Unhealthy Yet Avoidable – How Cognitive Bias Modification Alters Behavioral And Brain Responses To Food Cues In Obesity

Objective Obesity is associated with automatically approaching problematic stimuli, such as unhealthy food. Cognitive bias modification (CBM) could beneficially impact on problematic approach behavior. However, it is unclear which mechanisms are targeted by CBM in obesity: Candidate mechanisms include (1) altering reward value of food stimuli or (2) strengthening inhibitory abilities. Methods 33 obese people completed either CBM or sham training during fMRI scanning. CBM consisted of an implicit training to approach healthy and avoid unhealthy foods. Results At baseline, approach tendencies towards food were present in all participants. Avoiding vs. approaching food was associated with higher activity in the right angular gyrus (rAG). CBM resulted in a diminished approach bias towards unhealthy food, decreased activation in the rAG, and increased activation in the anterior cingulate cortex. Relatedly, functional connectivity between the rAG and right superior frontal gyrus increased. Analysis of brain connectivity during rest revealed training-related connectivity changes of the inferior frontal gyrus and bilateral middle frontal gyri. Conclusion Taken together, CBM strengthens avoidance tendencies when faced with unhealthy foods and alters activity in brain regions underpinning behavioral inhibition.


56
The way we process and react to food cues is linked to unhealthy eating and obesity. 57 Preferential processing and facilitated approach reactions towards food cues have been 58 demonstrated in obese compared to healthy-weight individuals 1-4 . This automatic and 59 biased processing may contribute to overconsumption of food 5,6 , especially in current 60 'obesogenic' environment. 61 Dual-process models address these automatic behavioral tendencies. The reflective-62 impulsive model 7 states that during automatic behavior, the fast impulsive system 63 overrules the slower reflective system. The former is guided by previously formed 64 associations -approach positive and avoid negative stimuli -while the latter relies on 65 explicit knowledge 7,8 . The incentive sensitization theory (e.g. 9 ) further states that through 66 repeated exposure, a reinforcer (e.g. tasty food) acquires incentive salience qualities via 67 the brain's reward system. In consequence, associated stimuli become attention-68 grabbing and the motivation to approach is increased 10 .

69
These theories can explain a paradox in maladaptive behaviors, where individuals 70 continue to behave in disadvantageous ways despite better knowledge and often against 71 personal intentions. Heavy drinkers, for example, were found to crave alcohol without 72 necessarily liking it 11,12 and were repeatedly found to display an approach bias towards 73 alcohol 8,13,14 . 74 Cognitive bias modification (CBM) aims at changing these automatic tendencies towards 75 problematic stimuli, subsequently improving maladaptive behavior. Promising effects 76 have been shown in alcohol-dependent subjects and smokers 8,15,16 , where a CBM 77 intervention decreased approach bias towards problematic stimuli while reducing 78 consumption thereof. Regarding eating behavior, research has produced mixed 79 findings 17 . Approach bias for and consumption of chocolate could be decreased through 80 CBM in normal-weight individuals 18 . In contrast, no CBM effects were found in normal-81 weight females across three studies for both unhealthy food and chocolate 19 . In a sample 82 of lean and obese participants, CBM reduced approach bias towards unhealthy food only 83 in obese individuals 4 . Discrepancies in these results, however, might be explained by 84 differences in samples and stimuli. 85 Despite the evidence for CBM's efficacy in the behavioral context, not much research has 86 been conducted regarding underlying neural mechanisms. One study investigated neural 87 correlates of CBM in hazardous drinkers, where half of the participants received CBM 88 while the other half received sham-training. CBM was associated with reducing activity in 89 the medial prefrontal cortex 20 . This structure, together with the nucleus accumbens and 90 the posterior cingulate gyrus, is engaged in approach behavior towards problematic 91 stimuli 21,22 .

92
We investigated neural correlates of CBM in obesity by applying an approach-avoidance 93 training 8 in obese individuals in the fMRI scanner, randomly assigning participants to a 94 training or a sham-training condition. The training was hypothesized to induce two effects: 95 decreased approach tendencies towards unhealthy, and increased approach tendencies 96 towards healthy foods. We hypothesized that CBM training could work through two 97 mechanisms: by a) changing rewarding values of food stimuli and activation in reward-98 related regions; b) increasing inhibitory abilities and changing activity of brain regions 99 engaged in inhibitory processing and cognitive control. We further explored whether CBM 100 training induces differences in task-independent resting-state functional connectivity.  and liking (see section: Selection of stimuli for details) on a cardboard with a scale (0-10).

