ALCOHOLISM: CLINICAL AND EXPERIMENTAL RESEARCH
Vol. 38, No. 3 March 2014
Different Trajectories of Adolescent Alcohol Use: Testing Gene–Environment Interactions Carmen S. van der Zwaluw, Roy Otten, Marloes Kleinjan, and Rutger C. M. E. Engels
Background: Transitions into heavy alcohol use often already take place during adolescence and are likely to be both genetically and environmentally determined. Therefore, in a 6-wave longitudinal study, we examined the effects of DRD2 Taq1A and OPRM1 A118G genotypes and the interaction with parental rule-setting on different groups of adolescent drinkers. Methods: Growth mixture modeling resulted in 3 distinct groups of adolescent drinkers: light drinkers (n = 346), moderate drinkers (n = 178), and heavy drinkers (n = 72). Results: Multinomial regression showed that moderate drinkers carried the OPRM1 G allele and received lower levels of parental rule-setting significantly more often than the light drinking group. No other significant main effects of DRD2, OPRM1, and rule-setting were found. The interaction between OPRM1 genotype and parental rule-setting significantly distinguished the heavy drinkers from the light (p < 0.001) and moderate groups (p = 0.055): Particularly, the alcohol use of OPRM1 G allele carriers was affected by parental rule-setting, while AA genotype carriers remained largely unaffected by parental rules. Conclusions: Findings showed that different trajectories of adolescent drinking are preceded by a gene–parenting interaction. These results concur with Belsky’s theory of plasticity (2009), as well as with Shanahan and Hofer’s typology of a controlling and restricting gene–environment interaction (2005). Key Words: Adolescence, Alcohol, Gene, Parenting, Development.
H
EAVY DRINKING IN adolescence has been associated with a variety of negative health outcomes, such as cognitive impairment, damage to brain structures, and symptoms of depression and anxiety (Townshend and Duka, 2005). In addition, adolescent heavy drinkers are at increased risk of developing alcohol disorders in (young) adulthood (Duncan et al., 1997; Enoch, 2006). In most Western societies, some experimentation with alcohol is considered normative behavior. In the Netherlands, 85% of 16-year-olds report lifetime alcohol use (Van Dorsselaer et al., 2010).1 A minority of youth, however, escalates quickly into heavy drinking behavior and maintains high rates of alcohol use throughout adolescence and young adulthood (Duncan et al., 1997; Enoch, 2006; Van Der Vorst et al., 2009). Considering the risks associated with heavy drinking, it is
From the Freelance researcher at Zwaluwstyle (CSZ), Nijmegen, the Netherlands; Eindhoven University of Technology (CSZ), Eindhoven, the Netherlands; and Behavioural Science Institute (RO, MK, RCMEE), Radboud University Nijmegen, Nijmegen, the Netherlands. Received for publication January 21, 2013; accepted August 29, 2013. Reprint requests: Roy Otten, Behavioural Science Institute, Radboud University Nijmegen, PO Box 9104, 6500 HE Nijmegen, the Netherlands; Tel.: +31 24 361 5787; Fax: +31 24 361 2776; E-mail: [email protected]. nl Copyright © 2013 by the Research Society on Alcoholism. DOI: 10.1111/acer.12291 The legal drinking age in the Netherlands is 16 for beer and wine and 18 for strong liquor. At the time of writing, the Dutch parliament just accepted a proposal to change the law to increase the legal drinking age to 18 for all liquors. 1
704
important to identify factors that discriminate between normative trajectories of drinking and more severe patterns of heavy alcohol use. As alcohol consumption is a complex behavior that is likely affected by multiple predictors, it has been argued that the focus should be on both genetic and environmental risk factors, and on how they interact, to better understand alcohol use patterns in adolescence (Rutter, 2002). Twin studies have consistently shown that approximately 50% of interindividual differences in alcohol use disorders can be accounted for by genetic factors (Goldman et al., 2005). Alcohol consumption results in increased dopamine activity in the ventral striatum, particularly the nucleus accumbens, generating a reinforcing effect (Pierce and Kumaresan, 2006). In addition, alcohol enhances levels of endogenous opiates (particularly beta-endorphin) that bind to opioid receptors, which in turn increases dopaminergic activity (Herz, 1997). It has been suggested that the opioid and dopaminergic systems underlie different developmental stages of alcohol use (Berridge et al., 2009; Robinson and Berridge, 2003). Opioid neurotransmission in subcortical structures has been predominantly associated with the hedonic effects of alcohol (“liking”), which may be most prominent in the early stages of drinking. Animal studies have shown the involvement of opioid brain sites and genotypes in “liking” reactions to sweet tastes and drugs (Berridge et al., 2009). In addition, the G allele of the A118G (Asn40Asp) single-nucleotide polymorphism in the OPRM1 gene (that encodes the mu-opioid receptor) has been related to enhanced positive feelings and craving following alcohol use Alcohol Clin Exp Res, Vol 38, No 3, 2014: pp 704–712
G 9 E AND DEVELOPMENT OF ADOLESCENT DRINKING
