Genetic Modification of the Relationship between Parental Rejection and Adolescent Alcohol Use

Abstract

Aims

Parenting practices are associated with adolescents' alcohol consumption, however not all youth respond similarly to challenging family situations and harsh environments. This study examines the relationship between perceived parental rejection and adolescent alcohol use, and specifically evaluates whether youth who possess greater genetic sensitivity to their environment are more susceptible to negative parental relationships.

Methods

Analyzing data from the National Longitudinal Study of Adolescent Health, we estimated a series of regression models predicting alcohol use during adolescence. A multiplicative interaction term between parental rejection and a genetic index was constructed to evaluate this potential gene-environment interaction.

Results

Results from logistic regression analyses show a statistically significant gene-environment interaction predicting alcohol use. The relationship between parental rejection and alcohol use was moderated by the genetic index, indicating that adolescents possessing more ‘risk alleles’ for five candidate genes were affected more by stressful parental relationships.

Conclusions

Feelings of parental rejection appear to influence the alcohol use decisions of youth, but they do not do so equally for all. Higher scores on the constructed genetic sensitivity measure are related to increased susceptibility to negative parental relationships.

INTRODUCTION

The prevalence of alcohol consumption continues to be high among adolescents in the USA. Recent studies reveal that nearly 70 percent of US students consume alcohol at some point during high school and over half have gotten drunk prior to graduation (Johnston et al., 2013). Alcohol use during these formative years has been associated with numerous negative consequences—even at low to moderate levels of use (Reimuller et al., 2011). It increases risk for perpetration of violence, victimization, risky sexual behaviors, poor academic performance, traffic accidents, cognitive/emotional impairment, and death (CDC, 2013). An adolescent's decisions to consume alcohol are often tied to the drinking behaviors modeled by their parents (Seljamo et al., 2006), but the connection between parenting and adolescent drinking may extend beyond direct imitation.

Studies show that alcohol consumption is more likely among adolescents whose parents express less interest in their lives, provide less monitoring, less warmth, and are less informed about their activities and peers (Sieving et al., 2000). Such lack of interest can lead to perceptions of parental rejection among adolescents (Creemers et al., 2011; Visser et al., 2012). Perceived parental rejection, defined as an adolescent's negative subjective assessment of their emotional connection with a parent, is a particular source of stress associated with maladaptive coping and alcohol use. However, not all youth who experience parental rejection will turn to alcohol, making it important to understand factors making adolescents most susceptible. Although perceived parental rejection may be considered comparable to other parenting concepts (i.e. inattention and lack of attachments/connection), we utilize this term to be consistent with a leading theory on deviance (general strain theory) and because our focus lies in the youth's satisfaction with the relationship rather than the degree of parental involvement—youth may feel detached (i.e. rejected) from a parent which does frequently interact with them.

Some evidence indicates the association between parenting and adolescent substance use is conditioned by an adolescent's genotype (Otten et al., 2012). For instance, parental monitoring and parental involvement have been linked to substance use among adolescents who possess ‘risk alleles’ in the promoter regions of the DRD4 and 5-HTTLPR genes, respectively (Brody et al., 2009; Otten et al., 2012). Other studies have found the association between stress inducing experiences and substance use varies by genotype (Ducci et al., 2007; Vaske et al., 2009), a finding in line with evidence of a moderating influence of genotype on the relationship between parenting and child antisocial behaviors (Widom and Brzustowitcz, 2006).

