Hedonic pathways

In contrast to homeostatic eating, hedonic eating is driven by the reward and motivational value of food; mediated by mesolimbic dopaminergic and opioid pathways^{(Reference Saper, Chou and Elmquist26)}. Foods high in fat and sugar, generate a significant release in dopamine in the nucleus accumbens, often prompting overconsumption^{(Reference Blundell and Macdiarmid27)}. Both systems interact to produce appropriate meal initiation, meal size regulation and meal termination. In obesity, both of these systems can be dysregulated, homeostatic satiety is delayed and hedonic responses to high-energy foods are heightened or blunted^{(Reference Ziauddeen, Farooqi and Fletcher28)}.

The mesolimbic dopaminergic system, comprised of the ventral tegmental area (VTA), nucleus accumbens and their innervations to the prefrontal cortex and amygdala mediates the reward and motivational properties of food^{(Reference Saper, Chou and Elmquist26)}. Dopamine release in the nucleus accumbens in response to palatable food induces the wanting aspect of appetitive eating behaviour, whilst opioidergic signalling mediates the subjective pleasure of eating^{(Reference Guillaumin, Viskaitis and Bracey29)}. Energy-dense foods generate dopamine responses disproportionate to their energy content, reinforcing food-seeking behaviour, observed particularly with high-fat and sugar foods^{(Reference Volkow, Wang and Baler30)}.

Another hypothesis postulates that striatal dopamine D2 receptor availability is reduced, a change also observed in substance use disorders and may drive compensatory overconsumption^{(Reference Ziauddeen, Farooqi and Fletcher28)}. The extent to which this reflects vulnerability in the primary reward system or as a consequence of chronic overconsumption remains contested^{(Reference Ziauddeen and Fletcher31)}. Individual variation in functioning of the reward system may explain why not all individuals with obesity display the same hedonic eating phenotype^{(Reference Ziauddeen, Alonso-Alonso and Hill32)}.

Roux-en-Y gastric bypass: an introduction

Roux-en-Y gastric bypass (RYGB) has historically been characterised as a procedure that works through mechanical restriction and the imposition of behavioural change, as highlighted by the landmark randomised trial by Olbers et al^{(Reference Olbers, Björkman and Lindroos40)}. Following the evolution of the evidence in this field this is mechanistically inaccurate. Weight loss and the behavioural changes observed following RYGB substantially exceed those produced by gastric band procedures, which offers a comparable degree of gastric restriction without influencing gut hormone response^{(Reference Abbott, Adams and Agrawal38)}. In the randomised trial by Olbers et al.,^{(Reference Olbers, Björkman and Lindroos40)} patients undergoing RYGB (n = 37) achieved significantly greater reductions in body fat compared to those who had vertical banded gastroplasty at one year post-operation (n = 46). The RYGB patients also achieved significantly greater loss of trunk fat (p < 0.001) and visceral fat (p = 0.016) at one year^{(Reference Olbers, Björkman and Lindroos40)}. This dissociation between the degree of mechanical restriction and the magnitude of the observed outcomes was one of the earliest indications that the dominant mechanism of action of RYGB is neurobiological and behavioural, rather than restrictive.

The overall effect is caused by a shift in the gut hormone profile, with RYGB, facilitating an accelerated delivery of undigested food to the distal aspect of the small intestine; this consequently causes L-cell stimulation causing a marked and sustained rise in circulating GLP-1 and PYY^{(Reference Guerrero-Hreins, Foldi and Oldfield39,Reference Miras and le Roux41,Reference le Roux, Aylwin and Batterham42)} . Simultaneously, ghrelin is suppressed, due to the removal of the gastric fundus from the alimentary stream. Within the framework of this review, the resultant changes experienced following bariatric procedures can be examined across two domains: homeostatic eating behaviour and non-homeostatic eating behaviour.