(Ray and Hutchison, 2004; van den Wildenberg et al., 2007). The OPRM1 G allele has also been associated with alcohol use in young adolescents (Pieters et al., 2011, 2012). Miranda and colleagues (2010) found that adolescent ORPM1 G allele carriers developed more alcohol-related problems and endorsed drinking for mood enhancement more strongly than those who were homozygous for the A allele. In addition, drinking to enhance positive affect mediated the association between OPRM1 genotype and alcohol-related problems. Dopaminergic projections in the reward system, on the other hand, have been related to more compulsive stages of alcohol use (“wanting”) and are thought to result from altered (sensitized) brain systems following heavy exposure to alcohol (Robinson and Berridge, 2003). For example, hyperdopaminergic mutant mice display higher wanting, but not liking for sweet tastes (Peci~ na et al., 2003). Additionally, the DRD2 risk allele (rs1800497T) has been associated with reduced dopamine receptor sensitivity and availability, which has been observed in persons with alcohol dependence (Lucht et al., 2010). As “wanting” processes are thought to only emerge after considerable amounts of alcohol consumption, associations between dopaminergic genes and alcohol use may not (yet) be present in adolescent drinkers. For example, Pieters and colleagues (2011) and van der Zwaluw and colleagues (2010) found no direct effects of DRD2 and DRD4 genotypes on adolescent drinking. Parental monitoring has proved to be an important environmental factor in controlling adolescent substance use, although effect sizes are often small (Dick et al., 2007). With regard to adolescent alcohol consumption, more convincing results have been found for alcohol-specific monitoring: Studies have consistently shown that higher levels of parental alcohol-specific monitoring significantly reduce adolescent drinking, both cross-sectionally and prospectively (Koning et al., 2010; van der Vorst et al., 2005; van der Zwaluw et al., 2008; Van Zundert et al., 2006). Moreover, a recent prevention study showed that young adolescents’ alcohol use can be tackled by encouraging parents to be more stringent toward adolescent alcohol use (Koning et al., 2009). Parent behavior has the strongest influence on young adolescents’ (drinking) behavior, and it has been shown that parent effects decrease when adolescents grow older and their drinking behavior changes (Power et al., 2005). However, there is less knowledge on whether parental rule-setting can postpone alcohol use initiation and prevent youth from engaging in heavy drinking trajectories (Van Der Vorst et al., 2009). From twin studies, it is known that the effects of parental monitoring on adolescents’ substance use may be dependent on the genetic makeup of the adolescent. Kendler and colleagues (2011), for example, found that when parental monitoring was low, genetic effects on adolescent alcohol use were much more profound than when parental monitoring was high (Dick et al., 2007). These studies provided the first evidence that a genetic diathesis for substance use may be controlled by successful parental monitoring, or, conversely,
705
triggered by low levels of parental control (Shanahan and Hofer, 2005). Nevertheless, it is still unclear how these interactions operate on a (neuro)biological level. With respect to specific genes involved, several studies have shown significant interactions between the OPRM1 G allele and parental monitoring on adolescents’ alcohol use (Miranda et al., 2013; Pieters et al., 2012). Similarly, interactions have been found between OPRM1 genotype and friends’ behavior on adolescents’ drinking. Chassin and colleagues (2012) showed that in 17- to 23-year-old females carrying the OPRM1 G allele, alcohol use disorder symptoms increased more in response to the influence of alcohol promoting peers than in females with the AA genotype. Miranda and colleagues (2013) found that OPRM1 G allele carriers with high levels of deviant peer affiliations showed the greatest risk of developing an alcohol use disorder. Regarding the DRD2 gene, it was found that early adolescents carrying the DRD2 risk (T) allele consumed significantly more alcohol when their parents were permissive toward alcohol use (Pieters et al., 2012; van der Zwaluw et al., 2010), although this interaction was not found by Creemers and colleagues (2011). The bulk of genetic research on alcohol use has focused on alcohol dependence, while alcohol use disorders generally are considered the last stage in a trajectory of increasingly heavier alcohol use that may already begin in adolescence (Dick et al., 2009). Given that an estimated 40% of adult alcoholics have initiated drinking heavily in adolescence (Enoch, 2006), prospective studies will be highly informative in elucidating how genetic factors relate to alcohol use trajectories among adolescents. With the present study, we aim to extend the current G 9 E literature on adolescent alcohol use by focusing on the development of drinking throughout adolescence, instead of concentrating on a single measurement moment or on the onset of use in early adolescents. A group approach will be applied, in which separate groups with different alcohol use profiles will be distinguished. In the current study, we will examine whether DRD2 and OPRM1 genotypes affect alcohol use trajectories throughout adolescence and whether this association is moderated by parental alcohol-specific rules. Growth mixture modeling will be applied to identify different trajectories of adolescent alcohol use. Previous studies have identified diverse numbers of alcohol use trajectories during adolescence, ranging from 2 (Li et al., 2001) to 5 different classes (Van Der Vorst et al., 2009). A general pattern of findings that can be distilled from these studies is the identification of an often fairly large group of adolescents who consume low levels of alcohol and whose alcohol use slightly increases over time, which seems to be the more normative development of drinking during adolescence, and a small group that already initially shows high levels of drinking that further increase or stabilize over time (Flory et al., 2004). We thus hypothesize to find at least 2 distinct developmental trajectories; 1 with low, slightly increasing levels of alcohol use, and 1 heavy drinking group. By including both opioid and dopaminergic genotypes, we aim to cover genetic associations with different stages of the
VAN DER ZWALUW ET AL.
