Adolescence represents the time during which the majority of individuals initiate alcohol use (Johnston, O'Malley, & Bachman, Reference Johnston, O'Malley and Bachman2001). Two-thirds of Dutch adolescents have had their first drink by the age of 12 (Verdurmen, Monshouwer, van Dorsselaer, ter Bogt, & Vollebergh, Reference Verdurmen, Monshouwer, van Dorsselaer, ter Bogt and Vollebergh2005), and adolescents' alcohol use increases especially between 12 and 15 years of age (Poelen, Scholte, Engels, Boomsma, & Willemsen, Reference Poelen, Scholte, Engels, Boomsma and Willemsen2005). Next to alcohol use, depressive feelings are common during adolescence (Hankin et al., Reference Hankin, Abramson, Moffit, Silva, McGee and Angell1998; Skitch & Abela, Reference Skitch and Abela2008), and the prevalence rate of depressive symptoms increases from early adolescence (McGee, Feehan, Williams, & Anderson, Reference McGee, Feehan, Williams and Anderson1992), especially among girls (Costello, Mustillo, Erkanli, Keeler, & Angold, Reference Costello, Mustillo, Erkanli, Keeler and Angold2003). Thus, both alcohol use and depressive feelings are likely to develop during early adolescence and can have detrimental consequences for physical and mental health in adulthood.
Increasing interest in the developmental relation between alcohol use and depressive mood in adolescents has emerged recently (Saraceno, Munafó, Heron, Craddock, & van den Bree, Reference Saraceno, Munafó, Heron, Craddock and van den Bree2009). It has been found that alcohol use and depressive feelings often co-occur among adolescents (Clark & Bukstein, Reference Clark and Bukstein1998; Verdurmen et al., Reference Verdurmen, Monshouwer, van Dorsselaer, ter Bogt and Vollebergh2005). There have been some studies that focused on identifying psychosocial and health risk factors of developing a combination of substance use and mood disorders. Shared risk factors were found on both the individual and the environmental levels and included among others low academic achievement, family dysfunction, externalizing traits, and poor health status (Diaz, Simantov, & Rickert, Reference Diaz, Simantov and Rickert2002; Gau et al., Reference Gau, Chong, Yang, Yen, Liang and Cheng2007; Lewinsohn et al., Reference Lewinsohn, Roberts, Seeley, Rohde, Gotlib and Hops1994; Miller-Johnson, Lochman, Coie, Terry, & Hyman, Reference Miller-Johnson, Lochman, Coie, Terry and Hyman1998; Pardini, White, & Stouthamer-Loeber, Reference Pardini, White and Stouthamer-Loeber2007). Several twin studies have indicated that both adolescent and adult alcohol use as well as depressive mood are influenced by genetic factors (Pagan et al., Reference Pagan, Rose, Viken, Pulkkinen, Kaprio and Dick2006; Sullivan, Neale, & Kendler, Reference Sullivan, Neale and Kendler2000) and that there is overlap in the genetic influences on both traits (Edwards et al., Reference Edwards, Sihvola, Korhonen, Pukkinen, Moilanen and Kaprio2011; Tambs, Harris, & Magnus, Reference Tambs, Harris and Magnus1997). We examined the co-development of adolescents' alcohol use and depressive feelings by conducting a prospective study among early adolescents, because alcohol use and depressive feelings start to develop during that period. In addition, we aimed to investigate possible shared genetic effects in the development of both alcohol use and depressive feelings.
A gene that has been associated with both alcohol use and depression is the mu-opioid receptor gene (OPRM1, located on chromosome 6q24–q25). The OPRM1 gene influences the affinity of mu-opioid receptors in the brain to bind β-endorphins and enkephalins. These opioid peptides are released after intake of alcohol (Town et al., Reference Town, Abdullah, Crawford, Schinka, Ordorica and Francis1999) and lead to a release of dopamine in particular brain reward areas (e.g., the nucleus accumbens and ventral tegmental area), known to result in feelings of reward and reinforcement. Functional genetic variants that change OPRM1 gene regulation or function might affect the vulnerability of the brain reward system to bind opioids, and as a consequence, the reward experience might weaken or strengthen. This implies that the experience of reward is dependent on the OPRM1 genotype. Thus far, the opioid system has been quite comprehensively studied in addiction, though less so in adolescent samples, and to an even lesser extent in depression.
