food addiction, seven met for overweight and 17 for obese BMI. The University of Michigan’s Institutional Review Board approved this research. Written informed consent was obtained for all participants.

2.2. Measures

2.2.1. Yale Food Addiction Scale 2.0

The YFAS 2.0 is a psychometrically sound measure for operationalizing indicators of a food addiction phenotype based on the DSM-5 criteria for substance-use disorders (35) and was utilized in the present study to categorize participants as individuals with or without food addiction. Twenty participants met criteria for moderate-to-severe YFAS 2.0 food addiction diagnostic score by endorsing 4 to 11 of the DSM-5 criteria for substance-use disorders when the substance is food, plus clinically significant impairment or distress (M=7.60, SD=2.06, range = 4–7). Participants without YFAS 2.0 food addiction (n=24) endorsed zero or one markers symptoms on the YFAS 2.0 (M=.17, SD=.38). The groups did not differ by age, income, or BMI (all ps>.34; Table 1).

Table 1.

Descriptive Characteristics for Participants by YFAS 2.0 Food Addiction

	Participants with YFAS 2.0 Food Addiction (n=20)	Participants without YFAS 2.0 Food Addiction (n=24)	Sig (p)
YFAS 2.0 Symptoms	M=7.60; SD=2.06	M=.17; SD=.38	<.001
BMI	M=34.57; SD=5.10	M=33.15, SD=5.81	.39
Hunger	M=55.35; SD=25.72	M=45.75; SD=25.62	.22
Age	M=29.90; SD=4.17	M=31.08; SD=3.93	.34
Income	Median=$30,000–39,999	Median=$30,000–39,999	.80
Race	White: 70.0% (n=14) African-American: 15.0% (n=3) Other: 15.0% (n=3)	White: 37.5% (n=9) African-American: 37.5% (n=9) Other: 25% (n=6)	.11

2.2.2. Hunger

Individuals rated their hunger immediately prior to the scan on a 100-point visual analog scale from 0 (not hungry at all) to 100 (never been hungrier). Hunger was not significantly associated with YFAS 2.0 food addiction categorization (p=.22), YFAS 2.0 symptoms, (p=.35) or BMI (p=.50). See Supplementary Materials Figure 1 for a boxplot distribution of hunger ratings by food addiction status.

2.2.3. BMI

BMI was calculated in the current study using bioelectrical impedance analysis via an InBody scanner and was not significantly associated with food addiction status (p=.39). Percent body fat, lean body mass, body fat mass, and visceral fat level were also assessed and did not differ by food addiction categorization (ps>.21). Thus, BMI was retained for use as a covariate in analyses.

2.2.4. fMRI Food Cue Reactivity Paradigm

Participants were asked to eat normally on the day of their appointment but refrain from eating or drinking anything besides water two hours prior to the appointment to encourage a state of moderate hunger during the scan. Prior to beginning the scan, participants confirmed that they adhered to the two-hour fast, and, on average, a moderate state of hunger was reported on the hunger VAS scale (Table 1). The food cue reactivity task was an event-related paradigm that assessed brain response to pseudo-randomized images of highly processed foods (n = 19) and minimally processed foods (n = 15). The food items were adapted from prior work examining which foods may be most associated with addictive-like eating indicators (30, 31) and varied on numerous nutritional characteristics, including caloric density, fat, sugar, carbohydrates, protein, and fiber (see Supplementary Materials Table 1 for the nutritional information of each food item). Household items (n = 17) were also included to provide a non-food comparison with similar color, complexity, and valence as the food cues. All cues were presented in color and on a solid white background. Stimuli were each presented for 4 seconds. Participants were asked to think about how much they wanted each item presented as if it were in front of them. Trials were separated by a jittered fixation cross (M duration = 4 secs, SD = 1.75 secs). Stimuli and fixation presentation durations were comparable to prior studies of food cue reactivity (36–38). Figure 1 details a sample trial sequence. In prior studies of drug cue reactivity in samples with substance-use disorders, reward-related neural responses seem to be elevated when the item is accessible (39, 40). Thus, prior to the task, participants were told they would get to consume some of the food items and take home some of the household items after the scan, which increased the accessibility of all cued conditions presented in the fMRI task. The paradigm lasted approximately 13.5 minutes and consisted of six runs, with 17 events in each run.