706
developmental process of alcohol (mis)use. We expect that the OPRM1 G allele and the DRD2 T allele will each be related to higher levels of alcohol use. Additionally, we will examine whether alcohol-specific parental rule-setting moderates the association between OPRM1/DRD2 genotypes and alcohol use trajectories. Because the aim of this study is to test whether parental rule-setting predicts (and thus precedes) youths’ engagement in a certain drinking trajectory, parental rule-setting was assessed only at Time 1 (T1). This will enable us to examine whether the interaction with parental rules is stronger for the early “liking”-phase of alcohol use or the more sensitized alcohol consumption stage of relatively heavy drinking. Given the study findings described above, it is hypothesized that the association between alcohol use and both DRD2 and OPRM1 genotypes will be moderated by parental alcohol-specific rule-setting. MATERIALS AND METHODS Sample Data were obtained from the Family and Health study that consisted of 6 yearly waves (for details on the sample selection, see also van der Vorst et al., 2005). In 2002, family addresses were obtained from Dutch municipality registers. Families were selected when they contained 2 adolescent children between the age of 13 and 16 (no twins). A total of 428 families took part in the Family and Health study at T1, in November and December 2002, and January of 2003. Families were annually visited at home by trained interviewers and answered extensive questionnaires. Saliva samples were collected in the fourth wave to enable genetic testing. The Medical Commission on Human Research in the Netherlands authorized the data collection, and informed written consent was obtained from adolescents and their parents. In the current study, participants were the adolescents of the families (n = 856 at T1). As the analyses include genetic testing, the 622 adolescents who were genotyped were included in the current study. Additionally, to refrain from population stratification, only adolescents born in the Netherlands were included in the analyses, resulting in a final sample of 596 adolescents. Participants (50% boys) were on average 14.3 years old at T1 (SD = 1.09) and approximately 5.4 years older at T6 (i.e., 19.7 years old). Although participants’ ages were not specifically requested at the waves following T1, they could be deducted from the time of assessment of the different waves. Participants’ estimated ages at the intermediate waves were 15.3 at T2, 16.3 at T3, 17.7 at T4, and 18.7 years at T5. Attrition analyses (cross-tabs and t-tests) showed no significant differences in sex, age, level of education, alcohol consumption, and parental rules (p > 0.05) between included and excluded adolescents. Measures Alcohol Use. At all waves, adolescents were asked, “How many glasses of alcohol did you drink in the past week?” The response had to be given in 4 subitems: number of glasses consumed at home, out, on weekdays, and during the weekend (Engels et al., 1999). The sum of these 4 subquestions represented the amount of alcohol-containing drinks consumed in the past week. To preclude skewness, the measure was recoded into 7 categories: 0 = 0 glasses, 1 = 1 or 2 glasses, 2 = 3 to 5 glasses, 3 = 6 to 10 glasses, 4 = 11 to 20 glasses, 5 = 21 to 30 glasses, and 6 = ≥ 31 glasses (cf. van der Zwaluw et al., 2008). Self-reports on alcohol use have proven to be reliable (Engels
et al., 2007), and the current measure has often been used in other studies (Bot et al., 2005; Koning et al., 2012). Rules. Parental alcohol-specific rule-setting was assessed at T1, with a 10-item questionnaire developed by van der Vorst and colleagues (2005). Adolescents were asked to answer on a scale from 1 (“not at all applicable to me”) to 5 (“completely applicable to me”) whether they were allowed to, for example, “drink alcohol at home when one of my parents is at home” or to “drink alcohol at a party with friends.” To generate a measure in which higher scores represented higher levels of parental rules, all items were recoded. Cronbach’s alpha was 0.93. Genes. High molecular weight DNA was isolated from saliva using Oragene containers (DNA genotek, Ottawa, ON, Canada). Both the DRD2 TaqI C>T polymorphism (rs1800497) and the OPRM1 118A>G polymorphism (rs1799971) were genotyped using Taqman analysis (OPRM1 assay ID: Taqman assay: C___8950074_1; reporter 1: VIC-A-allele, forward assay; Applied Biosystems, Nieuwerkerk a/d IJssel, The Netherlands. DRD2 assay ID: Taqman assay:C___7486676_10; reporter 1: VIC-A-allele, reverse assay; Applied Biosystems). Genotyping was carried out in a volume of 10 ll containing 10 ng of genomic DNA, 5 ll of Taqman Master Mix (29; Applied Biosystems), 0.125 ll of the Taqman assay, and 3.875 ll of water. Genotyping was performed on a 7500 Fast Real-Time PCR System, and genotypes were scored using the algorithm and software supplied by the manufacturer (Applied Biosystems). Due to low quantities of the homozygous variant genotypes (DRD2 TT: n = 27, OPRM1 GG: n = 6), these groups were combined with the heterozygous group, resulting in 2 groups of DRD2 (CC vs. CT and TT) and OPRM1 (AA vs. AG and GG) genotypes (cf. Pieters et al., 2012; van der Zwaluw et al., 2010). DRD2 genotype of 1 participant could not be determined, causing the analyses that included DRD2 genotype to be carried out at a sample size of n = 595. Allele frequencies were in Hardy–Weinberg equilibrium (DRD2: v2 = 2.47, p = 0.12; OPRM1: v2 = 0.12, p = 0.73). Strategy of Analyses Descriptives