A well-studied variant within the OPRM1 gene is a single nucleotide polymorphism (SNP; rs1799971) in exon 1 (Lerman et al., Reference Lerman, Wileyto, Patterson, Rukstalis, Audrain-Mcgovern and Restine2004). A transition in this particular exon from A to G on nucleotide 118 (A118G) results in an amino acid change from asparagine to aspartic acid in the receptor protein. This amino acid change might affect gene function. With regard to addiction, the idea is that the more active allelic variant would have a stronger affinity to bind β-endorphins, which would lead to greater release of dopamine and, hence, more feelings of reward. However, studies focusing on the OPRM1 gene in relation to alcohol use specifically report contrasting results. Gelernter, Kranzler, and Cubells (Reference Gelernter, Kranzler and Cubells1999) found no difference in frequencies for the OPRM1 A118G genotype among Caucasian controls and alcohol-dependent samples. Sander et al. (Reference Sander, Gsheidel, Wendel, Samochowiec, Smolka and Rommelspacher1998) also concluded that the OPRM1 gene does not satisfactorily explain alcohol dependence. However, Town et al. (Reference Town, Abdullah, Crawford, Schinka, Ordorica and Francis1999) and Schinka et al. (Reference Schinka, Town, Abdullah, Crawford, Ordorica and Francis2002) found that the OPRM1 A allele was significantly associated with increased risk for alcohol dependency, whereas other studies have indicated that individuals with the OPRM1 G allele reported higher subjective feelings of intoxication, stimulation, sedation, and positive mood across rising levels of blood alcohol concentration, as compared to those with the A allele (e.g., Ray & Hutchison, Reference Ray and Hutchison2004; Ray et al., Reference Ray, Miranda, Tidey, McGeary, MacKillop and Gwaltney2010). Five studies so far have examined the OPRM1 gene in relation to alcohol use among adolescents (Kleinjan, Poelen, Engels, & Verhagen, Reference Kleinjan, Poelen, Engels and Verhagen2012; Miranda et al., Reference Miranda, Ray, Justus, Meyerson, Knopik and McGeary2010; Pieters et al., Reference Pieters, van der Vorst, Burk, Schoenmakers, Van den Wildenberg and Smeets2011, Reference Pieters, Van Der Zwaluw, Van Der Vorst, Wiers, Smeets and Lambrichs2012; Van der Zwaluw, Otten, Kleinjan, & Engels, Reference Van der Zwaluw, Otten, Kleinjan and Engels2013). Miranda et al. found that adolescents with a G allele reported more alcohol-related problems, endorsed drinking to enhance positive effects more strongly, and were more likely to have a diagnosed alcohol abuse disorder than those that were homozygous for the A allele. Pieters et al. (Reference Pieters, van der Vorst, Burk, Schoenmakers, Van den Wildenberg and Smeets2011) showed that compared to AA homozygotes, G carriers showed a significant positive relation between having an attention bias for environmental alcohol-related cues as measured by an implicit association task and the frequency and quantity of alcohol consumed. Pieters et al. (Reference Pieters, Van Der Zwaluw, Van Der Vorst, Wiers, Smeets and Lambrichs2012) and Van der Zwaluw et al. (Reference Van der Zwaluw, Otten, Kleinjan and Engels2013) found that compared to OPRM1 AA carriers, in OPRM1 G carriers alcohol-specific parenting strategies are more robustly associated with alcohol consumption. Finally, Kleinjan et al. (Reference Kleinjan, Poelen, Engels and Verhagen2012) tested the increase in alcohol use in 428 adolescents aged 13 to 15 years over a 4-year time period with annual measurements. They did not find an effect of the OPRM1 gene on the development of alcohol use. In sum, the role of the OPRM1 gene in the development of problematic alcohol use is not clear, which is partly due to the differences in populations tested and the alcohol outcomes (phenotypes) utilized in these studies. In addition, the expressions of genes in relation to alcohol use may differ across the life course. Because previous prospective studies indicated that early initiation of alcohol use (<14 years of age) is associated with heavier alcohol use throughout adolescence and emerging adulthood (King & Chassin, Reference King and Chassin2007) and recent twin research indicated that the variance in early initiation of alcohol use (<15 years) is mainly explained by genetic factors (83% explained variance in males and 70% in females; Poelen et al., Reference Poelen, Derks, Engels, Van Leeuwe, Scholte and Willemsen2008), we adopted a developmentally appropriate and age-specific longitudinal model that focuses on exactly this phase in early adolescence. By means of five measurements with intervals of 4 months, we investigated whether the OPRM1 gene is associated with levels of alcohol use and with concurrent change at the onset of alcohol use.
