If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Red de Trastornos Adictivos, Universidad de Granada, GranadaPsychiatry Department, Bellvitge University Hospital, Bellvitge Biomedical Research Institute-IDIBELL, and Centro de Investigación Biomédica en Red de Salud Mental, Barcelona
Red de Trastornos Adictivos, Universidad de Granada, GranadaInstitute of Neuroscience F. Oloriz, Universidad de Granada, Granada, SpainSchool of Psychological Sciences and Monash Institute of Cognitive and Clinical Neurosciences, Monash University, Melbourne, Australia
The food addiction model proposes that obesity overlaps with addiction in terms of neurobiological alterations in the striatum and related clinical manifestations (i.e., craving and persistence of unhealthy habits). Therefore, we aimed to examine the functional connectivity of the striatum in excess-weight versus normal-weight subjects and to determine the extent of the association between striatum connectivity and individual differences in food craving and changes in body mass index (BMI).
Forty-two excess-weight participants (BMI > 25) and 39 normal-weight participants enrolled in the study. Functional connectivity in the ventral and dorsal striatum was indicated by seed-based analyses on resting-state data. Food craving was indicated with subjective ratings of visual cues of high-calorie food. Changes in BMI between baseline and 12 weeks follow-up were assessed in 28 excess-weight participants. Measures of connectivity in the ventral striatum and dorsal striatum were compared between groups and correlated with craving and BMI change.
Participants with excess weight displayed increased functional connectivity between the ventral striatum and the medial prefrontal and parietal cortices and between the dorsal striatum and the somatosensory cortex. Dorsal striatum connectivity correlated with food craving and predicted BMI gains.
Obesity is linked to alterations in the functional connectivity of dorsal striatal networks relevant to food craving and weight gain. These neural alterations are associated with habit learning and thus compatible with the food addiction model of obesity.
). In this context, recent theories have proposed a food addiction model of obesity, by which sensitization of the brain centers that represent reward and habits (i.e., the striatum) can lead to food craving, inability to cut down food intake, and weight gain (
The food addiction model is grounded in neuroimaging evidence showing that obese individuals display increased food cue-evoked activation in cortical-striatal regions that code food-related reward value (i.e., ventral striatum) and hedonic properties (i.e., insula/somatosensory cortices) and food choices (i.e., medial prefrontal/orbitofrontal cortex) (
). These regional alterations are plausibly associated with abnormalities in a broader network between the striatum and prefrontal regions representing food value. For example, neuroimaging studies have shown that functional connectivity between the ventral striatum and the medial prefrontal cortex correlates with external food sensitivity in healthy samples (
). Therefore, dorsal striatal neuroadaptations have been implicated in the transition between incentive-based and habit-based control of behavior. Hence, greater involvement of the dorsal striatum is predicted as food intake becomes addictive or compulsive (
Altogether, ventral and dorsal striatum networks are relevant to the pathophysiology of obesity and to the association between obesity and addiction-related characteristics, such as craving and persistent high-calorie food intake. We aimed to contrast the functional connectivity of ventral and dorsal striatum networks in excess-weight versus normal-weight participants and to determine the association between functional connectivity in the striatum and individual differences in food craving and weight gain. We applied a seed-based resting-state functional connectivity approach to assess ventral and dorsal striatum networks (
). Resting-state fluctuations reflect cognitive and emotional biases that contribute to shape individuals’ preferences; thus, striatal connectivity measures may have predictive validity in relation to craving and food intake (
). We hypothesized that excess-weight participants compared with normal-weight control subjects would show increased functional connectivity in the ventral and dorsal striatum. Increased dorsal striatal connectivity would be associated with greater food craving and weight gain.
Methods and Materials
Forty-two subjects with excess weight (BMI >25) and 39 subjects with normal weight enrolled in the study. Participants were recruited via general hospitals and community advertisement (i.e., local press, radio, and social media) and enrolled if they were aged 18 to 45 and had BMI >18. Exclusion criteria were 1) history of brain injury or diseases impacting the central nervous system; 2) history of substance use, major depression, or psychosis; 3) self-reported use of psychotropic medication; and 4) morbid obesity (BMI ≥ 40). The two groups had similar sociodemographic characteristics (Table 1). The Universidad de Granada Human Research Ethics Committee approved the study, and all participants provided informed consent.