and Pearson correlations (Spearman for genotype variables) were conducted in SPSS 19. Latent class growth analysis (LCGA) was employed to identify growth trajectories in adolescents’ alcohol use from T2 to T6 in Mplus 5 (Muthen and Muthen, 1998–2007). In LCGA, individual development is captured in 2 latent variables; the starting point (intercept) and the level of growth over time (slope). These latent variables were used to categorize adolescents’ alcohol use into distinct classes: adolescents with similar developmental trajectories were categorized into the same class. There are various ways to determine what the optimal solution is regarding the number of classes. First, statistical parameters, such as the Bayesian information criterion (BIC), Akaike’s information criterion (AIC) index, and the posterior probabilities (i.e., entropy; to what extent can the sample be categorized into different classes) should be examined. Second, the usability of the number of classes and class size should be taken into account (cf. Van Der Vorst et al., 2009). For example, trajectories that are too similar or too small (i.e., contain few individuals) may be less informative (Muthen and Shedden, 1999). After identifying the different alcohol use trajectories (based on adolescents alcohol use from T2 to T6), stepwise multinomial regression analysis was applied in Mplus (Muthen and Muthen, 1998–2007). Hereby, it could be examined whether DRD2 genotype, OPRM1 genotype, parental rule-setting at T1, and the interactions between DRD2/OPRM1 and parental rules differentiated between the different alcohol use trajectories. Prediction of class membership
G 9 E AND DEVELOPMENT OF ADOLESCENT DRINKING
707
Table 1. Pearson and Spearman Correlations 1 1. Parental rules T1 2. Alcohol T1 3. Alcohol T2 4. Alcohol T3 5. Alcohol T4 6. Alcohol T5 7. Alcohol T6 8. OPRM1 9. DRD2
0.51*** 0.39*** 0.28*** 0.17*** 0.19*** 0.11* 0.00 0.08
2
0.53*** 0.42*** 0.34*** 0.32*** 0.28*** 0.02 0.03
3
0.60*** 0.43*** 0.40*** 0.35*** 0.02 0.00
4
0.54*** 0.49*** 0.47*** 0.02 0.02
5
0.61*** 0.59*** 0.02 0.04
6
0.54*** 0.09* 0.01
7
0.05 0.04
8
9
0.01
T = Time (wave). OPRM1: 1 = AA genotype, 2 = AG/GG genotypes. DRD2: 1 = CC genotype, 2 = CT/TT genotypes. *p < 0.05; ***p < 0.001. Greater scores on parental rules indicate higher levels of parental monitoring. Italic correlations are Spearman correlations.
was carried out in 3 steps: (i) control variables gender, age, and alcohol use at T1 were entered into the regression; (ii) DRD2 and OPRM1 genotypes and parental rule-setting at T1 were added to the model; and (iii) the interaction terms (DRD2*rules and OPRM1*rules) were inserted.2 Our sample included 2 siblings of 428 participating families, and consequently, our data were nested within families. Therefore, in all Mplus analyses, we controlled for the dependence (complexity) of the data by employing the “TYPE = COMPLEX” procedure combined with the “CLUSTER” command (Muthen and Muthen, 1998–2007). In this way, standard errors of the parameter estimates are corrected for dependency, resulting in unbiased estimates.
RESULTS Descriptives and Correlations Mean alcohol-specific parental rule-setting at T1 was 3.61 (SD = 0.96). Average alcohol consumption in the past week was 2.32 glasses at T1 (SD = 3.71), which increased to 11.09 glasses at T6 (SD = 11.90). These descriptives on alcohol use and parental rule-setting concur with those found in other, similar samples (Bot et al., 2005; Koning et al., 2012). A total of 398 adolescents (67%) carried the DRD2 homozygous CC genotype, while 197 (33%) were carriers of the CT or TT genotypes. With regard to OPRM1 genotype frequencies, 479 (80%) carried the AA genotype, and 117 adolescents (20%) carried the AG or GG genotype. Parental rule-setting at T1 was negatively related to alcohol use at all waves ( 0.11 ≤ r ≤ 0.51, p < 0.02; see Table 1). Rules were not significantly related to DRD2 (ϱ = 0.08, p = 0.06) and OPRM1 (ϱ = 0.00, p = 0.92) genotypes, excluding the possibility that gene–environment correlations were obscuring the results. With the exception of OPRM1 genotype and alcohol use at T5 (ϱ = 0.09, p = 0.05), were DRD2 and OPRM1 genotypes not directly related to alcohol consumption ( 0.05 ≤ ϱ ≤ 0.02,
p > 0.32). The alcohol measures at all waves were positively and significantly correlated with each other (0.28 ≤ r ≤ 0.61, p < 0.001). Developmental Alcohol Trajectories To examine the number of different developmental alcohol trajectories among the adolescents, an LCGA was carried out. First, a 2-class solution was examined, which showed the following fit-indices: BIC = 10,412, AIC = 10,355, entropy = 0.78, followed by the 3-class solution (BIC = 10,356, AIC = 10,286, entropy = 0.88), and the 4-class solution (BIC = 10,292, AIC = 10,209, entropy = 0.93). Although parameters of model fit continued to decrease, indicating a better fit, the 4-class option also provided one rather small class containing
Vol. 38, No. 3 March 2014
Different Trajectories of Adolescent Alcohol Use: Testing Gene–Environment Interactions Carmen S. van der Zwaluw, Roy Otten, Marloes Kleinjan, and Rutger C. M. E. Engels
