Skip to main content
NCBI home page
JAMA Network logoLink to JAMA Network
. 2023 Aug 30;80(11):1131–1141. doi: 10.1001/jamapsychiatry.2023.2949

Brain Networks and Adolescent Alcohol Use

Sarah W Yip 1,2,✉, Sarah D Lichenstein 1, Qinghao Liang 3, Bader Chaarani 4, Alecia Dager 1, Godfrey Pearlson 1,5, Tobias Banaschewski 6, Arun L W Bokde 7, Sylvane Desrivières 7, Herta Flor 8,9, Antoine Grigis 9, Penny Gowland 10, Andreas Heinz 11, Rüdiger Brühl 12, Jean-Luc Martinot 13, Marie-Laure Paillère Martinot 14, Eric Artiges 15, Frauke Nees 6,8,16, Dimitri Papadopoulos Orfanos 17, Tomáš Paus 18,19,20, Luise Poustka 21, Sarah Hohmann 6, Sabina Millenet 6, Juliane H Fröhner 22, Michael N Smolka 22, Nilakshi Vaidya 23, Henrik Walter 12, Robert Whelan 24, Gunter Schumann 23,25, Hugh Garavan 4,26
PMCID: PMC10469292  PMID: 37647053

Key Points

Question

What functional brain connections are associated with alcohol use risk in adolescents, and are these connections different for male vs female adolescents?

Findings

This cohort study of 1359 adolescents identified common and distinct neurobiological substrates of alcohol use as well as sex divergence in model accuracies, such that models predicting female risk consistently outperformed models predicting male risk.

Meaning

The networks identified in this study and subsequently externally validated represent candidate targets for future prevention and intervention efforts.

This cohort study evaluates neuromarkers associated with alcohol use among adolescents.

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.

Introduction

Adolescence is a critical time for transitions from alcohol misuse and experimentation to disorder due to ongoing brain development during this time.1 This period is therefore considered an essential point for identification of brain circuits contributing to problems later in life, a phenomenon referred to as preaddiction.2 Independent of future adult drinking, high-risk alcohol use during adolescence is itself an adverse health behavior, associated with multiple negative sequelae, and represents a leading cause of disability and mortality.3 By identifying brain mechanisms associated with early risk, we aim to facilitate the future development of novel prevention and intervention approaches to help mitigate problematic alcohol use.

Contemporary theories of alcohol use risk emphasize the dual roles of top-down inhibitory systems involving prefrontal cortical structures and bottom-up subcortical reward systems. However, studies of the brain’s functional connectome through network neuroscience indicate that typical brain maturation involves complex functional reorganization of networks instead of simple linear maturation of individual regions.4,5,6,7,8 These insights have transformed understanding of basic neurodevelopment but are not well characterized within the context of alcohol use risk behaviors.

Sex differences in neurodevelopment have been widely noted, with female individuals displaying accelerated developmental trajectories relative to male individuals.9,10,11 Sex differences in alcohol use during adolescence are also commonly reported.12 Therefore, this study aimed to identify neural networks conferring vulnerability for alcohol use in adolescents with specific consideration of sex differences. We adopted a multivariate predictive modeling framework to identify and cross-validate (ie, test in novel data) brain networks associated with current and future alcohol use risk in a large sample (approximately 1500) of adolescents using neuroimaging data acquired during performance of different cognitive tasks relevant to current neurodevelopmental models: that is, a reward-processing task and an inhibitory control task. Based on contemporary theories of neurodevelopment13,14,15,16,17 (eAppendix in Supplement 1), we anticipated that networks identified during reward-related processes would be most relevant for predicting alcohol use behaviors in younger adolescents (approximately 14 years) vs older adolescents (approximately 19 years), and that networks identified during inhibitory-related processes would be most relevant for predicting alcohol use behaviors in older vs younger adolescents.

Methods

Overview

The IMAGEN consortium collected alcohol and neuroimaging data across 8 sites.18 Ethics committees at each site approved the study. A parent or guardian provided written consent, and adolescents provided verbal assent. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. Figure 1 provides a schematic overview of primary data collection and analysis steps. Baseline data were acquired at age 14 years. Follow-up data were acquired at age 19 years. Analyses focused on predicting alcohol use risk at age 19 years. Data acquired at baseline (age 14 years) were used to identify networks associated with future alcohol use risk, and data acquired at follow-up (age 19 years) were used to identify networks associated with current alcohol use risk.