Table 1Demographics and Clinical Characteristics of the Study Groups
Participants were involved in two assessment sessions. At baseline, they 1) were measured and weighed to calculate BMI via an automated scale; 2) underwent a functional magnetic resonance imaging (fMRI) scan; 3) rated their food craving immediately after fMRI scan; and 4) had a 30-minute diet counseling session with a professional dietitian who provided specific strategies to lose weight (i.e., excess-weight group only). At 12-week follow-up, excess-weight participants (n = 28, 67% of the original sample) were reassessed to calculate change in BMI relative to baseline. Twelve weeks is the standard benchmark to assess the outcome of weight loss interventions (
All participants were scanned at the same time of the day (4:00 to 6:00 PM) and after lunch, which is typically between 2:00 and 4:00 PM. Prescanner ratings of hunger (0–100 visual analog scale [VAS]) did not differ between groups (Table 1). Participants underwent a 6-minute resting-state scan. They were instructed to lie still with eyes closed. We used a 3.0 Tesla clinical magnetic resonance imaging scanner, equipped with an eight-channel phased-array head coil (Intera Achieva Philips Medical Systems, Eindhoven, The Netherlands). A T2*-weighted echo-planar imaging was obtained (repetition time = 2000 ms, echo time = 35 ms, field of view = 230 × 230 mm, 96 × 96 pixel matrix; flip angle = 90º, 21 4-mm axial slices, 1-mm gap, 180 whole-brain volumes). The sequence included four initial dummy volumes to allow the magnetization to reach equilibrium.
Participants viewed six photographs of highly appetizing foog., cheesecake, chocolate), and were instructed to rate their level of craving using VAS (range 0–100)." is correct as edited. -->d stimuli, all rich in sugar and fat content (e.g., cheesecake, chocolate), and were instructed to rate their level of craving using VAS (range 0–10). The dependent measure was the mean score of the six VAS ratings. To increase the task’s validity, all participants were pre-exposed to these foods in a catered tasting session conducted 1 week before the scan (Supplemental Figure S1).
Changes in BMI were computed by subtracting baseline BMI from follow-up BMI, and thus positive values reflected weight gain. This change index is a standard outcome measure in obesity treatment research (
) were also assessed (see detailed procedures in the Supplement).
Imaging Analyses and Preprocessing
The functional imaging data were processed and analyzed using MATLAB version R2008b (The MathWorks Inc., Natick, Massachusetts) and statistical parametric mapping software (SPM8; The Welcome Department of Imaging Neuroscience, London, United Kingdom). Preprocessing steps involved motion correction, spatial normalization, and smoothing using a Gaussian filter (full width at half maximum 8 mm). Data were normalized to the standard SPM echo planar imaging (EPI) template and resliced to a 2-mm isotropic resolution in Montreal Neurological Institute space. We compared both study groups for potential differences in movement for translations and rotations and found no significant differences (mean total movement [SD], normal-weight control subjects = .31 [.18], excess-weight participants = .31 [.29], p = .34).
), ventral and dorsal striatal subregions were distinguished using z < 7 mm as a marker for the ventral caudate (VC)/nucleus accumbens, z > 7 mm as a marker for dorsal caudate, and z = 2 as the boundary between the dorsal and ventral putamen per hemisphere. Respective seeds of interest were placed in VC (corresponding approximately to the nucleus accumbens) (x = ±9, y = 9, z = −8), ventral rostral putamen (VP) (x = ±20, y = 12, z = −3), dorsal caudate (DC) (x = ±13, y = 15, z = 9), and dorsal caudal putamen (x = ±28, y = 1, z = 3) using 3.5-mm-radius spheres. We included in the model two intermediate seeds (of no interest) located in the ventral caudate superior (x = ±10, y = 15, z = 0) and dorsal rostral putamen (x = ±25, y = 8, z = 6) to replicate the fine striatal parcellation method, and no seed placements were made in the globus pallidus, substantia nigra, or subthalamic nucleus considering the spatial data resolution and the smoothing.
First-level (single subject) maps were estimated in whole-brain SPM8 linear regression analyses for each striatal seed region (
) together with nuisance signals as predictors of interest and no interest. Nuisance signals included six head-motion parameters (three translations and three rotations) and time courses representing mean signal fluctuations in white matter, cerebrospinal fluid, and the entire brain. Separate first-level analyses were carried out for right and left hemisphere striatal regions. A high-pass filter (128 seconds) was used to remove low-frequency drifts. Contrast images were generated for each subject by estimating the regression coefficient between all brain voxels and each seed’s time series and were then included in separate two-sample models to assess within- and between-group effects. All imaging results were considered significant with a cluster of 1960 mm3 (245 voxels) at a height threshold of p < .005, which satisfied the familywise error rate correction of pfamilywise error < .05 according to Monte Carlo simulations using Alphasim within the REST toolbox (http://www.restfmri.net) (
) with a cluster connection radius of 5 mm, 12-mm full width at half maximum smoothness, and incorporating a gray matter mask volume of 128.190 voxels (2 × 2 × 2 mm). The selected threshold was deemed appropriate to control for type I error, as the targeted striatal networks have been previously defined and have anatomical specificity (
Associations With Food Craving and Prediction of BMI Change
We conducted correlations between the functional connectivity maps of the striatum and food craving in SPM8. We conducted these analyses in each group separately, as we were specifically interested in individual differences in this relationship within the excess-weight group. Analyses were considered significant at a height threshold of p < .005, 1960 mm3 (245 voxels) whole brain. A multiple regression analysis was conducted in SPSS version 22.0 (IBM Corp., Armonk, NY) to examine if functional connectivity associated with craving predicted BMI change. This analysis was thresholded at p ≤ .01 (at least 10% prediction).