Figure 1.

Example of Event-Related fMRI Cue Reactivity Trials

2.2.5. Self-Reported Wanting

Following the fMRI task, participants were shown the same images of the food and household items presented in the cue reactivity paradigm on a lab computer. They were asked how much they typically wanted each item on a visual analog scale ranging from 0 to 100.

2.3. Data Analytic Plan

2.3.1. fMRI data acquisition

MRI images were acquired using a 3T GE Signa scanner with an 8 channel head coil located at the University of Michigan fMRI Laboratory. Visual stimuli were displayed using an LCD screen by Nordic Neuro Labs (Bergen, Norway). Functional T2*-weighted BOLD images were acquired using a reverse spiral sequence of 40 contiguous axial 3 mm slices (TR = 2000ms, TE = 30ms, flip angle = 90°, FOV = 22cm). Slices were prescribed parallel to the AC-PC line, and images were reconstructed into a 64×64 matrix. One structural image set was acquired: 2D T1 Axial Overlay (TR = 3170, TE = 24, flip angle = 111°, FOV = 22cm, slice thickness = 3.0mm, 43 slices, matrix = 256*192. 3D SPGR was acquired axially (flip angle = 15°, FOV = 25.6cm, slick thickness = 1mm, 156 slices, matrix = 256*256).

2.3.2. fMRI data preprocessing

fMRI data were preprocessed and analyzed using SPM12 (Wellcome Department of Imaging Neuroscience; Institute of Neurology, University College of London, London UK). Images were first skull-stripped using the Brain Extraction Tool in FSL (FMRIB Analysis Group, Oxford, UK) and then manually realigned to the AC-PC line in SPM. During preprocessing in SPM, anatomical data were segmented and normalized using DARTEL, resulting in a sample-specific template and individual-level deformation fields for application to the normalization step during functional data preprocessing. Functional images were time-acquisition corrected to the slice obtained at 50% of the TR, realigned to the scan immediately preceding the anatomical T1 image, and smoothed with a 6mm FWHM isotropic Gaussian kernel. For time series analysis, a high-pass filter (128s) removed low frequency noise and signal drifts. The artifact detection tools (ART) toolbox in SPM12 was used for identifying motion spikes. Movements greater than 2mm in any direction were considered excessive and entered as regressors in first-level analyses.

In order to investigate whether highly processed and minimally processed food cues vary in their ability to engage neural responses, we specified eight contrasts to compare the two food categories to each other and to a household item cue control condition: 1) Highly Processed Foods > Minimally Processed Foods, 2) Minimally Processed Foods > Highly Processed Foods, 3) Highly Processed Foods > Household Items, 4) Household Items > Highly Processed Foods, 5) Minimally Processed Foods > Household Items, 6) Household Items > Minimally Processed Foods, 7) All Foods > Household Items, and 8) Household Items > All Foods. At the individual-level, T-maps were constructed for comparison of activation within each participant for the eight contrasts.