A link between the opioid system and depression has also been established. Kennedy, Koeppe, Young, and Zubieta (Reference Kennedy, Koeppe, Young and Zubieta2006) concluded that depressed patients have altered mu-opioid receptor availability compared to controls and show antidepressant effects of opioid peptides. Next to this, it has been shown that the mu-opioid system is involved in emotion regulation and hypothalamic–pituitary–adrenocortical (HPA) axis activity and stress responses (Kalin, Shelton, & Barksdale, Reference Kalin, Shelton and Barksdale1988; Zubieta et al., Reference Zubieta, Ketter, Bueller, Xu, Kilbourn and Young2003), both of which are involved in depression symptomatology. It has, for example, been found that cortisol function, which is a reliable indicator of the HPA axis, was elevated in 40%–60% of adults diagnosed with major depressive disorder (Parker, Schatzberg, & Lyons, Reference Parker, Schatzberg and Lyons2003).
Alterations in HPA axis functioning have also been consistently described in depressed adolescents compared to their control counterparts (see review and meta-analysis by Guerry & Hastings, Reference Guerry and Hastings2011; Lopez-Duran, Kovacs, & George, Reference Lopez-Duran, Kovacs and George2009). It was found that at-risk youth with dysphoria displayed hyperreactivity to psychosocial challenges, whereas controls revealed an appropriate stress response on these tasks (Hankin, Badanes, Abela, & Watamura, Reference Hankin, Badanes, Abela and Watamura2010).
The OPRM1 A118G variant was previously found to be related to HPA axis activity (Chong et al., Reference Chong, Oswald, Yang, Uhart, Lin and Wand2006), and Kertes et al. (Reference Kertes, Kalsi, Prescott, Kuo, Patterson and Walsh2011) found that the A118G variant and three other markers within the OPRM1 gene reached nominal significance with regard to lifetime depressive symptom scores in alcohol-dependent adults. It is therefore concluded that genetic variation within the OPRM1 A118G genotype could play a role at the interface of disturbed HPA axis activity, depression, and addiction. However, no studies have investigated the OPRM1 gene in relation to depressive symptoms among adolescents.
Alcohol use and depressive feelings both develop during early adolescence and seem to share a genetic liability, possibly in the form of OPRM1. Even though single gene effects are often modest, there is evidence that the OPRM1 polymorphism is associated with both internalizing and alcohol phenotypes. In most studies examining the OPRM1 gene, either alcohol use or depressive feelings have been investigated, and primarily in adult samples. The first aim of this study is to investigate the baseline levels (intercept) and the co-development of alcohol use and depressive feelings over time (slope), using a longitudinal five-wave design covering 2 years. It is hypothesized that alcohol use in this age group will be associated with the development of depressive feelings over time. Furthermore, OPRM1 is hypothesized to be part of the underlying mechanisms of the joint development of alcohol use and depressive feelings. Despite inconsistent findings in previous literature, we expected that the link between alcohol use and depressive feelings is partly conditional on the OPRM1 gene. Finally, because previous studies found differential associations between depression and alcohol use for boys and girls (Crum, Storr, Ialongo, & Anthony, Reference Crum, Storr, Ialongo and Anthony2008; Fleming, Mason, Mazza, Abbott, & Catalano, Reference Fleming, Mason, Mazza, Abbott and Catalano2008; Marmorstein, Reference Marmorstein2009; Needham, Reference Needham2007), interactions between sex and the OPRM1 genotype in relation to alcohol use and depression will be examined as well.
