Reward and Cognitive Control Connectivity Tracks Body Mass Index More Than Binge Eating Diagnosis
A new medRxiv preprint reports a clean split between brain signatures of body size and brain signatures of binge eating behavior. In a cohort of 61 women, functional connectivity centered on reward and cognitive control circuits predicted body mass index with moderate strength, yet those same connectivity fingerprints failed to distinguish women with binge eating disorder from women with subsyndromal binge eating or from controls without binge eating. The pattern is clinically provocative because it suggests that neural signals in these circuits may be indexing obesity-related biology more reliably than the diagnostic label of BED itself.
The study focused on two nodes with strong theoretical relevance to eating behavior: the nucleus accumbens for reward processing and the dorsolateral prefrontal cortex for cognitive control. Resting-state scans and task-based scans during a food bidding paradigm were used to derive connectivity fingerprints. Those fingerprints were then tested for their ability to predict group membership or BMI. The design is narrow enough to be mechanistically interpretable, but broad enough to ask whether BED has a reproducible network-level signature at all.
BMI Signals Emerged Across Reward and Control Circuits
BMI was the more tractable target. Both the reward-linked and control-linked seeds robustly predicted BMI, with correlation coefficients ranging from .33 to .46. The dlPFC seed carried predictive value across both resting-state and task conditions, with statistical support in resting state and during the bidding task. The NAcc seed was more selective, predicting BMI only when connectivity was measured during the task.
A single sentence captures the biological logic: heavier body mass appears to be reflected in network interactions between motivation and inhibition systems even when overt eating diagnosis does not separate cleanly. In practical terms, this fits with a model in which chronic excess weight is associated with broad changes in incentive valuation, executive control, or both, while BED as currently defined may remain too heterogeneous to map onto one stable circuit pattern. Task engagement also seems to matter; the food bidding context may have activated reward circuitry in a way that resting state alone could not capture.
Binge Eating Disorder Did Not Yield a Classifiable Connectivity Signature
The sharper result came from the negative finding. Functional connectivity failed to classify binge eating groups better than chance. Women with BED could not be separated reliably from those with subsyndromal BED or from women without binge eating using the tested connectivity fingerprints.
This is not a trivial null result. BED is clinically defined by behavior, distress, and frequency criteria, yet the brain measures selected here did not map onto that categorical boundary. One plausible explanation is that the neural architecture of binge eating is distributed, state-dependent, or diluted by the overlap between binge eating severity and body weight itself. Another is that the seed-based approach, while biologically grounded, may be too constrained to capture the full dimensional variability of reward valuation, impulse control, interoceptive processing, and habit formation that likely contribute to binge eating pathology.
The comparison groups sharpen the interpretation. The three cohorts were matched on age and BMI, which reduces the chance that body size alone explains the group-level contrast. Even with that control in place, the connectivity fingerprints still indexed BMI better than binge eating category. For a condition often discussed through the lens of impaired control over eating, the absence of a robust group classifier is a reminder that symptom-based diagnoses do not always correspond to a single neural signature.
Task-Based Food Decisions May Expose More Relevant Circuit Dynamics
The fact that the NAcc prediction emerged only during the food bidding task deserves attention. Resting-state connectivity can reveal intrinsic network organization, but task-based connectivity may better expose how reward circuits respond when food value is made explicit. A bidding task forces valuation, comparison, and action selection into the same frame, which may be closer to the everyday decisions that shape weight gain than a passive scan.
The dlPFC finding also fits with this interpretation. Cognitive control circuits are not static brakes; they fluctuate with context, effort, and competing reward signals. If BMI is related to the long-term balance between reward drive and self-regulatory capacity, then a prefrontal fingerprint that predicts BMI across both rest and task may be capturing a more trait-like architecture than one tied specifically to binge episodes.
A Preprint With Clear Limits and an Open Biological Question
This work remains a preprint and has not undergone peer review, so the findings should be treated as provisional. The sample was limited to 61 women, and the study examined only three groups defined by binge eating status and BMI matching. The analysis also relied on seed-based functional connectivity, which is informative but not exhaustive; other network-level methods could reveal features missed here.
The data availability statement indicates that group-level statistics are available, while individual connectivity vectors will be shared upon publication. That makes the study transparent in its current form, but the key clinical question is still unsettled: does BED have a reproducible neural fingerprint, or is its brain biology too variable to collapse into a single connectivity marker? For now, this preprint points in a narrower direction. Reward and control network connectivity appears to carry more information about BMI than about binge eating disorder, and any future biomarker work will need to explain why the weight signal is easier to detect than the behavioral diagnosis itself.