There remain several avenues, however, for expanding research on the linkage between genes, parenting, and adolescent alcohol use. First, studies have often overlooked the possibility of how accumulated genetic risk might affect the relationship between parenting and alcohol use. The use of cumulative genetic risk scores rather than individual genes continues to be controversial; however, this framework is gaining traction. Genetic risk indices are often used with biological outcomes and increasingly with behavioral outcomes (e.g. Drury et al., 2012) including those related to the use of substances (Belsky et al., 2013). These indices are particularly useful when effect sizes may be small and datasets underpowered (Belsky et al., 2013). Second, studies on gene-parenting interactions predicting substance use have not accounted for other sources of stress (e.g. school, personal victimization, discrimination), psychosocial traits (e.g. self-control and self-esteem), and social influences (e.g. peer substance use, social support, parental monitoring, religiosity) that can additionally increase or decrease one's risk of substance use during adolescence. One extant study suggests that parental rejection's relationship with alcohol use is not conditioned by genotype, but only considered two genes (DRD2 and DRD4) and did not incorporate a full array of sociological factors that may have influenced substance use decisions (Creemers et al., 2011). Thus, there is merit in continuing to use exploratory studies for delving into the relationships between parental rejection, genotypes, and substance use.

The current study examines how the relationship between perceived parental rejection and drinking is moderated by adolescents' genetic sensitivity to stressful experiences while statistically controlling for numerous factors found to affect adolescent alcohol use. A combination of risk alleles identified for polymorphic genes including MAOA, DAT1, DRD2, DRD4, and 5-HTTLPR is examined to determine whether the association between perceived parental rejection and adolescent alcohol use is stronger for those who are more genetically susceptible to stressful experiences. Each of these genes code for proteins related to neurotransmission and have been demonstrated to modify environmental factors, albeit not parental rejection specifically, affecting alcohol use. Functions of each gene and references to its connection to alcohol consumption are described in Table 1; however, we hypothesize that cumulative genetic risk, rather than an individual gene, modifies the parental rejection-alcohol use relationship.

Table 1.

Components of genetic sensitivity measure

Gene	Function	Direct or interactive linkages to alcohol use	Genetic sensitivity scoring
			Scored 0	Scored 1	Scored 2
DAT1	Codes for a dopamine transporter protein involved in dopamine reuptake.	Bhaskar et al. (2012) Le Strat et al. (2008) Vaske et al. (2009)	Homozygous, 9-repeat (4.9%)	Heterozygous 9-repeat, 10-repeat (34.0%)	Homozygous 10-repeat (61.1%)
DRD2	Codes for the dopamine receptor D2 which is expressed mainly in striatum.	Berggren et al. (2006) Meyers et al. (2013) Mota et al. (2013)	Homozygous A2 (54.6%)	Heterozygous A1, A2 (37.6%)	Homozygous A1 (7.9%)
DRD4	Codes for the dopamine receptor D4 which is highly expressed in frontal cortex.	Bau et al. (2001) Otten et al. (2012) Park et al. (2011)	No 7-repeat alleles (63.7%)	Heterozygous with one 7-repeat allele (32.0%)	Homozygous 7-repeat (4.3%)
5-HTTLPR	Codes for a serotonin transporter protein involved in serotonin reuptake.	Brody et al. (2009) Pascale et al. (2015) Vaht et al. (2014)	Homozygous L (long) (33.7%)	Heterozygous S,L (46.4%)	Homozygous S (short) (19.0%)
MAOA	Codes for the MAOA enzyme involved in dopamine and serotonin metabolism.	Ducci et al. (2007) Stogner and Gibson (2013) Widom and Brzustowitcz, (2006)	Females: both alleles are high activity (3.5 or 4-repeat) (36.9%) Males: high activity allele (56.6%)	Females: heterozygous with one low activity (2, 3, or 5-repeat) allele (44.8%) Males: low activity allele (43.5%)	Females: both alleles are low activity (18.3%)

Each individuals ‘genetic sensitivity’ score is the total of their score for the five risk genes.