RYGB: non-homeostatic eating behaviour

The evidence base for non-homeostatic eating after RYGB is significantly more complex, with the evolution of evidence over the last two decades highlighting the importance of directly measured data. The earliest evidence regarding this type of behaviour post-RYGB came from verbal and written reports, suggesting that RYGB produced a shift in food preference away from high-fat and high-sugar foods. Olbers et al.^{(Reference Olbers, Björkman and Lindroos40)} reported that 30% of RYGB patients reported actively avoiding fat, with the RYGB group reducing the proportion of dietary energy derived from fat significantly more than the banding gastroplasty patient group, at one year post-intervention. There was also a significant difference in the food groups consumed by each treatment group: the RYGB group had a significantly lower percentage energy intake in the desserts category compared to the banded gastroplasty group and consumed significantly more from the fruit and vegetable category^{(Reference Olbers, Björkman and Lindroos40)}. These observations were reinforced by direct measurement studies by Mathes and Spector^{(Reference Mathes and Spector47)}, who used laboratory food-selection assessments, reporting changes in food selection and taste responsiveness. Multiple studies have reported significant energy intake changes following RYGB, namely in fats and carbohydrates. Initially there is a decrease in percentage calories from fat in the first 12 months following RYGB, which ranges from −3.7 to −2%^{(Reference Coughlin, Bell and Bivins48–Reference Brolin, Robertson and Kenler50)}. This was further supported by a long-term follow-up study, displaying a −2% change in percentage calories from fat at 8 years post-operation^{(Reference Kruseman, Leimgruber and Zumbach51)}. The apparent shift in food preference was further supported by controlled animal studies, in which rodent models of RYGB demonstrated a reduced intake and preference for high-fat and high-sugar foods. Rodents on a high-fat diet pre-procedure consumed 1.6 ± 0.8 g/day versus 17.6 ± 0.5 g/day in the sham-procedure group at 15 days post-procedure^{(Reference Seyfried, Miras and Bueter52)}.

This emerging consensus, highlighting that patients following RYGB had a significantly altered food preference, was subsequently complicated by inconsistencies in the literature^{(Reference Miras, Jackson and Jackson53)}. A systematic review found the evidence for a genuine and durable shift in food preference to be mixed and noted that the studies reporting this shift relied predominantly on self-reported methodologies^{(Reference Redpath, Livingstone and Dunne54)}. Given the vulnerability of these studies to both reporter and researcher bias, the reported changes in preference is reflected by the limitations of the measurement tools as opposed to a true change in the type of food patients consumed.

Mechanistic studies provided more objective evidence that the reward value of food is reduced following RYGB. Using a progressive ratio task methodology, Miras et al.^{(Reference Miras, Jackson and Jackson53)} studied patients two weeks before and eight weeks post-RYGB alongside normal-weight controls. The median breakpoint for a sweet was reduced by 50% after RYGB, whereas the breakpoint for vegetables was unchanged, indicating a selective reduction in the reward value of energy-dense food rather than a reduction in motivation. Importantly, the patients that had the largest reduction in breakpoint had the largest decrease in BMI^{(Reference Miras, Jackson and Jackson53)}.

Functional MRI (fMRI) provides insight into the neural correlates of non-homeostatic mechanisms of eating behaviour following RYGB. Scholtz et al.,^{(Reference Scholtz, Miras and Chhina44)} compared BMI-matched patients following RYGB (n = 21) and gastric banding (n = 20) with BMI-matched controls (n = 20). Patients following RYGB showed significantly lower activation of the brain reward systems in response to high-calorie food images; this included the orbitofrontal cortex, amygdala, caudate, putamen and nucleus accumbens. This was accompanied by lower palatability and appeal-ratings for high-calorie foods in the RYGB group. Furthermore, the RYGB group had significantly higher postprandial anorexigenic gut hormones (GLP-1 and PYY), bile acids and dumping syndrome symptoms, indicating the altered gut hormone response as a mediator of the reduced food reward^{(Reference Scholtz, Miras and Chhina44)}.

The role of these gut hormones were then demonstrated directly using the somatostatin analogue octreotide to suppress postprandial gut hormones in patients more than five months post-RYGB. Ocreotide suppressed postprandial GLP-1, PYY and simultaneously increased appetitive food reward, increasing progressive ratio task breakpoint and reward system blood-oxygen level dependent (BOLD) signalling on fMRI. This provided direct evidence that the elevated postprandial gut hormone response is mechanistically responsible for the reduction in food reward following RYGB^{(Reference Goldstone, Miras and Scholtz55)}.

The decisive question is whether these measurable changes in food reward translate into observed changes in food choices. The directly measured evidence indicates that, for most patients, they do not. In the residential study by Livingstone and Price^{(Reference Livingstone, Redpath and Naseer46)}, explicit liking for sweet foods fell significantly at 3 months and implicit wanting for sweet foods fell at both 3 months and 12 months. These reductions however, were not matched by any reduction in the intake of sweet foods, and the macronutrient composition of the diet did not change^{(Reference Livingstone, Redpath and Naseer46)}. This dissociation between a measurable fall in food reward and an unchanged pattern of food intake is the central finding of this study, further casting doubt on the validity of the self-report by study participants. Furthermore, energy intake was underreported by 30% at 24 months, highlighting the unreliability of self-report in populations of people with normal weight and those living with obesity. This is further supported by Søndergaard Nielsen et al.,^{(Reference Søndergaard Nielsen, Rasmussen and Just Christensen56)} demonstrating that RYGB patients showed no macronutrient distribution change despite significant changes in self-reported food preferences.

