Alcohol use disorders are prevalent (SAMHSA, 2016) and represent significant public health concerns in the United States (Sacks, Gonzales, Bouchery, Tomedi, & Brewer, Reference Sacks, Gonzales, Bouchery, Tomedi and Brewer2015). Risk for alcohol use disorders generally peaks during young adulthood (Grant et al., Reference Grant, Goldstein, Saha, Chou, Jung, Zhang and Hasin2015). College students are more likely to drink heavily and meet alcohol use disorder criteria than their non-college-attending peers (Barnes, Welte, Hoffman, & Tidwell, Reference Barnes, Welte, Hoffman and Tidwell2010). College students also face a unique set of environmental pressures and sociodevelopmental contexts (Evans, Forney, Guido, Renn, & Patton, Reference Evans, Forney, Guido, Renn and Patton2010), such as leaving home and forming new peer groups. These changes and transitions usually create increased opportunity for alcohol involvement and exposure to stress, leading to elevated risk for alcohol problems among college students.
Numerous studies have demonstrated that both genetic predispositions and social environments contribute to alcohol use problems, and that genetic effects vary as a function of environmental experiences (i.e., gene–environment interaction; G × E), and vice versa (Young-Wolff, Enoch, & Prescott, Reference Young-Wolff, Enoch and Prescott2011). Despite the importance of alcohol use disorders as an important public health concern for all groups in the United States, a critical gap in the literature is the underrepresentation of racial/ethnic minority samples in genetic research (Dick, Barr, Guy, Nasim, & Scott, Reference Dick, Barr, Guy, Nasim and Scott2017; Oquendo, Canino, Lehner, & Licinio, Reference Oquendo, Canino, Lehner and Licinio2010). Participants included in gene identification efforts such as genome-wide association studies (GWAS) across disciplines are predominantly of European descent (~80%; Popejoy & Fullerton, Reference Popejoy and Fullerton2016). Although the proportion of racial/ethnic minority populations in GWAS has increased over the past decade, African Americans are still underrepresented in genetic research, with only about 3% of participants in GWAS being of African descent (Popejoy & Fullerton, Reference Popejoy and Fullerton2016). The cultural development and psychopathology framework emphasizes the importance of integrating culture into developmental psychopathology and sample diversity, and that findings based on studies of European Americans may not generalize to individuals from other backgrounds (Causadias, Reference Causadias2013). Given that the majority of previous genetic association studies and G × E research has focused on European Americans (Dick et al., Reference Dick, Barr, Guy, Nasim and Scott2017; Popejoy & Fullerton, Reference Popejoy and Fullerton2016), little is known about G × E in relation to alcohol use problems in other groups such as African Americans (Chartier, Karriker-Jaffe, Cummings, & Kendler, Reference Chartier, Karriker-Jaffe, Cummings and Kendler2017). Given differences in rates of alcohol use and related consequences across racial/ethnic groups (Chartier & Caetano, Reference Chartier and Caetano2010; Mulia, Greenfield, & Zemore, Reference Mulia, Ye, Greenfield and Zemore2009), as well as potential racial/ethnic differences in pathways of risk to alcohol and related problems (Akins, Smith, & Moshe, Reference Akins, Smith and Mosher2010; Meyers et al., Reference Meyers, Sartor, Werner, Koenen, Grant and Hasin2018), it is important to study G × E processes that impact risk for alcohol use disorders in all racial groups so that all can benefit from the research and intervention/prevention efforts. This also aligns with a cultural genomics approach by examining the interplay of genes and environments across and within different cultural groups (Causadias & Korous, Reference Causadias, Korous, Causadias, Telzer and Gonzales2018).
G × E: The Role of Peer Deviance and Interpersonal Traumatic Events
Numerous studies have suggested the importance of examining G × E effects in understanding the etiology of alcohol problems (Young-Wolff et al., Reference Young-Wolff, Enoch and Prescott2011). Affiliation with deviant peers and perceived peer alcohol and substance use are among the most robust predictors of alcohol use and related problems among adolescents (Hawkins, Catalano, & Miller, Reference Hawkins, Catalano and Miller1992). Having more substance-using friends is considered an indicator of greater access to alcohol and less social control against drinking, and thus is associated with greater risk for alcohol use and misuse and creates a context for greater manifestation of genetic risk toward alcohol problems (Dick & Kendler, Reference Dick and Kendler2012). College students are particularly susceptible to the influence of peers and social norms related to alcohol and substance use due to increased autonomy by leaving home and the pressure to fit in new social networks (Arnett, Reference Arnett2005). There is evidence from both latent (e.g., twin studies) and measured genetic studies that genetic effects on alcohol use outcomes are greater when peer deviance is high, for both adolescents (Cooke et al., Reference Cooke, Meyers, Latvala, Korhonen, Rose, Kaprio and Dick2015; Dick & Kendler, Reference Dick and Kendler2012) and young adults, including college students (O'Shea et al., Reference O'Shea, Thomas, Webb, Dick, Kendler and Chartier2017; Poelen, Scholte, Willemsen, Boomsma, & Engels, Reference Poelen, Scholte, Willemsen, Boomsma and Engels2007). Some studies indicated that peer influence on alcohol outcomes varied as a function of individuals’ genetic predisposition. For example, Mrug and Windle (Reference Mrug and Windle2014) found that friends’ alcohol use was associated with increased alcohol use among young adults who carried the risk allele of the dopamine D4 receptor gene (DRD4), but was not associated with alcohol use among those without the “risk” genotype. Other studies found no significant interaction effect between genotypes and peer deviance in relation to alcohol outcomes (Zaso, Maisto, Glatt, Belote, & Park, Reference Zaso, Maisto, Glatt, Belote and Park2017).
Experience of adverse life events, including potentially traumatic interpersonal events such as physical and sexual assaults, is prevalent among young adults, and college students specifically (American College Health Association, 2012), and has been associated with a host of negative psychosocial outcomes, including alcohol problems (Read et al., Reference Read, Colder, Merrill, Ouimette, White and Swartout2012). G × E research has demonstrated interaction effects between genotypes and adverse life events, such that experiences of adverse life events are associated with increased risk for alcohol use and related problems, particularly when coupled with “risk” genotypes (Dick & Kendler, Reference Dick and Kendler2012; Schmid et al., Reference Schmid, Blomeyer, Treutlein, Zimmermann, Buchmann, Schmidt and Laucht2010). For example, experience of negative life events in the past year was associated with drinking more frequently and heavy drinking among college students who carried the risk allele of the serotonin transporter gene (5-HTTLPR) but not among those without the “risk” genotype (Covault et al., Reference Covault, Tennen, Armeli, Conner, Herman, Cillessen and Kranzler2007; Kranzler et al., Reference Kranzler, Scott, Tennen, Feinn, Williams, Armeli and Covault2012). Some studies failed to find significant interaction effects between adverse life experiences and genotypes in relation to alcohol outcomes (Coley, Sims, & Carrano, Reference Coley, Sims and Carrano2017).