Functional Connectivity in Ventral and Dorsal Striatum
Within-group positive and negative functional connectivity maps of the ventral and dorsal striatum networks overlapped with previously described neurofunctional maps of these networks (
Excess-weight participants compared with normal-weight control subjects showed increased functional connectivity between the VC seed and the medial prefrontal cortex and between both ventral striatum seeds (VC and VP) and the parietal cortex, including posterior cingulate, angular gyrus, and somatosensory regions. Moreover, excess-weight participants showed increased anticorrelation between the VC seed and the dorsal anterior cingulate cortex and the posterior insula and between the VP seed and the lateral orbitofrontal cortex.
Excess-weight participants showed increased functional connectivity between the DC seed and the somatosensory cortex. Moreover, excess-weight participants showed increased anticorrelation between the DC seed and the cerebellum. There were no group differences in the dorsal caudal putamen seed at the selected threshold.
Correlations With Food Craving
In excess-weight participants, food craving was positively associated with functional connectivity between the DC seed and the somatosensory cortex (x = 46, y = −32, z = 62, t = 3.8, p < .005, 265 voxels). In normal-weight participants, in turn, food craving was positively associated with functional connectivity between the VP seed and the orbitofrontal cortex (x = −32, y = 50, z = −2, t = 4.8, p < .005, 459 voxels) (Figure 2). Fisher’s tests of between-group differences were not significant (DC-somatosensory: F = 1.32, p = .09; VP-orbitofrontal cortex: Fisher’s F = 1.06, p = .14). Additional correlations (i.e., functional connectivity patterns that do not map on between-group differences) are reported in Supplemental Table S2.
To further explore the link between BMI-related variation in striatum networks and food craving, we examined this association within subgroups with obesity (BMI ≥ 30, n = 21), overweight (BMI > 25 < 30, n = 21), and normal weight. Thus, we extracted the eigenvariate signal from the peak voxel in the orbitofrontal cortex cluster linked to the VP seed and food craving in normal-weight control subjects and the eigenvariate signal from the peak voxel in the somatosensory cortex cluster linked to the DC seed and food craving in excess-weight participants. We found that the positive association between VP-orbitofrontal functional connectivity and food craving was lower within participants with greater BMI (i.e., normal weight, r = .476, p = .001; overweight, r = .313, p = .044; obese, r = .141, p = .373) (Supplemental Figure S2). Conversely, the positive association between DC-somatosensory cortex functional connectivity and food craving was higher in participants with greater BMI (i.e., normal weight, r = .073, p = .527; overweight, r = .326, p = .035; obese, r = .378, p = .014).
Prediction of BMI Change
On average, BMI changed from 30.51 (SD = 3.63) to 29.95 (SD = 3.43) among excess-weight participants (1.96% change). The two networks associated with craving were included in a multiple regression model to predict the change in BMI. The functional connectivity between the VP seed and the orbitofrontal cortex was not associated with BMI change (FChange 1,53 = .001, p = .982, R2 = .000), yet inclusion of the functional connectivity between the DC seed and the somatosensory cortex showed statistically significant effects (FChange 1,53 = 6.787, p = .01, R2 = .114). DC-somatosensory functional connectivity was positively associated with BMI change, accounting for 11% of BMI change (Figure 3).
Sensitivity and Post Hoc Analyses
To examine if longitudinal associations between functional connectivity and BMI gain were driven by baseline correlations, we conducted bivariate correlations between the functional connectivity associated with food craving and BMI at baseline and BMI change. The functional connectivity between the DC seed and the somatosensory cortex was only associated with BMI change (r = .33, p = .01) but not with baseline BMI (r = −.11, p = .33).
To better delineate the implications of the connectivity findings, we ran post hoc correlations between the striatal connectivity patterns and cognitive measures of impulsivity and compulsivity. The functional connectivity between the DC seed and the somatosensory cortex was significantly associated with the measure of compulsivity (i.e., reversal learning perseveration error rate) in excess-weight participants (x = 52, y = −24, z = 44, t = 3.6, p < .005, 17,636 voxels). There were no significant correlations with impulsivity. Additional correlations between cognitive measures and functional connectivity patterns that did not overlap with the between-group differences in striatal connectivity are reported in the Supplemental Tables S3 and S4.