Figure 1. Schematic Overview of Data Collection and Analysis.

Figure 1.

Open in a new tab

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.

Neuroimaging data were processed using a validated functional connectivity pipeline (eMethods in Supplement 1) to generate individual participant functional connectivity matrices (hereafter referred to as connectomes). Connectomes provide a multivariate summary of an individual’s unique pattern of functional brain organization. While connectomes are relatively distinct across individuals (ie, able to uniquely identify individuals, a process referred to as neural fingerprinting),19,20 an individual’s connectome also varies as a function of cognitive task performance, or brain state. Moreover, recent work demonstrates that brain state manipulation (via cognitive task performance vs resting state) is optimal for identifying core individual difference factors, including predictive neuromarkers of substance use in adults.21,22 Our analyses focus on connectomes derived from neuroimaging data acquired during both reward and inhibitory tasks (Figure 1A). Connectomes were entered into connectome-based predictive models (CPMs) (Figure 1B) to identify networks associated with alcohol misuse.

Participants

Demographic characteristics are provided in the Table. An independent sample of US college students (n = 114; age 17-23 years at scan, mean [SD] age, 18.42 [0.77] years) was used as an external replication sample (in Supplement 1).

Table. Participant Characteristics.

Characteristic No. (%) Sex comparison
Full sample (N = 1359) Female (n = 729) Male (n = 630) Statistic P value
Baseline measures, mean (SD)
Age, y 14.42 (0.40) 14.43 (0.41) 14.41 (0.39) −0.97a .33
AUDIT total score 1.20 (2.23) 1.39 (2.31) 1.20 (2.14) −1.53a .13
Any lifetime alcohol use 666 (49.2) 367 (50.6) 299 (47.7) 1.10b .29
Any lifetime cigarette use 356 (26.3) 189 (26.0) 167 (46.9) 0.08b .78
Any lifetime hash/cannabis usec 84 (6.2) 30 (45.1) 54 (64.3) 11.57b <.001
Any lifetime stimulant used 8 (0.6) 5 (0.7) 3 (0.5) 0.25b .62
Follow-up measures, mean (SD)
Age 19.06 (0.75) 19.07 (0.75) 19.05 (0.74) −0.42a .67
AUDIT score 5.76 (4.54) 5.10 (4.32) 6.53 (4.66) 5.81a <.001
Impulsivity 2.20 (0.43) 2.18 (0.42) 2.22 (0.44) 1.61a .11
Neuroticism 1.74 (0.69) 1.93 (0.68) 1.53 (0.63) −11.18a <.001
Any lifetime alcohol use 1260 (92.7) 670 (91.9) 590 (93.7) 1.52b .22
Any lifetime cigarette use 917 (68.1) 480 (52.3) 437 (69.8) 1.61b .20
Any lifetime hash/cannabis usec 689 (51.2) 315 (43.7) 374 (59.7) 34.57b <.001
Any lifetime stimulant used 124 (9.2) 58 (8.0) 66 (10.5) 2.50b .11
Site
Nottingham 206 (15.2) 109 (15.0) 97 (15.4) 6.34b .50
Dresden 180 (13.2) 92 (12.6) 88 (14.0)
Paris 212 (15.6) 110 (15.1) 102 (16.2)
Berlin 118 (8.7) 71 (9.7) 47 (7.5)
Hamburg 190 (14.0) 96 (13.2) 94 (14.9)
London 172 (12.7) 103 (14.1) 69 (11.0)
Mannheim 160 (11.8) 84 (11.5) 76 (12.1)
Dublin 121 (8.9) 64 (8.8) 57 (9.0)
Open in a new tab

Abbreviation: AUDIT, Alcohol Use Disorders Identification Test.

a

t Value.

b

χ2 Value.

c

Based on European School Survey Project on Alcohol and Other Drugs (ESPAD) and Drug Abuse Screening Test (DAST).

d

Based on ESPAD response to lifetime cocaine, crack, and amphetamine use.

Alcohol Use

Alcohol use was assessed using the Alcohol Use Disorders Identification Test (AUDIT), a well-validated self-report measure of alcohol use risk with total scores ranging from 0 to 40.23 AUDIT scores at age 14 years were low with relatively little variance across the sample (mean [SD] score, 1.36 [2.42]) (eFigure 1 in Supplement 1). Variance in AUDIT scores at age 19 years was higher (mean [SD] score, 5.53 [4.47]), and scores increased significantly from age 14 years to 19 years (t870 = 28.94; P < .001). The Table and eFigure 2 in Supplement 1 present details on other substance use.