METHODS

Data and sample

Data from the National Longitudinal Study of Adolescent Health (Add Health), a longitudinal study of American youth enrolled in grades 7 through 12 during the 1994–1995 school year, were analyzed to examine our primary research question. The Add Health study used a multi-stage stratified sampling design to identify a sample of 80 high schools and 52 middle schools representative of schools within the USA in terms of region of country, urbanicity, size, type, and ethnic composition. Approximately 90,000 youth completed in-school self-report questionnaires and a subsample of 20,745 adolescents completed a more detailed in-home Wave I survey. Participants that had not graduated high-school were re-interviewed in their homes 1 year later as part of Wave II data collection. They were also interviewed 6 years later; genetic data were collected on a subsample of respondents at this time (Wave III). The analysis sample for the present study consists of a genotyped subsample of 1495 participants with non-missing data for all key variables. The initial study received approval the Institutional Review Board (IRB) on Research Involving Human Subjects at the University of North Carolina at Chapel Hill; youth respondents and their guardians provided written consent for participation. Secondary analysis of Add Health data was approved by the lead author's IRB under protocol number 13-05-19.

Measures

Alcohol use

Alcohol use was measured by asking respondents during the Wave II interview, ‘Have you had a drink of beer, wine, or liquor—not just a sip or a taste of someone else's drink—more than 2 or 3 times in your life?’ Those answering ‘yes’ were coded 1 whereas those answering ‘no’ were coded 0. Alcohol use rather than frequency or quantity of use is examined because individuals often err when reporting the frequency of how much they have drank (O'Malley et al., 1983), but should be able to self-report whether they have drank alcohol. Similarly, additional factors (e.g. body mass) may influence quantity that are unrelated to whether an individual uses. At the time of Wave II interviews (approximately 1 year after the predictor variables subsequently described were measured), 64.4% of the genetic sample reported having used alcohol more than two or three times.

Parental rejection

A measure of perceived parental rejection was constructed from adolescents' Wave I responses to eight parent-child relationship questions. Similar to Wright et al. (2008), we computed a maternal rejection measure as each respondent's average score for five Likert-type items which asked respondents to rate their agreement with statements on a five option scale ranging from ‘strongly agree’ to ‘strongly disagree.’ These statements assessed whether the respondent was satisfied with their relationship with their mother, felt she encouraged the respondent to be independent, considered the two to communicate well, considered the mother warm and loving, and how the mother responds to inappropriate behavior. Higher scores indicate more perceived parental rejection, or a more negative relationship perceived by the adolescent toward his/her mother. A paternal rejection measure was also created averaging three items that similarly assessed the respondent's perceived relationship with his/her father (maternal α = 0.836; paternal α = 0.887). The paternal and maternal measures are thus similar but not identical; while this is a limitation of the study, using these measures does allow for results more directly comparable to prior research (Wright et al., 2008). For respondents with two parents, the higher score was used to represent parental rejection since the most negative relationship is likely to be the most stressful. Each of the items included in the parental rejection measure is listed in Appendix 1.

Genetic sensitivity

A cumulative genetic index score was calculated for each adolescent. Following Belsky and Beaver (2011), each respondent's score represents a count of ‘risk alleles’ for five genes considered to be related to antisocial behavior and substance use. This includes the 10-repeat DAT1 alleles, A1 DRD2 alleles, 7-repeat DRD4 alleles, short (S) alleles for 5-HTTLPR, and low activity MAOA alleles. For each of the first four genes, no points were added to a subject's cumulative genetic risk score if he/she did not have any risk alleles, one if they had one risk allele, and two if both alleles were risk alleles. One was added for females with one low activity MAOA allele and two if both were low activity alleles. Since men have only one allele for the MAOA gene and males possessing one low activity allele are functionally equivalent to homozygous low activity allele females (Meyer-Lindenberg et al., 2006), two points were added to the genetic risk score for male participants with a low activity allele. As for females, the score of one was recorded given that heterozygous females show intermediate functioning (Meyer-Lindenberg et al., 2006). Cumulative genetic risk scores could range from 0 to 10 with higher scores indicating greater genetic sensitivity. Table 1 further details the genetic sensitivity measure and displays the portion of the sample possessing risk alleles for each gene. As part of the Add Health study, Hardy–Weinberg equilibrium was investigated for each demographic group. Deviations were only noted in one race (Blacks for DRD4 and 5-HTTLPR). Prior to continuing with the investigation, we evaluated whether a gene-environment correlation (rGE) was present and might confound results. A non-significant correlation between genetic sensitivity and perceived parental rejection was found (r = 0.021, P = 0.301), providing some confidence that an rGE was not present.

