Diabetes Care Volume 39, August 2016 1311

Sabine Frank,1 Jaana M. Heinze,2–4 Neuronal Food Reward Activity in Andreas Fritsche,2–4 Katarzyna Linder,2,4 CARE/EDUCATION/NUTRITION/PSYCHOSOCIAL CLIN Maximilian von Feilitzsch,5 Patients With Type 2 Alfred Konigsrainer,¨ 5 Hans-Ulrich Haring,¨ 2–4 Ralf Veit,1,3,4 and With Improved Glycemic Control Hubert Preissl2–4,6 After Diabetes Care 2016;39:1311–1317 | DOI: 10.2337/dc16-0094

OBJECTIVE Obesity and type 2 diabetes mellitus (T2DM) are associated with altered food- related neuronal functions. Besides weight loss, substantial improvement of glu- cose metabolism in patients with T2DM can be achieved by bariatric surgery. We aimedtotargettheneuronalandbehavioralcorrelatesofimprovedglycemic control after bariatric surgery.

RESEARCH DESIGN AND METHODS Two patient groups with T2DM were recruited. The treatment group (n = 12) consisted of patients who had undergone Roux-en-Y gastric bypass (RYGB) surgery, 1Institute for Medical and Behav- and a control group consisted of patients who did not undergo surgery (n = 12). ioural Neurobiology, University of Tubingen,¨ Tubingen,¨ Germany The groups were matched for age and current BMI. HbA1c was matched by using 2Department of Internal Medicine IV, University the presurgical HbA1c of the RYGB group and the current HbA1c of the nonsurgical Hospital, Tubingen,¨ Germany group. Neuronal activation during a food reward task was measured using func- 3Institute for Diabetes Research and Metabolic tional MRI (fMRI). Behavioral data were assessed through questionnaires. Diseases, Helmholtz Center Munich, University of Tubingen,¨ Tubingen,¨ Germany 4 ¨ RESULTS German Center for Diabetes Research, Munchen- Neuherberg, Germany 5 RYGB improved HbA1c from 7.07 6 0.50 to 5.70 6 0.16% (P < 0.05) and BMI from Department of General, Visceral and Transplant 52.21 6 1.90 to 35.71 6 0.84 kg/m2 (P < 0.001). Behavioral results showed lower Surgery, University Hospital, Tubingen,¨ Germany 6 wanting and liking scores as well as lower eating behavior–related pathologies for Department of Pharmacy and Biochemistry, Faculty of Science, University of Tubingen,¨ the patients after RYGB than for similar obese subjects without surgery but with Tubingen,¨ Germany impaired glycemic control. The fMRI analysis showed higher activation for the Corresponding author: Andreas Fritsche, andreas. nonsurgical group in areas associated with inhibition and reward as well as in the [email protected]. precuneus, a major connectivity hub in the brain. By contrast, patients after RYGB Received 21 January 2016 and accepted 26 April showed higher activation in the visual, motor, cognitive control, memory, and 2016. gustatory regions. Clinical trial reg. no. NCT01900483, clinicaltrials .gov. CONCLUSIONS This article contains Supplementary Data online In obese patients with diabetes, RYGB normalizes glycemic control and leads to at http://care.diabetesjournals.org/lookup/ food reward–related brain activation patterns that are different from those of suppl/doi:10.2337/dc16-0094/-/DC1. obese patients with less-well-controlled T2DM and without bariatric surgery. The © 2016 by the American Diabetes Association. Readers may use this article as long as the work is differences in food reward processing might be one factor in determining the properly cited, the use is educational and not for outcome of bariatric surgery in patients with T2DM. profit, and the work is not altered. 1312 Neuronal Correlates After RYGB in Diabetes Diabetes Care Volume 39, August 2016

Obesity, which is associated with vari- successfully prevents the development of surgery. We studied two groups with ous comorbidities, especially type 2 di- T2DM, particularly in patients with im- T2DM: One group underwent RYGB and abetes mellitus (T2DM), can shorten life paired glucose control but independent experienced substantially improved HbA1c, expectancy by up to 20 years (1). In the of presurgical BMI (19). Furthermore, a and the other group had a comparable past few decades, the prevalence of remission of T2DM after Roux-en-Y gas- BMI but did not undergo bariatric surgery. obesity and diabetes has increased dra- tric bypass (RYGB) surgery has been This design allowed us to compare brain matically (2,3) and has led to a growing reported in the majority of patients activation patterns in patients with T2DM interest in unraveling the underlying (20–22). Several independent factors, who mainly differed in glycemic control mechanisms of obesity and diabetes. including weight loss and diabetes dura- achieved by bariatric surgery. Besides altered behavioral and meta- tion, predict T2DM improvement (23). bolic processes, obesity (4) and T2DM The effectiveness of successful long- RESEARCH DESIGN AND METHODS (5) have been shown to be associated term weight loss and diabetes remission Study Population with altered neuronal mechanisms re- thus depends on various factors that In total, 24 obese, right hand–dominant lated to eating behavior and food and might also include neuronal food reward subjects with T2DM participated in the reward processes. When stimulated processes. study. Twelve underwent RYGB at least with food items, obese subjects show On a