Neuroimaging

Functional magnetic resonance imaging (fMRI) data were collected during a reward task (monetary incentive delay24) and during an inhibitory control task (stop signal25). Details on tasks, preprocessing, and connectivity analyses are in the eMethods and eFigures 3 and 4 in Supplement 1.

Connectome-Based Predictive Modeling

As shown in Figure 1B, CPM comprises the following steps: (1) feature selection, wherein regression is used to identify connectome features associated with a behavioral variable of interest (here, alcohol use severity, using a feature selection threshold of P < .05) in a training data set; (2) feature reduction, in which identified connections are summed to create a summary value for each individual in the training data set; (3) model building, wherein summary scores (independent variable) are linearly associated with the behavioral variable (dependent variable); (4) model application, in which resultant linear models are applied to novel connectomes in a testing data set to generate predictions; and (5) model evaluation, in which the predictive ability of the model is evaluated based on the correspondence between model-predicted and actual alcohol use risk scores. Unlike correlation or regression, CPM with built-in cross-validation protects against overfitting by testing the strength of the association in previously unseen data (step 5), increasing the likelihood of replication in future studies and thus applicability to other clinical samples.26

Connectomes for each task and time point were entered into CPMs using a nested cross-validation scheme, wherein 1 test site was left out, and 10-fold cross-validation was used for the left-in training sites. In this approach, data from a single site is left out. The predictive model is then generated using the remaining data from the other 7 sites. Model accuracy is then tested in the left-out site. This approach is the current gold standard for multisite neuroimaging data.27

Both sex-agnostic (both sexes combined) and sex-specific (female and male separately) models were conducted. Model results were Bonferroni corrected across the number of models run at each time point and considered significant at 2-tailed α < .008, as determined via permutation testing.

Network Anatomy and Virtual Lesioning

Consistent with best practice recommendations, we summarized the anatomy of identified networks associated with alcohol use risk via their spatial overlap with well-established large-scale neural networks (eFigure 5 in Supplement 1), hereafter referred to as canonical neural networks.28 Further, consistent with recommendations to maximize mechanistic insight, we used a post hoc virtual lesioning approach in which connections overlapping with all but a single canonical network (eg, default mode) were virtually lesioned, and the model was rerun.28 By doing this for each canonical network separately, we were able to determine the relative predictive accuracy, or feature importance, of connections within each of these established networks. Network masks are available for download.29

Results

Cross-Validated Brain-Behavior Models

Neural Networks Associated With Future Alcohol Use Risk

Among 1359 individuals in the study, the mean (SD) age was 14.42 (0.40) years, and 729 individuals (54%) were female. Analyses of baseline data (age 14 years) found sex divergence in the predictive accuracy of CPMs assessing future alcohol use severity. Specifically, sex agnostic models (reward ρ, 0.17; total edges, 1575; P < .001; inhibition ρ, 0.22; total edges, 1640; P < .001) and female-only models (reward ρ, 0.29; total edges, 3150; P < .001; inhibition ρ, 0.26; total edges, 2119; P < .001) were successful in predicting future alcohol use severity at age 19 years, but male-only models were not (reward ρ, 0.04; total edges, 359; P = .43; inhibition ρ, 0.11; total edges, 391; P = .02). The same pattern of results emerged in post hoc analyses controlling for baseline alcohol use severity and residual motion (eResults and eTable 1 in Supplement 1). Comparisons of model accuracies using Fisher r to z confirmed that the female-only models significantly outperformed male-only models (reward z = 4.0; P < .001; inhibition z = 2.47; P = .01).

Neural Networks Associated With Current Alcohol Use Risk

Analyses of follow-up data (age 19 years) further indicated sex divergence in the predictive accuracy of CPMs assessing current alcohol use severity. Specifically, both sex-agnostic models (reward ρ, 0.17; total edges, 2904; P < .001; inhibition ρ, 0.27; total edges, 2538; P < .001) and female-only models (reward ρ, 0.29; total edges, 2606; P < .001; inhibition ρ, 0.27; total edges, 2134; P < .001) were successful in predicting current severity across both task types. In contrast, male-only models were successful in predicting current severity using connectivity data acquired during inhibitory task performance (ρ, 0.18; total edges, 622; P < .001) but not using data acquired during reward task performance (ρ, 0.03; total edges, 906; P = 0.44), indicating task specificity of alcohol risk models in male individuals only.

