Objective <p>This study aimed to examine the interaction patterns between the reward and inhibitory control systems in children with overweight and obesity.</p> Methods <p>Resting-state fMRI data were collected at baseline from 38 children with overweight and obesity (OW/OB) and 68 children with normal weight (NW). We first examined the differences in seed-based functional connectivity (FC) between the two groups, focusing on eight predefined regions of interest in the reward and inhibitory control systems. Based on the FC results, we further applied the spectral dynamic causal modeling technique to assess the between-group differences in effective connectivity (EC). Finally, we employed a machine learning approach to determine whether these baseline core connections could predict eating behaviors at the one-year follow-up.</p> Results <p>Compared with the NW group, the FC between the left superior frontal gyrus (SFG) and left caudate was stronger while the FC between left SFG and right ventromedial prefrontal cortex (vmPFC) was weaker in the OW/OB group after controlling for age, sex, and head motion. After Bayesian contrasts, the OW/OB group exhibited stronger negative EC from the left caudate to left SFG, weaker negative EC from the left SFG to right vmPFC, and weaker positive EC from the right vmPFC to left SFG than NW group. The results indicated that the baseline caudate→caudate and vmPFC→SFG connectivity could predict changes in children’s eating behaviors one year later.</p> Conclusion <p>The current study provides novel evidence for the neural hierarchical basis of childhood obesity (especially the caudate→SFG, SFG→vmPFC and vmPFC→SFG connectivity), suggesting that interventions targeting reward processing and inhibition control may have important implications for childhood obesity.</p>