Based on recent evidence, these findings indicate that the earlier reports of a shift in food preference most likely reflected measurements taken too soon following surgery and based on self-report measures^{(Reference Mashkoori, Ibrahim and Shahsavan57)}. The reductions in food reward demonstrated by the progressive ratio task and fMRI are biologically valid and mechanistically attributable to the gut hormone response^{(Reference Scholtz, Miras and Chhina44,Reference Livingstone, Redpath and Naseer46)} . Despite this, in most patients they do not translate into a change in dietary composition. At best, in a subset of patients, a genuine reduction in both sweet-food preference and energy intake may occur and may serve as an additional mechanism of weight loss^{(Reference Livingstone, Redpath and Naseer46)}.

Sleeve gastrectomy

Sleeve gastrectomy (SG) is currently the most commonly performed bariatric procedure worldwide, involving resection of up to 80% of the stomach along the greater curvature, removing the ghrelin-rich gastric fundus. The primary mechanism of appetite modification differs from RYGB: ghrelin suppression is the dominant early effect, whilst GLP-1 and PYY elevation is attenuated relative to RYGB^{(Reference McCarty, Jirapinyo and Thompson58)}. The evidence on eating behaviour after SG is less extensive than for RYGB.

With respect to homeostatic eating behaviour, the primary mechanism of appetite modification after SG differs significantly from RYGB. Suppression of ghrelin, resulting from resection of the fundus, is the dominant and most consistent early hormonal effect, whereas the rise in GLP-1 and PYY is present but less heightened relative to RYGB^{(Reference McCarty, Jirapinyo and Thompson58)}.

For non-homeostatic eating behaviour, changes in taste perception have been reported following SG^{(Reference Mashkoori, Ibrahim and Shahsavan57)}. A cross-sectional study found altered taste perception at twelve months post-operatively, with reduced consumption of high-fat, high-sugar and ultra-processed foods^{(Reference Mashkoori, Ibrahim and Shahsavan57)}. As with the earlier RYGB literature, these findings rest substantially on self-report, with directly measured data not yet available for SG. A progressive ratio task study assessed the appetitive domain up to six years following SG compared to controls using the same methodology applied to RYGB. Patients were prospectively assessed pre-operatively, 1 year and 6 years post-operatively. Results displayed a non-significant difference between the median breakpoint at the pre-operative and 6 year timepoints (320 versus 60, p = 0.07). Additionally, there were no significant differences between visual analogue scale scores for hunger, fullness, creaminess or sweetness at 6 years follow-up (p > 0.05)^{(Reference Althukair, Abdeen and le Roux59)}. The extent to which any change in food preference after SG translates into altered food intake therefore remains uncertain at present.

Gastric banding

Adjustable gastric banding works through neuronal signalling. A silicone band is laparoscopically placed around the proximal aspect of the stomach, with weight loss and eating behaviour changes determined by vagal stretch receptor activation^{(Reference Burton and Brown60)}. Unlike RYGB and SG, gastric banding was thought to not alter the gut hormone response, as it produces neither the rise in GLP-1 and PYY nor the suppression of ghrelin^{(Reference Ionut, Burch and Youdim61)}. Dixon et al.^{(Reference Dixon, Dixon and O’Brien62)} investigated this with blind cross-over breakfast tests between 17 weight-stable gastric banding patients compared to 17 weight-stable controls, under two conditions; optimal band restriction and reduced restriction. The gastric bypass patients displayed significant differences in satiety visual analogue scale results compared to the controls.

Despite this, gastric banding does not produce the reduction in food reward or the changes in hedonic behaviour seen after RYGB. In the fMRI study conducted by Scholtz et al.,^{(Reference Scholtz, Miras and Chhina44)} banding patients retained significantly higher reward-system activation to high-calorie food compared to RYGB patients, despite comparable weight loss and hunger. Ultimately, in the absence of significant gut hormone changes and minimal impact on the hedonic pathways, gastric banding does not reproduce the eating behaviour changes that follow RYGB and SG. Longer-term outcomes are correspondingly inferior, with higher rates of suboptimal response or recurrent weight gain^{(Reference O’Brien, Hindle and Brennan63)}.

Obesity management medications

Prescribing rates of obesity medications have drastically increased in the last decade, primarily driven by the development of incretin therapies with significantly superior efficacy profiles compared to previous agents^{(Reference Li, Varghese and Shah64)}. Currently approved therapies in most countries are GLP-1 based agonists, orlistat and naltrexone/bupropion.