The same pattern of results emerged in post hoc analyses controlling for baseline alcohol use severity and residual motion (eTable 2 in Supplement 1). Comparisons of model accuracies indicated that the female-only model generated from reward data significantly outperformed the male-only model generated from reward data (z = 4.4; P < .001). There were no significant differences in predictive model performances between female-only and male-only models generated from inhibitory data (z = 1.67; P = .10).

Anatomy of Alcohol-Risk Networks

Figures 2 and 3 show the anatomy of the identified networks as a function of spatial overlap with canonical neural networks. Anatomy is summarized separately for positive and negative predictive connections. Positive predictive connections are those for which increased connectivity positively predicted alcohol use risk. Negative predictive connections are those for which decreased connectivity positively predicted alcohol use risk. By definition, positive and negative connections cannot overlap directly (as a single connection cannot be both a positive and negative predictor). Nonetheless, as summarized in Figures 2 and 3, positive and negative predictive connections may include connections within and between similar large-scale canonical neural networks. Figures 2 and 3 also show virtual lesioning results corresponding to the relative predictive accuracy or feature importance of connections.

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

Figure 2.

Open in a new tab

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.

Figure 3.

Open in a new tab

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).

Neural Networks Associated With Future Alcohol Use Risk

As shown in Figure 2, neural networks associated with future alcohol use using either reward or inhibitory control task data were similar among female adolescents, indicating the relevance of both processes in conferring vulnerability for alcohol use risk. For both tasks, 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 was further characterized by between-network connections within 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 networks with the highest individual feature weights—that is, amount of variance in future alcohol use associated with connectivity within a given canonical network—included frontoparietal, salience, motor-sensory, and cerebellar networks.

Neural Networks Associated With Current Alcohol Use Risk

As shown in Figure 3, female-only positive networks generated from both 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. Although the male-only networks identified using inhibitory data were sparser than the female networks, they featured connections within and between similar networks: the positive male-only network was dominated by connections between cerebellar and subcortical networks as well as by within-network cerebellar connections; the negative male-only network contained a high degree of cerebellar, subcortical, salience, and medial frontal connections as well as a high degree of within-network subcortical connections.

For female adolescents, post hoc virtual lesioning indicated that networks with the highest individual feature weights—that is, amount of variance in current alcohol use associated with connectivity within a given canonical network—included salience, motor-sensory, and cerebellar networks. For male adolescents, virtual lesioning indicated that networks associated with the highest amount of variance in current alcohol use included motor-sensory, cerebellar, and subcortical regions.

Network Change Over Time

For successful models, individual participant network summary scores were created as the sum of connectivity strengths within positive and negative networks and entered into paired t tests to characterize the development of the identified networks over time. Among female adolescents, connectivity within reward and inhibitory control networks predictive of future alcohol use significantly increased from age 14 years to age 19 years (reward t539 = −7.09; P < .001; inhibition t559 = −4.68; P < .001). Among both female and male adolescents, connectivity within inhibitory control networks predictive of current use also significantly increased from age 14 to 19 (female t = −5.61(df = 559), P < .001; males: t468 = −5.87; P < .001), while changes in connectivity within the reward network predictive of current use in female adolescents from age 14 years to age 19 years did not reach significance (t539 = −1.77; P = .08).

Network Overlap Across Sexes

Despite significant differences in predictive accuracies, networks identified as associated with alcohol use behaviors were relatively consistent across male and female adolescents. Consistent with this, cross-sex analyses demonstrated that models developed in female adolescents at age 19 years generalized to predict alcohol use in male adolescents at age 19 years (reward: ρ, 0.14; P = .001; inhibition: ρ, 0.15; P < .001), and that models developed in male adolescents at age 19 years generalized to predict alcohol use in female adolescents at age 19 years for inhibitory, but not reward, task data (reward: ρ, 0.07; P = .08 inhibition: ρ, 0.26; P < .001).

Specificity and Generalizability

Specificity Analyses

To determine the specificity of the identified anatomical connections to alcohol use, we repeated CPMs while controlling for variation in impulsivity and neuroticism. For all models, predictive accuracies were robust to these factors, indicating that the identified connections were specific to alcohol use and that model accuracies were not driven by a more general latent trait personality factor (eTables 3 and 4 Supplement 1).