For main effects, one-sample t-tests were utilized to evaluate the eight whole-brain contrasts in a single model. In order to assess whether individual differences emerged based on YFAS 2.0 food categorization, a mixed two-way ANOVA was run using a summary statistics approach in order to simultaneously account for the lack of independence of the betas that make up each contrast for each subject and the independence and unequal variance between the two participant groups (41). In this analysis, the within-subjects factor was the stimuli condition (highly processed food cues, minimally processed food cues, household item cues) and the between-subjects factor was participant group (YFAS 2.0 food addiction, control). Both hunger and BMI were entered as covariates in the ANOVA models, given that both variables have been associated with food cue responses in prior work (42–47). Whole-brain analyses were conducted after the binarized DARTEL-derived sample-specific gray matter mask was applied. An overall significance level of p<0.05 corrected for multiple comparisons across the gray matter-masked whole brain was calculated. This was accomplished by first estimating the intrinsic smoothness of the masked functional data with the spatial autocorrelation function (acf) option in the three-dimensional FWHM module in AFNI (Version AFNI_17.0.03). The acf parameters were then used in 10,000 Monte Carlo simulations of random noise at 3 mm³ through the gray matter masked data with the 3DClustSim module of AFNI. Simulation results indicated that activity surviving a threshold of P uncorrected < 0.001 with a minimum cluster of k ≥ 44 (one-sample t-tests) and at P uncorrected < 0.001 with k ≥ 50 (ANOVA models) was statistically significant corrected for multiple comparisons. Effect sizes (r) were calculated from the Z-value using the formula: Z/√n.

2.3.3. Self-Reported Wanting Ratings

Paired-sample t-tests were used to evaluate self-reported wanting ratings for highly and minimally processed foods and household items across all participants. One-way ANOVA tests were used to examine if individuals with versus without YFAS 2.0 food addiction differed in self-reported wanting for highly processed foods, minimally processed foods, and household items. Then, correlational analyses were conducted to investigate if self-reported wanting ratings were associated with neural responses, using parameter estimates extracted from SPM.

3. Results

Table 1 presents the descriptive statistics of the participant groups.

3.1. Main Effects of Highly Processed Versus Minimally Processed Foods and Foods Versus Household Items

3.1.1. Highly Processed versus Minimally Processed Foods

Whole brain analyses revealed that highly processed versus minimally processed foods resulted in lower BOLD response in the left fusiform gyrus (r=.78), right calcarine cortex (rs>.58), and right inferior temporal gyrus (r=.72) (Table 2). No regions had significantly greater response for highly processed versus minimally processed foods.

Table 2.

Main Effects for fMRI Cue Reactivity Contrasts

	x	y	z	# voxels in cluster	Activation cluster Z	Effect Size Z-value (r)
Minimally processed>Highly processed
Fusiform gyrus	−22	−78	−12	1198	5.15	.78
Calcarine cortex	18	−94	4	1035	5.60	.84
	0	−60	12	187	3.88	.58
Inferior temporal gyrus	44	−64	−8	140	4.76	.72
Household>Highly processed
Calcarine cortex	16	−92	4	4492	>8	>1.21
Inferior frontal gyrus	40	32	12	55	3.79	.57
Superior frontal gyrus	34	−8	60	54	3.99	.60
Minimally processed>Household _Middle temporal gyrus	−54	−52	0	95	4.22	.64
Household>Minimally processed
Calcarine cortex	−10	−94	−10	967	7.18	1.08
	16	−92	2	715	6.37	.96
Insula	−36	0	−2	437	4.61	.69
Inferior frontal gyrus	−48	40	10	187	4.39	.66
Middle frontal gyrus	44	46	2	183	3.67	.55
Inferior parietal lobe	−24	−66	42	53	3.67	.55
All Food>Household Middle temporal gyrus	−52	−54	2	72	3.66	.55
Household>All Food
Calcarine cortex	16	−92	4	1795	7.70	1.16
Lingual gyrus	−10	−92	−12	1690	7.62	1.15
Insula	−42	10	0	425	4.28	.65
	50	14	−4	110	3.77	.57
Inferior frontal gyrus	42	34	12	240	3.97	.60
	−44	44	6	152	3.85	.58
Middle occipital lobe	−24	−66	38	60	3.80	.57
Inferior parietal lobe	−54	−34	48	46	3.51	.53

3.1.2. Highly Processed and Minimally Processed Foods versus Household Items

Highly processed foods versus household items, resulted in less BOLD response in the right calcarine cortex (r>1.21), right inferior frontal gyrus (r=.57), and right superior frontal gyrus (r=.60). No regions demonstrated higher response to highly processed foods versus the household items. Minimally processed foods versus household items, resulted in greater BOLD response in the left middle temporal gyrus (r=.64) and less response in bilateral calcarine cortex (rs>−.96) left insula (r=.69), left inferior (r=.66) and right middle (r=.55) frontal gyri, and left inferior parietal lobe (r=.55). All foods versus household items resulted in greater activation in the left middle temporal gyrus (r=.55) and less activation in the right calcarine cortex (r=1.16), left lingual gyrus (r=1.15), bilateral insula (rs>.57), bilateral inferior frontal gyrus (rs>.58), left middle occipital lobe (r=.57), and left inferior parietal lobe (r=.53).

