Introduction

The worldwide lifetime prevalence of alcohol use is 80% and 10.7% for alcohol use disorder (AUD),¹ suggesting that only a portion of users develop AUD. In young adolescents, rates of weekly alcohol use vary between 9% and 16%,² and alcohol use in adolescence may result in physical injury, high-risk sexual behavior, long-term adverse neurocognitive effects, and an enhanced risk of developing AUD in adult life.³ Therefore, early identification of risk factors is crucial to develop interventions to prevent AUD. The most relevant risk factors are reward-related brain functioning, genetics, personality, and social factors.⁴ Personality and social environment factors, including parental drinking and parental provision of alcohol, are associated with alcohol misuse in early adolescence and in the development of adolescent alcohol use.^5,6 Peer and parental influences on alcohol misuse seem to decrease in adulthood, whereas environmental factors seem to change over time.⁷ In a recent study examining markers of binge drinking,⁸ the neuropsychosocial and the psychosocial model outperformed the neural model for predicting risky drinking behaviors. While functional and structural brain measures are associated with alcohol use,⁹ these factors demonstrated modest utility, and the development of a neuroimaging risk score to yield targeted interventions is ongoing.⁹ Still, a comprehensive review on the imaging genetics results of the IMAGEN study by Mascarell Maričić et al¹⁰ found various effects of genetics on intermediate phenotypes. Therefore, the aim of the current study is to quantify the risk for alcohol misuse and subsequent dependency with respect to such intermediate phenotypes and associated domains for the first time in a longitudinal design with participants aged 14 to 22 years in the IMAGEN sample.

Methods

In this cohort study, we used baseline data from participants from the longitudinal IMAGEN cohort¹¹ that started with 14-year-old students from 8 sites in Germany, the UK, France, and Ireland. All local ethics research committees approved the study in accordance with the Declaration of Helsinki.¹² This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. Verbal or written consent was obtained from the parent or guardian and the participants depending on their age. Participants were recruited in schools with a maximization of ethnic homogeneity of European descent and sample diversity regarding sociodemographic variables and behavioral and emotional functioning. Exclusion criteria were ineligibility for magnetic resonance imaging scans and serious medical conditions. Alcohol misuse trajectories were assessed at 4 consecutive time points (14, 16, 19, and 22 years) with the Alcohol Use Disorder Identification Test (AUDIT; self-report version) and modeled using latent growth curve models (LGCM). Exposure variables were preselected based on publications^{4 ,5,13-20} regarding risk and resilience factors associated with alcohol misuse in adolescents and young adults. We aimed to cover a broad variety of domains, namely personality, social, and brain functioning characteristics, and familial risk for substance misuse. All assessments are listed in Table 1 and described in the eMethods in Supplement 1.

Statistical Analyses

We used LGCM to analyze longitudinal within-participant as well as between-participant changes in alcohol misuse. LGCM analyses were conducted using Ωnyx (von Oertzen et al²¹) and descriptive analyses were conducted using SPSS version 28 (IBM).

The LGCM (Figure 1) contained individual changes in AUDIT scores (within-person variability) and exposure variables associated with alcohol misuse (between-person variability). The modeled intercepts and slopes represent the individual 8-year trajectory of alcohol misuse, and the exposure variables allow for analysis of interindividual associations of these variables with misuse trajectories.

To address domain-specific dependency of the exposure variables, the brain, social, and personality domains were treated as latent factors in the model. Estimate parameters of the 4 exposure variables (ie, brain, social, personality, and family history) were statistically compared analogously to Clogg and colleagues²² and corrected for multiple testing.

Freely estimated covariances were allowed between all exposure variables (3 latent factors and family history). Maximum likelihood was used to estimate parameters, and a linear trajectory of AUDIT scores was assumed, which was verified by comparing the model fit to an analogous quadratic model and a model without time trend using Akaike information criterion (AIC). Models with higher AIC scores were considered to fit the time trend of observed AUDIT scores better.²³ Missing data was handled with full information maximum likelihood. Small effect sizes were defined as an r value greater than 0.10 and less than 0.30.²⁴ A 2-sided P < .05 was considered statistically significant. For the multiple comparisons of the etsimate parameters, the Bonferroni correction was used to adjust the significance level to P < .004. Data analysis was conducted from July 2021 to September 2022.

Results

Sample characteristics including descriptive statistics on all analyzed variables are displayed in Table 2. The sample comprised 2240 participants (1110 female [49.6%] and 1130 male [50.4%]). Alcohol misuse (AUDIT) was measured in 2209 participants (98.6%) at study baseline (14 years), in 1698 (75.8%) participants at follow-up 1 (16 years; 511 participants [23.1%] dropped out from baseline), in 1511 participants (67.5%) at follow-up 2 (19 years; 729 participants [31.6%] dropped out from baseline) and in 1359 participants (60.7%) at follow-up 3 (22 years; 881 participants [38.5%] dropped out from baseline).

