*3.3. The E*ff*ect of Food Cues on Attention Bias and Brain Potentials*

ANOVA of the Stroop attention bias in response to food cues in the Food Stroop task revealed a significant 2-way group \* image type interaction (F (2,56) = 3.46, *p* = 0.04) (Figure 3). This interaction was statistically significant when controlling for VAS-hunger and PANAS scores (mean scores: 6.62 and 20.1, respectively). Post-hoc tests indicated a significant reduction of the Stroop bias following HCF compared to the NF image, which was observed in the FAOB but not in the other two groups (*p* = 0.05; Figure 3).

Nonparametrical permutation tests indicated differences between the three groups in brain potentials in response to food cues. There was an amplitude change induced by the food cues (HCF vs. NF), in the LPPb but not in the LPPa, in the frontal (F = 4.09 (*<sup>n</sup>* <sup>=</sup> 52), *p* = 0.02) and right occipital (F = 3.78 (*<sup>n</sup>* <sup>=</sup> 52), *p* = 0.02) but not in the left occipital ROI (F = 1.40 (*<sup>n</sup>* <sup>=</sup> 52), *p* = 0.26) (Figure 4). Post-hoc contrasts revealed that the FAOB group had a lower food cue-induced amplitude change in the LPPb compared with the NFAOB group, in the frontal (*t* = 2.84 (*<sup>n</sup>* <sup>=</sup> 37), *p* = 0.0058) and right posterior (*t* = 2.62 (*<sup>n</sup>* <sup>=</sup> 37), *p* = 0.0013) ROI. Neither the FAOB nor the NFAOB groups statistically differed from the H-C group.

**Figure 3.** Attention bias following food and nonfood cues. A graphical depiction of mean and SEM of the Stroop bias following HCF and NF images in the FAOB, NFAOB, and H-C participants. \* *p* < 0.05. SEM: standard error of mean. HCF: high-calorie food. NF: nonfood. FAOB: overweight and obese with food addiction. NFAOB: overweight and obese without food addiction. H-C: healthy controls.

**Figure 4.** Brain potentials in response to food cues. (**a**) Event related potential curves in the FAOB, NFAOB and H-C groups, following HCF compared with NF images, in the frontal and right posterior ROI (no group differences in the left posterior ROI, hence not shown). Curve's shades mark SEM; TOIs (LPPa, LPPb) are marked in gray. (**b**) Topographic plots of mean group amplitudes (upper row) and group contrasts (lower row) during the LPPb. Electrodes of the different ROI are marked in dark blue (**c**) Mean and SEM of the same in the frontal and right posterior ROI. \*\* *p* ≤ 0.01. FAOB: overweight and obese with food addiction. NFAOB: overweight and obese without food addiction. H-C: healthy controls. HCF: high-calorie food. NF: nonfood. ROI: region of interest. SEM: standard error of mean. TOI: times of interest. LPP: late positive potential.

Correlational analysis between the food cue-induced differences in both the Stroop bias and the LPPb amplitude revealed a negative correlation in the frontal ROI (*r* (*<sup>n</sup>* <sup>=</sup> 51) = −0.33, *p* = 0.02) and a positive correlation in the right posterior ROI (*r* (*<sup>n</sup>* <sup>=</sup> 51) = 0.30, *p* = 0.03; Figure 5). Therefore, the food cue-induced amplitude change was linearly associated with the Stroop bias change following HCF compared to NF images.

**Figure 5.** Association between the amplitude change and the Stroop bias change following HCF compared to NF images, for the participants altogether. (**a**) A topographic plot of the linear correlation magnitude between the food-cue induced LPPb amplitude and the Stroop bias changes. (**b**) Scatter plots of the same in the frontal and right posterior ROI. HCF: high-calorie food. NF: nonfood. LPP: late positive potential. ROI: region of interest.