Taken together, despite some inconsistencies in previous findings, there is substantial evidence that peer deviance and adverse life events are important risk factors associated with alcohol use and misuse, and that effects of these environmental risk factors on alcohol outcomes varied as a function of individuals’ genetic predispositions. Of note, the majority of prior G × E research employed a cross-sectional design, and thus there is limited understanding about how G × E effects unfold over time to predict trajectory of alcohol use behaviors (Li et al., Reference Li, Cho, Salvatore, Edenberg, Agrawal, Chorlian and Dick2017). In addition, most of the G × E studies using measured genotypes involved candidate genes, such as DRD4, 5-HTTLPR, and alcohol dehydrogenase related genetic markers (e.g., ADH1B). Findings from G × E studies with candidate genes have been inconsistent and controversial in general, due to the existence of publication bias, low statistical power, and a high false discovery rate (Duncan & Keller, Reference Duncan and Keller2011). Furthermore, because alcohol use and misuse are complex behaviors that are polygenic, that is, influenced by many genes of small effects (Clarke et al., Reference Clarke, Smith, Gelernter, Kranzler, Farrer, Hall and McIntosh2016; Gelernter et al., Reference Gelernter, Kranzler, Sherva, Almasy, Koesterer, Smith and Farrer2014), researchers have recommended moving beyond candidate genes by taking a genome-wide polygenic approach to better characterize genetic risk for alcohol use and related outcomes in G × E research (Dick et al., Reference Dick, Agrawal, Keller, Adkins, Aliev, Monroe and Sher2015). For example, Salvatore et al. (Reference Salvatore, Aliev, Edwards, Evans, Macleod, Hickman and Dick2014) found that polygenic risk for alcohol problems interacted with parental knowledge and peer deviance in predicting alcohol use in a sample of European adolescents.
Studying G × E and Alcohol Problems in African Americans
We focused on African American college students here for several important reasons. African Americans, on average, are more likely to abstain from alcohol use and report lower rates of alcohol consumption and heavy drinking than their European American peers (Chartier & Caetano, Reference Chartier and Caetano2010). Despite their lower levels of alcohol use, African Americans experience higher or similar levels of negative social and health consequences related to alcohol use compared to European Americans (Mulia et al., Reference Mulia, Ye, Greenfield and Zemore2009). This same pattern of racial disparity is also observed among college students (Clarke, Kim, White, Jiao, & Mun, Reference Clarke, Kim, White, Jiao and Mun2013; O'Malley & Johnston, Reference O'Malley and Johnston2002). In addition, there are significant differences in the environmental conditions and stressors experienced by African Americans compared to European Americans, such as lower socioeconomic status and experience of stressful life events such as racial discrimination (Roberts, Gilman, Breslau, Breslau, & Koenen, Reference Roberts, Gilman, Breslau, Breslau and Koenen2011; Williams, Mohammed, Leavell, & Collins, Reference Williams, Mohammed, Leavell and Collins2010). Moreover, prior research documents the links between cultural factors such as religiosity and acculturation strategy and lower levels of alcohol consumption in African Americans (e.g., Bazargan, Sherkat, & Bazaragan, Reference Bazargan, Sherkat and Bazaragan2004; Klonoff & Landrine, Reference Klonoff and Landrine1999). Finally, there are important differences in genetic diversity, allele frequencies, and linkage disequilibrium patterns between European Americans and African Americans (Campbell & Tishkoff, Reference Campbell and Tishkoff2008; Gabriel et al., Reference Gabriel, Schaffner, Nguyen, Moore, Roy, Blumenstiel and Altshuler2002; Rosenberg et al., Reference Rosenberg, Huang, Jewett, Szpiech, Jankovic and Boehnke2010), implying potential differences in the effects of specific genes on phenotypes such as alcohol use. Together, these differences in environmental, cultural processes, and genetic factors between European Americans and African Americans suggest that the G × E processes linking to alcohol outcomes may vary across groups, and it is important to take a cultural genomic approach to understand the G × E processes within cultural groups (Causadias & Korous, Reference Causadias, Korous, Causadias, Telzer and Gonzales2018).
However, African Americans have been very much underrepresented in genetic research in general, including twin studies, gene-identification efforts (e.g., genome-wide association studies), and G × E research, limiting our understanding of the etiological factors that contribute to alcohol problems and related psychiatric outcomes in African Americans (Chartier et al., Reference Chartier, Karriker-Jaffe, Cummings and Kendler2017; Dick et al., Reference Dick, Barr, Guy, Nasim and Scott2017). Lack of representation of African Americans in these genetic studies also has implications for understanding genetic effects and improving predictive accuracy in smaller developmental psychopathology studies, as genetic variants identified using largely European descent samples may not generalize to individuals of other ancestral groups (Martin et al., Reference Martin, Gignoux, Walters, Wojcik, Neale, Gravel and Kenny2017). Moreover, the heterogeneity in pathways of risk to alcohol problems within and across populations needs to be taken into account, and findings from research conducted in European Americans may not be generalizable to African Americans.
Some studies have found differences in specific G × E effects between European Americans and African Americans. For example, two studies found that the protective effects of ADH1B variants against different alcohol phenotypes were reduced in risky environments, such as affiliation with drinking peers (Olfson et al., Reference Olfson, Edenberg, Nurnberger, Agrawal, Bucholz, Almasy and Kuperman2014) or exposure to childhood adversity (Sartor et al., Reference Sartor, Wang, Xu, Kranzler and Gelernter2014) in European Americans but not African Americans. Other studies have examined G × E effects within African American populations. For example, Desalu, Zaso, Kim, Belote, and Park (Reference Desalu, Zaso, Kim, Belote and Park2017) found a significant interaction effect between an ADH1B variant and perceived peer drinking in a sample of African American college students, such that carriers of the ADH1B protective allele reported lower frequency of drinking and fewer negative consequences than noncarriers, only when their perceived peer drinking was low. However, systematic study of measured G × E effects in African Americans in a hypothesis-driven fashion is lacking, and more research is needed in this area to better understand the role of genetic and environmental influences in the pathways of risk to alcohol problems in African Americans (Dick et al., Reference Dick, Barr, Guy, Nasim and Scott2017).