3.2. Differences in Neural Food Cue Reactivity between Individuals with YFAS 2.0 Food Addiction and Controls

3.2.1. Highly Processed Versus Minimally Processed Foods

Figure 2 illustrates the significant interaction between participant group and neural response in the right superior frontal gyrus to highly versus minimally processed food cues (r=.57). Descriptively, individuals with YFAS 2.0 food addiction exhibited a modest elevated BOLD response in the right superior frontal gyrus to highly processed food cues but robust diminished activations in this region to minimally processed food cues. Participants in the control group showed the opposite pattern of responses, with large diminished activation for highly processed food cues and elevated responses to minimally processed food cues.

Figure 2.

BOLD activity in the superior frontal gyrus (20, 26, 50; Z=3.78; k=54) during the fMRI food cue reactivity task (highly processed (HP) versus minimally processed (MP) foods) with greater activation in participants with YFAS 2.0 food addiction versus controls, with the bar graphs of parameter estimates from that peak. Since the same parameter estimates that created the peak were extracted from it, the plotted values are for descriptive purposes only and do not include error bars (Kriegeskorte et al., 2009).

3.3. Self-Reported Wanting and Associations with Neural Responses

Across all participants, highly processed foods were associated with greater self-reported wanting than minimally processed foods (p=.02) and household items (p<.001). Minimally processed foods were related to elevated self-reported wanting ratings compared to household items (p<.001). There were no differences by YFAS 2.0 food addiction categorization for self-reported wanting for the highly or minimally processed foods or household items (ps > .10). In addition, there were no significant associations of wanting with any of the significant peak activations that emerged from the main effects or ANOVA fMRI analyses (ps >.13).

4. Discussion

This study investigated whether YFAS 2.0 food addiction was differentially associated with neural responses to food cues in a sample of overweight and obese women. In an fMRI food cue reactivity task, there was a significant interaction between participant group and neural response in the right superior frontal gyrus to highly versus minimally processed food cues. This interaction appeared to be driven by the robust diminished activation in the superior frontal gyrus to minimally processed foods in persons with YFAS 2.0 food addiction, whereas control participants exhibited elevated activation in this region. There were no regions where control participants showed greater activation to highly processed food cues than persons with YFAS 2.0 food addiction. There were also no differences by YFAS 2.0 food addiction in self-reported wanting ratings for highly processed or minimally processed food cues or household item cues, and no associations of these ratings with neural activation. Lastly, across all participants, there was more widespread activation to household object cues, relative to food cues.

4.1. Unique Neural Responses to Highly Processed Food Cues in Food Addiction

The superior frontal gyrus is an executive control region that has broadly been implicated in stimulus selection, working memory, and decision making, which are processes that may be influenced by dopaminergic responses to reward appraisal (48, 49). Interestingly, the functioning of executive control regions has been shown to be impaired by overeating, such that reward-related decision making (enhanced subjective valuation based on expected reinforcement (50)), may be altered (51). Given that YFAS 2.0 food addiction is characterized by compulsive consumption of highly processed foods (12), it may be that this pattern of eating behavior broadly disrupted reward-related decision making for food and household item cues. Persons with YFAS 2.0 food addiction, compared to controls, may have exhibited modest, increased activation in the superior frontal gyrus for highly processed food cues due to enhanced reward-related decision making for these items. Notably, these participants also exhibited robust diminished activation in this region relative to controls for minimally processed food cues, perhaps because these foods have demonstrated little association with addictive-like eating (30, 31). In contrast, participants without YFAS 2.0 food addiction exhibited the opposite pattern of responses in the superior frontal gyrus for highly and minimally processed food cues compared to those with addictive-like eating, suggested greater reward-related decision making associated with minimally processed food cues. Thus, the present findings suggest that individuals with YFAS 2.0 food addiction may exhibit differential responses to food cues in a region associated with reward-related decision making, which may be maintained by chronic compulsive consumption of highly processed foods (51).