Statistical controls

Demographic variables included were age, race (0 = white, 1 = nonwhite), and sex (0 = female, 1 = male). The analytic sample is 48.0% male, 34.8% nonwhite, and has an average age of 15.8. Additionally, income served as a control variable and was measured as total household income in thousands of dollars; families netting more than $999,000 annually were scored as 999. To minimize other potential concerns of confounding, additional control variables were included based on findings from prior research examining risk factors for adolescent substance use. Each measure was replicated from a previous study (for details, see the referenced study). These include a four-item measure of negative school experiences (α = 0.694; Johnson and Morris, 2008), a two-item measure of perceived discrimination (Stogner and Gibson, 2013), a four-item measure of low self-control (α = 0.742; Stogner et al., 2014), a three-item measure of peer substance use (α = 0.756; Wright et al., 2008), a ten-item depression scale (α = 0.809; Johnson and Morris, 2008), and a two-item measure of adolescents' perceptions of neighborhood quality (Stogner and Gibson, 2013). Several binary response items were also included to control for whether respondents reported having been violently victimized, were physically abused, lived in a household receiving government aid, or had an angry temperament as characterized by their parents. With the exception of physical abuse as a child (measured during Wave III when respondents were asked to recall maltreatment prior to sixth grade), all risk factors were measured during the Wave I interview.

Four additional variables were included which have been related to lower alcohol use among adolescences. First, religiosity was measured as the average of three Wave I items (α = 0.800; Rostosky et al., 2003). Social support was measured by averaging respondents reports to seven items assessing perceived care from others in their life (α = 0.784; Johnson and Morris, 2008). Self-esteem was measured using seven items that capture respondents' overall perceptions of themselves (α = 0.849; Stogner and Gibson, 2013; Stogner, 2015) and a parental supervision measure was calculated as the average of seven Wave I items (Stogner and Gibson, 2013). Table 2 shows descriptive statistics for all variables.

Table 2.

Descriptive statistics (n = 1495)

	Mean	SD	Minimum	Maximum
Main variables
Alcohol initiation	0.644	0.479	0	1
Parental rejection	2.055	0.812	1	5
Genetic risk	4.200	1.527	0	10
Demographic controls
Age	15.579	1.672	12	20
Gender (1 = male)	0.480	0.500	0	1
Race (1 = nonwhite)	0.348	0.477	0	1
Income	47.320	53.758	0	999
Risk factors
Peer Substance Use	0.815	0.872	0	3
Negative school experiences	1.017	0.739	0	4
Discrimination	5.664	1.775	2	10
Violent victimization	0.209	0.407	0	1
Neighborhood disadvantage	4.529	1.863	2	10
Parental abuse/neglect	0.550	0.498	0	1
Government financial aid	0.399	0.490	0	1
Low self-control	2.201	0.623	1	5
Anger	0.318	0.466	0	1
Depression	0.661	0.464	0	2.70
Protective factors
Religiosity	1.716	0.995	0	3
Self-esteem	1.931	0.602	1	5
Social support	4.042	0.572	1.43	6
Parental supervision	2.003	1.694	0	7

Analysis

Logistic regression models were estimated to predict adolescents' self-reported alcohol use, and a step-wise process was used to arrive at a model that includes a multiplicative interaction term between parental rejection and the genetic sensitivity index, demographics, and other control variables. Supplemental analyses were conducted for those who did consume alcohol by using two additional dependent measures: an estimated number of alcoholic beverages consumed per month and a nine-item alcohol-related problem scale. As genetic risk did not moderate the relationship between parental rejection and these outcomes among alcohol consumers (i.e. interaction terms were nonsignificant), we focus our presentation of results on drinking among adolescents rather than the quantity of drinks or the problems that result from drinking.