As shown in the Table, substance use rates were low at age 14 years but were higher at age 19 years (eg, approximately 64% lifetime cigarette use). Post hoc analyses indicated significant associations between substance use and connectivity within the male-only inhibition model such that connectivity was greater among males with nonalcohol substance use. Follow-up analyses indicated that associations between alcohol use and connectivity remained significant within each group (substance-use positive and substance-use negative) separately (eTables 3-5 in Supplement 1).

Generalizability

The ultimate test of statistical reliability is whether a model produces consistent results in novel individuals and settings. For this reason, in our analyses above, we adopted a rigorous leave-one-site-out cross-validation scheme, in which models were tested in previously unseen testing sites. As a further test of the robustness of the effects, we sought to determine whether the identified neural networks would also be associated with alcohol use risk in an entirely separate sample of 114 adolescents from the Brain and Alcohol Research in College Students (BARCS) study30,31 who were recruited independently from the IMAGEN consortium (eResults in Supplement 1).

BARCS differed from IMAGEN in multiple ways, including a different within-scanner inhibitory control task (go/no-go task), different imaging acquisition parameters, a smaller sample size, recruitment from the US (vs Europe), and different alcohol use risk assessments. Despite these differences, connections identified in IMAGEN as associated with current alcohol use risk replicated to predict risky alcohol use (in this case, defined as the largest number of alcoholic drinks consumed in 24 hours) in this external test sample (ρ, 0.20; total edges, 113; P = .03) (Figure 4). Additional details and analyses related to BARCS are in the eResults in Supplement 1.

Figure 4. External Replication of Current Severity Network.

Figure 4.

Open in a new tab

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).

Discussion

This study leveraged data from the IMAGEN consortium to identify neural networks associated with alcohol use risk using an advanced connectome-based approach. Using data from 2 cognitive tasks with relevance to adolescent risk behaviors—a reward task and an inhibitory control task—we identified both common and distinct neurobiological substrates of early alcohol use risk among female and male adolescents, as well as sex divergence in model accuracies. For female adolescents, models generated using neuroimaging data from both task types were successful in identifying reliable neural signatures of future and current alcohol use risk. In contrast, for male adolescents, only the model generated using inhibitory control task data was successful in predicting current alcohol use risk, and no reliable signature of future risk was identified (ie, brain features identified in training data did not predict alcohol use in previously unseen testing data). Despite these differences, connections identified as commonly associated with current alcohol use risk across male and female adolescents not only generalized to held-out testing sites within IMAGEN, but also predicted alcohol behaviors in an entirely independent cohort of adolescents. Thus, the identified alcohol use risk network may be considered as a robust neuromarker that may be targeted in future prevention and intervention efforts.

Among female adolescents, identified connections included subcomponents of multiple well-established resting state networks, including medial frontal, frontoparietal, salience, motor-sensory, and cerebellar networks. Connectivity within these networks generally increased from age years 14 to age 19 years, concurrent with increases in alcohol use. In addition, despite similar features overall, the relative predictive weights of network subcomponents associated with future (age 14 years) vs current (age 19 years) alcohol use risk varied over time, such that medial frontal and salience networks were primarily associated with future alcohol use risk, whereas cerebellar and motor-sensory networks emerged as primarily associated with current alcohol use risk. These findings are consistent with recent data demonstrating altered cerebellar growth trajectories among adolescents with heavy alcohol consumption32 as well as with much broader literature implicating cerebellar and motor systems in substance-use pathophysiology33,34,35,36,37 (eDiscussion in Supplement 1). Together, these data suggest that targeting prefrontal cortical regions may be particularly relevant to prevention efforts in younger adolescents and that targeting cerebellar and motor-sensory regions may be more relevant for intervention efforts in older adolescents who have already initiated drinking. Consistent with this, network connectivity at age 19 years generalized to predict alcohol use in our external sample of older young adults.