The Present Study
The goal of this study was to examine whether and how genetic risk for alcohol problems and environmental risk factors (i.e., peer deviance and interpersonal traumatic events) independently and interactively influence alcohol use disorder symptoms across the college years in a sample of African Americans. We characterized individuals’ genetic risk for alcohol problems by using a genome-wide polygenic score approach (International Schizophrenia Consortium, 2009). This approach considers contributions of many common genetic variants of small magnitude across the genome to examine genetic risk for alcohol dependence and reflects the polygenic nature of behavioral outcomes such as alcohol dependence (Plomin, Haworth, & Davis, Reference Plomin, Haworth and Davis2009). The calculation of reliable polygenic scores requires a genome-wide association study (GWAS) with an independent discovery sample of large sample size and matching ancestry with the target sample (Martin et al., Reference Martin, Gignoux, Walters, Wojcik, Neale, Gravel and Kenny2017). For the purpose of this study, we calculated polygenic scores for alcohol dependence using GWAS estimates from the largest published GWAS on alcohol dependence among African Americans to date (Gelernter et al., Reference Gelernter, Kranzler, Sherva, Almasy, Koesterer, Smith and Farrer2014).
Despite the recommended use of polygenic scores in characterizing genetic risk, there is concern about its predictive power given prior research generally showed that polygenic scores, particularly those for alcohol outcomes, only account for a very small amount of variance of the alcohol phenotypes (Clarke et al., Reference Clarke, Smith, Gelernter, Kranzler, Farrer, Hall and McIntosh2016; Salvatore et al., Reference Salvatore, Aliev, Edwards, Evans, Macleod, Hickman and Dick2014). Family history of alcohol problems is a robust predictor for individuals’ alcohol use disorder and related outcomes (Kendler et al., Reference Kendler, Edwards, Myers, Cho, Adkins and Dick2015; Powers, Berger, Fuhrmann, & Fendrich, Reference Powers, Berger, Fuhrmann and Fendrich2017). Because alcohol use disorders are heritable (Verhulst, Neale, & Kendler, Reference Verhulst, Neale and Kendler2015), family history of alcohol problems has been considered to indicate genetic risk for alcohol problems (Slutske et al., Reference Slutske, D'Onofrio, Turkheimer, Emery, Harden, Heath and Martin2008), although it can also influence individuals’ alcohol outcomes through environmental pathways (Chassin, Pillow, Curran, Molina, & Barrera, Reference Chassin, Pillow, Curran, Molina and Barrera1993; Leonard & Eiden, Reference Leonard and Eiden2007). Thus, we also considered family history of alcohol problems as an additional proxy/indicator of genetic risk for alcohol problems.
We examined the following two hypotheses:
Hypothesis 1: Polygenic risk for alcohol dependence, family history of alcohol problems, perceived peer deviance, and interpersonal traumatic events would be associated with greater risk for alcohol use disorder symptoms in college.
Hypothesis 2: The effects of peer deviance and interpersonal traumatic events on alcohol use disorder symptoms would be moderated by alcohol dependence genome-wide polygenic risk scores and family history of alcohol problems. Specifically, we hypothesized that the effects of peer deviance and interpersonal traumatic events on alcohol use disorder symptoms would be stronger for individuals who had higher alcohol dependence genome-wide polygenic risk scores and/or had a family history of alcohol problems.
Method
Sample
Data came from the Spit for Science (S4S) Study, an ongoing longitudinal study of how genetic and environmental factors impact substance use and related mental health outcomes across the college years and beyond (Dick et al., Reference Dick, Nasim, Edwards, Salvatore, Cho, Adkins and Kendler2014). Incoming students aged 18 or older in a large public urban university in the mid-Atlantic region were invited to participate in an online survey at the beginning of the fall semester of their freshman year and provide a saliva sample for DNA using the Oragene collection kit. Freshmen students who did not participate during the fall also had the opportunity to enroll in the study during the spring semester. Participants subsequently complete a follow-up online survey in the spring of each year while attending college and beyond graduation. All surveys were administered using the REDCap software (Harris et al., Reference Harris, Taylor, Thielke, Payne, Gonzalez and Conde2009). The self-report questionnaires assess a wide range of psychosocial and environmental factors, including alcohol and substance use, mental health, family and peer influences, and stressful life events. Informed consent was obtained from all participants, and all study procedures were approved by the university's institutional review board.
Data collection for S4S began in the fall of 2011, and five cohorts of incoming freshman students have been enrolled in the study currently (n = 12,025). Participation rates have been high, with 67% of the eligible incoming students participating in S4S, and the demographic characteristics of the S4S participants closely map onto the diverse university population (Dick et al., Reference Dick, Nasim, Edwards, Salvatore, Cho, Adkins and Kendler2014). The sample for the current study consisted of the first three cohorts of participants who enrolled in S4S in the fall of 2011, 2012, or 2013 and provided DNA samples for genotyping. These participants have been followed up throughout their college years and thus have completed up to five assessments across college. The current study included 1,339 participants who were classified as of African ancestry based on genetic ancestry principal component analysis of the larger S4S sample. Of these 1,339 African American participants, 220 were missing data on alcohol use disorder symptoms across all five assessments across college, either because they had never used alcohol or because they chose not to answer related questions. Thus, the final analytic sample included 1,119 African American participants (74% female); 9.3% of this sample completed all five assessments of alcohol use disorder symptoms, 18.9% completed four assessments, 17.2% completed three assessments, 26.5% completed two assessments, and 28.2% completed one assessment. Age at enrollment for the sample ranged from 18 to 24 years (M = 18.44, SD = 0.42).
Measures
Alcohol use disorder symptoms
Participants reported on their alcohol use behaviors at every assessment. Alcohol use disorder symptoms was operationalized as the number of alcohol use disorder criteria (e.g., Have you had times when you ended up drinking more, or longer, than you intended?) ever endorsed according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2013), using questions adapted from the Semi-Structured Assessment for the Genetics of Alcoholism (Bucholz et al., Reference Bucholz, Cadoret, Cloninger, Dinwiddie, Hesselbrock, Nurnberger and Schuckit1994). Sum scores were created and prorated to account for missingness (when at least 50% of the items were available), with a possible range of 0 to 11.
Peer deviance
At each assessment, participants answered six questions about their friends’ behaviors over the past 12 months (Kendler, Jacobson, Myers, & Eaves, Reference Kendler, Jacobson, Myers and Eaves2008). Specifically, participants reported on how many of their friends had smoked cigarette, got drunk, had problems with school, had been drunk in school, had been in trouble with the law, and had smoked marijuana. Response options for each question ranged from 0 (none) to 4 (all). Items were summed and prorated to account for missingness (when at least 50% of the items were available). The Cronbach's α reliability ranged from 0.81 to 0.87 across assessments.