RESULTS

Table 3 displays logit coefficients and odd ratios from five logistic regression models predicting alcohol use. Model A shows parental rejection has a statistically significant association with alcohol use when controlling for demographic variables (b = 0.436, OR = 1.547). Age (b = 0.256, OR = 1.292), sex (b = 0.141, OR = 1.151), and race (b = −0.299, OR = 0.742) are significantly associated with alcohol use, indicating that males, whites, and older youth were more likely to report using alcohol (a similar model replacing the overall parental rejection measure with maternal rejection yielded a substantively similar result as did a model utilizing paternal rejection; as such, we focus on overall parental rejection). Model B includes a more extensive set of control variables to examine whether parental rejection remains significantly associated with adolescent alcohol use. Parental rejection has a non-significant association with alcohol use (b = 0.080, OR = 1.084). Of the demographic controls, age remains a significant predictor of alcohol use (b = 0.185, OR = 1.203). Of the additional control variables, having drug using friends (b = 1.249, OR = 3.485), negative school experiences (b = 0.224, OR = 1.251) and perceived discrimination (b = 0.104, OR = 1.109) increased the likelihood of adolescent alcohol use. Religiosity (b = −0.257, OR = 0.774) and parental supervision (b = −0.089, OR = 0.914) significantly decreased the likelihood of adolescent alcohol use.

Table 3.

Logistic regression models predicting adolescent alcohol use (n = 1495)

	Model A		Model B		Model C		Model D		Model E
	b (SE)	OR	b (SE)	OR	b (SE)	OR	b (SE)	OR	b (SE)	OR
Main variables
Parental rejection	0.436* (0.072)	1.547	0.080 (0.103)	1.084	0.110 (0.107)	1.117	0.112 (0.107)	1.119	−0.408 (0.285)	0.665
Genetic index	–	–	–	–	0.004 (0.048)	1.004	0.014 (0.048)	1.014	0.012 (0.048)	1.012
Genetic index × parental rejection	–	–	–	–	–	–	0.125* (0.060)	1.133	–	–
DAT1 × parental rejection	–	–	–	–	–	–	–	–	0.120 (0.138)	1.127
DRD2 × parental rejection	–	–	–	–	–	–	–	–	−0.035 (0.134)	0.965
DRD4 × parental rejection	–	–	–	–	–	–	–	–	0.001 (0.148)	1.001
5-HTTLPR × parental rejection	–	–	–	–	–	–	–	–	0.213† (0.124)	1.237
MAOA × parental rejection	–	–	–	–	–	–	–	–	0.215* (0.106)	1.239
Demographic controls
Age	0.256* (0.030)	1.292	0.185* (0.043)	1.203	0.187* (0.045)	1.206	0.188* (0.045)	1.207	0.189* (0.045)	1.207
Gender (1 = male)	0.141* (0.104)	1.151	−0.099 (0.132)	0.906	−0.099 (0.140)	0.905	−0.102 (0.139)	0.903	−0.096 (0.141)	0.909
Race (1 = nonwhite)	−0.299* (0.11)	0.742	−0.093 (0.143)	0.911	−0.098 (0.152)	0.906	−0.908 (0.153)	0.913	−0.076 (0.153)	0.926
Income	0.000 (0.010)	1.000	−0.001 (0.001)	0.999	−0.001 (0.001)	0.999	−0.001 (0.001)	0.997	−0.001 (0.001)	0.999
Risk factors
Negative school experiences	–	–	0.224* (0.108)	1.251	0.200† (0.112)	1.221	0.194† (0.112)	1.214	0.186 (0.113)	1.205
Perceived discrimination	–	–	0.104* (0.038)	1.109	0.112* (0.040)	1.118	0.116* (0.040)	1.124	0.117* (0.040)	1.124
Violent victimization	–	–	0.292 (0.189)	1.339	0.290 (0.196)	1.337	0.293 (0.197)	1.341	0.302† (0.198)	1.352
Neighborhood disadvantage	–	–	−0.050 (0.037)	0.952	−0.046 (0.037)	0.955	−0.052 (0.038)	0.949	−0.057 (0.039)	0.945
Government financial aid	–	–	−0.206 (0.141)	0.814	−0.240 (0.147)	0.786	−0.241 (0.147)	0.786	−0.229 (0.147)	0.795
Parental abuse	–	–	0.238† (0.127)	1.269	0.291* (0.131)	1.338	0.295* (0.132)	1.343	0.286* (0.132)	7.331
Low self-control	–	–	−0.004 (0.108)	0.996	0.024 (0.111)	1.025	0.030 (0.111)	1.030	0.025 (0.112)	1.026
Anger	–	–	0.205 (0.144)	1.228	0.260† (0.148)	1.298	0.267† (0.149)	1.305	0.274† (0.149)	1.316
Peer Substance Use	–	–	1.249* (0.109)	3.485	1.285* (0.113)	3.617	1.290* (0.114)	3.636	1.298* (0.113)	3.663
Depression	–	–	−0.183 (0.179)	0.833	−0.207 (0.185)	0.813	−0.205 (0.185)	0.815	−0.202 (0.186)	0.817
Protective factors
Religiosity	–	–	−0.257* (0.066)	0.774	−0.276* (0.069)	0.759	−0.274* (0.069)	0.761	−0.277* (0.069)	0.758
Self-esteem	–	–	0.033 (0.141)	1.033	0.048 (0.145)	1.049	0.043 (0.146)	1.043	0.042 (0.146)	1.043
Social support	–	–	−0.257† (0.157)	0.773	−0.196 (0.161)	0.822	−0.202 (0.161)	0.817	−0.208 (0.162)	0.812
Parental supervision	–	–	−0.089* (0.043)	0.914	−0.090* (0.044)	0.914	−0.091* (0.045)	0.913	−0.090* (0.045)	0.914
Constant			−2.011		−2.513		−2.548			−1.427
Pseudo R2			0.226		0.234		0.236			0.238