Further, recent research has specifically investigated the function of the superior frontal gyrus in the context of substance-use disorders, which may also aid in understanding how this region may be relevant in the context of the neural correlates of food cue reactivity among those with YFAS 2.0 food addiction. Prior work has observed elevated responses in the superior frontal gyrus for drug cues have been associated with cue-induced craving for persons with a substance-use disorder and indicative of chronic, relapsing patterns of drug consumption (52, 53). In addition, the superior frontal gyrus may be particularly susceptible to addiction-related neuroplastic changes, as this region has been associated with the modulation of drug cravings (54). While the magnitude of the elevated responses in the superior frontal gyrus to highly processed food cues was modest for individuals with YFAS 2.0 food addiction, this significantly contrasted a large diminished response to highly processed food cues in control participants. In parallel with prior work in in substance-use disorders (52, 53), modest, increased responses to highly processed food cues in this region among individuals with YFAS 2.0 may suggest elevated cue-induced craving responses for these foods. Interestingly, this was not mirrored in the self-reported wanting ratings. Highly processed foods were reported to be wanted similarly by all participants, suggesting that neural responses to highly processed food cues differentiated addictive-like eating to a greater degree than subjective self-reported wanting ratings.

In addition, the superior frontal gyrus exhibited diminished responses to minimally processed food cues in persons with YFAS 2.0 food addiction but elevated activations in the control group. Our results suggest that participants with YFAS 2.0 food addiction showed less engagement in this region implicated in craving for minimally processed foods. Notably, individuals with a substance-use disorder seem to exhibit decreased engagement of this and other regions implicated in craving and reward to non-drug stimuli (55, 56), suggesting selective, drug-specific responding. Thus, minimally processed foods may not mimic these drug-specific responses for individuals with YFAS 2.0 food addiction, which parallels self-report findings in prior work showing that these foods have demonstrated little association with addictive-like eating behaviors (30, 31). The current findings are also consistent with a prior neuroimaging study of subclinical food addiction (26), where participants with increased symptoms of YFAS food addiction exhibited significantly diminished responses in the caudate, a region implicated in reward motivation, in response to a tasteless cue compared to a highly processed food cue. Thus, individuals with food addiction may exhibit decreased activation in a region implicated in drug craving for minimally processed foods. In contrast, those without food addiction exhibited less craving-related responses for highly processed, compared to minimally processed, foods. This may complement self-report studies demonstrating that persons without food addiction endorse being less likely to experience problems with highly processed foods (30, 57).

Overall, the present findings provide further evidence that food addiction represents a unique phenotype within overweight and obesity, marked by modest elevated responses to highly processed food cues and robust diminished neural responses to minimally processed food cues in a region implicated in cue-induced craving, when compared to individuals with overweight/obesity but without food addiction. Notably, this also provides support for highly processed foods being more implicated in addictive-like neural responses for persons with food addiction, relative to minimally processed foods, as observed in prior self-report studies (30, 31, 57). These data collectively may suggest that intervention approaches used to reduce cue-induced craving in substance-use disorders may also be effective for the treatment of addictive-like consumption of highly processed foods, such as cue-exposure treatment, which has been effective at reducing responses in the superior frontal gyrus for drug cues (58).