Inhibitory control (vs reward) brain states were more relevant for predicting alcohol use in older adolescents; however, this effect was specific to male adolescents. This finding is consistent with work demonstrating that the accuracy of generalizable brain-behavior models of core individual difference factors (eg, intelligence) differs as a function of both sex and brain state,22,38,39 but for the first time to our knowledge demonstrates similar sex specificity within the context of alcohol use risk. Together, these data suggest that interventions focusing on inhibitory control processes may be particularly effective in combating current alcohol use risk in male adolescents but that both inhibitory and reward-related processes are likely of relevance to current alcohol use behaviors in female adolescents. Sex-specific tailoring of interventions has been proposed previously but is seldom—if ever—grounded in neurobiology. By identifying robust markers of risky alcohol use during development, we hope to support the development of such neurobiologically informed prevention and intervention efforts.40

Strengths and Limitations

This study has several strengths, including its unique longitudinal sample and the use of a wholly data-driven approach with the incorporation of rigorous leave-one-site-out internal validation and external validation using an independent sample. This study also has several limitations, including the absence of a consistent measure of alcohol use severity across internal and external validation samples and the absence of more frequently acquired neuroimaging data that could be used to track trajectories. Further, our external replication sample was primarily White European, so it remains to be seen whether these predictive networks will generalize to more diverse populations.

Conclusion

These data provide a connectome-wide assessment of neural networks subserving alcohol use risk and identify a dimensional neural signature of alcohol use risk in adolescents. External validation supported these findings.

Supplement 1.

eAppendix

eMethods

eFigure 1. Alcohol Use Disorders Identification Test (AUDIT) scores at baseline (age 14)

eFigure 2. Alcohol Use Disorders Identification Test (AUDIT) scores at 5-year follow-up (age 19)

eFigure 3. Outlier detection

eFigure 4. Data inclusion/exclusion for CPM models

eResults

eTable 1. Motion control and sensitivity analyses of models built using neuroimaging data acquired at age 14

eTable 2. Motion control and sensitivity analyses of models built using neuroimaging data acquired at age 19

eTable 3. Specificity analyses of models built using neuroimaging data acquired at age 14

eTable 4. Specificity analyses of models built using neuroimaging data acquired at age 19

eTable 5. Effects of substance-use initiation on network strengths at age 19

eFigure 5. Canonical neural networks

eDiscussion

eReferences

Click here for additional data file. (1.2MB, pdf)
Supplement 2.

Data sharing statement

Click here for additional data file. (135.6KB, pdf)