*P < 0.05; ^†P < 0.10.

Model C reports the direct association between the genetic sensitivity index and adolescent alcohol use while controlling for demographic and other control variables. The genetic sensitivity index had a non-significant direct association with alcohol use (b = 0.004, OR = 1.004). With few exceptions, its inclusion did not substantively alter associations between control variables and alcohol use. However, physical abuse became significant and positively related to alcohol use (b = 0.291, OR = 1.338) and negative school experiences became non-significant (b = 0.200, OR = 1.221).

Model D includes the interaction between parental rejection and genetic sensitivity. This multiplicative interaction term was created as the product of mean-centered parental rejection scores and the genetic sensitivity index. This interaction term significantly predicted alcohol use (b = 0.125, OR = 1.133), indicating that parental rejection has a stronger influence on alcohol use for those possessing increased genetic vulnerability. Age (b = 0.188, OR = 1.207), having drug using friends (b = 1.290, OR = 3.636), perceived discrimination (b = 0.116, OR = 1.124), physical abuse (b = 0.295, OR = 1.343), religiosity (b = −0.274, OR = 0.761), and parental supervision (b = −0.091, OR = 0.913) retained their significant associations with alcohol use.

Figure 1 plots the conditional predicted probabilities of alcohol use by perceived parental rejection and genetic sensitivity. Probabilities of alcohol use are shown for individuals at varying numbers of risk alleles across a range of parental rejection scores after all other variables included in Model D are statistically held at their respective means. Alcohol use among adolescents with none to a few genetic risk alleles is largely unaffected by perceived parental rejection. The predicted probability of alcohol use for an adolescent with no risk alleles and who reported the highest perceived parental rejection was only 0.11 greater than for an adolescent who reported the lowest perceived parental rejection. Those who possess a higher number of genetic risk alleles are highly sensitive to perceived parental rejection. For example, the predicted probability of alcohol use is very low among those with nine risk alleles who perceive very low parental rejection (0.18), but it is highly likely (0.99) for those with the same number of risk alleles who perceive high parental rejection.

Fig. 1.

Predicted probabilities of adolescents' reported alcohol use by level of parental rejection and genetic vulnerability.