Method
Procedure and participants
A total of 804 adolescents (51.50% female) were recruited from 14 schools in the Netherlands, with an average age of 13.09 years (range = 11.59–15.60, SD = 0.48) at Time 1 (T1). Across the five waves (T1–T5), 783 (97.4), 736 (91.7%), 696 (86.6%), 684 (85.1%), and 636 (79.1%) adolescents participated, respectively. Because the current study focuses on a specific developmental period (i.e., 12–13 years old), adolescents younger than 12 and older than 14 were omitted from the analyses, leaving a sample of 739 adolescents (mean age = 13.03, SD = 0.39). Most of the adolescents were of Dutch origin (97.3%). The participants were in the first grade of secondary school at T1, and 11.6% of the participating adolescents were in university preparatory training, 30.2% in senior general secondary education, and 57.9% in preparatory vocational training and junior general secondary training at T1. At T1, saliva samples were collected for DNA extraction (Oragene, DNA Genotek Inc.). Active, informed consent for gene analysis was obtained from the adolescents as well as their parents. During each wave, the participants filled out an online or paper-and-pencil questionnaire during school hours. A total of 12.9% of the sample completed a paper questionnaire because their school did not have the necessary facilities to allow for the online completion of questionnaires. No differences were found between completers of online or paper questionnaires on demographic variables or on alcohol use or depressive feelings on any of the five waves. Students were explicitly instructed that all questions were about their regular patterns and not exceptional situations such as holidays, unless otherwise stated. The time between each of the five waves was approximately 4 months. The research design for this study was evaluated and ethically approved by an independent medical ethical committee (METiGG, Utrecht, The Netherlands).
We conducted power analyses using the software Quanto (Gauderman & Morrison, Reference Gauderman and Morrison2006). For the power calculation we applied the gene-only design option for continuous outcomes with independent individuals. To detect a small effect with an R 2 of .01 to .015, with 80% power (α = 0.05), the sample size required is between 518 and 781. With our sample size of 739 adolescents, we should be able to detect a small effect size of OPRM1.
Materials
Genotyping
The OPRM1 118A > G polymorphism (rs1799971) was genotyped using Taqman analysis (Taqman assay: C_8950074_1; reporter 1: VIC-A-allel, forward assay; Applied Biosystems, Nieuwekerk a/d IJssel, The Netherlands). The probe of the A allele was labeled with VIC and the probe of the G allele was labeled with Fam (see Oroszi et al., Reference Oroszi, Anton, O'Malley, Swift, Pettinati and Couper2009). Genotyping was carried out in a volume of 10 µl containing 10 ng of genomic DNA, 5 µl of Taqman Mastermix (2X, Applied Biosystems), 0.125 µl of the Taqman assay, and 3.875 µl 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). Generally, all genotyping assays have been validated earlier, and 5% duplicates and blanks were taken along as quality controls during genotyping. Of the 739 adolescents included in our analyses, 11 could not be genotyped with regard to the OPRM1 genotype (1.5%).
Questionnaires
Alcohol use
During each wave, adolescents' alcohol use was assessed with a single item asking about how often they had consumed alcohol during the past 4 weeks. They responded on a 6-point scale, ranging from 1 (have not been drinking) to 6 (every day; Engels & Knibbe, Reference Engels and Knibbe2000). Because of the low frequencies in the last two categories (5 or 6 days a week and every day), these categories were summed into one.
Depressive feelings
Depressive feelings were assessed at each wave with a translated version of the Center for Epidemiologic Studies Depression Scale (Radloff, Reference Radloff1977). The scale consists of 20 items about how often adolescents felt a certain way or engaged in a certain behavior during the past week. Examples of items are “I felt depressed” and “I enjoyed life.” The adolescents were asked to rate the items on a 4-point scale, ranging from 1 (rarely or never) to 4 (often or always). Cronbach αs for this scale were 0.79 (T1), 0.84 (T2), 0.84 (T3), 0.85 (T4), and 0.82 (T5). An average score for depressive feelings was computed at each wave.
Alcohol use of parents
Parental alcohol use at T1 was assessed with a single item per parent in which adolescents were asked how often their father or mother had consumed alcohol during the past 4 weeks. They responded on a 6-point scale, ranging from 1 (have not been drinking) to 6 (every day; Engels & Knibbe, Reference Engels and Knibbe2000).
Alcohol use of friends
Alcohol use of friends was assessed by asking, “How many of your friends drink alcohol?” This item could be scored on a 5-point scale, ranging from 1 (none of them) to 5 (all of them).
Attrition analyses
Attrition analyses were conducted in order to examine whether adolescents who remained participants in the fifth wave of the study (n = 636; 79.1%) differed from adolescents who dropped out (n = 168; 20.9%). The t tests showed no significant differences (p > .05) in age, educational level, ethnicity, and OPRM1 genotype at T1. Participants lost to follow-up were more likely to be boys, χ2 (1, N = 804) = 6.34; p < .05, and to show a higher frequency of alcohol use, t (725) = –3.86, p < .001. No differences were found for depressive feelings, t (771) = –1.67, p = .09. All participants are included in the analyses regardless of missing data via the full information maximum likelihood estimation (see also the following section).
