ImportanceThe extent to which neuroanatomical variability associated with early substance involvement, which is associated with subsequent risk for substance use disorder development, reflects preexisting risk and/or consequences of substance exposure remains poorly understood.ObjectiveTo examine neuroanatomical features associated with early substance use initiation and to what extent associations may reflect preexisting vulnerability.Design, Setting, and ParticipantsCohort study using data from baseline through 3-year follow-up assessments of the ongoing longitudinal Adolescent Brain Cognitive Development Study. Children aged 9 to 11 years at baseline were recruited from 22 sites across the US between June 1, 2016, and October 15, 2018. Data were analyzed from February to September 2024.ExposuresSubstance use initiation through 3-year follow-up (ie, age <15 years).Main Outcomes and MeasuresSelf-reported alcohol, nicotine, cannabis, and other substance use initiation and baseline magnetic resonance imaging (MRI)–derived estimates of brain structure (ie, global and regional cortical volume, thickness, surface area, sulcal depth, and subcortical volume). Covariates included family (eg, familial relationships), pregnancy (eg, prenatal exposure to substances), child (eg, sex and pubertal status), and MRI (eg, scanner model) variables.ResultsAmong 9804 children (mean [SD] baseline age, 9.9 [0.6] years; 5160 boys [52.6%]; 213 Asian [2.2%], 1474 Black [15.0%], 514 Hispanic/Latino [5.2%], 29 American Indian [0.3%], 10 Pacific Islander [0.1%], 7463 White [76.1%], and 75 other [0.7%]) with nonmissing baseline neuroimaging and covariate data, 3460 (35.3%) reported substance use initiation before age 15. Initiation of any substance or alcohol use was associated with thinner cortex in prefrontal regions (eg, rostral middle frontal gyrus, β = −0.03; 95% CI, −0.02 to −0.05; P = 6.99 × 10−6) but thicker cortex in all other lobes, larger globus pallidus and hippocampal volumes, as well as greater global indices of brain structure (eg, larger whole brain volume, β = 0.05; 95% CI, 0.03 to 0.06; P = 2.80 × 10−8) following Bonferroni or false discovery rate multiple testing correction. Cannabis use initiation was associated with lower right caudate volume (β = −0.03; 95% CI, −0.01 to −0.05; P = .002). Post hoc examinations restricting to postbaseline initiation suggested that the majority of associations, including thinner prefrontal cortex and greater whole brain volume, preceded initiation.Conclusions and RelevanceIn this cohort study of children, preexisting neuroanatomical variability was associated with substance use initiation. In addition to putative neurotoxic effects of substance exposure, brain structure variability may reflect predispositional risk for initiating substance use earlier in life with potential cascading implications for development of later problems.