This review focuses on the neurobiological evidence base of eating behaviour modification with obesity pharmacotherapy. While the mechanistic effects of bariatric surgery are well characterised through objective behavioural and experimental studies, corresponding evidence for pharmacotherapy remains comparatively limited. As the evidence base for bariatric surgery is substantially more mature than for pharmacotherapy, we therefore use insights from these interventions as a mechanistic framework to interpret the emerging evidence on obesity medications, with a particular focus on how these interventions differentially influence appetitive and consummatory aspects of eating behaviour over time.

The distinction between appetitive and consummatory behaviour is particularly important when it comes to self-report questionnaires. These primarily capture appetitive behaviour, whilst measurement of consummatory behaviour is primarily through direct measures, 24-hour intake or food frequency questionnaires (FFQ), however, still rely upon an element of self-reporting. Therefore, conclusions from these measurement modalities cannot be assumed to directly apply to the other. This is a significant limitation in interpreting the existing evidence for both surgical and pharmacological treatments; as both appetitive and consummatory measurement modalities are subject to bias.

Homeostatic eating behaviour

To date Blundell et al.^{(Reference Blundell, Finlayson and Axelsen73)} has provided the most rigorous early data for GLP-1 receptor agonists, demonstrating that semaglutide 1 mg significantly reduced ad libitum energy intake by 24% at 12 weeks, after a standardised breakfast across subsequent daytime meals compared to placebo, measured in a controlled buffet setting. The most significant change occurred at lunchtime with a reduced energy intake of −1255kJ during an ad libitum lunch meal, compared to placebo. Critically, the measured effects were independent of nausea-mediated energy reduction, establishing semaglutide as an appetite suppressant as opposed to an emetic. This is further supported by van Can et al.^{(Reference Van Can, Sloth and Jensen74)}, demonstrating increased mean postprandial satiety and reduced ad libitum energy intake by 16% compared to placebo.

Martin et al.,^{(Reference Martin, Carmichael and Carnell75)} also demonstrated that tirzepatide reduced energy intake by a mean of 2194.9kJ versus placebo at three weeks post-treatment during an ad libitum lunch meal test; with results also showing significant reductions in overall appetite, food craving, perceived hunger and reactivity to food cues.

At least in the short term, total energy intake is reduced by 24–47% in patients using GLP-1 receptor agonists, with even greater energy intake reductions of 59–72% observed in GIP/GLP-1 receptor agonists^{(Reference Blundell, Finlayson and Axelsen73,Reference Martin, Carmichael and Carnell75–Reference Friedrichsen, Breitschaft and Tadayon77)} .

Non-homeostatic eating behaviour

Blundell et al.^{(Reference Blundell, Finlayson and Axelsen73)} examined food preference directly using the Leeds Food Preference Task, presenting participants with images of food with varying fat content and sweetness to measure both linking and implicit wanting. Semaglutide 1 mg produced significantly lower explicit liking for high-fat, non-sweet foods compared with placebo (p = 0.0016), with no significant difference in liking for other food categories. Implicit wanting followed the same pattern, being significantly lower for high-fat, non-sweet foods (p = 0.0203). Energy intake from high-fat, non-sweet foods was approximately 35% lower with semaglutide than placebo (p = 0.0184) by assessment using covert evening snack boxes. These findings indicate that the shift in food preference is directed away from fatty, energy-dense foods, rather than away from sweet foods.

The observed changes in food preferences may be underpinned by emerging evidence that GLP-1 receptor agonists change the reward value of food. This has primarily been shown through the use of fMRI studies. Van Ruiten et al.^{(Reference van Ruiten, Veltman and Schrantee78)} demonstrated that exenatide reduced brain responses to anticipation and consumption of energy-dense foods compared to placebo in patients with type 2 diabetes. Martin et al.^{(Reference Martin, Carmichael and Carnell75)} investigated the hedonic drive following treatment with both liraglutide and tirzepatide; conducting a randomised 6-week Phase 1 trial using fMRI alongside eating behaviour instruments. The neuroimaging findings displayed reduced BOLD activation in the medial frontal and cingulate gyri, orbitofrontal cortex and hippocampus in response to high-fat and high-sugar food pictures. These regions are strongly associated with hedonic food valuation – paralleling MRI findings from RYGB studies^{(Reference Scholtz, Miras and Chhina44)}.

At the time of writing there is currently no prospective, covertly measured, multi-timepoint assessment of food intake and food choices in GLP-1 receptor agonist users, largely due to the novelty of this drug class. This is a significant gap in the evidence base that would allow direct comparison of self-reported preference changes with measurable dietary behaviour.