## **4. Discussion**

In the current study, we aimed to investigate psychobiological indices of food addiction in overweight and obesity. The results provide novel and unique neurobehavioral and psycho-cognitive markers characterizing FA versus no FA in overweight and obese participants. Our hypotheses were partially confirmed; during rest, the FAOB group showed greater left-brain asymmetry than that of the NFAOB group. This neurobiological signature is in line with approach motivation tendencies [82], indicating greater vulnerability to approach a motivationally salient cue relevant to the addictive condition [83]. The greater left-brain asymmetry at rest in the FAOB group may indicate a neuro-marker of repeatedly and compulsively approaching highly rewarding food, similarly to those observed in substance addiction [36]. Conversely, the greater right-brain dominance in the NFAOB group in comparison with the FAOB and (to a lesser extent) the H-C groups, may protect them from developing FA symptoms. The lack of addictive features in the NFAOB group may be reflected in their right frontal asymmetry. An extensive body of research points to inhibitory control deficits in obesity, which is associated with impulsivity and a lack of delayed discounting of food-related reward [84,85]. To overcome impulsivity toward food reward, the NFAOB may have developed heightened compensatory mechanisms in their right brain hemisphere, which is absent in healthy adults and the FAOB group, the latter who may be lacking the capacity to consistently control their food intake.

Our findings are in line with past research indicating left PFC asymmetry in chronic overeaters [31], in adults with high hedonic hunger (i.e., a drive for a food reward in the absence of hunger) [86], and in obesity [87]. In the current study, we also found left-brain asymmetry in the occipital ROI, and a general trend of reduced left alpha asymmetry in the whole hemisphere, possibly indicating a widespread left hemispheric dominance in the FAOB group, extending to brain areas in the sensory association cortex [77]. Specifically, brain responses with robust asymmetry to the left parietal and occipital brain areas may be related to strategic and tactical aspects of goal pursuit [88], such as reward valuation and integration [77,89], as well as hedonic valuation of food [77]. A widespread asymmetry that includes temporal and occipital electrodes has been observed in healthy individuals [76] and in

psychiatric patients [90], but this is the first study that shows this electrophysiological marker in food addiction. Therefore, additional studies are needed to help uncover the significance of these findings.

In the Food Stroop task, participants viewed pictures of highly rewarding food, as well as nonfood items, before the Stroop word assignment. We hypothesized that the FAOB participants will show greater cue-reactivity in response to highly rewarding food cues, reflected in greater Stroop bias and heightened ERP responses, than that of NFAOB and H-C participants. This is based on previous literature indicating greater AB in the Food Stroop task in obese compared with lean participants [91,92]. Our results refuted previous findings; at earlier stages of cognitive processing (300–450 ms following picture presentation), the HCF pictures elicited heightened emotional reaction similarly in all groups. Thereafter, during the LPPb (at 450–495 ms) the FAOB group seems to have inhibited their emotional response to the HCF cues, differently from the NFAOB group, who displayed clear electrophysiological difference in response to HCF versus NF images. This neurobiological difference between the groups was behaviorally reflected in their differential performance on the Stroop word assignment, which started immediately thereafter. Indeed, in the Food Stroop task, the FAOB group showed a lower Stroop bias following images of HCF, suggesting that the lack of electrophysiological response to HCF images may result from an inhibitory process. The negative correlation between the neuronal response to HCF vs. NF images and the performance during the Stroop task further implies an inhibitory process of affective response, starting at the neuronal level and reducing attention bias on the Stroop task. The positive correlation between the occipital response to HCF vs. NF images and the performance during the Stroop task implies that increased sensory response to these images (without prefrontal inhibition) may induce the increased attention bias on the Stroop task.