4.2. The Absence of Widespread Reward Responses to Highly Processed Food Cues in Main Effects Analyses

Mesocorticolimbic reward regions have been elevated in response to drug cues for persons with a substance-use disorder versus healthy controls (21–24, 59) and highly processed food cues for persons with obesity, relative to those of healthy weight (18, 19). Further, in the only other fMRI study of YFAS food addiction, although participants endorsed only subclinical levels, participants exhibited greater reward responses when anticipating the highly processed food reward (26). Thus, it may have been expected that similar reward activity would have been observed in the current sample. However, main effect analyses revealed greater activation to household item cues compared to food cues in mid-insula and insula/fronto-operculum regions that have been implicated in interoceptive awareness, craving, and visuospatial-attention in prior studies of drug (60, 61) and food (36, 62, 63) cue reactivity. Thus, the elevated activation in these regions to household (relative to food cues) was unexpected and contradictory to study hypotheses. The mid-insula and insula/fronto-opercular peaks are implicated in functions other than drug/food cue reactivity, including general awareness, motor planning, and responding to uncertainty (15, 64–67). Thus, activation in these regions in response to household versus food cues may be reflecting these non-food related processes. Yet, given the unexpected nature of these findings and the contrast to prior literature (36, 60–63) studies of food cue reactivity, However, the elevated activation main effect findings contrast prior fMRI drug and food cue reactivity studies and this study’s hypotheses. Despite the lack of precedent in the literature for the control stimuli to trigger greater activation in regions associated with reward than food cues, prior studies may provide insight into limitations of the current methodological approach that may have contributed to the unexpected main effects.

Although we did not directly assess ambivalence in the current study, this has been found to play a key role in predicting neural responses to cues in addictive disorders. In drug cue reactivity studies of substance-use disorders, ambivalence has been an important moderator of neural responses, with greater ambivalence being related to diminished reward-related activation for drug cues (68, 69). The current sample and experimental paradigm may have increased ambivalence in a manner that may have influenced the findings in a similar manner. Rather than including a control group of healthy individuals of normal weight akin to prior studies that have found widespread reward activation to food cues (17, 18, 20, 70–72), the present sample investigated women all whom had overweight or obesity and half with clinical levels of YFAS 2.0 food addiction. Prior studies of individuals with overweight/obesity have found significant ambivalence towards highly processed foods, marked by simultaneous craving for their hedonic nature but a desire to avoid consuming them due to their obesogenic properties (73, 74). Ambivalence may have been especially exacerbated by the current approach of emphasizing the availability of the foods after the scan, which has successfully elevated the reward potential of drugs of abuse in prior studies of active drug users (75, 76). However, unlike active substance users who are freely using their drug of choice, women with overweight and obesity may not be freely consuming highly processed foods without ambivalence, given the greater likelihood of weight loss attempts (77) and weight discrimination (78, 79) in this population. Ambivalence during the scan may also be evident by the contrast of diminished reward-related neural responses to highly processed food cues across all participants, despite higher self-reported wanting for these foods. Highlighting the impending availability of the food items during the fMRI task instructions, and perhaps subsequently increasing ambivalence, may have disrupted the elevated typical hedonic wanting for these foods that was self-reported by participants. Thus, methods that effectively override ambivalence and prime food wanting, such as food administration during the scan, as done in the prior study of subclinical YFAS food addiction (26), may warrant consideration in future studies elucidating the rewarding nature of highly processed foods in women with overweight/obesity and/or YFAS 2.0 food addiction.

Ambivalence may also have contributed to the unexpected findings regarding greater activation the mid-insula and insula/fronto-operculum regions to household items compared to food cues. Unlike food, household items (e.g., candles, pens, notebooks) have little association with negative weight-related consequences. Thus, unlike food, household items may have been less likely to trigger ambivalent feelings and may have been pleasant to consider taking home after the scan. This may also provide insight into the unexpected finding that individuals with YFAS 2.0 food addiction exhibited elevated responses in the superior frontal gyrus to the household item cues compared to control participants. A key feature of the food addiction phenotype is persistent attempts to cut back on highly processed foods, and thus, ambivalence may be higher for participants with YFAS 2.0 food addiction relative to control participants. Household items may have been more reinforcing for persons with YFAS 2.0 food addiction (relative to control participants) because these items may have been less likely to trigger heightened food-related ambivalence associated with addictive-like eating. However, these results were entirely unexpected, and future research is necessary to evaluate if differences in ambivalence towards food and household items, particularly for those with YFAS 2.0 food addiction, may have contributed to the current findings.