Interpersonal traumatic events
Participants reported whether they had experienced different categories of potentially traumatic events within the past 12 months prior to the completion of the survey based on an abbreviated Life Events Checklist (Gray, Litz, Hsu, & Lombardo, Reference Gray, Litz, Hsu and Lombardo2004). Response options were 0 (no) and 1 (yes) for each category. Types of potentially traumatic events assessed were physical assaults, sexual assaults, other unwanted sexual experiences, transportation accidents, and natural disasters. The current study focused only on physical assaults, sexual assaults, and other unwanted sexual experiences because prior research showed that these interpersonal potentially traumatic events were important risk factors for risky health behaviors while accidental traumatic events were not related (Moore et al., Reference Moore, Overstreet, Kendler, Dick, Adkins and Amstadter2017). Scores were summed across three categories to create a summary variable indicating interpersonal potentially traumatic events. Scores ranged from 0 to 3.
Genotyping and alcohol dependence genome-wide polygenic scores
Participants’ DNA samples were genotyped on the Affymetrix BioBank Array that contains 653k both common and rare genetic variants. Genotyping was performed at Rutgers University Cell and DNA Repository, with quality control performed locally following procedures from the Psychiatric Genomics Consortium. Imputation was conducted using the HapMap 1000 genomes Phase 3 reference panel. Single nucleotide polymorphisms (SNPs) with a genotyping rate <0.95 or that violated Hardy–Weinberg equilibrium (p < 10−6) or with minor allele frequency <0.01 were excluded from analysis. We used genome-wide association estimates for alcohol dependence from Gelernter et al. (Reference Gelernter, Kranzler, Sherva, Almasy, Koesterer, Smith and Farrer2014), the largest published GWAS of alcohol dependence for African Americans to date, to calculate alcohol dependence genome-wide polygenic scores for participants in our sample. We used the score procedure in PLINK (Purcell et al., Reference Purcell, Neale, Todd-Brown, Thomas, Ferreira, Bender and Sham2007), which computes a linear function of the number of scored alleles an individual possesses weighted by the associated GWAS t statistic. Matching SNPs were pruned for linkage disequilibrium based on HapMap 1000 Genomes Phase 3 reference panel genotype data for African ancestry, with clumping based on Gelernter et al. (Reference Gelernter, Kranzler, Sherva, Almasy, Koesterer, Smith and Farrer2014) GWAS p values using a 500kb physical distance and a linkage disequilibrium threshold of r 2 ≥ .25. Given that there are no set criteria for establishing a threshold to create maximally informative scores (Evans, Visscher, & Wray, Reference Evans, Visscher and Wray2009), we calculated a series of polygenic scores in our sample that included SNPs meeting decreasingly stringent p value thresholds (p < .0001, p < .001, p < .01, p < .05, p < .10, p < .20, p < .30, p < .40, and p < .50) in the discovery GWAS (Gelernter et al., Reference Gelernter, Kranzler, Sherva, Almasy, Koesterer, Smith and Farrer2014). The percent variance accounted for (R2) in alcohol use disorders symptoms in our sample using these polygenic scores ranged from <0.01% to 0.25%.
Family history of alcohol problems
During the assessment at the fall semester of their freshman year, participants answered questions about alcohol problems for four types of relatives: mother, father, aunts/uncles/grandparents, and siblings. For example, the question for mothers was “Do you think your biological mother has ever had problems with alcohol (by problems with alcohol we mean that her alcohol use caused problems at home, at work, with her health, or with the police, or that she received alcohol treatment)?” This question was repeated for “biological father,” “aunts/uncles/grandparents,” and “biological siblings.” Response options for each question were 0 (no) and 1 (yes). Following the approach used in prior research (Kendler et al., Reference Kendler, Edwards, Myers, Cho, Adkins and Dick2015), standardized mean scores of the responses for all four categories of relatives were calculated to indicate family history of alcohol problems. Scores for family history of alcohol problems ranged from –0.12 to 3.13 in our sample.
Covariates
Genetic ancestry principal components were considered as covariates to account for potential population stratification (Hellwege et al., Reference Hellwege, Keaton, Giri, Gao, Velez Edwards and Edwards2017). Because our sample included college students from three different cohorts, we considered cohort (indexed using two dummy-coded variables) as a covariate to account for potential cohort effects. We also included gender (1 = male, 0 = female) and age as control variables, given prior evidence that alcohol use problems and related environmental risk factors may differ between males and females and across age (Grant et al., Reference Grant, Goldstein, Saha, Chou, Jung, Zhang and Hasin2015; Schulte, Ramo, & Brown, Reference Schulte, Ramo and Brown2009).
Analytic plan
We began with preliminary analyses to examine descriptive statistics and intercorrelations between key study variables. We then conducted multilevel growth curve modeling to examine the trajectory of alcohol use disorder symptoms across college years and the effects of genetic risk, peer deviance, and interpersonal traumatic events in predicting the trajectory of alcohol use disorder symptoms, using Mplus version 7.3, with maximum information likelihood with robust standard errors as estimator. We used this estimator because it provides standard errors that are robust to nonnormality, and the distribution of alcohol use disorder symptoms in our sample was positively skewed. Multilevel growth curve models are useful in handling nested data structures (e.g., repeated measures nested within individuals) and are flexible in accounting for missing values within individuals (Raudenbush & Bryk, Reference Raudenbush and Bryk2002).
A series of multilevel growth models were conducted. First, we estimated an unconditional growth model for alcohol use disorder symptoms (Model 1). We specified both linear and quadratic growth curve models to examine the nature of the growth trajectory of alcohol use disorder symptoms over time, given that we had up to five assessments of alcohol use disorder symptoms in our sample that allowed for the examination of quadratic growth. Time values were specified as 0, 0.5, 1.5, 2.5, and 3.5, for the five assessments across college years, respectively, to represent the number of years from the baseline assessment; time was centered at the first assessment point, and the intercept can be interpreted as the group mean at the beginning of college. Second, after establishing a baseline growth model, we examined the main effects of alcohol dependence genome-wide polygenic risk scores or family history of alcohol problems on the trajectory of alcohol use disorder symptoms by adding them as time-invariant predictors of the intercept and slope, while including gender, cohort, and genetic ancestry principal components (time invariant) and age (time varying) as control variables for the intercept and slope (Model 2). Third, peer deviance and interpersonal traumatic events were included as time-varying covariates to examine their effects on alcohol use disorder symptoms across time (Model 3). Fixed effects of peer deviance and interpersonal traumatic events represent their average effects on alcohol use disorder symptoms across time. Random effects of peer deviance and interpersonal traumatic events examined variations in effects of these variables across individuals, and the fixed effects of peer deviance and interpersonal traumatic events showed the main effects of these variables adjusted for between-person variance in these effects and in alcohol use disorder symptoms. Fourth and finally, in the case that random effects of peer deviance and interpersonal traumatic events were statistically significant, which would indicate that these associations vary across individuals, a subsequent model was fit to examine whether alcohol dependence genome-wide polygenic risk scores or family history of alcohol problems moderate effects of peer deviance and interpersonal traumatic events on alcohol use disorder symptoms (Model 4).
