Retention and data exclusion challenges for representative longitudinal neuroimaging in the understanding of addiction

Head motion during resting-state functional magnetic resonance imaging (rsfMRI) poses a major challenge for neuroimaging research, often leading to data quality concerns and participant exclusions, particularly among pediatric and clinical populations. Although necessary for ensuring reliable data, motion-related exclusions may inadvertently bias samples by disproportionately excluding certain sociodemographic groups. Using data from the Adolescent Brain Cognitive Development (ABCD) Study, we employed both frequentist and Bayesian approaches to examine how head motion-related exclusions and participant retention shape sample composition over time. At baseline (ages 9-10), Black, Hispanic, and Asian youth were significantly more likely than White youth to be excluded due to excessive head motion; these disparities were not observed at the two-year follow-up (ages 11-13). In contrast, disparities in retention persisted; Black participants were less likely to return for follow-up, even after accounting for socioeconomic factors and motion. Together, these findings highlight how both motion-related exclusions and differential retention can systematically influence the representativeness of longitudinal neuroimaging samples, with important implications for the generalizability of research in addiction neuroscience.