4.3. Limitations and Future Directions

While the present work was novel in exploring a food addiction phenotype within overweight and obesity, there were several limitations that should be considered in the future. First, while this approach highlighted neural responses that may indicate elevated craving for highly processed foods in persons with YFAS 2.0 food addiction, additional behavioral indexes during the fMRI scan (e.g., craving/wanting ratings) of highly processed foods may have increased engagement with the cue reactivity paradigm to a greater degree than imaginal wanting while passively viewing the cues. Second, there was no assessment of ambivalence, which may have moderated main effects. Ambivalence towards highly processed foods emerges from the current work as an important construct to measure in future studies with samples of individuals all with elevated BMI and/or YFAS 2.0 food addiction. Third, emphasizing the availability of the cued food items may have increased ambivalence, particularly in those with YFAS 2.0 food addiction, in a manner that diminished individual differences in more classically observed reward regions (e.g., striatum). As such, future research may implement methodology that has been effective in enhancing reward reactivity for drug cues (e.g., preloading prior to cue reactivity task) (80, 81). Relatedly, since the present study was conducted in a sample of women, future research is needed to determine whether the findings may generalize to men. Further, the current approach cannot rule out the possibility that the findings may be driven by an interactive effect between food addiction and overweight/obesity, as no participants with normal BMI and with or without food addiction were included. Thus, additional work examining the associations of food addiction with food cue reactivity may consider recruiting a more heterogeneous sample.

4.4. Conclusions

The current findings suggest that women with overweight or obesity and YFAS 2.0 food addiction seem to exhibit unique neural responses to highly processed and minimally processed food cues compared to women with only overweight or obesity. Females with YFAS 2.0 food addiction, relative to those without, exhibited modestly elevated responses to highly processed food cues and diminished neural responses to minimally processed food cues in a region implicated in cue-induced craving in prior studies of drug cue reactivity in active substance users (21, 23, 52, 53, 82). Thus, persons with food addiction may experience greater cue-induced craving for highly processed foods, which may contribute to the greater addictive-like consumption of these foods that have been reported in prior self-report studies (30, 31, 57). The present work provided preliminary evidence that the food addiction phenotype is differentially associated with reward-related neural cue reactivity for highly processed foods. This novel study represents an important step in elucidating mechanisms that may drive an addictive-like process to highly processed foods for some individuals and how this may contribute to the understanding of obesity.

Supplementary Material

1

Supplementary Materials Table 1. Nutritional Characteristics for Foods Included in the Food Cue Reactivity Task

Note: g=grams; mg=milligrams

Supplementary Materials Figure 1. Boxplot Distributions of Hunger Ratings by Food Addiction Categorization

Table 3.

Differential Associations of Neural Activation by Food Addiction Categorization

	x	y	z	# voxels In cluster	Activation cluster Z	Effect Size Z-value (r)
FA>Control, Highly processed>Minimally processed Superior frontal gyrus	22	28	50	88	3.79	.57

Highlights.

• Functional magnetic resonance imaging was used to investigate whether neural responses to highly and minimally processed food cues differed based on YFAS 2.0 food addiction status in a sample of women with overweight or obesity.

• Persons with, relative to without, food addiction exhibited modest elevated responses to highly processed food cues and robust diminished neural responses to minimally processed food cues in a region associated with cue-induced craving (superior frontal gyrus).

• Greater neural response in this region has been observed in persons with a substance-use disorder in response to relevant drug cues.

• Within overweight and obesity, individuals with food addiction may experience greater craving-related neural activation for highly processed foods, which may contribute to the addictive-like consumption of these foods that has been reported in prior self-report studies.