Results
Preliminary analysis
We excluded 220 participants who were missing data on alcohol use disorder symptoms across all five assessments during college, either because they had never used alcohol or because they chose not to answer related questions, from analyses. Compared to these excluded participants, those included in the analytic sample were more likely to be female, t (1,331) = 2.48, p = .01, scored higher for family history of alcohol problems, t (1,302) = 2.84, p < .01, reported more interpersonal traumatic events at spring semester of the freshmen, t (1,326) = 2.14, p = .03, and sophomore, t (715) = 2.88, p < .01, years, and reported higher levels of peer deviance across college years (all p < .01). There were no significant differences in alcohol dependence genome-wide polygenic risk scores between those excluded and those included in the final analytic sample.
Out of the whole analytic sample, 418 (37.4%) students completed assessment of alcohol use disorder symptoms during the spring semester of their senior year. Compared to those who did not completed assessments during their senior year, these students reported lower peer deviance at their baseline assessment, t (1,106) = –4.48, p < .001, and were more likely to be female, t (1,112) = –2.98, p < .01. There were no significant differences in alcohol dependence genome-wide polygenic risk scores, family history of alcohol problems, alcohol use disorder symptoms at other assessments, and interpersonal traumatic events between those who completed follow-up assessments during their senior year and those who did not.
Descriptive statistics and correlations between the key study variables are summarized in Table 1. Participants in our sample on average endorsed more than one alcohol use disorder symptom at the beginning of college, and the mean of alcohol use disorder symptoms increased across the college years. Alcohol use disorder symptoms across assessments were correlated (r ranged from .20, a small correlation, to .62, a medium correlation), suggesting that alcohol use disorder symptoms were only moderately stable across college years. Peer deviance and interpersonal traumatic events were positively correlated (r ranged from .10, a small correlation, to .44, a medium correlation) with alcohol use disorder symptoms across assessments, confirming that they are important factors to consider in relation to alcohol use disorder symptoms in college. Alcohol dependence genome-wide polygenic risk scores were generally not correlated with alcohol use disorder symptoms, peer deviance, and interpersonal traumatic events, except that alcohol dependence genome-wide polygenic risk scores at p < .05 or larger thresholds were positively correlated (small effect sizes; r < .20) with alcohol use disorder symptoms and peer deviance assessed during senior year. Family history of alcohol problems was positively correlated with alcohol use disorder symptoms, peer deviance, and interpersonal traumatic events across assessments (small effect sizes, r ≤ .20), but was not correlated with alcohol dependence genome-wide polygenic risk scores across different p value thresholds.
Table 1. Descriptive statistics and bivariate correlations between key study variables
Note: Bolded coefficients are significant at p < .05. AUDsx, alcohol use disorder symptoms. Y1f, Year 1 Fall; y1s, Year 1 Spring; y2s, Year 2 Spring; y3s, Year 3 Spring; and y4s, Year 4 Spring. Peer, peer deviance. Trauma, interpersonal traumatic events. AD_GPS, alcohol dependence genome-wide polygenic scores. Coefficients for AD_GPS across different p value thresholds (ranged from p < .0001 to p < .50) are presented. FH_alc, family history of alcohol problem.
Trajectory of alcohol use disorder symptoms across college years
Because we had five repeated assessments for alcohol use disorder symptoms, we started with estimating an unconditional growth model with linear and quadratic effects of time. Results indicated that both the linear (B = 0.44, p < .001) and the quadratic slopes (B = –0.06, p = .025) were statistically significant. In addition, the unconditional growth model specifying both linear and quadratic slopes demonstrated significantly better model fit than a growth model only specifying a linear slope (Δχ2 = 14.27, df = 4, p < .001). Thus, a quadratic growth model was established as the baseline model for subsequent analyses. The intercept for the trajectory of alcohol use disorder symptoms is 1.62, indicating that African American college students in our sample on average had 1.62 alcohol use disorder symptoms at the beginning of college. The positive linear slope indicated that, on average, alcohol use disorder symptoms increased over time during college. However, the negative quadratic slope suggested that the rate of increase in alcohol use disorder symptoms slowed down over college years (Table 2, Model 1). The variance components for both the intercept (σ2 = 1.68, p < .001) and linear slope (σ2 = 0.92, p = .078, marginally significant) were statistically significant, indicating that there was significant variability among individuals regarding their alcohol use disorder symptoms when enrolled in college as well as their linear rate of increase in alcohol use disorder symptoms over time. However, the variance component of the quadratic slope was not statistically significant (σ2 = 0.04, p = .242), suggesting that the decline in rate of increase in alcohol use disorder symptoms over college years did not differ among individuals.
Table 2. Growth curve modeling of alcohol use disorder symptoms across college years: Effects of alcohol dependence genome-wide polygenic scores, peer deviance, and interpersonal traumatic events
Note: Estimates of unstandardized coefficients are presented for fixed effects. Estimates of (residual) variance components are presented for random effects. Est, estimates. SE, standard error. AD_GPS, alcohol dependence genome-wide polygenic scores. Coefficients for AD_GPS at p < .10 are presented in the table. We note that patterns of associations are consistent across AD_GPS p value thresholds, and detailed results for AD_GPS across all p value thresholds are available upon request. Trauma, interpersonal traumatic events. +p < .10; *p < .05; **p < .01; ***p < .001. Coefficients for covariates are not presented for clarity of presentation.