The linear model showed an acceptable and better²³ model fit in comparison with a quadratic model (−2 log likelihood ratio = 15.85; AIClinear = 77 694.85; AICquadratic = 77 710.70 and a better model fit in comparison with a model with no time trend (−2 log likelihood ratio = 222.73; AIClinear = 77 694.85; AICno-time-trend = 77 917.57) (Table 3). Regarding the factor paths (Table 3 and the eTable in Supplement 1), we found a significant negative association of the social domain with the intercept of alcohol misuse (β = −0.29; SE = 0.03; P < .001) and a significant positive association of the personality domain with the intercept (β = 0.19; SE = 0.03; P < .001), both of which had small effect sizes. Furthermore, there were significant positive associations of the social domain (β = 0.13; SE = 0.02; P < .001) and personality domains (β = 0.07; SE = 0.02; P < .001) with the slope of developmental trajectories of alcohol misuse, although these were small effect sizes. Family history also showed significant negative associations with the alcohol misuse intercept (β = −0.04; SE = 0.02; P = .045) and slope (β = −0.03; SE = 0.01; P = .01), but corresponding r coefficients suggested correlations beneath the threshold for a small effect size (Table 3). The latent factor, brain functioning, showed no association with intercept or slope of alcohol misuse. The 4 intercept estimates as well as the 4 slope estimates differed significantly (also after Bonferroni correction for multiple testing). Thus, social domain scores showed significantly larger-magnitude associations with the intercept and slope of alcohol misuse trajectories compared with the personality domain scores, which showed significantly larger-magnitude associations compared with the brain domain scores and family history of substance abuse. Figure 2A illustrates that higher social scores (higher socioeconomic status and lower frequency of life events related to family and sexuality) were associated with higher scores of alcohol misuse at baseline but were also associated with a greater slope over 8 years. Figure 2B shows that higher risk-taking aspects of personality were associated with higher scores of alcohol misuse at baseline, as well as a greater increase (slope) over 8 years.

Discussion

In this cohort study, we aimed to investigate the differential association of various domains with alcohol misuse trajectories from adolescence to young adulthood. The results of our final model indicate that psychosocial resources—indexed by lower frequency of life events related to family and sexuality, together with higher socioeconomic status—were initially associated with a lower general risk for alcohol misuse (intercept) but with increased alcohol misuse over the course of 8 years (slope). At the same time, personality characteristics—specifically higher impulsivity, risk-taking and extraversion scores—were associated with a higher risk of alcohol misuse on average, as well as a higher risk for alcohol misuse development over 8 years (slope). This finding suggests an influence of personality features on the general risk regarding alcohol misuse and on the development of alcohol misuse from adolescence to young adulthood. The absence of family history was significantly associated with a lower general risk and a decrease in alcohol misuse over time, with estimates below the threshold for a small effect size, which cannot be interpreted as meaningful. In earlier analyses of IMAGEN data,^10,11 we saw a differential association of reward-related brain functioning at baseline with alcohol misuse 2 years later depending on family history of substance abuse.¹³ This finding suggests that family history of substance abuse and reward-system functioning may influence the course of adolescent alcohol misuse, specifically in certain combinations. Therefore, a possible mediating role of familial risk of the influence of reward processing on the course of alcohol misuse should be further investigated. Overall, the present analysis suggests a comparably large general association of social and personality factors with adolescent alcohol misuse that outweighs reward-system functioning and the individual familial risk. Thus, the risk of developing an alcohol addiction might primarily be influenced by social factors that in large parts are associated with the adolescent’s family as well as personality factors. Also, it has been shown that social and language functional magnetic resonance imaging tasks discriminate well on a neural level, suggesting that a broader range of brain processes might be associated with binge drinking in late adolescence.⁸

Strengths and Limitations

The strength of the present study lies in our use of rich longitudinal data collected by the IMAGEN project to track the differential association of a diverse range of established factors with alcohol abuse over a broad neurodevelopmental window in a large sample. There were limitations to the study as well. First, limitations concern a selective bias due to dropouts that might have contributed to an underestimation of alcohol misuse.²⁵ Second, alcohol misuse was only obtained via self-report; no external validation was administered. Third, as of yet, we have no information about whether participants with risky drinking trajectories will develop AUD in later life. Fourth, statistically the model showed an acceptable, but not a good fit, which is still highly informative for the field and future analyses.

Conclusions

This cohort study found an association of social and personality factors with adolescent alcohol misuse compared with reward-related brain functioning and family history. This finding suggests that known risk factors for adolescent drinking may contribute differently to future alcohol misuse. This approach may inform more individualized preventive interventions.