neuronal level, bariatric surgery 6 months prior and had experienced higher activation in gustatory, reward, results in altered neuronal responses to substantial weight loss and improved and frontal control regions than lean food cues, particularly in the reward, glycemic control (RYGB group). The – subjects (6 8). Patients with T2DM gustatory, and homeostatic regions other 12 did not undergo bariatric sur- have increased responses to food pic- (11,24–26). A comparison of neuronal gery (nonsurgical group). Patient char- tures in gustatory and reward regions, responses to food items between pa- acteristics are shown in Table 1. Patients 6 which are modulated by dietary tients postsurgery and a lean and an were matched for age (50.0 2.7 vs. 6 self-care (5). obese control group showed that se- 50.7 3.3 years, P = 0.88) and current In terms of behavior, food reward can BMI (35.71 6 0.84 vs. 37.81 6 1.38 verely obese women who underwent 2 be divided into a wanting and a liking RYGB at least 1 year earlier showed a kg/m , P =0.21).HbA1c was matched by component. Liking is explained as the normalization of obesity-associated al- using the presurgical level of the RYGB hedonic reaction to the of a re- group with the current level of the non- terations in brain activity (25). This neu- ward, whereas wanting is described as surgical group (7.07 6 0.50 vs. 7.04 6 ronal normalization is also supported by a the incentive salience linked with the 0.37%, P = 0.97). The study protocol was partial reversal of hypothalamic dysfunc- motivation toward an item. Animal re- approved by the ethics committee of tion after glucose intake in postoperative search has shown that wanting and the medical faculty of the University of patients without diabetes (27). liking are associated with different Tubingen,¨ and all subjects gave written These studies mainly focused on the neuronal circuits (9). Finlayson et al. informed consent. effect of weight loss and changes in the (10) used behavioral studies to demon- activity of eating-related and strate that this concept can be used in Study Design did not specifically address the effect of human research, and several brain im- Subjects arrived at the study site after changes in glycemic control. In the pre- aging studies have shown that food fasting for at least 3 h and taking their sent cross-sectional study, we aimed reward–associated brain structures are usual as instructed. They to determine whether neuronal food specifically involved in the processing of completed questionnaires to assess their – wanting and liking (11–13). reward related activation patterns de- current , mood, and eating be- Although various treatment options pend on improvement of glycemic control havior traits (the Three Factor Eating for obesity exist, long-term weight loss in a population with T2DM after bariatric Questionnaire with the scales cognitive maintenance is difficult to achieve, and the majority of patients regain weight Table 1—Patient characteristics within the first year after losing weight T2DM group by a conventional weight loss therapy (14,15). The most successful, as well as RYGB Nonsurgical control P value most invasive, therapy for weight loss is No. patients (female/male) 12 (10/2) 12 (6/6) 0.08 bariatric bypass surgery (16). A recent Age (years) 50.0 6 2.67 50.7 6 3.29 0.88 review summarized the hormonal, neu- BMI (kg/m2)35.716 0.84 37.81 6 1.38 0.21 ronal, and metabolic consequences of RYGB-induced change in BMI 216.5 6 1.53 bariatric surgery and emphasized that HbA1c (%) 5.70 6 0.16 7.04 6 0.37 0.005 the interplay of these factors leads to RYGB-induced change in HbA1c 21.37 6 0.48 changes in feeding behavior (17). Long- Years since diabetes diagnosis 11.17 6 1.91 6.46 6 1.09 0.043 term follow-up studies on weight devel- Months since RYGB surgery 17.73 6 2.68 opment after bariatric surgery showed Oral diabetes medication 2 8 0.005 an excess weight loss of .50% and, in Insulin therapy 0 2 0.078 some cases, .70% after 10 years (18). In addition to being an effective form of Data are mean 6 SEM. Medication is given as the number of subjects taking oral medication or undergoing insulin therapy. weight loss treatment, bariatric surgery care.diabetesjournals.org Frank and Associates 1313

restraint, disinhibition, experienced hun- Analyses of the fMRI data during vi- (wanting, liking), and food (high calorie, ger [28] and the Power of Food Scale sual stimulation were performed using low calorie). Analyses were controlled measuring the power of food that is Statistical Parametric Mapping 8 soft- for age, BMI, and hunger. Results were available, present, or tasted [29]). The ware (www.fil.ion.ucl.ac.uk/spm). Data considered significant at P , 0.05 family- Beck Depression Inventory (30) was were preprocessed, beginning with slice wise error-corrected on cluster level as used to address depressive symptoms. timing and realignment of the images to a function of an uncorrected primary A blood sample was taken to determine the mean image. To account for suscep- threshold level of P , 0.001 (33). the HbA1c level. Presurgery BMI and tibility by movement artifacts, unwarp- Statistical analyses of behavioral data HbA1c (measured by using the same ing of time series was performed. The were performed using SPSS version 22 method as for the presurgery HbA1c) anatomical T1-weighted image was software (IBM Corporation, Armonk, were retrieved from clinical records. coregistered to the mean functional im- NY). Differences in eating behavior For the functional MRI (fMRI) mea- age. Normalization into Montreal Neu- traits, descriptive data, and blood mea- surement, subjects were positioned in rological Institute (MNI) space (3-mm surements between the two groups the scanner with their heads fixed in a isotropic voxel size) and Gaussian spatial were analyzed using two-sample t tests. 