Main effects of genetic risk, peer deviance, and interpersonal traumatic events
Main effects of genetic risk
Main effects of alcohol dependence genome-wide polygenic risk scores and family history of alcohol problems were examined in separate models, by adding them as time-invariant (between-person level) predictors of the intercept and linear slope to the unconditional growth model, while controlling for age, sex, cohort, and genetic ancestry principal components. We did not include predictors for the quadratic slope because results from the unconditional growth model indicated that there was no significant variation in the quadratic slope among individuals to predict. Controlling for covariates, alcohol dependence genome-wide polygenic risk scores were not significantly associated with alcohol use disorder symptoms at the baseline assessment (B = –2.17, SE = 1.66, p = .19, 95% confidence interval; CI [–4.91, 0.562]) or the rate of increase (B = 1.31, SE = 0.77, p = .09, 95% CI [0.05, 2.57]) in alcohol use disorder symptoms over time (Table 2, Model 2a). Family history of alcohol problems was positively associated with the intercept (B = 0.42, SE = 0.10, p < .001, 95% CI [0.26, 0.60]), suggesting that individuals with a family history of alcohol problems had more alcohol use disorder symptoms at the beginning of college. Family history of alcohol problems accounted for 7.7% of the variance in the students’ alcohol use disorder symptoms at baseline. However, family history of alcohol problems was not associated with rate of increase in alcohol use disorder symptoms across college years (B = 0.08, SE = 0.05, p = .15, 95% CI [–0.01, 0.16]; Table 3, Model 2b.
Table 3. Growth curve modeling of alcohol use disorder symptoms across college years: Effects of family history of alcohol problems, peer deviance, and interpersonal traumatic events
Note: Estimates of unstandardized coefficients are presented for fixed effects. Estimates of (residual) variance components are presented for random effects. Est, estimates. SE, standard error. FH_alc, family history of alcohol problems. Trauma, interpersonal traumatic events. +p < .10; *p < .05; **p < .01; ***p < .001. Coefficients for covariates are not presented for clarity of presentation.
Main effects of peer deviance and interpersonal traumatic events
Main effects of peer deviance and interpersonal traumatic events were examined simultaneously by adding them as time-varying covariates (within-person level) to the model where alcohol dependence genome-wide polygenic risk scores and all covariate variables were included. Results indicated that both peer deviance (B = 0.14, SE = 0.01, p < .001, 95% CI [0.12, 0.16]) and interpersonal traumatic events (B = 0.44, SE = 0.07, p < .001, 95% CI [0.32, 0.55]) were associated with more alcohol use disorder symptoms (Table 2, Model 3). In addition, random effects of peer deviance (σ2 = 0.01, p = .01) and interpersonal traumatic events (σ2 = 0.38, p = .01) were both significant, suggesting variations in the effects of these variables on alcohol use disorder symptoms among individuals.
Interactions between genetic risk, peer deviance, and interpersonal traumatic events
Given results from Model 3 that effects of peer deviance and interpersonal traumatic events varied across individuals, we examined whether alcohol dependence genome-wide polygenic risk scores or family history of alcohol problems moderated these effects. Specifically, we examined these moderation effects by regressing the random slopes of peer deviance and interpersonal traumatic events on alcohol dependence genome-wide polygenic risk scores or family history of alcohol problems. In order to reduce potential bias in testing G × E effects, we followed recommendations by Keller (Reference Keller2014) to include interaction terms between alcohol dependence genome-wide polygenic risk scores and all covariates (i.e., age, sex, cohort, and genetic ancestry principal components), as well as interactions between covariates and environmental factors (i.e., peer deviance and interpersonal traumatic events) as control variables in testing the moderation effects of alcohol dependence genome-wide polygenic risk scores. Neither alcohol dependence genome-wide polygenic risk scores (Table 2, Model 4a) nor family history of alcohol problems (Table 3, Model 4b) predicted the random slopes of peer deviance (B = 0.01, SE = 0.24, p = .97, 95% CI [–0.38, 0.40] and B = 0.01, SE = 0.02, p = .57, 95% CI [–0.02, 0.03] for alcohol dependence genome-wide polygenic risk scores and family history of alcohol problems, respectively) and interpersonal traumatic events (B = –0.22, SE = 1.70, p = .90, 95% CI [–3.02, 2.57] and B = –0.02, SE = 0.10, p = .87, 95% CI [–0.17, 0.14] for alcohol dependence genome-wide polygenic risk scores and family history of alcohol problems, respectively), suggesting that they did not moderate the effects of peer deviance and interpersonal traumatic events on alcohol use disorder symptoms.
Discussion
In this study, we examined the independent and interactive effects between genetic risk for alcohol problems, indexed by alcohol dependence genome-wide polygenic risk scores and family history of alcohol problems, peer deviance, and interpersonal traumatic events on the trajectory of alcohol use disorder symptoms across college years in a sample of African American students. We found that alcohol use disorder symptoms generally increased across college years, with the rate of increase decline over time. Alcohol dependence genome-wide polygenic risk scores did not predict trajectory of alcohol use disorder symptoms, while family history of alcohol problems was associated with alcohol use disorder symptoms at the beginning of college but not associated with the rate of change in symptoms across college. Perceived peer deviance and interpersonal traumatic events were associated with more alcohol use disorder symptoms across college. No significant G × E effects were found such that neither alcohol dependence genome-wide polygenic risk scores nor family history of alcohol problems moderated the effects of peer deviance and interpersonal traumatic events on alcohol use disorder symptoms.
There is a large literature suggesting that peer deviance and adverse life events are salient environmental risk factors for alcohol use and misuse in adolescents and young adults, among European Americans as well as other racial/ethnic minority groups (Hawkins et al., Reference Hawkins, Catalano and Miller1992; Keyes, Hatzenbuehler, & Hasin, Reference Keyes, Hatzenbuehler and Hasin2011; Read et al., Reference Read, Colder, Merrill, Ouimette, White and Swartout2012). Consistent with our hypothesis, results of our longitudinal analysis built on prior research to show that perceived peer deviance and experience of interpersonal traumatic events such as physical and sexual assaults are associated with increased alcohol use disorder symptoms over time across college years. Furthermore, findings indicated that there was a significant amount of variance in the effects of peer deviance and interpersonal traumatic events on alcohol use disorder symptoms. These findings suggest that African American college students are not equally susceptible to the adverse influences of peer deviance and interpersonal traumatic events, and identifying factors that may buffer or exacerbate these effects will have important implications for prevention and intervention efforts targeting alcohol problems among college students.