Summary

A longitudinal study utilizing data from the IMAGEN cohort (n=2240, ages 14-22) investigated the interplay of psychosocial, personality, neurological, and familial factors in predicting alcohol misuse trajectories. Latent growth curve modeling (LGCM) analyzed the longitudinal data, incorporating measures of personality, social environment, brain function, and family history of substance abuse as predictors of alcohol misuse, assessed via the Alcohol Use Disorder Identification Test (AUDIT). The study addressed limitations of previous research by employing a comprehensive longitudinal design and examining the relative contributions of diverse risk factors.

Methods

Baseline data from the IMAGEN cohort, encompassing participants of European descent from various socioeconomic backgrounds, were analyzed. Ethical approvals and informed consent were obtained. Exclusion criteria included ineligibility for MRI and serious medical conditions. Alcohol misuse was assessed at four time points (ages 14, 16, 19, and 22) using the AUDIT. LGCMs were employed to model alcohol misuse trajectories, incorporating pre-selected exposure variables representing personality, social factors, brain function, and familial substance abuse risk. Data analyses were conducted using Ωnyx and SPSS.

Statistical Analyses

Longitudinal changes in alcohol misuse were analyzed using LGCM, modeling both within- and between-person variability. Brain, social, and personality domains were treated as latent factors to account for inter-domain dependencies. Model fit was assessed using AIC, comparing linear, quadratic, and no-time-trend models. Maximum likelihood estimation was used, handling missing data via full information maximum likelihood. Statistical significance was determined using a two-sided P < .05, adjusted for multiple comparisons using the Bonferroni correction.

Results

The linear LGCM exhibited superior fit compared to quadratic and no-time-trend models. Social domain scores (socioeconomic status and life events) showed significant negative associations with initial alcohol misuse but positive associations with the trajectory slope. Personality domain scores (impulsivity, risk-taking, extraversion) positively correlated with both initial misuse and trajectory slope. Family history showed a negative association with both intercept and slope, although effect sizes were small. Brain functioning showed no significant association with alcohol misuse trajectories.

Discussion

Psychosocial factors, particularly life events and socioeconomic status, were associated with lower initial alcohol misuse risk but greater increases over time. Personality characteristics, primarily impulsivity and risk-taking, were associated with both higher initial risk and steeper trajectory slopes. Familial risk exhibited a negative association, though effect size was minimal. These findings suggest that social and personality factors exert a stronger influence on alcohol misuse trajectories than brain function and familial risk. Future research should explore potential mediating roles of family history on the impact of reward processing on alcohol misuse.

Strengths and Limitations

Strengths include the longitudinal design, diverse sample, and comprehensive assessment of risk factors. Limitations include potential selection bias due to participant attrition, reliance on self-reported alcohol use, and the absence of long-term AUD outcomes. Additionally, although the model demonstrated acceptable fit, it wasn't considered a highly precise fit.

Conclusions

Social and personality factors emerged as significant predictors of adolescent alcohol misuse trajectories, surpassing the influence of brain function and family history. This suggests a need for individualized preventive interventions targeting these psychosocial and personality characteristics.

Summary

A study investigated the risk factors for alcohol misuse and subsequent dependence in adolescents and young adults (ages 14-22) using longitudinal data from the IMAGEN cohort. The research utilized latent growth curve models (LGCM) to analyze alcohol misuse trajectories assessed via the Alcohol Use Disorder Identification Test (AUDIT) across four time points. Multiple domains—personality, social factors, brain functioning, and family history of substance misuse—were considered as potential predictors.

Methods

The study employed a longitudinal cohort design with participants of primarily European descent. Ethical approvals were obtained, and informed consent was secured. Participants were recruited from schools, maximizing ethnic homogeneity while maintaining sociodemographic diversity. Exclusion criteria included ineligibility for MRI scans and serious medical conditions. Alcohol misuse trajectories were assessed using the AUDIT at four time points (ages 14, 16, 19, and 22). A wide array of exposure variables representing personality, social factors, brain functioning, and family history were included.

Statistical Analyses

Longitudinal data were analyzed using LGCMs. The models incorporated individual changes in AUDIT scores (within-person variability) and exposure variables (between-person variability). Personality, social, and brain functioning domains were treated as latent factors to account for inter-correlations. Model fit was assessed using the Akaike Information Criterion (AIC). Statistical significance was determined using a Bonferroni correction for multiple comparisons.

Results

The analysis included 2240 participants (approximately 50% female). Significant participant attrition occurred across the four time points. The linear LGCM provided a better fit than quadratic or no-time-trend models. Social factors negatively predicted initial alcohol misuse but positively predicted subsequent increases in misuse. Personality factors (impulsivity, risk-taking, extraversion) positively predicted both initial misuse and subsequent increases. Family history showed a weaker negative association with both. Brain functioning showed no significant association.