The LPP component is commonly found in obesity research, and it indicates selective and motivated attention to rewarding cues [93], specifically highly rewarding food [94]. Greater LPP component response in addicted vs. non-addicted individuals, following a visual presentation of a cue associated with the addiction, has been shown in cannabis use and may be a neurobiological marker of addiction [36]. High LPP has also been found in response to an acute stressor, when cortisol levels are high [78], implying vigilant attention to a threat. In our sample, the strong inhibition of emotional-motivational reaction to the HCF cues in the FAOB group may point to the hypervigilance-avoidance hypothesis [95,96]. Research is pointing to an interaction between emotional valence and executive control demands in tasks involving attention and cognitive interference [44,97]. Attentional avoidance in a state of emotional vigilance has been observed in dieters who attempt to attentionally avoid pictures of the food they desire [64,98], and in addicted patients exposed to the stimuli they are trying to abstain from [99]. Accordingly, the HCF images in our study may have elicited in the FAOB group a strong affective response. At the early stages of information processing, the FAOB group, similarly to the other two groups, showed an initial heightened emotional response to the appetitive cues, which was thereafter extensively inhibited in the FAOB group, possibly when experiencing hypervigilance with triggers of an emotionally-laden problematic behavior associated with their condition.

The hypervigilance-avoidance hypothesis has been shown in adults with social phobia, who respond faster in a Stroop paradigm with images of socially challenging situations, following an anxiety-inducing task [100,101]. Similarly, overeating of highly rewarding food has been postulated to function as a relief from the physical tension associated with hypervigilance [102], suggesting an addictive cycle whereby compulsivity develops to relief from the psychophysiological tension associated with the condition. Following this hypothesis, individuals with overweight or obesity and FA may vigilantly detect highly rewarding food cues in their environment, propelling a negative affect and a negative urgency to impulsively consume that food. They may try to counteract their tendency to approach highly rewarding food (to relieve their hypervigilance) by exercising cognitive avoidance, up to the point where they disinhibit their restraint and lose their control over eating. At the point of disinhibition, consumption of the food may function to relieve the physical and emotional tension associated with generalized or cue-specific hypervigilance [84]. This hypothesis is in line with the

theoretical understanding of impulsivity and loss of control of eating seeing in FA [5]. These behaviors and their neurobiological precursors may also be reflected in left-brain asymmetry [33].

The FAOB differed from the NFAOB (and the H–C) in both binge-eating (BE) and depressive symptoms and in symptoms of emotional and uncontrollable eating. These results are in-line with our hypotheses and replicate previous research comparing individuals with and without FA [103,104]. BED is related to FA [105], but research about distinctions and similarities between the two conditions is in its infancy. FAOB has been suggested to be an extreme form of BED [6]. However, in our sample, only 13 participants out of the 30 (i.e., 43%) in the FAOB group showed BE symptoms [106]. Moreover, BE symptom scores were not correlated with participants' performance on the Food Stroop task, nor with brain asymmetry scores or ERPs, and the three groups in our study differed in FA symptoms and BMI even when controlling for binge-eating scores. These are novel and important findings, pointing to FA as a unique clinical construct, characterized by distinct psycho-neurobiological markers, above and beyond BE symptoms. Future studies may employ the parameters addressed in the current study to directly compare two cohorts of overweight/obese adults: one with FA and the other with BE symptoms/BED.

BE and FA may escalate depressive symptoms [5,104], and there is ample evidence to support the co-occurrence of obesity and depression [2]. In our work, despite greater depressive symptoms in the FAOB group compared with the other two groups, symptoms level did not reach clinical significance but more of a melancholic state [107]. It is possible that the uncontrollable compulsion to eat, low self-esteem [19], self-inefficacy in controlling one's eating and weight [108], and the impairment in the quality of life in overweight/obesity with FA [109], contributed to greater depressive symptoms in the FAOB group.

