3.1 Study characteristics

Tables 1 and 2 provide a summary of the participant characteristics, methodological design, and key findings of each eligible study that reports fMRI based predictive outcomes (Table 1) and response outcomes (Table 2) respectively.

Table 1 fMRI predictive outcomes: summary of study characteristics and main findingsTable 2 fMRI response outcomes: summary of study characteristics and main findings

From the 57 eligible papers, 7 reported fMRI findings on predictive outcomes only, 41 reported findings on response outcomes only, and 9 reported both predictive and response outcomes. Studies varied in the population sampled (50 studies for adults with overweight or obesity, 4 for children and adolescents with obesity, 3 for adults with binge eating), type of intervention (13 pharmacological, 16 surgical, 7 psychological, and 21 lifestyle), and the fMRI food cue related tasks employed (50 studies used passive-viewing cue reactivity, and 7 adapted cognitive tasks with food-related stimuli).

Across the eligible studies, sample sizes ranged from 5 to 148 (M = 33; SD = 24), yielding data from 1885 participants in total. Females represented the majority of participants (71%), although 12 of our studies reported male participants being the majority or the only sample (e.g., [21]). The mean age for the fifty-seven studies was 38.7 years, and the BMI for obese and binge eating populations had a mean of 35.9 kg/m2.

3.2 Neuroimaging findings summary

Figure 2 illustrates the conceptual framework used for fMRI predictive outcomes (Fig. 2 A) and fMRI response outcomes (Fig. 2B). The figure also provides a summary of the main findings for each type of outcome, participant group (adults with overweight and obesity, children and adolescents with obesity, and adults with binge eating) and intervention (pharmacological, surgical, psychological, and lifestyle).

Fig. 2The alternative text for this image may have been generated using AI.

Conceptual framework for (A) fMRI predictive outcomes and (B) fMRI response outcomes, and summary of findings per type of outcome, participants and intervention. In brackets is the number of studies for each type of outcome (predictive or response), participants (adults with overweight/obesity, children and adolescents with obesity, adults with binge eating) and intervention (pharmacological, surgical, psychological, lifestyle); the tally includes 9 studies reporting both predictive and response outcomes. fMRI contrasts included food vs. non-food stimuli, and/or high-energy-density vs. low-energy-density food stimuli. (A) fMRI predictive outcomes refers to studies using fMRI brain activation to food cues at baseline to predict clinical outcomes post-intervention. Clinical outcomes varied across studies, with weight/BMI loss and binge eating behaviour being the most used metric for participants with overweight/obesity and binge eating respectively. Findings are summarised by noting the direction of activation at baseline (higher ↑ or lower ↓) in the brain regions associated with better clinical outcomes post-intervention; (B) fMRI response outcomes refers to studies measuring change in fMRI brain activation to food cues from baseline to post-intervention. Findings are summarised by noting the direction of change in activation (increase ↑ or decrease ↓) from baseline to post-intervention in the brain regions associated with better clinical outcomes post-intervention

3.3 Predictive outcomes3.3.1 Adults with overweight and obesity

Fourteen of the eligible studies reported predictive outcomes for adults with overweight and obesity (Sample size: M = 48.10, SD = 36.17, 80% females; Age: M = 44.11, SD = 8.17; BMI: M = 35.87, SD = 3.99). These studies included pharmacological (e.g., GLP-1 agonist) (Sample size: M = 28, SD = 11.31, 51% females; Age: M = 54.65, SD = 6.58; BMI: M = 34.8, SD = 3.96), surgical (e.g., RYGB) (Sample size: M = 69, SD = 2.83, 74% females; Age: M = 37.85, SD = 0.92; BMI: M = 43.6, SD = 1.27), psychological (e.g., cognitive strategy) (Sample size: 148, 77% females; Age: M = 27.9, SD = 6.90; BMI: M = 32, SD = 4.75) and lifestyle (e.g., diet) (Sample size: M = 36.78, SD = 22.51, 84% females; Age: M = 44.97, SD = 5.64; BMI: M = 34.82, SD = 2.33) interventions.

Pharmacological interventions included GLP-1 agonists [32] and serotonin agonists [24]. Farr et al. [24] reported that baseline neural sensitivity to food cues in reward-related (amygdala) and visual (occipital) regions may serve as predictive biomarkers of weight-loss and caloric-intake reduction following Lorcaserin (serotonin agonist) treatment. Ten Kulve et al. [32] on the other hand, did not show significant predictive relationships between neural activity and weight-loss post-intervention with GLP-1 agonists.

Surgical interventions included RYGB, LAGB, and SG procedures [22, 25]. Across these interventions, lower baseline activity in parietal (e.g., supramarginal gyrus, inferior parietal lobule) and occipital/precuneus regions generally predicted clinical outcomes post-surgery, including greater weight-loss and reduced reward-related eating [25]. On the other hand, in RYGB specifically, higher baseline activation in the posterior cingulate gyrus (PCG) to high- vs. low-energy-density food cues also predicted greater weight-loss [22]. This association with weight-loss was interpreted by the authors as potentially reflecting RYGB’s modulatory effects on neural circuits involved in motor planning and goal-directed eating behaviour.