Background: Transitions into heavy alcohol use often already take place during adolescence and are likely to be both genetically and environmentally determined. Therefore, in a 6-wave longitudinal study, we examined the effects of DRD2 Taq1A and OPRM1 A118G genotypes and the interaction with parental rule-setting on different groups of adolescent drinkers. Methods: Growth mixture modeling resulted in 3 distinct groups of adolescent drinkers: light drinkers (n = 346), moderate drinkers (n = 178), and heavy drinkers (n = 72). Results: Multinomial regression showed that moderate drinkers carried the OPRM1 G allele and received lower levels of parental rule-setting significantly more often than the light drinking group. No other significant main effects of DRD2, OPRM1, and rule-setting were found. The interaction between OPRM1 genotype and parental rule-setting significantly distinguished the heavy drinkers from the light (p < 0.001) and moderate groups (p = 0.055): Particularly, the alcohol use of OPRM1 G allele carriers was affected by parental rule-setting, while AA genotype carriers remained largely unaffected by parental rules. Conclusions: Findings showed that different trajectories of adolescent drinking are preceded by a gene–parenting interaction. These results concur with Belsky’s theory of plasticity (2009), as well as with Shanahan and Hofer’s typology of a controlling and restricting gene–environment interaction (2005). Key Words: Adolescence, Alcohol, Gene, Parenting, Development.
H
EAVY DRINKING IN adolescence has been associated with a variety of negative health outcomes, such as cognitive impairment, damage to brain structures, and symptoms of depression and anxiety (Townshend and Duka, 2005). In addition, adolescent heavy drinkers are at increased risk of developing alcohol disorders in (young) adulthood (Duncan et al., 1997; Enoch, 2006). In most Western societies, some experimentation with alcohol is considered normative behavior. In the Netherlands, 85% of 16-year-olds report lifetime alcohol use (Van Dorsselaer et al., 2010).1 A minority of youth, however, escalates quickly into heavy drinking behavior and maintains high rates of alcohol use throughout adolescence and young adulthood (Duncan et al., 1997; Enoch, 2006; Van Der Vorst et al., 2009). Considering the risks associated with heavy drinking, it is
From the Freelance researcher at Zwaluwstyle (CSZ), Nijmegen, the Netherlands; Eindhoven University of Technology (CSZ), Eindhoven, the Netherlands; and Behavioural Science Institute (RO, MK, RCMEE), Radboud University Nijmegen, Nijmegen, the Netherlands. Received for publication January 21, 2013; accepted August 29, 2013. Reprint requests: Roy Otten, Behavioural Science Institute, Radboud University Nijmegen, PO Box 9104, 6500 HE Nijmegen, the Netherlands; Tel.: +31 24 361 5787; Fax: +31 24 361 2776; E-mail: [email protected]. nl Copyright © 2013 by the Research Society on Alcoholism. DOI: 10.1111/acer.12291 The legal drinking age in the Netherlands is 16 for beer and wine and 18 for strong liquor. At the time of writing, the Dutch parliament just accepted a proposal to change the law to increase the legal drinking age to 18 for all liquors. 1
704
important to identify factors that discriminate between normative trajectories of drinking and more severe patterns of heavy alcohol use. As alcohol consumption is a complex behavior that is likely affected by multiple predictors, it has been argued that the focus should be on both genetic and environmental risk factors, and on how they interact, to better understand alcohol use patterns in adolescence (Rutter, 2002). Twin studies have consistently shown that approximately 50% of interindividual differences in alcohol use disorders can be accounted for by genetic factors (Goldman et al., 2005). Alcohol consumption results in increased dopamine activity in the ventral striatum, particularly the nucleus accumbens, generating a reinforcing effect (Pierce and Kumaresan, 2006). In addition, alcohol enhances levels of endogenous opiates (particularly beta-endorphin) that bind to opioid receptors, which in turn increases dopaminergic activity (Herz, 1997). It has been suggested that the opioid and dopaminergic systems underlie different developmental stages of alcohol use (Berridge et al., 2009; Robinson and Berridge, 2003). Opioid neurotransmission in subcortical structures has been predominantly associated with the hedonic effects of alcohol (“liking”), which may be most prominent in the early stages of drinking. Animal studies have shown the involvement of opioid brain sites and genotypes in “liking” reactions to sweet tastes and drugs (Berridge et al., 2009). In addition, the G allele of the A118G (Asn40Asp) single-nucleotide polymorphism in the OPRM1 gene (that encodes the mu-opioid receptor) has been related to enhanced positive feelings and craving following alcohol use Alcohol Clin Exp Res, Vol 38, No 3, 2014: pp 704–712
G 9 E AND DEVELOPMENT OF ADOLESCENT DRINKING
(Ray and Hutchison, 2004; van den Wildenberg et al., 2007). The OPRM1 G allele has also been associated with alcohol use in young adolescents (Pieters et al., 2011, 2012). Miranda and colleagues (2010) found that adolescent ORPM1 G allele carriers developed more alcohol-related problems and endorsed drinking for mood enhancement more strongly than those who were homozygous for the A allele. In addition, drinking to enhance positive affect mediated the association between OPRM1 genotype and alcohol-related problems. Dopaminergic projections in the reward system, on the other hand, have been related to more compulsive stages of alcohol use (“wanting”) and are thought to result from altered (sensitized) brain systems following heavy exposure to alcohol (Robinson and Berridge, 2003). For example, hyperdopaminergic mutant mice display higher wanting, but not liking for sweet tastes (Peci~ na et al., 2003). Additionally, the DRD2 risk allele (rs1800497T) has