We found that individuals with excess weight display increased functional connectivity between the ventral striatum and medial prefrontal and parietal cortices and between the dorsal striatum and the somatosensory cortices. They also show reduced functional connectivity between the ventral striatum and the dorsal anterior cingulate cortex, the insula, and the lateral orbitofrontal cortex. Dorsal striatal connectivity positively correlates with food craving and predicts BMI gains after 12 weeks. These findings are consistent with the food addiction model, which proposes that obesity is associated with neural adaptations in the striatum and particularly within the dorsal striatum network (
Ventral striatum findings are consistent with previous studies showing functional connectivity differences between obesity and normal weight via complementary approaches. A recent study using a region-of-interest based approach showed increased functional connectivity between the ventral caudate and the medial frontal cortex in excess-weight individuals (
). Therefore, more studies are needed to establish the neuropharmacologic underpinnings of these findings and the overlap between obesity and different types of addiction (e.g., psychostimulant vs. opiate). Increased connectivity between the ventral striatum and the posterior cingulate cortex and the angular gyrus is consistent with Kullmann et al. (
), which suggested that abnormal connectivity in the brain default network may contribute to overeating through an imbalance between reward-affective and cognitive processes. Decreased connectivity between the ventral caudate and the dorsal anterior cingulate cortex/insula accords with previous findings showing abnormal recruitment of brain regions involved in interoception in obese subjects (
). Moreover, the link between the dorsal striatum network and food craving was significant in the excess-weight group and greater within participants with higher BMI values, in agreement with the notion of an addictive dimensionality of obesity (
). The finding that functional connectivity between the dorsal caudate and the somatosensory cortex longitudinally predicts weight gain also accords with previous neuroimaging findings showing that increased dorsal caudate activation is associated with BMI gains in obese subjects (
). Our functional connectivity approach extends these findings, demonstrating that a broad network linking the dorsal caudate with sensory/gustatory regions is implicated in future weight gain. Notably, this network is implicated in the coding of the energetic value of foods (
) and in habit learning, as indicated by correlations with reversal learning. Therefore, these findings may contribute to explain how ingrained preferences for high-calorie food can trigger eating habits beyond homeostatic needs (
). Future studies are needed to formally evaluate this mechanism.
We conclude that excess weight is associated with striatocortical functional connectivity alterations that are consistent with a food addiction model. Specifically, the associations with food craving and BMI gain during diet resonate with the DSM substance use disorders’ criteria of craving and inability to cut down drug seeking (i.e., in this case food seeking) behavior (
). Our results should be appraised in the context of limitations. The cross-sectional design does not allow us to determine if striatum network alterations are a consequence of excessive weight or premorbid vulnerability factors. Further longitudinal research is necessary to understand the causal link between the brain reward system and obesity. The spatial resolution of fMRI did not allow us to differentiate the connectivity of medial versus lateral territories of the dorsal striatum. This dissociation is potentially relevant for human obesity, as preclinical data have shown that habit-based behaviors are supported by the anterior dorsolateral striatum (
). Our correlation analyses showed significant correlations between dorsal striatal connectivity and food craving within the excess-weight group, but correlations did not significantly differ between groups (i.e., Fisher’s test) and thus the specificity of this finding must be interpreted with caution. Our longitudinal analyses testing the clinical meaningfulness of functional connectivity were conducted in a subgroup representing only 67% of the original sample. Nonetheless, this attrition rate is consistent with that reported in meta-analytic research on obesity interventions (
), and participants in the follow-up subgroup did not significantly differ from the main sample in antecedent variables.
Acknowledgments and Disclosures
This study was funded by the Project Grant NEUROCOBE (Grant No. HUM-6635) from the Andalusian Council of Innovation, Science and Industry (Principal Investigator: Antonio Verdejo-Garcia). Dr. Contreras-Rodriguez is currently funded by a Sara Borrell postdoctoral fellowship (Grant No. CD14/00246) from the Carlos III Health Institute.
The authors report no biomedical financial interests or potential conflicts of interests.
An alarming number of people worldwide are overweight or obese, posing significant health risks and leading to substantial individual and societal costs. The wide availability of highly palatable and calorically dense foods along with increasingly sedentary lifestyles make it easy to take in more calories than we burn. This leads to a calorie surplus that, over time, produces weight gain. It seems a simple solution to this epidemic should be to eat less and/or exercise more. However, the reality, as many dieters know firsthand, is not so simple.