12-channel head coil. An intercom sys- smoothing (full width at half maximum 6 Group differences in the hunger state tem was used for communication. Dur- mm) were then performed. Data were (pre- and postmeasurement) were ana- ing the scanning session, subjects were high-pass (cutoff 128 s) filtered and lyzed by using a repeated-measures stimulated with 40 food pictures (20 global AR(1) autocorrelation corrected. ANOVA with the within factor time and high calorie and 20 low calorie) within For each subject, a general linear the between factor group. Repeated mea- the wanting (run 1) and liking (run 2) model was applied for the high- and surements of HbA1c levels and BMI for tasks. During these two runs, subjects low-caloric food wanting and liking the RYGB group (pre- and postsurgery) were asked to first rate each picture conditions. For each condition, a sepa- were analyzed using paired t tests. for wanting (“How much do you want to rate regressor was modeled by using a An ANOVA was performed with the fac- eat this food now?”) and then for liking canonical hemodynamic response func- tors group and task and the covariates (“How much do you like this food in gen- tion that included time derivatives. age and BMI to test for group and task- eral?”) on a 5-point Likert scale using Movement parameters were modeled related differences. Group differences in an fMRI–compatible button box. Each as confounders. medication and sex distribution were picture was presented for 3 s, with an in- Because imaging results also depend calculated by x2 tests. Results of behav- terstimulus interval of 1–12 s. The task on vascularity effects and neuronal vas- ioral analyses were considered signifi- was programmed with Presentation ver- cularity depends on age, medication, cant at P , 0.05 (Table 1). sion 10.2 software (www.neurobs.com). and diabetes (31), we controlled the After completion of the task, an anatom- analyses for vascularization. We imple- RESULTS ical scan was recorded and hunger ratings mented the method by Tsvetanov et al. Weight, Glycemic Control, and obtained. (31) by using a scaling of the first-level Medication contrast images by the individual rest- All subjects had preexisting T2DM (du- Imaging Procedures and Analyses ing state fluctuation amplitude before ration since diagnosis: 6.46 6 1.09 vs. Whole-brain fMRI blood oxygen level– group analyses. Resting state measure- 11.17 6 1.91 years for RYGB vs. nonsur- dependent data were obtained with a 3-T ments were analyzed using Data Pro- gical, respectively) and were severely fMRI scanner (MAGNETOM Trio, A Tim Sys- cessing Assistant for Resting-State fMRI obese before surgery. The two study tem; Siemens Healthcare, Erlangen, Ger- running on Statistical Parametric Map- groups were matched for age, postsur- many) equipped with a 12-channel head ping 8 software (32). Functional images gical BMI, and presurgical HbA1c. coil. During the stimulation paradigm, each were realigned and coregistered to the The HbA1c level significantly improved session consisted of 150 scans (repeti- T1 structural image. Images were nor- after surgery in the RYGB group from tion time 2 s, echo time 30 ms, matrix malizedintoMNIspace(3-mmisotro- 7.07 6 0.50 to 5.70 6 0.16% (P , 64 3 64, flip angle 908, voxel size 3.3 3 pic voxel size) and smoothed with a 0.05). BMI also improved from 52.21 6 3 3.3 3 3.2 mm , slice thickness 3.2 mm, three-dimensional isotropic Gaussian 1.90 to 35.71 6 0.84 kg/m2 (P , 0.001). 0.8-mm gap, 30 slices [images acquired kernel (full width at half maximum Medication was documented as the in ascending order]). Resting state mea- 6 mm). A temporal filter (0.01–0.08 Hz) number of subjects taking oral diabetes- surements had already been taken before- was applied to reduce low-frequency specific drugs and/or undergoing insulin hand with 176 scans (repetition time 2 s, drifts and high-frequency physiologi- therapy. Diabetes medication was signifi- echo time 30 ms, matrix 64 3 64, flip angle cal noise. The resulting amplitude cantly lower in the RYGB group than 3 908, voxel size 3.3 3 3.3 3 3.6 mm , slice of low-frequency fluctuations maps in the nonsurgical group (Table 1). thickness 3.6 mm, 0.9-mm gap, 26 slices were then used to scale the contrast [images acquired in ascending order]). images obtained during the stimulation Behavioral Results At the end of the scanning period, high- protocol. Behavioral results of the wanting and resolution T1-weighted anatomical liking ratings during the scanning ses- images (magnetization-prepared rapid Statistical Second-Level Analyses sions revealed higher ratings in both gradient-echo 160 slices, matrix 256 3 For the fMRI data, block design 2 3 2 3 2 tasks for the nonsurgical group than 224, 1 3 1 3 1mm3)ofthebrainwere full factorial analyseswereused,including for the RYGB group (P , 0.001) and gen- acquired. the factors group (nonsurgical, RYGB), task erally higher liking than wanting ratings 1314 Neuronal Correlates After RYGB in Diabetes Diabetes Care Volume 39, August 2016