Strategy of analyses
The relation between the independent variables and the co-development of alcohol use and depressive feelings was examined with a latent growth curve model (cf. Kleinjan et al., Reference Kleinjan, Poelen, Engels and Verhagen2012). First, we assessed the single growth curves of alcohol use and depressive feelings from T1 to T5 by estimating the initial level (intercept) and the rate of change over time (slope) for both alcohol use and depressive feelings. To correct for the skewed distribution of the alcohol frequency, we conducted our analyses using the maximum likelihood robust estimator. Second, we assessed a dual growth model in order to assess the simultaneous change in alcohol use and depressive feelings, using Mplus (Muthén & Muthén, Reference Muthén and Muthén1998–2007). OPRM1 was included as a predictor variable to assess whether it was predictive of initial values or growth over time of alcohol use and depression. In this approach, effects of the predictor variables on the intercepts and slopes of alcohol use are controlled for the development of depression and vice versa. In a final step, we examined whether the associations between initial values and growth effects were different for OPRM1 A homozygotes compared to OPRM1 G-allele carriers. These analyses are based on χ2 difference testing between two identical models, where one model is subjected to constraints of one of the parameters, in this case OPRM1, whereas the other model is not. This was done by constraining all paths of interest (i.e., the six associations between the intercepts and the slopes) to be equal and testing whether the model fit (△χ2) was significantly better for the model in which all paths were allowed to differ between OPRM1 A homozygotes and OPRM1 G-allele carriers, compared to the model in which all paths were constrained to be equal. To examine which specific paths between the intercepts and slopes of alcohol use and depressive feelings differ for A homozygotes and G-allele carriers, we constrained each path separately while all other paths were unconstrained and compared this model to the model in which the path of interest, as well as all other paths, were unconstrained. Because the χ2 value cannot be used for standard χ2 difference testing when using the maximum likelihood robust estimator, the Satorra–Bentler scaled χ2 difference test was used (Satorra & Bentler, Reference Satorra and Bentler2001). All models were controlled for the variables sex, age, educational level, ethnicity, and alcohol use of parents and friends. Possible interactions with sex and OPRM1 were additionally examined. Full information maximum likelihood estimation was applied to make use of all available data. The model fit was investigated by the following global fit indices: χ2, the comparative fit index (CFI; good fit > 0.90), and the root mean square error of approximation (RMSEA; good fit < 0.08; Hu & Bentler, Reference Hu and Bentler1999).
Results
Descriptives
The means and standard deviations for the four measures of alcohol use and depressive feelings are presented in Table 1. At T1 56.6% of adolescents indicated that they had tried alcohol at least once, this was 53.4% at T2, 57.8% at T3, 61.2% at T4, and 57.2% at T5. Alcohol use in the past 4 weeks was reported by 14.3% at T1, 14.2% at T2, 17.7% at T3, 27.0% at T4, and 24.1% at T5. Descriptive findings on the OPRM1 SNP revealed that 79.5% of the adolescents had the AA genotype, 19.1% the AG genotype, and 0.8% the GG genotype. No deviations from Hardy–Weinberg equilibrium were detected (p = .18). To maximize the power of the analyses, OPRM1 genotype was dummy coded into 1 (AA) and 2 (AG and GG).
Table 1. Means (standard deviations) of model variables for the total sample and separate for OPRM1 AA carriers and AG/GG carriers
Note: Alcohol use: min = 1, max = 5. Depressive feelings: min = 1, max = 4.
Correlations between model variables
Correlations between the model variables are presented in Table 2. These findings show some significant positive correlations between the measures of alcohol use and depressive feelings. There was a positive correlation between OPRM1 and alcohol use at T2 and a negative correlation between OPRM1 and depression at T5.
Table 2. Pearson correlations between the model variables
Note: Correlations with p < .05 are in italic; correlations with p < .01 are in bold.
Model findings
First, we tested the latent growth model for alcohol use and depressive feelings separately and without predictors. The model for alcohol use showed a good fit to the data, χ2 (df = 9, p < .001) = 35.63, CFI = 0.90, RMSEA = 0.06. The means of the intercept and slope were both significant (respectively, 2.96, p < .001; variance = 0.16, p < .001; and 0.38, p < .001; variance = 0.01, p < .01), suggesting that the participants scored greater than zero on alcohol use at baseline, that alcohol generally increased over time, and that participants differed around the means. The association between the intercept and the slope was significant (β = –0.41, p < .001). This suggests that higher initial levels of alcohol use are related to less growth in alcohol use over time. Quadratic trends were also examined, but they were not significant.