been associated with reduced dopamine receptor sensitivity and availability, which has been observed in persons with alcohol dependence (Lucht et al., 2010). As “wanting” processes are thought to only emerge after considerable amounts of alcohol consumption, associations between dopaminergic genes and alcohol use may not (yet) be present in adolescent drinkers. For example, Pieters and colleagues (2011) and van der Zwaluw and colleagues (2010) found no direct effects of DRD2 and DRD4 genotypes on adolescent drinking. Parental monitoring has proved to be an important environmental factor in controlling adolescent substance use, although effect sizes are often small (Dick et al., 2007). With regard to adolescent alcohol consumption, more convincing results have been found for alcohol-specific monitoring: Studies have consistently shown that higher levels of parental alcohol-specific monitoring significantly reduce adolescent drinking, both cross-sectionally and prospectively (Koning et al., 2010; van der Vorst et al., 2005; van der Zwaluw et al., 2008; Van Zundert et al., 2006). Moreover, a recent prevention study showed that young adolescents’ alcohol use can be tackled by encouraging parents to be more stringent toward adolescent alcohol use (Koning et al., 2009). Parent behavior has the strongest influence on young adolescents’ (drinking) behavior, and it has been shown that parent effects decrease when adolescents grow older and their drinking behavior changes (Power et al., 2005). However, there is less knowledge on whether parental rule-setting can postpone alcohol use initiation and prevent youth from engaging in heavy drinking trajectories (Van Der Vorst et al., 2009). From twin studies, it is known that the effects of parental monitoring on adolescents’ substance use may be dependent on the genetic makeup of the adolescent. Kendler and colleagues (2011), for example, found that when parental monitoring was low, genetic effects on adolescent alcohol use were much more profound than when parental monitoring was high (Dick et al., 2007). These studies provided the first evidence that a genetic diathesis for substance use may be controlled by successful parental monitoring, or, conversely,
705
triggered by low levels of parental control (Shanahan and Hofer, 2005). Nevertheless, it is still unclear how these interactions operate on a (neuro)biological level. With respect to specific genes involved, several studies have shown significant interactions between the OPRM1 G allele and parental monitoring on adolescents’ alcohol use (Miranda et al., 2013; Pieters et al., 2012). Similarly, interactions have been found between OPRM1 genotype and friends’ behavior on adolescents’ drinking. Chassin and colleagues (2012) showed that in 17- to 23-year-old females carrying the OPRM1 G allele, alcohol use disorder symptoms increased more in response to the influence of alcohol promoting peers than in females with the AA genotype. Miranda and colleagues (2013) found that OPRM1 G allele carriers with high levels of deviant peer affiliations showed the greatest risk of developing an alcohol use disorder. Regarding the DRD2 gene, it was found that early adolescents carrying the DRD2 risk (T) allele consumed significantly more alcohol when their parents were permissive toward alcohol use (Pieters et al., 2012; van der Zwaluw et al., 2010), although this interaction was not found by Creemers and colleagues (2011). The bulk of genetic research on alcohol use has focused on alcohol dependence, while alcohol use disorders generally are considered the last stage in a trajectory of increasingly heavier alcohol use that may already begin in adolescence (Dick et al., 2009). Given that an estimated 40% of adult alcoholics have initiated drinking heavily in adolescence (Enoch, 2006), prospective studies will be highly informative in elucidating how genetic factors relate to alcohol use trajectories among adolescents. With the present study, we aim to extend the current G 9 E literature on adolescent alcohol use by focusing on the development of drinking throughout adolescence, instead of concentrating on a single measurement moment or on the onset of use in early adolescents. A group approach will be applied, in which separate groups with different alcohol use profiles will be distinguished. In the current study, we will examine whether DRD2 and OPRM1 genotypes affect alcohol use trajectories throughout adolescence and whether this association is moderated by parental alcohol-specific rules. Growth mixture modeling will be applied to identify different trajectories of adolescent alcohol use. Previous studies have identified diverse numbers of alcohol use trajectories during adolescence, ranging from 2 (Li et al., 2001) to 5 different classes (Van Der Vorst et al., 2009). A general pattern of findings that can be distilled from these studies is the identification of an often fairly large group of adolescents who consume low levels of alcohol and whose alcohol use slightly increases over time, which seems to be the more normative development of drinking during adolescence, and a small group that already initially shows high levels of drinking that further increase or stabilize over time (Flory et al., 2004). We thus hypothesize to find at least 2 distinct developmental trajectories; 1 with low, slightly increasing levels of alcohol use, and 1 heavy drinking group. By including both opioid and dopaminergic genotypes, we aim to cover genetic associations with different stages of the
VAN DER ZWALUW ET AL.
