Adolescence is a critical time in experimenting with alcohol use due to continued brain development during this stage. This period is considered an essential point for identifying brain circuits that contribute to problems later in life, a phenomenon known as pre-addiction .
Regardless of future alcohol use in adulthood, adolescent alcohol use is itself an adverse health behavior, and represents a leading cause of disability and mortality. By identifying brain mechanisms associated with early risk, the authors’ goal was to facilitate the development of new prevention and intervention approaches to help mitigate problematic alcohol use.
Contemporary theories of alcohol risk emphasize the dual functions of top-down inhibitory systems involving prefrontal cortical structures and bottom-up subcortical reward systems. However, typical brain maturation involves complex functional reorganization of networks rather than simple linear maturation of individual regions.
Sex differences in neurological development have been widely described. Women show accelerated development trajectories relative to men. Sex differences in alcohol consumption during adolescence are also commonly reported. Therefore, this study aimed to identify the neural networks that confer vulnerability to adolescent alcohol use with specific consideration of sex differences.
In a large sample of adolescents, neuroimaging data acquired during the performance of different cognitive tasks were used: a reward processing task and an inhibitory control task. It was hypothesized that networks identified during reward-related processes would be more relevant for predicting alcohol consumption behaviors in younger adolescents (~14 years) versus older adolescents (~19 years), and that networks identified during reward-related processes inhibition would be more relevant to predict alcohol consumption behaviors in older adolescents compared to younger adolescents.
| Methods |
The IMAGEN consortium collected alcohol and neuroimaging data at 8 sites. Baseline data were acquired at age 14 years, and follow-up data at age 19 years. The analysis focused on predicting the risk of alcohol consumption at age 19. Baseline data were used to identify networks associated with future risk of alcohol use, and follow-up data were used to identify networks associated with current risk of alcohol use.
Neuroimaging data were processed using a validated method to generate functional connectivity matrices of individual participants (hereafter referred to as connectomes). Connectomes provide a multivariate summary of the unique pattern of functional organization of an individual’s brain. While connectomes are relatively distinct between individuals (i.e., capable of uniquely identifying individuals, a process known as neural fingerprinting ), they also vary as a function of cognitive task performance or individual brain state. Analyzes focused on connectomes derived from neuroimaging data acquired during reward and inhibitory tasks.
An independent sample of university students aged 17-23 years was used for the study . Alcohol consumption was assessed using the AUDIT, a validated self-report measure of alcohol risk with total scores ranging from 0 to 40. Data were collected using functional magnetic resonance imaging (fMRI) during a reward task and during an inhibitory control task.
The connectome-based predictive model (CPM) comprises the following steps: (1) feature selection, in which regression is used to identify connectome features associated with a behavioral variable of interest (here, alcohol consumption severity) on a training data set; (2) feature reduction, in which the identified connections are summed to create a summary value for each individual in the training data set; (3) model building, where summary scores (independent variables) are linearly associated with the behavioral variable (dependent variable); (4) model application, in which the resulting linear models are applied to novel connectomes in a test data set to generate predictions; and (5) evaluation of the model, in which its predictive capacity is evaluated, based on the correspondence between what is predicted by the model and the actual alcohol consumption risk scores.
Sex-independent (both sexes combined) and sex-specific (female and male separately) models were performed.
| Results |
> Brain behavior models with cross validation
Of the 1359 individuals in the study, with a mean age of 14.4 years, 729 (54%) were female. Analyzes of baseline data (age 14 years) found gender divergence in the predictive accuracy of MPCs assessing the severity of future alcohol use. Specifically, sex-independent models and female-only models were successful in predicting severity of future alcohol use at age 19, but male-only models were not.
The same pattern of results emerged in post hoc analyzes that controlled for severity of initial and residual alcohol use. Female models significantly outperformed male models.
Analyzes of follow-up data (age 19 years) found gender divergence in the predictive accuracy of MPCs assessing the severity of current alcohol use. Specifically, both sex-independent models and female-only models were successful in predicting current severity on both task types.
In contrast, for men, the models were successful in predicting current severity using connectivity data acquired during performance of the inhibitory tasks, but not with data acquired during performance of the reward task, indicating task specificity. in risk models of alcohol consumption in male individuals.
The same pattern of results emerged in post hoc analyzes controlling for severity of initial and residual alcohol consumption. The female model generated from the reward data significantly outperformed the male model generated from the same data. There were no significant differences in predictive model performance between female and male models generated from inhibitory data.