The present study has strengths and limitations. This study is the first to find neurocognitive markers and psycho-behavioral correlations in overweight/obesity with FA, using brain asymmetry indices and ERP in a Food Stroop task. In the present study, participants' hunger and metabolism were carefully controlled for 24 h prior to, and on the day of, the study, to reduce the chance of confounding variables, such as metabolic hunger [47], biasing the results. Future work may regress these and other potential confounders on the neurocognitive parameters we applied in the current study; this was not performed here and may be a shortcoming of the present work. The present study has a limitation in terms of sample size, particularly in the NFAOB and H-C groups. Moreover, participants' recruitment in the present study poses several limitations to the generalizability of our findings; we did not recruit participants with the co-presence of obesity and SUD [110], and our study lacks a subgroup of FA who shows lean body mass [111], limiting the conclusions to FA in overweight/ obesity. Moreover, we used the original version of the YFAS, which is based on the DSM-IV, since participants' recruitment started prior to the publication of the most updated version, the YFAS version 2 [112]. Therefore, we did not distinguish between mild, moderate, and severe FA symptoms in participants' recruitment. We, therefore, suggest these limitations be addressed in future FA studies. Lastly, our research setting has possibly impacted the participants in the study. Future ecological momentary assessment studies may help examine overweight/obesity with FA in a different, more natural setting, to avoid possible confounding factors of conducting research in the lab.

#### **5. Conclusions**

Our study uniquely demonstrated that overweight and obese adults with FA show markers of left-brain asymmetry at rest, relative to overweight/obese adults without FA. In addition, the overweight/obese participants with FA show markers of a hypervigilant inhibition of emotional reaction to food triggers that may elicit excessive cravings, evident in a lower LPPb response to HCF images and reduced AB in a Food Stroop task. Our results are in line with psychobiological markers of SUD and behavioral addiction, and they introduce novel understandings of overweight/obesity with FA. Neurocognitive training and neuro-modulatory treatment, such as transcranial magnetic stimulation (TMS), may help rebalance hemispheric symmetry in obesity with FA. Future studies may also address potential therapies to help individuals with FA cope better with environmental stimuli relevant to their condition.

**Author Contributions:** Conceptualization, R.A.-F., U.A., A.Z., and L.K.; Methodology, R.A.-F., U.A., A.Z.; software, R.A.-F. and U.A.; validation, U.A.; formal analysis: R.A.-F., L.K., and U.A.; investigation, R.A.F, L.K. and G.B.; resources: A.Z.; Data curation, U.A. and L.K.; writing – initial draft preparation, R.A.-F.; writing - review and editing, U.A., L.K., G.B., and A.Z.; visualization, U.A.; supervision, A.Z.; project administration, R.A.-F., L.K. and G.B.; funding acquisition: U.A. and A.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by MAGNET program, Israel Innovation Authority, #61450.

**Acknowledgments:** We thank Noga Cohen at the Edmond J. Safra Brain Research Center for the Study of Learning Disabilities at the University of Haifa, Israel, for reviewing and commenting on this paper.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

**Appendix A**

**Figure A1.** Procedures' timeline.


**Table A1.** Examples of YFAS items and scoring [45].

other options available to me at home.


**Table A2.** Number of participants included in the analysis.

FAOB: overweight and obese with food addiction. NFAOB: overweight and obese without food addiction. H-C: healthy controls. RT: reaction time. ERP: event-related potentials.

**Figure A2.** Event-related components in response to food (HCF) and nonfood (NF) images. Data is averaged for all participants together, regardless of group affiliation. (**a**) A butterfly plot of potential curves that include all electrodes. (**b**) Global mean field potential (GMFP) computed sample-wise as the root mean squared deviation (RMSD) from the mean amplitude of all electrodes. The different event-related components and time windows of interest (marked gray) were defined according to the peaks in GMFP and are in line with the literature [65,79]; from left to right: P100 (95–135 ms), N200 (155–195 ms), P300 (240–280 ms), and the Late Positive Potential (LPP) subdivided into LPPa (300–450 ms) and LPPb (450 495). (**c**) Topographic plots of the mean potentials during the different components following HCF and NF images. The frontal, right posterior, and left posterior regions of interest (electrodes marked in dark blue) were defined to capture differences in brain potentials induced by the food images. (**d**) Event-related potential curves in response to HCF and NF images. \*\*\* *p* ≤ 0.001. HCF: high-calorie food. NF: nonfood.

**Figure A3.** Topographical power maps of the raw resting state EEG activity. Topographical maps of the average resting state power in the experimental groups. No group differences were found using a nonparametric F test in any of the electrodes.