117
This picture set was independent of the one used in the fMRI task. It included healthy and 118 unhealthy food pictures of comparable healthiness and liking to the fMRI picture set. 119 2.3. fMRI task 120 We used a training version of the approach-avoidance task (AAT; described in 4 , for task 121 details see Figure 1A), which measures and modifies 4 automatic approach and avoidance 122 tendencies towards unhealthy and healthy food pictures (30 healthy/30 unhealthy).

123
Participants react with push and pull movements of a joystick to the format of presented 124 food pictures (e.g. push-vertical/pull-horizontal). The AAT consisted of three main phases: 125 a pre-phase, a training or a sham-training phase, and a post-phase. In the pre-, post-and 126 sham-training phases food pictures were presented equally often in push and pull 127 formats. In the training phase 90% of unhealthy pictures appeared in a push format, and 128 90% of healthy pictures appeared in a pull format ( Figure 1B). The amount of approach 129 and avoidance responses was equal in both groups -any behavioral effects would 130 therefore be related to the pictures' content. We randomly assigned participants to a 131 training or a sham-training group. Participants were not informed and not aware that 132 training would take place, and the experiment introduction was performed by a blinded 133 experimenter.

134
Further, we tested whether potential changes in automatic action tendencies are specific 135 to pictures included in the training phase or generalized to the entire category (healthy 136 vs. unhealthy). Therefore, only a subset of pictures used in pre-and post-phases was 137 used for training (randomly chosen set of 20 out of 30 pictures for each participant).

138
The AAT lasted around 40 minutes and was symmetrically divided into four runs, each 139 including 110 trials (independent of pre-, training, or post-phases). Participants were 140 offered breaks between runs to relax and close their eyes. Food images for the AAT and the picture-sorting task were selected from the food-pics 143 database 23 and categorized into healthy and unhealthy according to 4 . Only images that 144 were clearly identified as healthy or unhealthy were included 4 .   No subject had to be excluded due to outliers or error rate. Outliers were defined as mean 163 reaction times below or above 2 standard deviations from the group mean. The task was 164 performed with high accuracy (mean = 97%, SD=3.23%).

165
During the pre-phase, bias scores significantly differing from zero would reflect baseline 166 behavioral tendencies. Further, to ensure that no baseline group differences were 167 present, we compared bias scores of training and the sham-training group. Analyses were 168 carried out using one-sample and independent-samples t-tests, respectively.

169
Changes from pre to post were analyzed using a 2x2x2 rmANOVA. Group (training/sham-170 training) was used as a between-subject factor, and image category (healthy/unhealthy) 171 and time (pre/post) as within-subject factors. We followed up by testing if bias scores 172 significantly differed from zero in the post-phase with t-tests. were entered into a GLM and convolved with a double-gamma hemodynamic function.

184
Contrast files were entered into second-level analysis, where we compared subjects as 185 groups. BMI and age were entered into the analysis as covariates of no interest. By 186 entering BMI as a covariate, we investigated general group effects dependent on training 187 condition only and accounted for between-subject differences that could potentially be 188 caused by differences in BMI. Age was entered as a covariate since groups differed in 189 this respect. Results were thresholded at a whole-brain voxel-wise level with a threshold   To investigate whether CBM was related to changes in task-related functional 206 connectivity, we conducted a psychophysiological interactions (PPI) analysis. It compares 207 brain connectivity changes from a specified seed in the brain between two different 208 experimental conditions. We placed the seed region in the right angular gyrus (6mm 209 radius sphere) and tested whether connectivity of the seed in the unhealthy avoidance 210 condition differed between experimental phases and groups. 211 2.6.2.3. Resting-state fMRI data analysis 212 We acquired resting-state data to investigate whether effects of CBM are transferrable to 213 functional changes outside the AAT. These data can be used to analyze resting-state 214 functional connectivity -addressing how different brain regions interact. Resting-state 215 connectivity analysis helps to understand how all brain regions generally interact with 216 each other (degree centrality, DC), but also how specific a priori defined brain regions 217 correlate with other brain areas (seed-based connectivity analysis, SCA).

