Neural Decision-Making and Affective Dynamics in Weight Regain after Metabolic and Bariatric Surgery: A Multimodal...

To determine whether weight regain after metabolic bariatric surgery reflects a systematic reorganization of neural decision signals elicited by food cues and longitudinal affective states, and to develop an interpretable, trackable, and stratifiable multimodal computational psychiatry framework.

This study was based on a multi-timepoint longitudinal cohort of patients undergoing primary metabolic bariatric surgery, integrating a standardized food-choice task, 64-channel EEG, psychometric measures, and longitudinal affective-state data. Food stimuli were first standardized across five domains: subjective experience, health/naturalness attributes, reward-control attributes, nutritional composition, and visual features. EEG data underwent predefined quality-control procedures, after which N2, P3, centroparietal positivity, feedback-related negativity, feedback P3, and frontal theta indices were extracted. Food-choice behavior was decomposed using a hierarchical drift-diffusion model into drift rate, boundary separation, starting point, and non-decision time. A multimodal RT-by-frontal-theta interaction model was then fitted, and GIMME dynamic network modelling was used to characterize shared, subgroup-specific, and individual pathways within affect-reward-control states.

Food stimuli were decomposed into 39 auditable variables, forming a continuous representational space spanning health, reward, control, nutritional, and visual-salience dimensions, with high reliability across domains. EEG quality control preserved clinically expected heterogeneity while excluding recordings at high risk of biasing ERP and time-frequency estimates. Behavioral and DDM results showed that the weight-regain group had longer mean reaction times than the maintained-weight-loss group (0.726 s vs. 0.651 s), together with lower drift rate (0.93 vs. 1.27), lower boundary separation (1.20 vs. 1.37), lower starting point (0.42 vs. 0.51), and longer non-decision time (0.31 s vs. 0.27 s). ERP findings indicated that weight-regain-related differences spanned early control/conflict monitoring, posterior stimulus evaluation, evidence accumulation, and feedback processing, with posterior P3/CPP and feedback-related components contributing the dominant late-stage separation. The multimodal interaction model showed an RT-by-frontal-theta interaction of approximately 0.92% points in the primary adjusted model (95% CI, 0.36-1.49), with a consistent direction across alternative model specifications. GIMME identified stable affect-reward-control dynamic pathways and separated three psychological dynamic subtypes: affect-reactive, reward-sensitized, and control-resilient.

Weight regain after metabolic bariatric surgery is not merely an endpoint of body-weight change, but a dynamic risk process shaped jointly by food-cue evaluation, evidence-accumulation efficiency, frontal control modulation, and longitudinal affect-reward state transitions. This multimodal computational framework reframes postoperative weight regain as a mechanistically interpretable neurobehavioral phenotype, providing a translational basis for earlier risk stratification, individualized follow-up, and interventions targeting emotion, craving, and control.