> Anatomy of alcohol consumption risk networks
Positive predictive connections are those for which greater connectivity positively predicted risk of alcohol use. Negatively predictive connections are those for which decreased connectivity positively predicted risk of alcohol use. By definition, positive and negative connections cannot directly overlap (since a single connection cannot be both a positive and a negative predictor). Regardless, positive and negative predictive connections may include connections within and between similar large-scale canonical neural networks.
Neural networks associated with future alcohol consumption using data from reward or inhibitory control tasks were similar among adolescents, indicating the relevance of both processes in conferring vulnerability to alcohol consumption risk. For both tasks, positive connections include a high degree of frontal, motorsensory, and medial cerebellar connections within the network, as well as substantial connections with cerebellar and subcortical networks.
Post hoc virtual lesions indicated that networks with the highest loadings for individual characteristics, that is, variation in future alcohol consumption associated with connectivity within a given canonical network, included fronto-parietal, sensory motor, and cerebellar networks. .
The all-female positive networks generated from both tasks were dominated by sensory-motor, cerebellar, subcortical, and salience network connections, while the negative networks were largely characterized by sensory-motor, salience, and subcortical connections.
For adolescent girls, post hoc virtual lesions indicated that networks with the highest loadings for individual characteristics, that is, variation in current alcohol consumption associated with connectivity within a given canonical network, included saliency, sensory, and motor and cerebellar. For males, virtual lesions indicated that the networks associated with the greatest amount of variance in current alcohol consumption included sensory-motor, cerebellar, and subcortical networks.
For successful models, individual participants’ network summary scores emerged from the sum of connectivity strengths within positive and negative networks. Among adolescent girls, connectivity within networks predictive of future alcohol use increased significantly from age 14 to age 19. In both female and male adolescents, connectivity within inhibitory control networks predicting current use also increased significantly from ages 14 to 19, while changes in connectivity within reward networks did not reach significance.
Despite significant differences in predictive accuracy, networks identified as associated with alcohol use behaviors were relatively consistent between male and female adolescents.
> Specificity and generalization
To determine the specificity of the anatomical connections identified with alcohol consumption, MPCs were repeated controlling for variation in impulsivity and neuroticism. For all models, predictive accuracies were robust to these factors, indicating that the connections identified were specific to alcohol use and that model accuracy was not driven by a more general latent personality trait factor.
Rates of substance use were low at age 14, but were highest at age 19. Post hoc analyzes indicated significant associations between substance use and connectivity within the male inhibition model, such that connectivity was greater among men with nonalcoholic substance use. Follow-up analyzes indicated that associations between alcohol consumption and connectedness remained significant within each group separately.
| Discussion |
This study used data from the IMAGEN consortium to identify neural networks associated with alcohol consumption risk using an advanced connectome-based approach. Common and distinct neurobiological substrates of early alcohol use risk among male and female adolescents were identified, as well as sexual divergence in model accuracy.
For adolescent girls, models generated using neuroimaging data from both types of tasks were able to identify reliable neural signatures of current and future 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. The identified alcohol consumption risk network can be considered as a robust neuromarker that can be targeted in future prevention efforts.
Among adolescent girls, the connections identified included subcomponents of multiple well-established resting networks, such as the medial frontal, frontoparietal, salience, sensory-motor, and cerebellar networks. Connectivity within these networks generally increased from ages 14 to 19, concurrent with increases in alcohol consumption.
Furthermore, despite having similar characteristics, the relative predictive weights of network subcomponents associated with future (age 14) versus current (age 19) alcohol use risk varied over time, such that the frontal networks medial and salience networks were primarily associated with the risk of future alcohol consumption, while the cerebellar and sensory-motor networks were primarily associated with the risk of current alcohol consumption.
These findings are consistent with recent data demonstrating alterations in cerebellar growth trajectories among adolescents with heavy alcohol use, as well as with much broader literature implicating the cerebellar and motor systems in the pathophysiology of substance use. Taken together, these data suggest that targeting regions of the prefrontal cortex may be particularly relevant for prevention efforts in younger adolescents and that targeting cerebellar and sensory motor regions may be more relevant for intervention efforts in older adolescents. who have already started drinking.
Inhibitory control (vs. reward) brain states were more relevant in predicting alcohol consumption in older adolescents; however, this effect was specific to males. These data suggest that interventions focused on inhibitory control processes may be particularly effective in combating current risk of alcohol use in adolescent boys, but that both inhibitory and reward-related processes are likely to be relevant for adolescent boys. Current alcohol consumption behaviors in women.
| Conclusion |
These data provide a connectome-wide assessment of neural networks serving alcohol use risk and identify a dimensional neural signature of adolescent alcohol use risk. External validation supported these findings.