Regarding psychological interventions, only one eligible study informative of predictive biomarkers was included in this review. Stice and colleagues [30] employed a food-specific attention training intervention and reported heightened food cue reactivity in cognitive control (IFG) and evaluative (precuneus) regions as relevant predictors of poorer intervention outcomes, such as increased body fat.

Finally, lifestyle interventions ranged from comprehensive programs combining diet and exercise (e.g., [29]), to those focused on specific weight-loss diets (e.g., [28]). Across these interventions, greater baseline activation in reward (NAc, putamen, insula), cognitive control (DLPFC, IFG, ACC, IPL), and perception related (occipital/lingual/fusiform) networks predicted post-intervention weight-loss, whereas greater baseline activity in OFC/medial frontal regions, which are also implicated in reward valuation, generally predicted poorer outcomes. Additional findings highlighted that the predictive role of control-related regions (DLPFC) may depend on neuroendocrine signalling (GLP-1), suggesting brain–hormone interactions as key moderators of weight-loss outcomes [26]. Moreover, evidence indicates that early adaptation of reward circuitry to dietary changes, rather than baseline activity alone, provides a stronger predictive marker of long-term success [14].

3.3.2 Children and adolescents with obesity

From our eligible studies, only Schur et al. [35] reported predictive outcomes for children with obesity (Sample size: 37, 38% females; Age: M = 10.5, SD = 0.9; BMI: M = 29.5, SD = 7). They examined how pre-treatment neural responses to food cues relate to changes in BMI following Family-Based Therapy (FBT) for 6 months. Results demonstrated that children with greater reductions in reward-related (e.g., orbitofrontal cortex, striatum, insula) activation to high-energy-density food cues following a meal at baseline achieved larger decreases in BMI z-score post-intervention, whereas sustained activation in these regions from pre- to post-meal predicted poorer weight-loss. Exploratory analyses suggested that stronger pre-meal activation in reward (vmPFC) and visual regions (occipital pole) predicted greater BMI reduction, while increased activation in contextual and evaluative processing regions (precuneus) was associated with weaker treatment response. Although these predictive outcomes were significant at the end of the 6-month FBT treatment, baseline cue reactivity in appetite regulation regions did not predict long-term BMI changes at 1-year follow-up. We note that the generalisability of these findings to this population is limited given it reflects results from a single study.

3.3.3 Adults with binge eating

From our eligible studies, only Fleck et al. [15] reported predictive outcomes for a population with binge eating (Sample size: 33, 100% females; Age: M = 38.6, SD = 9.05; BMI: M = 36.6, SD = 6.5). In this study, they report that activity in the vmPFC and subgenual ACC at baseline was inversely associated with self-reported binge eating days per week following a 12-week Lisdexamfetamine (LDX) treatment. Other than neural predictive markers, they also reported that LDX led to reductions of binge episodes, with successful remission (defined as no binge eating episodes in the previous four weeks) in 87% of the sample. We note that the generalisability of these findings to this population is limited given it reflects results from a single study.

3.4 Response outcomes3.4.1 Adults with overweight and obesity

Forty-four of the eligible studies reported response outcomes for adults with overweight and obesity (Sample size: M = 29.39, SD = 18.46, 68% females; Age: M = 41.22, SD = 10.79; BMI: M = 36.25, SD = 5.34). These studies included pharmacological (e.g., GLP-1 agonist) (Sample size: M = 28.85, SD = 20.55, 58% females; Age: M = 48.60, SD = 12.23; BMI: M = 35.07, SD = 5.32), surgical (e.g., RYGB) (Sample size: M = 32, SD = 19.81, 73% females; Age: M = 35.84, SD = 6.42; BMI: M = 41.49, SD = 3.17), psychological (e.g., cognitive strategy) (Sample size: M = 30.75, SD = 14.86, 92% females; Age: M = 33.2, SD = 9.96; BMI: M = 31.60, SD = 4.57) and lifestyle (e.g., diet) (Sample size: M = 24, SD = 15.02, 68% females; Age: M = 43.27, SD = 8.47; BMI: M = 34.14, SD = 2.66) interventions.

Pharmacological interventions included GLP-1 agonists (e.g., [46]), serotonin agonists (e.g., [24]), suppression of satiety hormones (e.g., to explore mechanistic changes following RYGB, [48]), leptin repletion (e.g., [49]), intranasal oxytocin [21], and combined naltrexone and bupropion [67]. Common findings across these pharmacological interventions included modulated activity in reward and incentive valuation regions, with reduced striatal and insula responses to energy-dense food cues, and increased prefrontal activation (DLPFC, IFG), consistent with enhanced cognitive control.

Surgical interventions included RYGB (e.g., [70]), LSG (e.g., [55]) and LAGB (e.g., [39]). Common findings across these surgical interventions included decreased activation in reward-related regions (striatum, OFC) in response to high-calorie foods. Shifts were also observed toward increased activation in control-related regions (PFC, ACC) and reduced visual-perceptual cortex engagement when viewing palatable foods.

