The nucleus accumbens (NAc) and its prefrontal connections are implicated in the aetiology of chronic low back pain (CLBP). Animal and human studies suggest that the NAc and its connections play a critical role in the transition from acute to CLBP. However, whole-brain connectivity in individuals with longstanding CLBP has not been systematically investigated. Using a functional connectomics approach, we examined whether the 2 NAc subregions-shell and core-exhibit different whole-brain connectivity between CLBP patients and healthy controls (HCs
total N = 197). The identified connections were correlated with CLBP intensity (multiple comparisons corrected), and their reproducibility was validated in 2 independent cohorts. These clinically relevant and reproducible connections were further leveraged to classify CLBP using machine learning. Compared with HC (n = 41), individuals with CLBP (n = 39) exhibited hyperconnectivity between the NAc shell and core and the prefrontal cortex (PFC). Although several NAc-PFC connections were linked to higher CLBP intensity, only the connections between the left NAc shell and core to the right dorsolateral PFC were reproduced in validation cohorts (total CLBP n = 53
HC n = 64). Nucleus accumbens-right dorsolateral PFC connections achieved 84% classification accuracy using logistic regression. The machine learning analyses demonstrate how knowledge-based feature selection can reliably detect CLBP. Overall, we report that NAc-PFC connectivity consistently distinguishes people with CLBP from HC and suggest an abnormal interaction between the NAc and brain regions involved in motivation, decision-making, and pain regulation.