706
developmental process of alcohol (mis)use. We expect that the OPRM1 G allele and the DRD2 T allele will each be related to higher levels of alcohol use. Additionally, we will examine whether alcohol-specific parental rule-setting moderates the association between OPRM1/DRD2 genotypes and alcohol use trajectories. Because the aim of this study is to test whether parental rule-setting predicts (and thus precedes) youths’ engagement in a certain drinking trajectory, parental rule-setting was assessed only at Time 1 (T1). This will enable us to examine whether the interaction with parental rules is stronger for the early “liking”-phase of alcohol use or the more sensitized alcohol consumption stage of relatively heavy drinking. Given the study findings described above, it is hypothesized that the association between alcohol use and both DRD2 and OPRM1 genotypes will be moderated by parental alcohol-specific rule-setting. MATERIALS AND METHODS Sample Data were obtained from the Family and Health study that consisted of 6 yearly waves (for details on the sample selection, see also van der Vorst et al., 2005). In 2002, family addresses were obtained from Dutch municipality registers. Families were selected when they contained 2 adolescent children between the age of 13 and 16 (no twins). A total of 428 families took part in the Family and Health study at T1, in November and December 2002, and January of 2003. Families were annually visited at home by trained interviewers and answered extensive questionnaires. Saliva samples were collected in the fourth wave to enable genetic testing. The Medical Commission on Human Research in the Netherlands authorized the data collection, and informed written consent was obtained from adolescents and their parents. In the current study, participants were the adolescents of the families (n = 856 at T1). As the analyses include genetic testing, the 622 adolescents who were genotyped were included in the current study. Additionally, to refrain from population stratification, only adolescents born in the Netherlands were included in the analyses, resulting in a final sample of 596 adolescents. Participants (50% boys) were on average 14.3 years old at T1 (SD = 1.09) and approximately 5.4 years older at T6 (i.e., 19.7 years old). Although participants’ ages were not specifically requested at the waves following T1, they could be deducted from the time of assessment of the different waves. Participants’ estimated ages at the intermediate waves were 15.3 at T2, 16.3 at T3, 17.7 at T4, and 18.7 years at T5. Attrition analyses (cross-tabs and t-tests) showed no significant differences in sex, age, level of education, alcohol consumption, and parental rules (p > 0.05) between included and excluded adolescents. Measures Alcohol Use. At all waves, adolescents were asked, “How many glasses of alcohol did you drink in the past week?” The response had to be given in 4 subitems: number of glasses consumed at home, out, on weekdays, and during the weekend (Engels et al., 1999). The sum of these 4 subquestions represented the amount of alcohol-containing drinks consumed in the past week. To preclude skewness, the measure was recoded into 7 categories: 0 = 0 glasses, 1 = 1 or 2 glasses, 2 = 3 to 5 glasses, 3 = 6 to 10 glasses, 4 = 11 to 20 glasses, 5 = 21 to 30 glasses, and 6 = ≥ 31 glasses (cf. van der Zwaluw et al., 2008). Self-reports on alcohol use have proven to be reliable (Engels
et al., 2007), and the current measure has often been used in other studies (Bot et al., 2005; Koning et al., 2012). Rules. Parental alcohol-specific rule-setting was assessed at T1, with a 10-item questionnaire developed by van der Vorst and colleagues (2005). Adolescents were asked to answer on a scale from 1 (“not at all applicable to me”) to 5 (“completely applicable to me”) whether they were allowed to, for example, “drink alcohol at home when one of my parents is at home” or to “drink alcohol at a party with friends.” To generate a measure in which higher scores represented higher levels of parental rules, all items were recoded. Cronbach’s alpha was 0.93. Genes. High molecular weight DNA was isolated from saliva using Oragene containers (DNA genotek, Ottawa, ON, Canada). Both the DRD2 TaqI C>T polymorphism (rs1800497) and the OPRM1 118A>G polymorphism (rs1799971) were genotyped using Taqman analysis (OPRM1 assay ID: Taqman assay: C___8950074_1; reporter 1: VIC-A-allele, forward assay; Applied Biosystems, Nieuwerkerk a/d IJssel, The Netherlands. DRD2 assay ID: Taqman assay:C___7486676_10; reporter 1: VIC-A-allele, reverse assay; Applied Biosystems). Genotyping was carried out in a volume of 10 ll containing 10 ng of genomic DNA, 5 ll of Taqman Master Mix (29; Applied Biosystems), 0.125 ll of the Taqman assay, and 3.875 ll of water. Genotyping was performed on a 7500 Fast Real-Time PCR System, and genotypes were scored using the algorithm and software supplied by the manufacturer (Applied Biosystems). Due to low quantities of the homozygous variant genotypes (DRD2 TT: n = 27, OPRM1 GG: n = 6), these groups were combined with the heterozygous group, resulting in 2 groups of DRD2 (CC vs. CT and TT) and OPRM1 (AA vs. AG and GG) genotypes (cf. Pieters et al., 2012; van der Zwaluw et al., 