The model for depressive feelings also showed a good fit to the data, χ2 (df = 10, p < .001) = 29.78, CFI = 0.97, RMSEA = 0.05. The means of both intercept and slope were significant (respectively, 5.26, p < .001; variance = 0.08, p < .001 and –0.17, p < .001; variance = 0.004, p < .001), suggesting that the participants scored greater than zero on depressive feelings at baseline, that depressive feelings generally decreased over time, and that participants differed around the means. The association between the intercept and the slope was significant (β = –0.28, p < .01). This suggests that higher initial levels of depressive feelings are associated with less growth in depressive feelings over time. It is important to note here that the negative slope does not mean that there is no individual growth in depression. In the latent growth curve approach, it is not assumed that all participants start at the same level of depression at baseline and have the same rate of change over time; instead, individual growth is examined for each participant. Finally, the possibility of a quadratic trend was examined, but this was not significant.
Second, we tested the dual growth model of alcohol use and depressive feelings. This model fits the data well, χ2 (df = 40, p < .001) = 97.62, CFI = 0.95, RMSEA = 0.04. The association between the initial values of alcohol use and depressive feelings was significant (β = 0.17, p < .01). This means that higher baseline levels of alcohol use were associated with higher baseline levels of depressive feelings. The association between the initial values of alcohol use and the change in depressive feelings over time was not significant (β = –0.12, p = .14). In addition, the relationship between the initial values of depressive feelings and change in alcohol use over time was not significant (β = –0.06, p = .48). The slopes of alcohol use and depressive feelings were positively related (β = 0.30, p < .05), indicating that increases of depression co-occur with increases in alcohol use. Table 3 provides a further clarification of this association. For this purpose, individual scores on the intercepts and slopes of alcohol and depression were transferred from Mplus to SPSS. For depressive feelings, the sample was divided into a group containing individuals for whom the slope was negative (decliners: 60%) and a group for whom the slope was positive (increasers: 40%). For alcohol use, we used a median split to distinguish between those who showed no or moderate increases (55.6%) and those who showed a high increase (44.4%). A median split was chosen because only 12% of our sample showed a decrease in alcohol use over time. Table 3 shows that for the total sample, 77.6% of all individuals who increased in depression also showed a high increase in alcohol over time. In contrast, 77.5% of all individuals who declined in depression over time showed no or a moderate increase in alcohol use over time.
Table 3. Increase in alcohol use for negative and positive depression slopes
Note: For depressive feelings the sample was divided into a group containing individuals for whom the slope was negative and a group for whom the slope was positive. For alcohol use a median split was used to distinguish between low and high increase in use over time.
Third, the predictors were added to the dual growth model of alcohol use and depressive feelings. The model including the genetic and control variables fits the data well, χ2 (88) = 154.28, p < .001, CFI = 0.96, RMSEA = 0.03. The control variables sex and perceived alcohol use of friends predicted initial levels of alcohol use (see Table 4). Initial values were higher for males and for participants whose friends drink more frequently. Higher initial levels of depressive feelings were predicted by being female, having a lower educational level, being of Dutch descent, and having more drinking friends. Girls also showed a higher increase in depression over time than boys. When controlled for sex, age, education level, ethnicity, and perceived drinking behavior of father, mother, and friends, the OPRM1 genotype showed a main effect on both the slopes of alcohol use (R 2 change = .020) and depressive feelings (R 2 change = .022). The OPRM1 G allele was associated with less increase in both alcohol use and depressive feelings over time. No effect of OPRM1 on the initial values was found. Finally, no interaction effect of sex with OPRM1 was found on the intercepts or slopes (not shown in Table 4). Figure 1 shows the growth curves for alcohol use and depressive feelings separately for OPRM1 AA carriers and G carriers.
Figure 1. Dual growth models of alcohol use and depressive feelings separated for adolescents who are AA carriers and AG/GG carriers (standardized estimates). *p < .05, **p < .01, ***p < .001.
Table 4. Standardized estimates for control variables and OPRM1 on the intercepts and slopes of alcohol use and depressive feelings in the dual growth model
Note: Sex: 0 = male and 1 = female; ethnicity: 0 = Dutch and 1 = not Dutch.
*p < .05. **p < .01. ***p < .001.