(P = 0.002). No interaction effect of the In an exploratory analysis, we tested visual stimulation of food items during factors group and task was observed whether improved glycemic control the wanting and liking tasks triggers an (Fig. 1). Hunger scores were low and in- (change in HbA1c) is associated with spe- increase in self-related processes associ- creased over time (P = 0.019) but did not cific changes in brain activation within ated with palatable food, which has to be significantly differ between the two the RYGB group. The analysis with an actively modulated by inhibitory process- groups (P = 0.115) (Supplementary Fig. uncorrected threshold level of P , es. Similarly, Bruce et al. (39) reported 1). Mood parameters showed no group 0.01 revealed a significant correlation that patients who had undergone gastric differences. of the HbA1c reduction with activity in the banding surgery showed higher inferior In general, the RYGB group showed orbitofrontal cortex (OFC) (Supplementary frontal activity during the processing of lower scores in eating behavior–related Fig. 2). Greater HbA1c reduction was food pictures before than after surgical traits than did the nonsurgical group. associated with higher activity in the intervention. In addition, changes in In particular, the RYGB group scored OFC during food reward processing. food reward–related brain regions were significantly lower in cognitive restraint For the main effect task, activation dif- reported soon after bariatric surgery (24) (Three Factor Eating Questionnaire) and ferences were detected in the somatosen- and after a weight loss of at least 8% (40). power of food (Power of Food Scale) sory cortex (postcentral gyrus), showing The globus pallidus is proposed to directly (Supplementary Table 1). higher activation in the wanting task than initiate reward-related signals (41). in the liking task (Supplementary Fig. 3 Changed food reward–related activation Neuronal Differences and Table 2). patterns, therefore, might be a direct con- We found significant main effects for sequence of the surgery, even without the factors group and task but not for CONCLUSIONS weight loss. Likewise, an improvement the factor food. No significant interac- The aim of this study was to investigate of glycemic control can be observed im- tions between factors were observed. the effect of a food reward–related task mediately after surgery before weight For the main effect group, the non- (wanting, liking) on brain functions in loss is measurable (21). Hence, the meso- surgical subjects displayed higher activ- two groups of similarly obese subjects limbic might play a crucial ity in regions associated with inhibition with T2DM: one group after RYGB sur- role in diabetes remission after bariatric (inferior frontal gyrus) and reward (globus gery and one group without surgery. surgery. pallidus) and in a general connectivity hub The RYGB group showed substantially The RYGB subjects with normalized (precuneus) than subjects in the RYGB improved glycemic control in the normal glycemic control showed increased ac- group (Fig. 2A and Table 2). In compari- range, whereas the nonsurgical group tivity in a large number of brain areas son, the RYGB surgical group showed had impaired glycemic control. compared with the nonsurgical control higher activation in visual (primary vi- On a behavioral level, we observed group. A large cluster was found in the sual, fusiform gyrus), frontal control re- lower scores in questionnaires examin- visual area, including the fusiform gyrus. gions (frontal middle gyrus, anterior ing eating behavior–related pathol- The fusiform gyrus is consistently acti- cingulate cortex), somatosensory cortex, ogies in the RYGB group, which has vated in visual food processing (42), and motor regions (supplementary motor also been reported in previous studies higher activity in patients after RYGB area, supramarginal gyrus), memory- (25,34–36). Furthermore, the lower than in an obese control group without related areas (hippocampus), and gusta- wanting and liking scores after bariat- surgery has also been reported (25). In tory regions (anterior insula, operculum) ric surgery are in line with previous addition, the fusiform gyrus was found than the nonsurgical group (Fig. 2B and findings of a reduced desire to eat after to be more highly activated after stimu- Table 2). RYGB (11). lation with high-caloric food pictures in Neuronal activation patterns showed lean subjects after ingestion of water, distinct differences between the groups whereas glucose ingestionresultedin and between the two task conditions. an enhanced activity of the fusiform gy- Obese patients with T2DM without sur- rus to low-caloric stimuli in lean sub- gery and with impaired glycemic