Contrary to our prediction, polygenic risk for alcohol dependence as measured in our sample did not predict alcohol use disorder symptoms. Although we observed some significant bivariate correlations between alcohol dependence genome-wide polygenic risk scores and alcohol use disorder symptoms (see Table 1), alcohol dependence genome-wide polygenic risk scores was not associated with alcohol use disorder symptoms in the multilevel growth model where genetic ancestry principal components (along with other covariates) were controlled to account for potential bias due to population stratification. These findings emphasize the importance of taking into account population stratification in genetic analysis (Hellwege et al., Reference Hellwege, Keaton, Giri, Gao, Velez Edwards and Edwards2017). That important bias due to population stratification existed within our sample of African ancestry suggests that there is a significant degree of genetic diversity within the African American population (Campbell & Tishkoff, Reference Campbell and Tishkoff2008), which may have implications for variations in alcohol problems and related psychiatric outcomes within this group. In addition, the majority of the GWAS discovery sample was ascertained based on severe substance use problems (Gelernter et al., Reference Gelernter, Kranzler, Sherva, Almasy, Koesterer, Smith and Farrer2014). This may have implications for the current study such that genetic risk for alcohol dependence derived in a densely affected clinical sample may not translate to alcohol use disorder symptoms in a college-aged community sample (Savage et al., Reference Savage, Salvatore, Aliev, Edwards, Hickman, Kendler and Dick2018).
Matching ancestry between the GWAS discovery and the target sample, and an extremely large discovery sample size, are two critical factors for deriving nonbiased, well-powered polygenic scores (Martin et al., Reference Martin, Gignoux, Walters, Wojcik, Neale, Gravel and Kenny2017). Although we tried to maximize the predictive power of alcohol dependence genome-wide polygenic risk scores by using estimates from the largest GWAS on alcohol dependence for African Americans published to date (Gelernter et al., Reference Gelernter, Kranzler, Sherva, Almasy, Koesterer, Smith and Farrer2014), the sample size of the discovery GWAS sample was still relatively small (n = 3,318), particularly in comparison to other larger GWAS studies that have been more successful in characterizing genetic risk for psychiatric outcomes (e.g., Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2014). Thus, our null finding of the effect of polygenic risk for alcohol dependence on alcohol use disorder symptoms is likely due to low predictive power of the alcohol dependence genome-wide polygenic risk scores calculated to use in our study. Consortium efforts are ongoing to increase the sample size in GWAS of alcohol dependence and related phenotypes for African Americans (Agrawal, Edenberg, & Gelernter, Reference Agrawal, Edenberg and Gelernter2016). Future research is needed to replicate current findings with better powered polygenic scores, such as using estimates from larger GWAS conducted by the Psychiatric Genetics Consortium Substance Use Disorders Working group (Agrawal et al., Reference Agrawal, Edenberg and Gelernter2016) when these results become available.
Consistent with our hypothesis, family history of alcohol problems was associated with more alcohol use disorder symptoms at the beginning of college; however, it was not associated with the rate of increase in alcohol use disorder symptoms during college. This is consistent with prior research that showed associations between family history of alcohol problems and alcohol outcomes among college students (Kendler et al., Reference Kendler, Edwards, Myers, Cho, Adkins and Dick2015; LaBrie, Migliuri, Kenney, & Lac, Reference LaBrie, Migliuri, Kenney and Lac2010; Powers et al., Reference Powers, Berger, Fuhrmann and Fendrich2017). We note that previous studies on the associations between family history of alcohol problems and college students’ alcohol outcomes tend to be cross-sectional. Our findings from longitudinal analysis suggest that, while family history of alcohol problems is associated with students’ baseline risk for alcohol problems when starting college, other psychosocial factors may be more important in predicting trajectory of alcohol use disorder symptoms over the college years.
Contrary to our hypothesis, the effects of peer deviance and interpersonal traumatic events in predicting alcohol use disorder symptoms were not moderated by genetic risk for alcohol problems in our sample of African American college students. Nonsignificant interaction effects were observed across the two indicators of genetic risk (i.e., alcohol dependence genome-wide polygenic risk scores and family history of alcohol problems). Although there is substantial evidence that effects of peer deviance and adverse life events on alcohol use and related outcomes vary as a function of individuals’ genotypes, particularly for adolescents, G × E effects in young adulthood have been less studied. It is possible that G × E processes may operate differently in young adulthood (and in college specifically) compared to adolescence, given the substantial changes and transitions in the environmental and socioemotional context during this developmental period (Arnett, Reference Arnett2005), as well as prior evidence that genetic and environmental influences on alcohol use outcomes change across development (Edwards & Kendler, Reference Edwards and Kendler2013; Meyers et al., Reference Meyers, Salvatore, Vuoksimaa, Korhonen, Pulkkinen, Rose and Dick2014). Furthermore, G × E effects have been understudied in African American populations (Dick et al., Reference Dick, Barr, Guy, Nasim and Scott2017), and our hypotheses regarding G × E effects were derived from prior evidence based primarily on research conducted with samples of European Americans. It is possible that the patterns of G × E effects previously observed in European Americans are not generalizable to African Americans (Olfson et al., Reference Olfson, Edenberg, Nurnberger, Agrawal, Bucholz, Almasy and Kuperman2014; Sartor et al., Reference Sartor, Wang, Xu, Kranzler and Gelernter2014). It is also possible that, other cultural risk and protective factors relevant to African Americans, such as racial discrimination and religiosity (Causadias, Reference Causadias2013; Causadias & Korous, Reference Causadias, Korous, Causadias, Telzer and Gonzales2018; Obasi, Wilborn, Cavanagh, Yan, & Ewane, Reference Obasi, Wilborn, Cavanagh, Yan, Ewane, Causadias, Telzer and Gonzales2018), are potential moderators of genetic risk for alcohol problems and related outcomes within this population (Brody et al., Reference Brody, Beach, Chen, Obasi, Philibert, Kogan and Simons2011). Alternatively, our null finding of G × E effects could be due to low predictive power of our alcohol dependence genome-wide polygenic risk scores as discussed above. Thus, we reserve caution in concluding from our null G × E finding that genetic risk for alcohol dependence does not moderate environmental risk on alcohol problems, until these findings are replicated with well-powered polygenic scores.