Fourth and finally, we conducted a multigroup analysis to test whether the effects that we found in the total sample differed for homozygous A-allele carriers compared to G-allele carriers. The Satorra–Bentler χ2 difference test indicated that the model differed for OPRM1 A homozygotes and OPRM1 G carriers, χ2 (6) = 10.52, p < .01 (Satorra & Bentler, Reference Satorra and Bentler2001). The fit of the model, where the paths were allowed to differ for the OPRM1 genotypes, was good, χ2 (165) = 154.28, p < .001, CFI = 0.96, RMSEA = 0.04. Figure 2 depicts the standardized estimates of the dual growth model, separately for homozygous A carriers and G-allele carriers. The most important finding is reflected in the link between the slopes of alcohol use and depressive feelings. As depicted, the slopes were positively associated, but only in the AA carriers (β = 0.41, p < .05) and not in the G carriers (β = 0.09, p = .65). In the G carriers, the initial values of alcohol use were negatively related to the slopes, and the same pattern was found for depressive feelings. In AA carriers the intercept of alcohol use was not related to the slope, and the intercept and slope of depression were also not related. The Satorra–Bentler χ2 difference test established support for this path difference, χ2 (1) = 10.52, p < .001. Finally, no interaction effects of sex with OPRM1 were found on the relations between the intercepts or slopes.
Figure 2. Plotted dual growth curves of alcohol use and depressive feelings separated for adolescents who are AA carriers and AG/GG carriers, controlled for sex, age, education level, and parental and peer alcohol use.
To more closely examine the association between alcohol use and depressive symptoms in AA carriers over time, we conducted a post hoc piecewise growth model. In a piecewise growth model, different phases of development are captured by more than one slope growth factor. We divided the five measurements into two developmental phases. Phase 1 includes the first three measurements (T1–T3), Phase 2 includes in the last three measures (T3–T5). The piecewise growth model showed a good fit, χ2 (112) = 145.48, p < .05, CFI = 0.98, RMSEA = 0.03. We found that in AA carriers, the slope of depressive feelings in developmental Phase 1 was marginally positively associated with the slope of alcohol use in Phase 2 (β = 0.21, p = .09). The slope of alcohol use in Phase 1 was not associated with the slope of depressive feelings in Phase 1 or Phase 2 (β = 0.16, p = .32 and β = 0.07, p = .63). During Phase 2, the slopes of depressive feelings and alcohol use are marginally positively related (β = 0.22, p = .07). Within GG carriers, none of the slopes in the different phases were associated.
Discussion
The present study aimed at examining how the development of alcohol use and depressive feelings is related among a sample of early adolescents in a developmentally sensitive period, and how the OPRM1 genetic variant affects this co-development. We found that higher levels of alcohol use at baseline were associated with higher levels of depressive feelings at baseline and that a greater increase in alcohol use was associated with a greater increase in depressive feelings. This is in accord with previous studies that showed alcohol use and depressive feelings to co-occur (Clark & Bukstein, Reference Clark and Bukstein1998; Verdurmen et al., Reference Verdurmen, Monshouwer, van Dorsselaer, ter Bogt and Vollebergh2005).
Homozygous OPRM1 118A carriers showed the strongest increases in alcohol use and the least decrease in depressive feelings over time, and moreover, the increase in alcohol use and depressive feelings was associated only in this group. In AG and GG carriers no associations were found between early alcohol use and depressive feelings. It has been suggested that the molecular mechanism for the 118A allele in association with alcoholism involves hyposensitivity of the endogenous mu-opioid receptor system and that this leads to increased consumption of alcohol in order to compensate for this intrinsic opioid response deficit (Town et al., Reference Town, Abdullah, Crawford, Schinka, Ordorica and Francis1999). This is in accord with Bond et al. (Reference Bond, LaForge, Tian, Melia, Zhang and Borg1998), who demonstrated that the 118G receptor isoform binds endogenous β-endorphin approximately three times more tightly than 118A, giving rise to heightened feelings of reward. In the present study, we did not study alcohol dependency but frequency of alcohol use in early adolescence (early onset). It seems that at ages 12–14, adolescents who are homozygous for the 118A allele increase more in frequent alcohol use in a 2-year time window than adolescents with the 118G allele. A possible explanation for this finding may be that the lack of an intrinsic opioid response to alcohol in young AA carriers may give them less reason to hold back on drinking once started because physiological reactions might be less prominent (e.g., unusual body sensations and the feeling of losing control). In G carriers alcohol use seemed to be relatively stable over time, and the development of alcohol use and depressive feelings seemed to be largely independent of each other. In this group, the initial scores on both phenotypes were reversely correlated with the slope, meaning that higher levels at baseline were associated with less increase of the respective phenotypes over time. Adolescent G carriers, who might experience a stronger response to alcohol use because of the stronger mu-opioid binding potential, seem to start out at somewhat higher levels of alcohol use, but these levels change relatively little over time and possible changes seem independent of depressive feelings.