control jects. Obese subjects showed no such showed substantially higher activations differential reaction pattern in the fusi- in brain areas associated with inhibi- form gyrus (43). This region, therefore, tion (inferior frontal gyrus) and reward appears to be sensitive to postprandial (globus pallidus) compared with their glycemic and hormonal actions. In this counterparts who experienced improved regard, it is important that food-related glycemic control after RYGB. Also, in the fusiform gyrus activity is specifically nonsurgical subjects, we found higher ac- modulated by insulin (44). Thus, the tivity in the precuneus, a central hub higher activity in the fusiform gyrus in the brain with strong and multiple con- may indicate an improved insulin sensi- nections to various brain areas and, thus, tivity of the brain region after bariatric involvement in high-level cognitive surgery with improved glycemic control. Figure 1—Mean wanting and liking rating for functions, including episodic memory, Because impaired insulin action in the food pictures within the scanning sessions for the nonsurgical and RYGB groups. Error self-related processing, and aspects of human brain has negative consequences bars represent SEM. *P , 0.01, **P # consciousness (37,38). One may specu- for metabolism, improvement of brain 0.001. late that in the nonsurgical subjects, the insulin action might be one reason for care.diabetesjournals.org Frank and Associates 1315

Figure 2—A: Regions with higher activation in the nonsurgical group than in the RYGB group. B: Regions with higher activity in the RYGB group than in the nonsurgical group. Results are significant at P , 0.05 family-wise error-corrected after an initially uncorrected threshold of P , 0.001. The color bar represents T values. ACC, anterior cingulate cortex; FMG, frontal middle gyrus; SMA, supplementary motor area.

improved metabolic control (45). How- behavior–related changes after improved to be sensitive to HbA1c change within ever, the current study cannot distinguish glycemic control. the RYGB group. between cause and consequence in this In an exploratory analysis, change in Independent of the groups, the main regard. the HbA1c level in the RYGB group was effect task revealed higher activation in The increased activity in the anterior associated with increased activation in the somatosensory cortex for wanting insula and the rolandic operculum may the OFC. This area is part of the second- than for liking, suggesting that each of also be associated with food evaluation. ary gustatory cortex and is involved in the two food-reward components has a The anterior insula is part of the primary the integration of perceived stimuli, different evaluation process. In a previ- gustatory cortex (46). Scholtz et al. (40) evaluative processes, and reward. ous study, we investigated a leaner 2 also observed higher activation in the Larger HbA1c reduction and, therefore, (mean BMI 27 kg/m ) nonsurgical study anterior insular cortex with food stimuli improvement of glycemic control are population and found higher activation after bariatric bypass surgery. In addi- associated with higher OFC activity. for wanting than for liking in a large tion, we observed increased activity This goes along with our findings in brain network, including the visual, re- in the hippocampus, a major memory- the RYGB group of increased activity in ward, memory, food processing, emo- related area. In conjunction with the the primary gustatory cortex (insula, tion, and frontal control regions (12). other areas, the hippocampus may operculum) (Fig. 2B), which seems to Therefore, one might speculate that be responsible for the establishment be more sensitive after surgery. Accord- the different task effect compared with of new memory traces for food in re- ing to the present exploratory analyses, healthy subjects is associated with se- lation to the metabolic and eating the secondary gustatory cortex seems vere obesity and/or diabetes. 1316 Neuronal Correlates After RYGB in Diabetes Diabetes Care Volume 39, August 2016

Table 2—Significant differences in brain activation for the nonsurgical and RYGB groups during the wanting and liking rating tasks MNI coordinates Cluster size Effect Brain region xyz(voxels)§ T value Task-related activation Nonsurgical . RYGB Lateral globus pallidus 21 27 28105.60 Inferior frontal gyrus 33 26 19 22 5.58 239 29 16 16 4.88* Precuneus 33 255 34 44 5.18 RYGB . nonsurgical Fusiform gyrus/visual cortex 39 279 25 2,975 10.61 Supplementary motor area/anterior cingulate cortex 0 23 52 315 6.96 Supramarginal gyrus 254 237 34 32 6.75 Rolandic operculum 239 2716706.09 Frontal middle gyrus 239 38 22 88 6.07 Somatosensory cortex (postcentral gyrus) 242 240 61 110 5.75 39 237 64 42 5.50 Primary motor cortex (precentral gyrus) 254 2 40 126 5.38 Hippocampus 39 213 223 11 6.04 Anterior insula 233 29 4 60 5.33 33 29 1 23 4.23* Wanting . liking Somatosensory cortex (postcentral gyrus) 39 234 58 70 5.21 §Results are significant with P , 0.05 family-wise error-corrected on cluster level after an initially uncorrected threshold of P , 0.001. *P , 0.001 uncorrected. Result included to show the bilateral activation pattern.