It is interesting to note that family history of alcohol problems, but not alcohol dependence genome-wide polygenic risk scores, was associated with alcohol use disorder symptoms in our sample. This is consistent with prior research that found family history of alcohol dependence to be much more predictive of alcohol problems than polygenic risk scores (Yan et al., Reference Yan, Aliev, Webb, Kendler, Williamson, Edenberg and Dick2014). We note that alcohol dependence genome-wide polygenic scores were not significantly correlated with family history of alcohol problems in our sample. The lack of significant correlation in our sample of African Americans could be due to our relatively poor measurement of polygenic risk for alcohol dependence and/or limitations and potential bias of our self-report measure of family history of alcohol problems. Given the complexity of the genetic architecture underlying alcohol use disorders, although a useful and promising strategy, polygenic risk scores may be not of much predictive utility until GWAS of larger sample sizes are available to better characterize genetic risk for alcohol problems and better powered polygenic scores can be derived (Clarke et al., Reference Clarke, Smith, Gelernter, Kranzler, Farrer, Hall and McIntosh2016; Salvatore et al., Reference Salvatore, Aliev, Edwards, Evans, Macleod, Hickman and Dick2014). This is particularly true for populations of non-European descent whom have been historically underrepresented in genetic research (Dick et al., Reference Dick, Barr, Guy, Nasim and Scott2017). In the meantime, consistent with prior research (Yan et al., Reference Yan, Aliev, Webb, Kendler, Williamson, Edenberg and Dick2014), our findings suggest that family history of alcohol problems can be a useful alternative way to index genetic risk for alcohol problems and help further our understanding of how risks unfold across development and in conjunction with the environment. We do acknowledge, however, that family history of alcohol problems can represent both genetic and environmental pathways of risk for alcohol problems (Chassin et al., Reference Chassin, Pillow, Curran, Molina and Barrera1993; Slutske et al., Reference Slutske, D'Onofrio, Turkheimer, Emery, Harden, Heath and Martin2008), and it is important to keep this in mind when interpreting the influence of family history of alcohol problems on individuals’ alcohol use outcomes.
Our study has several notable strengths. The longitudinal design with repeated measures of alcohol use disorder symptoms, peer deviance, and interpersonal traumatic events allowed us to examine the associations across the college years. We focused on African Americans, an understudied population in genetic and G × E research. In addition, moving beyond candidate genes, we employed a genome-wide polygenic approach to characterize genetic risk and conducted hypothesis-driven tests of G × E effects. Finally, we followed Keller's recommendations (Reference Keller2014) to adjust our G × E models with all cross-term interactions (among covariates, genome-wide polygenic scores, and environmental factors) to account for potential confounding and gene–environment correlations.
Despite these strengths, findings from this study need to be interpreted in light of several limitations. First, our measure of polygenic risk for alcohol dependence is underpowered and thus might not be a good indicator of genetic risk. This is currently a general limitation of the alcohol research field (particularly for minority populations). Efforts to expand inclusion of non-European samples in genetic research are essential, which has implications for culture and developmental psychopathology studies. Second, all of our phenotypic measures were self-report and thus may be subject to reporting bias. For example, our measure of family history of alcohol problems was based on participants’ perceptions of alcohol problems of their first-degree (i.e., biological parents and siblings) and second-degree (i.e., grandparents, uncles, and aunts) relatives, which may be biased because participants may not know the alcohol problems of their relatives well. Data collected directly from participants’ relatives would be more informative in measuring family history of alcohol problems. Third, our measure of alcohol use disorder symptoms represented lifetime alcohol use disorder symptoms and was not specific to recent alcohol use disorder symptoms. This means that, in our study, alcohol use disorder symptoms could either remain stable or increase over time and we were not able to model all possible forms of change (e.g., decrease) in alcohol use disorder symptoms across college years. Future studies that measure alcohol use disorder symptoms at a specific time frame (e.g., past year) would be important to replicate our findings.
We focused on peer deviance and interpersonal traumatic events because they are well-documented environmental risk factors for alcohol problems (Hawkins et al., Reference Hawkins, Catalano and Miller1992; Read et al., Reference Read, Colder, Merrill, Ouimette, White and Swartout2012) and have been well studied in G × E studies using samples of European Americans (Dick & Kendler, Reference Dick and Kendler2012). Although our study represents a first step in integrating culture to developmental psychopathology using a cultural genomic approach by focusing on a sample of African American college students (Causadias, Reference Causadias2013; Causadias & Korous, Reference Causadias, Korous, Causadias, Telzer and Gonzales2018), we note that we did not assess culture directly. There are other environmental stressors (e.g., experiences of racial discrimination and daily microaggressions) that are salient risk factors for alcohol and related problems among African Americans (Williams et al., Reference Williams, Mohammed, Leavell and Collins2010). There is also neurobiological evidence linking stress dysregulation and alcohol and drug use in racially diverse samples such as African American communities (Obasi et al., Reference Obasi, Wilborn, Cavanagh, Yan, Ewane, Causadias, Telzer and Gonzales2018). Equally important is the need for more investigations on the unique and shared cultural protective factors (e.g., racial/ethnic identity, cultural socialization, and religiosity) that buffer the effect of adversity among African American individuals on shaping development and psychopathology (Causadias, Reference Causadias2013; García Coll, Akerman, & Cicchetti, Reference García Coll, Akerman and Cicchetti2000). Thus, future studies are needed to use a multiple-level analysis to examine the interplay between cultural risk and protective factors, genetic predispositions, and neurobiological systems in influencing alcohol problems and related outcomes in African Americans and other racial/ethnic groups (Brody et al., Reference Brody, Beach, Chen, Obasi, Philibert, Kogan and Simons2011; Causadias & Korous, Reference Causadias, Korous, Causadias, Telzer and Gonzales2018; Obasi et al., Reference Obasi, Wilborn, Cavanagh, Yan, Ewane, Causadias, Telzer and Gonzales2018).
In conclusion, our research extended the literature by taking a longitudinal design and a genome-wide polygenic approach to study how genetic risk for alcohol problems may interact with salient environmental risk factors (i.e., peer deviance and interpersonal traumatic events) to predict alcohol use disorder symptoms among African American college students. No significant main effect of alcohol dependence polygenic risk scores or G × E effect were found for predicting alcohol use disorder symptoms, suggesting the need for larger sample sizes in gene discovery efforts to better characterize genetic risk for alcohol problems (Agrawal et al., Reference Agrawal, Edenberg and Gelernter2016), with a particular need for increasing representation of populations of non-European descent in these efforts (Dick et al., Reference Dick, Barr, Guy, Nasim and Scott2017). Our findings indicated peer deviance and experience of interpersonal traumatic events as salient risk factors that elevate risk for alcohol problems among African American college students, suggesting that these could be important targets for prevention and intervention efforts aimed at reducing college drinking problems. For example, providing support to help students who have experienced physical or sexual assaults cope with related emotional distress may be an effective strategy to reduce their risk for alcohol problems. Our findings also suggest that family history of alcohol problems could be a useful alternative indicator of genetic risk for alcohol problems. Examination of how genetic risk, indexed by family history of alcohol problems or well-powered polygenic risk scores, interacts with culturally relevant environmental risk and protective factors (e.g., discrimination) in predicting alcohol use disorders over time in populations of non-European descent would be an important future direction of research in this area.