For psychological interventions, these included cognitive strategy training (e.g., [42]), food cue exposure therapy (e.g., [48]), and food-specific go/no-go training (e.g., [68]). Across these interventions, cognitive reappraisal and related strategies reduced activity in reward regions (striatum, vmPFC) and increased activity in prefrontal cognitive control regions (DLPFC, superior/middle frontal gyri).

Finally, regarding lifestyle interventions, these ranged from comprehensive programs including both diet and exercise (e.g., [29]), to interventions focused on specific aspects of diet (e.g., effects of reduced fat vs. reduced carbohydrates, [41]) or exercise (e.g. [40],). Overall, lifestyle interventions consistently modulated striatal and OFC activity, with reductions in neural activation to food cues linked to higher adherence and greater weight-loss.

3.4.2 Children and adolescents with obesity

Three eligible studies reported response outcomes for children and adolescents with obesity (Sample size: M = 28, SD = 14.73, 70% females; Age: M = 14.17, SD = 4.37; BMI: M = 28.6, SD = 0.7). These studies included lifestyle (e.g., diet) (Sample size: M = 19.5, SD = 0.71, 82% females; Age: M = 16, SD = 4.24; BMI: M = 28.6, SD = 0.7) and psychological (e.g., FBT) (Sample size: 45, 45% females; Age: M = 10.5, SD = 0.9; BMI: z-scored, M = 2.21, SD = 0.36) interventions.

Regarding lifestyle interventions, two studies reported findings for this population. Hinton et al. [71] analysed the effects of Mandolean® training, a behavioural intervention designed to slow eating and regulate portion sizes. They found that Mandolean® training attenuated striatal (putamen) and temporo-occipital responses to food cues, relative to standard care. Leidy et al. [72] examined the effects of having a normal or high-protein breakfast, versus breakfast-skipping in a population that would usually skip this meal. They found that having breakfast, compared to skipping it, was associated with lower food cue–related activation in the amygdala, hippocampus, and midfrontal gyri, whereas consuming a high versus normal-protein breakfast was associated with further reductions on hippocampal and parahippocampal activation. Combined, these studies suggest that lifestyle interventions in children and adolescents with obesity modulate food cue reactivity in regions implicated in reward and salience processing, typically resulting in reduced engagement of these brain regions following behavioural or dietary modification.

Regarding psychological interventions, Roth et al. [73] looked at changes in neural responses to food cues following Family-Based Therapy (FBT). They found children with obesity who exhibited less meal-induced suppression of activation to high- vs. low-energy-density food cues in the striatum, OFC, amygdala, SN/VTA, and insula showed greater BMI z-score reduction after FBT. This reduced suppression also correlated with increased meal-induced GLP-1, PYY, and ghrelin. Similar patterns were seen in the OFC, occipital pole, and parahippocampal gyrus. These findings suggest that although changes in satiety-related hormones were consistent with weight-loss, persistent attention and motivation for high-calorie foods after eating may remain, potentially signalling susceptibility to weight regain post-intervention.

3.4.3 Adults with binge eating

Three eligible studies reported response outcomes for a binge eating population (Sample size: M = 27.67, SD = 5.51, 100% females; Age: M = 32.83, SD = 7.46; BMI: M = 35.28, SD = 8.35). These studies included surgical (e.g., RYGB) (Sample size: 28, 100% females; Age: M = 35.5, SD = 11.35; BMI: M = 42.9, SD = 3.35) and pharmacological (e.g., LDX) (Sample size: M = 27.5, SD = 7.78, 100% females; Age: M = 31.51, SD = 10.03; BMI: M = 31.48, SD = 7.25) interventions.

Regarding surgical interventions, Baboumian et al. [74] reported lower activation in the dmPFC to high-energy-density food cues and increased activation for low-energy-density cues after RYGB surgery, for both binge eaters and non-binge eaters. Binge eating specific results included reduced middle occipital gyrus activation to high-energy-density (vs. low-energy-density) cues after RYGB surgery, compared to non-treatment controls. Based on these results, reduced middle occipital gyrus activation to high-energy-density cues appears to be a binge eating specific response biomarker after surgery, while dmPFC changes seem to represent a more general obesity-related neural adaptation to RYGB. Indeed, these dmPFC changes align with a reduced salience of palatable foods, commonly reported after bariatric surgery in an obese population [57,58,59].

Regarding pharmacological interventions, two studies in our eligible cohort investigated LDX as a treatment for binge eating. Fleck et al. [15] found that long-term (chronic) LDX administration decreased activation to food cues in the globus pallidus, and that changes in activation to food cues of the vmPFC and the thalamus were positively associated with changes in binge eating severity measures (BES and YBOCS-BE). Schneider et al. [75] reported that acute LDX administration decreased thalamic responses to food cues and attenuated functional connectivity between the thalamus and the insula, while also reporting associated reductions in immediate food intake. Taken together, these findings highlight the thalamus as a central target of LDX pharmacological modulation in binge eating, with reduced activation and connectivity in this region being associated with improvements in binge eating behaviour.