218
To investigate connectivity in an exploratory fashion, but also in a priori defined regions, We investigated neural correlates of pre-training approach and avoidance tendencies with 284 a one-sample t-test, as in this stage groups did not differ in any way. Contrasts included 285 food approach and food avoidance, together and separately for healthy and unhealthy 286 food. A contrast corresponding to general food avoidance (push>pull independent of 287 picture category) revealed significant clusters in the right angular gyrus (rAG) and the 288 cuneus ( Figure 3A). Food avoidance activations were driven by the unhealthy food 289 category (push>pull for unhealthy food). For the general approach for food (pull>push),

290
we found a significant cluster in the left postcentral gyrus ( Figure 3B, Table 2). We found 291 no further significant results and did not find group differences for above-mentioned 292 contrasts.  We consistently found the rAG to be associated with unhealthy food avoidance and the 310 effects of CBM, and therefore performed PPI analysis with the rAG as the seed. We 311 compared connectivity differences for unhealthy food avoidance between pre-and post-312 phases. This analysis showed a significant cluster in the rSFG/rMFG and in the right 313 caudate/putamen, indicating that connectivity between the rAG and these structures 314 increased post-training in the training group ( Similar to SCA, DC describes task-independent connectivity changes within the brain.

324
These changes, however, are general and not specific to chosen ROIs. In our study this 325 analysis did not produce any significant results. in reward-related brain regions, or b) increasing inhibitory abilities and affecting brain 333 regions engaged in inhibitory processing and cognitive control. We found that all 334 participants showed faster approach than avoidance reactions towards healthy and 335 unhealthy food images, suggesting that approaching food is an automatic process. This 336 was paralleled by our findings on the neural level, where the rAG showed higher activation 337 for avoiding food -a potentially conflicting situation. The rAG is a part of the 338 temporoparietal junction (TPJ), which is often related to both processing of social cues 339 and attentional processes 26,27 . CBM specifically affected the training group, where 340 approach tendencies towards unhealthy food were successfully decreased. This was 341 related to a lower activation in the rAG after training. Additionally, we observed group-342 specific changes in resting-state connectivity between inhibitory regions, such as the 343 MFG or the IFG 28-31 , and in task-related connectivity between the rAG and the right 344 caudate/putamen (dorsal striatum). Avoiding food thus appears to be a potentially 345 conflicting situation, requiring activation of inhibitory and conflict resolution brain 346 mechanisms. CBM seems to decrease this demand by means of strengthening 347 connectivity between inhibitory brain regions. Further, we found no evidence for altered 348 reward valuation of food stimuli after CBM in both behavioral and imaging data.

349
As mentioned, avoiding food was related to higher activity of the rAG -a part of the TPJ. 350 Bzdok and colleagues showed that the right TPJ links two brain networks integrating 351 external (sensory) vs. internal (memory, social-oriented stimuli) information 26 . It is 352 conceivable that a conflict between external instruction (avoid unhealthy food), and 353 internal impulse (approach unhealthy food) increases activity of the rAG in order to solve 354 this conflict. This is consistent with studies showing that the rAG is directly engaged in 355 resolution of stimulus-response conflicts, but also attentional reorientation and response 356 inhibition 32-37 .

357
Though approach behavior towards unhealthy food pictures decreased, approach 358 behavior towards healthy food pictures remained unchanged. This is in line with previous 359 findings, where decreasing approach behavior towards unhealthy food was the main 360 training effect 4,38 . In our study, decreasing approach tendencies towards unhealthy food 361 was related to decreased brain activation in the rAG, suggesting that training makes 362 avoiding food a less conflicting and more automatic behavior. Further, we found no 363 conclusive evidence for generalization effects. While generalization was repeatedly 364 observed in other contexts (e.g. 8 ), results in the obesity context have been mixed 4,38 .

365
For unhealthy food avoidance, we found increased post-training task-related connectivity 366 between the rAG and the dorsal striatum, which is related to stimulus-response learning, 367 executive attention and exerting cognitive control [39][40][41][42][43] . Increased connectivity between the 368 dorsal striatum and rAG was previously related to explicit usage of learned stimulus-