Overall, in the current study, patients between the groups cannot be ruled out. with diabetes. Thus, we controlled the who underwent RYGB surgery are char- To our knowledge, only one study com- analyses for the individual vascularity as acterized by reduced eating pathologies pared food-related neuronal processes proposed by Tsvetanov et al. (31). associated with a reduced need for in- between an obese group and an equally In summary, in obese patients with hibitory processes, higher cognitive con- obese group with diabetes (5) wherein a T2DM, we show distinct neuronal differ- trol, and increased gustatory activity. higher activation was reported in gusta- ences between those who underwent Thus, brain areas involved in the control tory and reward regions for patients with bariatric surgery with improved/normalized of temptations to eat are altered in food diabetes than for the control subjects. In glycemic control and those who did not reward processes after bariatric surgery, the current study, we also found higher undergo surgery with worse glycemic which is especially noteworthy because activation in a reward-associated region control. On the basis of the BMI-matched the surgery changes eating behavior (pallidus) as well as in inferior frontal re- sample, the observed neuronal effects based not only on smaller stomach size gions associated with inhibition. In addi- do not depend on BMI differences. In- but also on neuronal and psychological tion, we found a rather large network of stead, differences in food reward– processes related to eating behavior. higher activation in the RYGB group, in- associated brain functions might be This finding suggests a general change cluding gustatory, frontal control, visual, based on substantial weight loss and an in the reaction to food-related tempta- motor, and memory-related regions. Thus, improvement in glycemic control. Those tions after RYGB in T2DM. the results of the current study clearly dif- neuronal activation patterns might be The use of bariatric surgery as a dia- fer from those of the previous study, in- one factor in improved eating patholo- betes and weight loss treatment has to dicating that the improvement of HbA1c gies and in gaining long-lasting remission be considered carefully because some induced by the surgery might be causal. of diabetes. patients have shown a relapse of diabe- Several limitations of the study re- tes (47,48) and weight regain (20). A main to be mentioned. Because of the driving factor for successful diabetes re- cross-sectional study design, we cannot Acknowledgments. The authors thank Maike mission and weight loss maintenance assess a causal role of changes in glycemic Borutta (Institute for Diabetes Research and could be bariatric surgery–associated control on neuronal function. In addition, Metabolic Diseases, Helmholtz Center Munich, neuronal changes as reported in the cur- the tasks were not counterbalanced, University of Tubingen)¨ for assistance during the rent study. However, further studies are which might have affected the main ef- measurements and Shirley Wurth¨ for language needed to determine the effect of neu- fect task. Groups were not matched for editing and proofreading of the manuscript. Funding. This work was partially supported by ronal plasticity on success or nonsuccess individual duration since T2DM diagnosis. grants from the Bundesministerium fur¨ Bildung of bariatric surgery in relation to diabe- Diabetes-related medication in the RYGB und Forschung (German Federal Ministry of Ed- tes relapse and weight regain. group was lower than in the nonsurgical ucation and Research) to the German Center Whether the observed differences in group at the time of measurement. More- for Diabetes Research (DZD e.V., 01GI0925) and the Competence Network for Adiposity neuronal activation are due to either the over, Duarte et al. (49) showed that func- (01GI1122F). change in HbA1c and BMI in the RYGB tional brain activation can be influenced Duality of Interest. No potential conflicts of group or the difference in current HbA1c by impaired vascular coupling in patients interest relevant to this article were reported. care.diabetesjournals.org Frank and Associates 1317