References

  • 1.Koob GF, Volkow ND. Neurobiology of addiction: a neurocircuitry analysis. Lancet Psychiatry. 2016;3(8):760-773. doi: 10.1016/S2215-0366(16)00104-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.McLellan AT, Koob GF, Volkow ND. Preaddiction—a missing concept for treating substance use disorders. JAMA Psychiatry. 2022;79(8):749-751. doi: 10.1001/jamapsychiatry.2022.1652 [DOI] [PubMed] [Google Scholar]
  • 3.Tapert SF, Eberson-Shumate S. Alcohol and the adolescent brain: what we’ve learned and where the data are taking us. Alcohol Res. 2022;42(1):07. doi: 10.35946/arcr.v42.1.07 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Di Martino A, Fair DA, Kelly C, et al. Unraveling the miswired connectome: a developmental perspective. Neuron. 2014;83(6):1335-1353. doi: 10.1016/j.neuron.2014.08.050 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Yip SW, Potenza MN. Application of Research Domain Criteria to childhood and adolescent impulsive and addictive disorders: implications for treatment. Clin Psychol Rev. 2016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Cai L, Dong Q, Niu H. The development of functional network organization in early childhood and early adolescence: a resting-state fNIRS study. Dev Cogn Neurosci. 2018;30:223-235. doi: 10.1016/j.dcn.2018.03.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Cao M, Wang JH, Dai ZJ, et al. Topological organization of the human brain functional connectome across the lifespan. Dev Cogn Neurosci. 2014;7:76-93. doi: 10.1016/j.dcn.2013.11.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Wu K, Taki Y, Sato K, et al. Topological organization of functional brain networks in healthy children: differences in relation to age, sex, and intelligence. PLoS One. 2013;8(2):e55347. doi: 10.1371/journal.pone.0055347 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Giedd JN, Blumenthal J, Jeffries NO, et al. Brain development during childhood and adolescence: a longitudinal MRI study. Nat Neurosci. 1999;2(10):861-863. doi: 10.1038/13158 [DOI] [PubMed] [Google Scholar]
  • 10.Kaczkurkin AN, Raznahan A, Satterthwaite TD. Sex differences in the developing brain: insights from multimodal neuroimaging. Neuropsychopharmacology. 2019;44(1):71-85. doi: 10.1038/s41386-018-0111-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Lenroot RK, Gogtay N, Greenstein DK, et al. Sexual dimorphism of brain developmental trajectories during childhood and adolescence. Neuroimage. 2007;36(4):1065-1073. doi: 10.1016/j.neuroimage.2007.03.053 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.White AM. Gender differences in the epidemiology of alcohol use and related harms in the United States. Alcohol Res. 2020;40(2):01. doi: 10.35946/arcr.v40.2.01 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Galvan A. Adolescent development of the reward system. Front Hum Neurosci. 2010;4:6. doi: 10.3389/neuro.09.006.2010 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Crone EA, Dahl RE. Understanding adolescence as a period of social-affective engagement and goal flexibility. Nat Rev Neurosci. 2012;13(9):636-650. doi: 10.1038/nrn3313 [DOI] [PubMed] [Google Scholar]
  • 15.Casey BJ. Beyond simple models of self-control to circuit-based accounts of adolescent behavior. Annu Rev Psychol. 2015;66:295-319. doi: 10.1146/annurev-psych-010814-015156 [DOI] [PubMed] [Google Scholar]
  • 16.Galvan A, Hare TA, Parra CE, et al. Earlier development of the accumbens relative to orbitofrontal cortex might underlie risk-taking behavior in adolescents. J Neurosci. 2006;26(25):6885-6892. doi: 10.1523/JNEUROSCI.1062-06.2006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Bunge SA, Wright SB. Neurodevelopmental changes in working memory and cognitive control. Curr Opin Neurobiol. 2007;17(2):243-250. doi: 10.1016/j.conb.2007.02.005 [DOI] [PubMed] [Google Scholar]
  • 18.Schumann G, Loth E, Banaschewski T, et al. ; IMAGEN consortium . The IMAGEN study: reinforcement-related behaviour in normal brain function and psychopathology. Mol Psychiatry. 2010;15(12):1128-1139. doi: 10.1038/mp.2010.4 [DOI] [PubMed] [Google Scholar]
  • 19.Finn ES, Shen X, Scheinost D, et al. Functional connectome fingerprinting: identifying individuals using patterns of brain connectivity. Nat Neurosci. 2015;18(11):1664-1671. doi: 10.1038/nn.4135 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Horien C, Shen X, Scheinost D, Constable RT. The individual functional connectome is unique and stable over months to years. Neuroimage. 2019;189:676-687. doi: 10.1016/j.neuroimage.2019.02.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Finn ES, Scheinost D, Finn DM, Shen X, Papademetris X, Constable RT. Can brain state be manipulated to emphasize individual differences in functional connectivity? Neuroimage. 2017;160:140-151. doi: 10.1016/j.neuroimage.2017.03.064 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Lichenstein SD, Scheinost D, Potenza MN, Carroll KM, Yip SW. Dissociable neural substrates of opioid and cocaine use identified via connectome-based modelling. Mol Psychiatry. 2021;26(8):4383-4393. doi: 10.1038/s41380-019-0586-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Saunders JB, Aasland OG, Babor TF, de la Fuente JR, Grant M. Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption–II. Addiction. 1993;88(6):791-804. doi: 10.1111/j.1360-0443.1993.tb02093.x [DOI] [PubMed] [Google Scholar]