Figure 2 is similarly included for illustrative purposes; it presents the predicted probability of alcohol initiation by parental rejection level across a range of genetic risk scores. As shown, differences in perceived parental rejection only alter the predicted probability of alcohol initiation slightly when no risk alleles are present. As shown on the right side of Fig. 2, parental rejection has a notable effect on the likelihood of alcohol initiation for those with a large number of risk alleles. Figure 2 also shows that those with more risk alleles are less likely to initiate alcohol use when parental rejection is below the mean (indicating a higher quality of parental attachment) compared to those who possess fewer ‘risk alleles’. In this way, genetic risk may be more accurately considered genetic susceptibility to parental influence.

Fig. 2.

Predicted probability of reporting alcohol use by level of parental rejection across the range of genetic risk.

In order to determine whether an individual gene was driving the interaction depicted in Model D, the analysis was repeated removing the genetic risk-parental rejection variable and introducing five gene-parental rejection interaction variables. Each of these was created as the number of risk alleles for that specific gene multiplied by a mean-centered parental rejection score. Coefficients associated with control variables remained substantively equivalent to those in Table 3 with age, perceived discrimination, parental abuse, deviant peer association, religiosity, and supervision maintaining their relationship with early alcohol use. Of the five genes, it appears that MAOA, 5-HTTLPR, and DAT1 are most responsible for the genetic modification of the risk resulting from perceived parental rejection; however, only the coefficient for the MAOA-parental rejection interaction meet the traditional threshold for significance (P = 0.044). Its coefficient and odds ratio (b = 0.215, OR = 1.239) were similar in magnitude to that of 5-HTTLPR (b = 0.213, OR = 1.237) which was found to be only marginally significant (P = 0.088). A Clogg comparison of coefficients test did not demonstrate the two to be significantly different, nor was the DAT1 interaction term significantly different from either of these two. In regards to the link between perceived parental rejection and alcohol consumption, this suggests that each may contribute to genetic susceptibility.

DISCUSSION

In line with existing studies (e.g. Horton and Gil, 2008; Visser et al., 2012), no direct association between parental rejection and alcohol use was identified after statistically controlling for other risk factors. However, this work builds on extant research by depicting how genetic variability moderates the association between perceived parental rejection and alcohol use among adolescents. When perceiving more parental rejection, adolescents possessing more genetic risk alleles were more likely to report alcohol use than those with fewer of these alleles. Adolescents least likely to report alcohol use when perceiving low parental rejection were also those with the most genetic sensitivity. Thus, those with higher genetic sensitivity scores were both more likely to consume alcohol when experiencing negative relationships and less likely to consume alcohol when no negative parental relationship exists. As low scores on the parental rejection scale represent a positive parental relationship, these results might be interpreted as indicating that higher scores on the genetic risk index are associated with both more negative outcomes as a result of parental rejection and more positive outcomes as a result of high perceived parental affection.

Perceived parental rejection seems to influence these individuals' behavior far more than those with a limited number of risk alleles. These findings are in line with studies showing that gene-environment interactions predict substance use and those that provide evidence of differential genetic susceptibility to environment stimuli (Belsky and Beaver, 2011; Simons et al., 2013). However, our results run counter to others who have examined the interaction between particular genes and parental rejection when predicting alcohol use (Creemers et al., 2011). The important distinction may be this study's inclusion of a genetic sensitivity measure operationalized as a count of risk alleles from multiple polymorphic genes linked to substance use and antisocial behavior. It appears likely that the influence of experienced parental rejection on substance use behavior is conditioned more so by the combination of genetic sensitivity alleles than any individual gene.

Practitioners and policymakers should view this finding conservatively given the need for replication and extension. However, the results highlight that genetic risk does not generally alter substance use propensity independent of the environment. Much like many other biosocial studies, this study notes that genetic risk is associated with an increased propensity for problematic behavior when paired with an environmental issue. In this case, the work suggests that environmental interventions targeting the parent-child relationship will lessen early alcohol initiation risk significantly for those at heightened risk. However, given the broader positive benefits of a nurturing parental relationship, interventions such as parenting classes, family conflict resolution programs, and support to single-parent households are warranted regardless of genetic risk; however, their efficacy as related to alcohol use is likely greater for the subgroup at heightened risk.

