Brain Networks and Adolescent Alcohol Use

Abstract

Importance: Alcohol misuse in adolescence is a leading cause of disability and mortality in youth and is associated with higher risk for alcohol use disorder. Brain mechanisms underlying risk of alcohol misuse may inform prevention and intervention efforts.

Objective: To identify neuromarkers of alcohol misuse using a data-driven approach, with specific consideration of neurodevelopmental sex differences.

Design, setting, and participants: Longitudinal multisite functional magnetic resonance imaging (fMRI) data collected at ages 14 and 19 years were used to assess whole-brain patterns of functional organization associated with current and future alcohol use risk as measured by the Alcohol Use Disorder Identification Test (AUDIT). Primary data were collected by the IMAGEN consortium, a European multisite study of adolescent neurodevelopment. Model generalizability was further tested using data acquired in a single-site study of college alcohol consumption conducted in the US. The primary sample was a developmental cohort of 1359 adolescents with neuroimaging, phenotyping, and alcohol use data. Model generalizability was further assessed in a separate cohort of 114 individuals.

Main outcomes and measures: Brain-behavior model accuracy, as defined by the correspondence between model-predicted and actual AUDIT scores in held-out testing data, Bonferroni corrected across the number of models run at each time point, 2-tailed α < .008, as determined via permutation testing.

Results: Among 1359 individuals in the study, the mean (SD) age was 14.42 (0.40) years, and 729 individuals (54%) were female. The data-driven, whole-brain connectivity approach identified networks associated with vulnerability for future and current AUDIT-defined alcohol use risk (primary outcome, as specified above, future: ρ, 0.22; P < .001 and present: ρ, 0.27; P < .001). Results further indicated sex divergence in the accuracies of brain-behavior models, such that female-only models consistently outperformed male-only models. Specifically, female-only models identified networks conferring vulnerability for future and current severity using data acquired during both reward and inhibitory fMRI tasks. In contrast, male-only models were successful in accurately identifying networks using data acquired during the inhibitory control-but not reward-task, indicating domain specificity of alcohol use risk networks in male adolescents only.

Conclusions and relevance: These data suggest that interventions focusing on inhibitory control processes may be effective in combating alcohol use risk in male adolescents but that both inhibitory and reward-related processes are likely of relevance to alcohol use behaviors in female adolescents. They further identify novel networks of alcohol use risk in youth, which may be used to identify adolescents who are at risk and inform intervention efforts.

Figure 1.. Schematic Overview of Data Collection and Analysis

A, Functional magnetic resonance imaging (fMRI) data were collected at ages 14 and 19 years while participants completed the monetary incentive delay task (reward) and the stop signal task (inhibitory control). Whole-brain connectomes were computed using the 268-node Shen atlas, and node-by-node pairwise Pearson r connectivity correlations were computed using the mean time course for each of the 268 nodes. For each individual and task, r values were transformed using Fisher r to z and combined to create a symmetric 268 × 268 matrix summarizing whole-brain patterns of connectivity for each participant for each task type (ie, reward and inhibitory control connectomes for each individual). Connectomes were then entered into connectome-based predictive modeling analyses predicting scores at age 19 years on the Alcohol Use Disorders Identification Test (AUDIT) to identify models associated with future alcohol use severity (age 14 years fMRI data predicting age 19 years AUDIT data) and current alcohol use severity (age 19 years fMRI data predicting age 19 years AUDIT data). Finally, model features were directly mapped back to neuroanatomy to provide mechanistic insight. B, Connectome-based predictive modeling comprises the following steps: (1) feature selection, in which linear regression is used to identify connectome features that are associated with a target behavioral variable (here, alcohol use severity) in a training data set; (2) feature reduction, in which identified connections are summed to create a single summary value for each individual in the training data set; (3) model building, in which summary scores (the independent variable) are linearly associated with the behavioral variable (the dependent variable); and (4) model application, in which resultant linear models are applied to novel connectomes in a testing data set to generate behavioral predictions. Finally, the predictive ability of the model is evaluated based on the correspondence between model-predicted and actual alcohol use risk scores.

Figure 2.. Anatomy and Virtual Lesioning of Future Severity Networks

Network anatomy is summarized based on spatial overlap with well-established large-scale neural networks and cells are shaded according to the percentage of the total network accounted for by connections between each canonical network pair. The relative importance of connections within each of these established networks is assessed using a virtual lesioning approach, with model performance (ρ) for each model on the x-axis. For female-only future severity networks identified using reward (A) and inhibitory control task (B) data, positive connections included a high degree of within-network medial frontal, motor-sensory, and cerebellar connections, as well as substantial between-network connections with the cerebellar and subcortical networks. The network identified using reward data (A) was further characterized by between-network connections with the medial frontal and motor-sensory networks. For both tasks, negative connections included a high degree of between-network motor-sensory connections. Post hoc virtual lesioning indicated that connectivity within frontoparietal, salience, motor-sensory, and cerebellar networks accounted for the most variance in future alcohol use.

Figure 3.. Anatomy and Virtual Lesioning of Current Severity Networks

Network anatomy is summarized based on spatial overlap with well-established large-scale neural networks and cells are shaded according to the percentage of the total network accounted for by connections between each canonical network pair. The relative importance of connections within each of these established networks is assessed using a virtual lesioning approach, with model performance (ρ) for each model on the x-axis. Female-only positive networks generated from reward (A) and inhibitory control (B) tasks were dominated by motor-sensory, cerebellar, subcortical, and salience network connections, whereas negative networks were largely characterized by motor-sensory, salience, and subcortical connections. For the male-only current severity network identified using inhibitory control data (C), the positive network was dominated by connections between cerebellar and subcortical networks as well as by within-network cerebellar connections, and the negative contained a high degree of cerebellar, subcortical, salience, and medial frontal connections as well as a high degree of within-network subcortical connections. Post hoc virtual lesioning indicated that connectivity within salience, motor-sensory, and cerebellar accounted for the greatest amount of variance in current alcohol use for the female-only networks (A and B). For male individuals, virtual lesioning indicated that motor-sensory, cerebellar, and subcortical connectivity accounted for the highest amount of variance in current alcohol use (C).

Figure 4.. External Replication of Current Severity Network

A, Degree maps illustrate the number of connections from each node for the sex-agnostic current severity network identified from inhibitory control task data. B, Cells are shaded according to the percentage of the total network accounted for by connections between each canonical network pair. The positive network is dominated by connections between cerebellar and subcortical networks as well as within-network connections of the motor-sensory and cerebellar networks. The negative network is characterized by a high degree of within-network subcortical connections as well as connections between subcortical, cerebellar, and salience networks. C, Network strength of the sex-agnostic current severity network identified from inhibitory control data in the IMAGEN data set replicated to predict risky alcohol use in an entirely separate sample (n = 114).