Patterns of structural brain differences in cocaine use disorder: Integrating multilevel frequentist, Bayesian, and...

Cocaine Use Disorder (CUD) has been associated with differences in gray matter (GM) measures, but prior meta-analyses have yielded inconsistent results due to methodological variability, publication bias, and selective reporting. To address these gaps, this study integrates multilevel frequentist, Bayesian, and coordinate-based approaches to provide a comprehensive, anatomically resolved synthesis of structural neuroimaging findings in CUD.

Following PRISMA guidelines (PROSPERO: CRD42024585260), 47 studies were included that compared GM indices between individuals with CUD and healthy controls. Multilevel random-effects meta-analyses were conducted to quantify cortical and subcortical group differences while accounting for clinically relevant moderators. Bayesian hierarchical models were used to conduct sensitivity analyses addressing potential reporting bias, and a coordinate-based Activation Likelihood Estimation (ALE) analysis evaluated spatial convergence across voxel-based morphometry studies.

The analyses indicated that CUD showed lower GM measures in prefrontal and temporal cortical regions, particularly the ITG and the rostral middle frontal gyrus. Subcortical analyses showed lower GM volume in the amygdala, hippocampus, thalamus, pallidum, nucleus accumbens, brainstem, and cerebellum. Longer duration of cocaine use and greater clinical severity were associated with larger negative effect-size estimates for cortical measures. Bayesian regularization yielded consistent cortical effect-size estimates across sensitivity analyses, whereas the coordinate-based ALE analysis identified no clusters surviving family-wise error correction.

Together, results from this meta-analysis identify consistent patterns of GM differences across cortico-subcortical regions in individuals with CUD. These findings provide an integrated neuroanatomical reference framework that may inform future longitudinal studies and the evaluation of candidate neuroimaging biomarkers.