  • 24.Knutson B, Westdorp A, Kaiser E, Hommer D. fMRI visualization of brain activity during a monetary incentive delay task. Neuroimage. 2000;12(1):20-27. doi: 10.1006/nimg.2000.0593 [DOI] [PubMed] [Google Scholar]
  • 25.Eagle DM, Baunez C, Hutcheson DM, Lehmann O, Shah AP, Robbins TW. Stop-signal reaction-time task performance: role of prefrontal cortex and subthalamic nucleus. Cereb Cortex. 2008;18(1):178-188. doi: 10.1093/cercor/bhm044 [DOI] [PubMed] [Google Scholar]
  • 26.Shen X, Finn ES, Scheinost D, et al. Using connectome-based predictive modeling to predict individual behavior from brain connectivity. Nat Protoc. 2017;12(3):506-518. doi: 10.1038/nprot.2016.178 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Sripada C, Rutherford S, Angstadt M, et al. Prediction of neurocognition in youth from resting state fMRI. Mol Psychiatry. 2020;25(12):3413-3421. doi: 10.1038/s41380-019-0481-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Yip SW, Kiluk B, Scheinost D. Toward addiction prediction: an overview of cross-validated predictive modeling findings and considerations for future neuroimaging research. Biol Psychiatry Cogn Neurosci Neuroimaging. 2020;5(8):748-758. doi: 10.1016/j.bpsc.2019.11.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.NeuroImaging Tools & Resources Collabotory. Accessed August 2, 2023. https://www.nitrc.org/frs/download.php/13635/yip_jama_psych.zip
  • 30.Worhunsky PD, Dager AD, Meda SA, et al. A preliminary prospective study of an escalation in ‘maximum daily drinks’, fronto-parietal circuitry and impulsivity-related domains in young adult drinkers. Neuropsychopharmacology. 2016;41(6):1637-1647. doi: 10.1038/npp.2015.332 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Meda SA, Hawkins KA, Dager AD, et al. Longitudinal effects of alcohol consumption on the hippocampus and parahippocampus in college students. Biol Psychiatry Cogn Neurosci Neuroimaging. 2018;3(7):610-617. doi: 10.1016/j.bpsc.2018.02.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Sullivan EV, Brumback T, Tapert SF, et al. Disturbed cerebellar growth trajectories in adolescents who initiate alcohol drinking. Biol Psychiatry. 2020;87(7):632-644. doi: 10.1016/j.biopsych.2019.08.026 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Sullivan EV, Harris RA, Pfefferbaum A. Alcohol’s effects on brain and behavior. Alcohol Res Health. 2010;33(1-2):127-143. [PMC free article] [PubMed] [Google Scholar]
  • 34.Shokri-Kojori E, Tomasi D, Wiers CE, Wang GJ, Volkow ND. Alcohol affects brain functional connectivity and its coupling with behavior: greater effects in male heavy drinkers. Mol Psychiatry. 2017;22(8):1185-1195. doi: 10.1038/mp.2016.25 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Manuweera T, Kisner MA, Almira E, Momenan R. Alcohol use disorder-associated structural and functional characteristics of the insula. J Neurosci Res. 2022;100(11):2077-2089. doi: 10.1002/jnr.25113 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Goldfarb EV, Scheinost D, Fogelman N, Seo D, Sinha R. High-risk drinkers engage distinct stress-predictive brain networks. Biol Psychiatry Cogn Neurosci Neuroimaging. 2022;7(8):805-813. doi: 10.1016/j.bpsc.2022.02.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Yalachkov Y, Kaiser J, Naumer MJ. Sensory and motor aspects of addiction. Behav Brain Res. 2010;207(2):215-222. doi: 10.1016/j.bbr.2009.09.015 [DOI] [PubMed] [Google Scholar]
  • 38.Greene AS, Gao S, Scheinost D, Constable RT. Task-induced brain state manipulation improves prediction of individual traits. Nat Commun. 2018;9(1):2807. doi: 10.1038/s41467-018-04920-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Jiang R, Calhoun VD, Cui Y, et al. Multimodal data revealed different neurobiological correlates of intelligence between males and females. Brain Imaging Behav. 2020;14(5):1979-1993. doi: 10.1007/s11682-019-00146-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Volkow ND, Koob GF, McLellan AT. Neurobiologic advances from the brain disease model of addiction. N Engl J Med. 2016;374(4):363-371. doi: 10.1056/NEJMra1511480 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplement 1.

eAppendix

eMethods

eFigure 1. Alcohol Use Disorders Identification Test (AUDIT) scores at baseline (age 14)

eFigure 2. Alcohol Use Disorders Identification Test (AUDIT) scores at 5-year follow-up (age 19)

eFigure 3. Outlier detection

eFigure 4. Data inclusion/exclusion for CPM models

eResults

eTable 1. Motion control and sensitivity analyses of models built using neuroimaging data acquired at age 14

eTable 2. Motion control and sensitivity analyses of models built using neuroimaging data acquired at age 19

eTable 3. Specificity analyses of models built using neuroimaging data acquired at age 14

eTable 4. Specificity analyses of models built using neuroimaging data acquired at age 19

eTable 5. Effects of substance-use initiation on network strengths at age 19

eFigure 5. Canonical neural networks

eDiscussion

eReferences

Click here for additional data file. (1.2MB, pdf)
Supplement 2.

Data sharing statement

Click here for additional data file. (135.6KB, pdf)

Articles from JAMA Psychiatry are provided here courtesy of American Medical Association

RESOURCES