Our results are specific to low or moderate alcohol use among adolescents. Although low levels of use can result in negative consequences for adolescents, future research should extend these findings by examining onset and frequency of alcohol use, as well as binge drinking. Since gene-environmental interplay varies at different points in the life-course (Skelton et al., 2013), it is inappropriate to assume that identical processes influence adult alcohol consumption. Therefore, while composite genetic risk conditions the relationship between parental relationships and alcohol use for adolescents, a time when genetic effects on substance use are believed to be weakest, it may play a stronger role or even no role at all in childhood or early adulthood. The assessment of this relationship during childhood or adulthood should be the subject of future study. Research should also evaluate whether findings in this study extend to the use of other illegal substances such as marijuana, tobacco, and cocaine. Future research should also examine whether the relationship between perceived rejection from one parent and alcohol use may be complicated by the relationship the adolescent has with the other parent and/or additional family characteristics. For instance, the influence of perceived parental rejection may be conditional upon having one or both parents living in a household. In two parent households, a strong perception of rejection from one parent may be buffered by strong attachment to the other. In single-parent households, such a buffer may not exist, although some youth may form strong connections to other adults. Future research should consider the compounding influences of household structural dynamics in determining when genetic sensitivity to perceived parental rejection matters most.

Finally, noting the potential for epigenetic modification to affect and be affected by substance use (Jasiewicz et al., 2015), future investigations evaluating the relationship between genes, parental behavior, and alcohol use should attempt to explore epigenetic issues. Briefly, epigenetics refers to processes which alter genetic activity other than the base pair sequencing. This typically involves acetylation (or deacetylation) of histones or the methylation of a cytosine base (Powledge, 2009). Using cocaine as an example, Walsh et al. (2012) have recently highlighted the role these modifications may play in dependence. However, studies which champion the exploration of epigenetics as related to substance use (Randle et al., 2015), theoretically tie epigenetic modifications more closely to addiction than initiation or initial habituation (as was this study's primary focus). Further, adverse and stressful environments, such as a lack of nurturing relationship with parents, may also be linked to epigenetic changes.

Our results provide support for how genetic sensitivity to environment moderates the link between perceived parental rejection and alcohol use among adolescents using two waves of prospective longitudinal data to control for numerous family, peer, and individual risk and protective factors. Although it is important for parents to remain attentive to the quality and quantity of time spent with their offspring beyond childhood, it is also important to teach adolescents to affectively cope with feelings of rejection. As our study has shown, this is particularly relevant for reducing underage drinking among those possessing a heightened genetic sensitivity to perceived stressful environmental stimuli.

FUNDING

This research uses data from Add Health, a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwistle for assistance in the original design. Information on how to obtain the Add Health data files is available on the Add Health website (http://www.cpc.unc.edu/addhealth). No direct support was received from grant P01-HD31921 for this analysis.

CONFLICT OF INTEREST STATEMENT

None declared.

APPENDIX 1: ITEMS WITHIN PARENTAL REJECTION MEASURE

All items were scored as:

• 1 = strongly agree, 2 = agree, 3 = neither agree nor disagree, 4 = disagree, 5 = strongly disagree.

Questions asked about mothers:

1. Most of the time, your mother is warm and loving toward you.

2. Your mother encourages you to be independent.

3. When you do something wrong that is important, your mother talks about it with you and helps you understand why it is wrong.

4. You are satisfied with the way your mother and you communicate with each other.

5. Overall, you are satisfied with your relationship with your mother.

Questions asked about fathers:

1. Most of the time, your father is warm and loving toward you.

2. You are satisfied with the way your father and you communicate with each other.

3. Overall, you are satisfied with your relationship with your father.