Author Contributions. S.F. contributed to the and wanting as a function of metabolic state and confounding effects of vascular reactivity eval- study design, experiments, data research, and BMI. Soc Cogn Affect Neurosci 2015;10:561–568 uated by joint fMRI and MEG in 335 adults. Hum writing of the manuscript. J.M.H. contributed to 14. Purcell K, Sumithran P, Prendergast LA, Brain Mapp 2015;36:2248–2269 the writing, review, and editing of the manuscript. Bouniu CJ, Delbridge E, Proietto J. The effect 32. Chao-Gan Y, Yu-Feng Z. DPARSF: a MATLAB A.F. contributed to the study design, discussion, of rate of weight loss on long-term weight man- toolbox for “pipeline” data analysis of resting- and review and editing of the manuscript. K.L. and agement: a randomised controlled trial. Lancet state fMRI. Front Syst Neurosci 2010;4:13 R.V. contributed to experiments and review and Diabetes Endocrinol 2014;2:954–962 33. Woo CW, Krishnan A, Wager TD. Cluster- editing of the manuscript. M.v.F. and A.K. contrib- 15. van der Baan-Slootweg O, Benninga MA, extent based thresholding in fMRI analyses: pit- uted to the subject recruitment and review and Beelen A, et al. Inpatient treatment of children falls and recommendations. Neuroimage 2014; editing of the manuscript. H.-U.H. contributed to and adolescents with severe obesity in the 91:412–419 the discussion and review and editing of the Netherlands: a randomized clinical trial. JAMA 34. Ullrich J, Ernst B, Wilms B, Thurnheer M, manuscript. H.P. contributed to the study design, Pediatr 2014;168:807–814 Schultes B. Roux-en Y gastric bypass surgery re- data research, and review and editing of the 16. Sjostr¨ om¨ L, Peltonen M, Jacobson P, et al. duces hedonic hunger and improves dietary manuscript. H.P. is the guarantor of this work Bariatric surgery and long-term cardiovascular habits in severely obese subjects. Obes Surg and, as such, had full access to all the data in the events. JAMA 2012;307:56–65 2013;23:50–55 study and takes responsibility for the integrity of 17. Manning S, Pucci A, Batterham RL. Roux-en-Y 35. Laurenius A, Larsson I, Bueter M, et al. the data and the accuracy of the data analysis. gastric bypass: effects on feeding behavior and Changes in eating behaviour and meal pattern underlying mechanisms. J Clin Invest 2015;125: following Roux-en-Y gastric bypass. Int J Obes References 939–948 2012;36:348–355 1. Fontaine KR, Redden DT, Wang C, Westfall 18. O’Brien PE, McPhail T, Chaston TB, Dixon JB. 36. Dixon JB, le Roux CW, Rubino F, Zimmet P. AO, Allison DB. Years of life lost due to obesity. Systematic review of medium-term weight loss Bariatric surgery for type 2 diabetes. Lancet JAMA 2003;289:187–193 after bariatric operations. Obes Surg 2006;16: 2012;379:2300–2311 2. Ng M, Fleming T, Robinson M, et al. Global, 1032–1040 37. Vogt BA, Vogt L, Laureys S. Cytology and regional, and national prevalence of overweight 19. Carlsson LM, Peltonen M, Ahlin S, et al. functionally correlated circuits of human pos- and obesity in children and adults during 1980- Bariatric surgery and prevention of type 2 dia- terior cingulate areas. Neuroimage 2006;29: 2013: a systematic analysis for the Global Burden betes in Swedish obese subjects. N Engl J Med 452–466 of Disease Study 2013. Lancet 2014;384:766–781 2012;367:695–704 38. Cavanna AE. The precuneus and conscious- 3. Unwin N, Gan D, Whiting D. The IDF diabetes 20. Courcoulas AP, Goodpaster BH, Eagleton JK, ness. CNS Spectr 2007;12:545–552 atlas: providing evidence, raising awareness and et al. Surgical vs medical treatments for type 2 39. Bruce JM, Hancock L, Bruce A, et al. promoting action. Diabetes Res Clin Pract 2010; diabetes mellitus: a randomized clinical trial. Changes in brain activation to food pictures af- 87:2–3 JAMA Surg 2014;149:707–715 ter adjustable gastric banding. Surg Obes Relat 4. Carnell S, Gibson C, Benson L, Ochner CN, 21. Mingrone G, Panunzi S, De Gaetano A, et al. Dis 2012;8:602–608 Geliebter A. Neuroimaging and obesity: current Bariatric surgery versus conventional medical ther- 40. Scholtz S, Miras AD, Chhina N, et al. Obese knowledge and future directions. Obesity Rev apy for type 2 diabetes. N Engl J Med 2012;366: patients after gastric bypass surgery have lower 2012;13:43–56 1577–1585 brain-hedonic responses to food than after gas- 5. Chechlacz M, Rotshtein P, Klamer S, et al. 22.SchauerPR,KashyapSR,WolskiK,etal. tric banding. Gut 2014;63:891–902 Diabetes dietary management alters responses Bariatric surgery versus intensive