2010). DRD2 genotype of 1 participant could not be determined, causing the analyses that included DRD2 genotype to be carried out at a sample size of n = 595. Allele frequencies were in Hardy–Weinberg equilibrium (DRD2: v2 = 2.47, p = 0.12; OPRM1: v2 = 0.12, p = 0.73). Strategy of Analyses Descriptives and Pearson correlations (Spearman for genotype variables) were conducted in SPSS 19. Latent class growth analysis (LCGA) was employed to identify growth trajectories in adolescents’ alcohol use from T2 to T6 in Mplus 5 (Muthen and Muthen, 1998–2007). In LCGA, individual development is captured in 2 latent variables; the starting point (intercept) and the level of growth over time (slope). These latent variables were used to categorize adolescents’ alcohol use into distinct classes: adolescents with similar developmental trajectories were categorized into the same class. There are various ways to determine what the optimal solution is regarding the number of classes. First, statistical parameters, such as the Bayesian information criterion (BIC), Akaike’s information criterion (AIC) index, and the posterior probabilities (i.e., entropy; to what extent can the sample be categorized into different classes) should be examined. Second, the usability of the number of classes and class size should be taken into account (cf. Van Der Vorst et al., 2009). For example, trajectories that are too similar or too small (i.e., contain few individuals) may be less informative (Muthen and Shedden, 1999). After identifying the different alcohol use trajectories (based on adolescents alcohol use from T2 to T6), stepwise multinomial regression analysis was applied in Mplus (Muthen and Muthen, 1998–2007). Hereby, it could be examined whether DRD2 genotype, OPRM1 genotype, parental rule-setting at T1, and the interactions between DRD2/OPRM1 and parental rules differentiated between the different alcohol use trajectories. Prediction of class membership
G 9 E AND DEVELOPMENT OF ADOLESCENT DRINKING
707
Table 1. Pearson and Spearman Correlations 1 1. Parental rules T1 2. Alcohol T1 3. Alcohol T2 4. Alcohol T3 5. Alcohol T4 6. Alcohol T5 7. Alcohol T6 8. OPRM1 9. DRD2
0.51*** 0.39*** 0.28*** 0.17*** 0.19*** 0.11* 0.00 0.08
2
0.53*** 0.42*** 0.34*** 0.32*** 0.28*** 0.02 0.03
3
0.60*** 0.43*** 0.40*** 0.35*** 0.02 0.00
4
0.54*** 0.49*** 0.47*** 0.02 0.02
5
0.61*** 0.59*** 0.02 0.04
6
0.54*** 0.09* 0.01
7
0.05 0.04
8
9
0.01
T = Time (wave). OPRM1: 1 = AA genotype, 2 = AG/GG genotypes. DRD2: 1 = CC genotype, 2 = CT/TT genotypes. *p < 0.05; ***p < 0.001. Greater scores on parental rules indicate higher levels of parental monitoring. Italic correlations are Spearman correlations.
was carried out in 3 steps: (i) control variables gender, age, and alcohol use at T1 were entered into the regression; (ii) DRD2 and OPRM1 genotypes and parental rule-setting at T1 were added to the model; and (iii) the interaction terms (DRD2*rules and OPRM1*rules) were inserted.2 Our sample included 2 siblings of 428 participating families, and consequently, our data were nested within families. Therefore, in all Mplus analyses, we controlled for the dependence (complexity) of the data by employing the “TYPE = COMPLEX” procedure combined with the “CLUSTER” command (Muthen and Muthen, 1998–2007). In this way, standard errors of the parameter estimates are corrected for dependency, resulting in unbiased estimates.
RESULTS Descriptives and Correlations Mean alcohol-specific parental rule-setting at T1 was 3.61 (SD = 0.96). Average alcohol consumption in the past week was 2.32 glasses at T1 (SD = 3.71), which increased to 11.09 glasses at T6 (SD = 11.90). These descriptives on alcohol use and parental rule-setting concur with those found in other, similar samples (Bot et al., 2005; Koning et al., 2012). A total of 398 adolescents (67%) carried the DRD2 homozygous CC genotype, while 197 (33%) were carriers of the CT or TT genotypes. With regard to OPRM1 genotype frequencies, 479 (80%) carried the AA genotype, and 117 adolescents (20%) carried the AG or GG genotype. Parental rule-setting at T1 was negatively related to alcohol use at all waves ( 0.11 ≤ r ≤ 0.51, p < 0.02; see Table 1). Rules were not significantly related to DRD2 (ϱ = 0.08, p = 0.06) and OPRM1 (ϱ = 0.00, p = 0.92) genotypes, excluding the possibility that gene–environment correlations were obscuring the results. With the exception of OPRM1 genotype and alcohol use at T5 (ϱ = 0.09, p = 0.05), were DRD2 and OPRM1 genotypes not directly related to alcohol consumption ( 0.05 ≤ ϱ ≤ 0.02,
p > 0.32). The alcohol measures at all waves were positively and significantly correlated with each other (0.28 ≤ r ≤ 0.61, p < 0.001). Developmental Alcohol Trajectories To examine the number of different developmental alcohol trajectories among the adolescents, an LCGA was carried out. First, a 2-class solution was examined, which showed the following fit-indices: BIC = 10,412, AIC = 10,355, entropy = 0.78, followed by the 3-class solution (BIC = 10,356, AIC = 10,286, entropy = 0.88), and the 4-class solution (BIC = 10,292, AIC = 10,209, entropy = 0.93). Although parameters of model fit continued to decrease, indicating a better fit, the 4-class option also provided one rather small class containing