Discussion

Psychosocial factors (socioeconomic status, life events) and personality traits were strongly associated with alcohol misuse trajectories. Family history showed a smaller, negative association. Brain functioning was not a significant predictor. The findings suggest that social and personality factors may be primary drivers of adolescent alcohol misuse, potentially outweighing the influence of reward-related brain functioning and family history. Future research should investigate the interplay of these factors.

Strengths and Limitations

Strengths include the longitudinal design, large sample size, and diverse range of predictor variables. Limitations include potential selection bias due to attrition, reliance on self-reported alcohol misuse, and the lack of information on whether risky drinking patterns translate to AUD in later life. The model showed acceptable but not optimal fit.

Conclusions

This study highlights the significant roles of social and personality factors in adolescent alcohol misuse, relative to brain functioning and family history. These findings underscore the potential for individualized preventive interventions targeting these specific risk factors.

Summary

A study examined factors influencing alcohol misuse from adolescence to young adulthood. Researchers used data from the IMAGEN cohort, a longitudinal study of 14- to 22-year-olds. The study focused on the interplay between personality, social environment, brain function, and family history in predicting alcohol misuse trajectories. Statistical analyses, including latent growth curve models, revealed significant associations between social and personality factors, and alcohol misuse. While brain function and family history showed some associations, their impact was less pronounced than that of social and personality traits.

Methods

The study utilized baseline data from the IMAGEN cohort, encompassing adolescents from various European countries. Ethical approvals were obtained, and participants provided consent. Researchers focused on a relatively homogenous sample to minimize confounding variables. Alcohol misuse was tracked over four time points (ages 14, 16, 19, and 22) using the AUDIT (Alcohol Use Disorder Identification Test). The analysis included personality, social, and brain functioning characteristics, along with family history of substance misuse.

Statistical Analyses

Longitudinal data analysis employed latent growth curve models (LGCMs). These models accounted for individual changes in AUDIT scores (within-person variability) and the influence of exposure variables (between-person variability). The analysis considered personality, social, and brain functioning as latent factors to account for their interrelationships. Statistical significance was determined using a two-sided p-value < .05, with adjustments for multiple comparisons.

Results

The study involved 2240 participants. The linear LGCM provided a better fit than alternative models. Key findings indicated significant associations between social factors (socioeconomic status, life events) and alcohol misuse, both at baseline and across time. Personality traits (impulsivity, risk-taking, extraversion) also showed significant positive associations with alcohol misuse trajectories. Family history of substance abuse showed a weaker, negative association. Brain functioning did not demonstrate significant associations.

Discussion

The findings highlight the substantial influence of social and personality factors on adolescent alcohol misuse, exceeding the influence of brain function and family history. Social factors appear to impact both the initial risk and the trajectory of alcohol misuse. Personality traits seem to similarly influence both initial risk and the development of alcohol misuse over time. Future research should explore the potential mediating role of family history and reward-system functioning on alcohol misuse.

Strengths and Limitations

Strengths included the use of longitudinal data from a large sample and consideration of diverse factors. However, limitations included potential bias due to participant dropout and reliance on self-reported alcohol use. The model fit, while acceptable, was not optimal. Further, the study lacks data on whether risky drinking patterns translated to AUD in later life.

Conclusions

This study emphasizes the importance of social and personality factors in adolescent alcohol misuse. The findings suggest that preventive interventions should tailor strategies to address these key factors, potentially leading to more effective prevention of future alcohol misuse.

Summary

A big study looked at what makes kids and young adults start drinking too much alcohol. They followed a group of kids for eight years, from age 14 to 22.

Methods

The scientists used information from a giant study called IMAGEN. They asked kids questions about their drinking habits, and also about their personalities, families, and friends. They used special math to see how all these things were connected to drinking.

Statistical Analyses

They used special ways to figure out how kids' drinking changed over time. They looked at how things like personality and family affected how much kids drank.

Results

They found that kids with good friends and families were less likely to drink too much at first, but that those same factors could make them drink more as they got older. Kids who were more impulsive or outgoing tended to drink more, too. Family history of drinking also made a small difference. Brain scans didn't seem to help as much in predicting alcohol use.

Discussion

The study showed that friends, family, and personality are really important in how much kids and young adults drink. It's important to focus on helping kids develop these healthy aspects of their lives to help prevent problems with alcohol.

Strengths and Limitations

The study was really big and followed people for a long time, which is good. But, some kids dropped out of the study, and they only used questionnaires to ask about drinking, not other methods to assess alcohol use.

Conclusions

This study shows that helping kids develop healthy relationships and personalities is important in preventing alcohol problems. This could lead to better ways to help kids make healthy choices.