medical ther- 41. Hong S, Hikosaka O. The globus pallidus to food pictures in brain regions associated with apy in obese patients with diabetes. N Engl J sends reward-related signals to the lateral motivation and emotion: a functional magnetic Med 2012;366:1567–1576 habenula. Neuron 2008;60:720–729 resonance imaging study. Diabetologia 2009;52: 23. Panunzi S, Carlsson L, De Gaetano A, et al. 42. van der Laan LN, de Ridder DT, Viergever 524–533 Determinants of diabetes remission and glyce- MA, Smeets PA. The first taste is always with 6. Scharmuller¨ W, Ubel¨ S, Ebner F, Schienle A. mic control after bariatric surgery. Diabetes the eyes: a meta-analysis on the neural corre- Appetite regulation during food cue exposure: a Care 2016;39:166–174 lates of processing visual food cues. Neuro- comparison of normal-weight and obese women. 24. Ochner CN, Kwok Y, Conceição E, et al. Se- image 2011;55:296–303 Neurosci Lett 2012;518:106–110 lective reduction in neural responses to high 43. Heni M, Kullmann S, Ketterer C, et al. Dif- 7. Stoeckel LE, Weller RE, Cook EW 3rd, Twieg calorie foods following gastric bypass surgery. ferential effect of glucose ingestion on the neu- DB, Knowlton RC, Cox JE. Widespread reward- Ann Surg 2011;253:502–507 ral processing of food stimuli in lean and system activation in obese women in response 25. Frank S, Wilms B, Veit R, et al. Altered brain overweight adults. Hum Brain Mapp 2014;35: to pictures of high-calorie foods. Neuroimage activity in severely obese women may recover af- 918–928 2008;41:636–647 ter Roux-en Y gastric bypass surgery. Int J Obes 44. Guthoff M, Grichisch Y, Canova C, et al. In- 8. Rothemund Y, Preuschhof C, Bohner G, et al. 2014;38:341–348 sulin modulates food-related activity in the cen- Differential activation of the dorsal striatum by 26. Ochner CN, Laferrere` B, Afifi L, Atalayer D, tral nervous system. J Clin Endocrinol Metab high-calorie visual food stimuli in obese individ- Geliebter A, Teixeira J. Neural responsivity to 2010;95:748–755 uals. Neuroimage 2007;37:410–421 food cues in fasted and fed states pre and post 45. Heni M, Kullmann S, Preissl H, Fritsche A, 9. Berridge KC. ‘Liking’ and ‘wanting’ food re- gastric bypass surgery. Neurosci Res 2012;74: Haring¨ HU. Impaired insulin action in the human wards: brain substrates and roles in eating dis- 138–143 brain: causes and metabolic consequences. Nat orders. Physiol Behav 2009;97:537–550 27. van de Sande-Lee S, Pereira FR, Cintra DE, Rev Endocrinol 2015;11:701–711 10. Finlayson G, King N, Blundell JE. Is it possi- et al. Partial reversibility of hypothalamic dys- 46. Frank S, Kullmann S, Veit R. Food related ble to dissociate ‘liking’ and ‘wanting’ for foods function and changes in brain activity after body processes in the insular cortex. Front Hum Neu- in humans? A novel experimental procedure. mass reduction in obese subjects. Diabetes rosci 2013;7:499 Physiol Behav 2007;90:36–42 2011;60:1699–1704 47. Poirier P, Auclair A. Role of bariatric surgery 11. Ochner CN, Stice E, Hutchins E, et al. Re- 28. Pudel D, Westenhofer¨ J. Fragebogen zum in diabetes. Curr Cardiol Rep 2014;16:444 lation between changes in neural responsivity Eßverhalten (FEV). Handanweisung.Gottingen,¨ 48. Sjostr¨ om¨ L, Lindroos AK, Peltonen M, et al.; and reductions in desire to eat high-calorie Germany, Hogrefe, 1989 Swedish Obese Subjects Study Scientific Group. foods following gastric bypass surgery. Neuro- 29. Lowe MR, Butryn ML. Hedonic hunger: a new Lifestyle, diabetes, and cardiovascular risk fac- science 2012;209:128–135 dimension of appetite? Physiol Behav 2007;91: tors 10 years after bariatric surgery. N Engl J 12. Frank S, Veit R, Sauer H, et al. 432–439 Med 2004;351:2683–2693 depletion reduces food-related reward activity 30. Beck AT, Ward CH, Mendelson M, Mock J, 49. Duarte JV, Pereira JM, Quendera B, et al. independent of BMI. Neuropsychopharmacol- Erbaugh J. An inventory for measuring depres- Early disrupted neurovascular coupling and ogy 2016;41:1551–1559 sion. Arch Gen Psychiatry 1961;4:561–571 changed event level hemodynamic response 13. Jiang T, Soussignan R, Schaal B, Royet JP. 31. Tsvetanov KA, Henson RN, Tyler LK, et al. function in type 2 diabetes: an fMRI study. Reward for food odors: an fMRI study of liking The effect of ageing on fMRI: correction for the J Cereb Blood Flow Metab 2015;35:1671–1680