Impulsivity-Related Predictors of Adolescent Substance Use Initiation

Background and Aims: Neurodevelopmental models regard impulsivity as a central risk factor for adolescent substance use and subsequent adverse health outcomes. However, prior research has been constrained by cross-sectional designs, relatively small sample sizes, and/or the use of a narrow subset of impulsivity measures. More recent studies using mutli-site consortia, while larger in sample size, have focused on largely data-driven substance use prediction from all available variables, spanning sociodemographic factors to brain imaging, but without strong theory, leaving critical questions on the neurodevelopmental impulsivity theory of adolescent substance use unanswered. Here we aimed to comprehensively test the extent to which measures of impulsivity predict adolescent substance use with large, multi-site data and multiple questionnaire and behavioral impulsivity measures, while following best practices in reproducible analyses of observational data. Method: We leveraged the large, multi-site ABCD dataset (n = 11,868) with baseline and up to three years of longitudinal follow up and assessed impulsivity with three questionnaire (CBCL Externalizing, UPPS-P Impulsive Behavioral Scale, Behavioral Activation System) and three behavioral (Delay Discounting, Stop Signal, and Flanker Task) impulsivity assessments. First, we identified correlations among questionnaire and behavioral impulsivity measures. Next, we assessed their prospective longitudinal and concurrent predictive utility regarding two adolescent substance use outcomes (substance use prior to age 15, and perceived harm of substance use). All analyses used pre-established and pre-registered discovery and validation subsamples to directly test reproducibility. Finally, we used simulation to inform how associations between impulsivity and adolescent substance use and impulsivity vary across sampling strategies (population versus high-risk cohorts) and sample sizes. Findings: Correlations between standard questionnaire and behavioral measures of impulsivity were consistently small and did not exceed our pre-registered threshold (r-equivalent of 0.08). Questionnaires significantly outperformed behavioral measures of impulsivity in predicting both future substance use initiation and harm perception, with negligible incremental benefit from behavioral assessments. Predictions of substance use based on impulsivity were generally modest (e.g. AUC’s ~0.6 - 0.76), exhibiting high sensitivity to both sample size and the base rate of substance use. Large samples (n > 1,000) were needed to have adequate power for impulsivity measures to predict substance use initiation.Conclusion: These results support the involvement of impulsivity in the onset of adolescent substance use, particularly as indicated through questionnaires, but clearly indicate that these factors alone, as currently measured, are insufficient for robust predictions. In non-enriched community samples or population-based datasets, large sample sizes are needed for reproducible impulsivity prediction of adolescent substance use.