Further, because young adolescents are novice drinkers, the effect of the activity and function of the OPRM1 A118G SNP may be different compared to later stages of adolescence in which drinking patterns become more entrenched. When adolescents get older, drinking patterns become more established and drinking motives may change. Young adolescents who have little alcohol experience will probably drink because of social reasons. When more alcohol has been consumed, other drinking motives such as coping and enhancement motives will play a larger role (Kuntsche Knibbe, Gmel, & Engels, Reference Kuntsche, Knibbe, Gmel and Engels2006). Coping motives were found to be particularly associated with problematic alcohol use, such as heavy episodic drinking and alcohol-related problems (e.g., Kuntsche, Knibbe, Gmel, & Engels, Reference Kuntsche, Knibbe, Gmel and Engels2005). Therefore, because of the stronger binding potential of the receptors in G-allele carriers, adolescents with the 118G allele might be more at risk for excessive alcohol use and alcohol-related problems later in adolescence because the G allele generates more enhancement effects of alcohol. This is in line with Miranda et al. (Reference Miranda, Ray, Justus, Meyerson, Knopik and McGeary2010), who found in a sample of older adolescents that those with a G allele reported more drinking to enhance positive affect and were more likely to report alcohol-related problems than those who were homozygous for the A allele. In this later stage of adolescence, sex differences in the relations between alcohol use and depression might also become more pronounced, seeing that boys tend to drink more excessively and girls tend to increase more in depressive feeling during adolescence (e.g., Costello et al., Reference Costello, Mustillo, Erkanli, Keeler and Angold2003).
The activity and function of the OPRM1 A118G SNP in depression in humans has yet to be largely established (Garroick et al., Reference Garriock, Tanowitz, Kraft, Dang, Peters and Jenkins2010). Up until now, the OPRM1 A118G has not been investigated in relation to depressive symptoms in normative adult or adolescent populations, though a few pharmacogenetic studies on the OPRM1 gene in relation to antidepressant response have been conducted (Garroick et al., Reference Garriock, Tanowitz, Kraft, Dang, Peters and Jenkins2010; Perlis, Fijal, Dharia, Heinloth, & Houston, Reference Perlis, Fijal, Dharia, Heinloth and Houston2010). However, these studies did not yield significant associations for the A118G variant and treatment response. Therefore, assumptions on gene-based processes in relation to depression are highly speculative. Despite this, the mu-opioid system is likely to be involved in depression because previous studies have found that the binding potential of the mu-opioid receptor was significantly lower in women with major depressive disorder relative to nondepressed women (Kennedy et al., Reference Kennedy, Koeppe, Young and Zubieta2006). Further, it was shown that depressed women who did not respond to antidepressant treatment exhibited lower binding potential for the mu-opioid receptor than depressed women who responded positively to psychopharmacology. The link between being a homozygous OPRM1 A-allele carrier and increases in depressive symptoms may thus be partly explained by the lower binding potential that characterizes the OPRM1 +118A genotype.
Our results form an addition to the literature on the co-occurrence of alcohol use and depressive feelings in several ways. Most studies that focused on the link between alcohol use and depressive symptoms concentrated on more severe forms of psychopathology, such as clinical depression, mostly in individuals who present excessive alcohol use patterns. Our findings provide evidence that in a normative group of early adolescents, increases in depressive feelings, even at relatively low levels, are associated with increases in alcohol use over time but only in a subgroup with a specific genotype, namely, AA carriers of the OPRM1 +118 gene. When probing the association between alcohol use and depressive feelings in AA carriers by looking at the associations between the slopes in two different developmental phases, we found a marginally significant positive association between the slope of depressive feelings for the first three measures and the slope of alcohol use for the last three measures. This finding tentatively indicates that the early increase in depressive feelings might influence later increases in alcohol use in AA carriers. It is possible that this association reflects the vulnerability to a subtle process that eventually leads to and accelerates a reinforcing interplay between depressive symptoms and alcohol use. Because adolescents who had their first drink between the ages of 11 to 14 years are more likely to develop an alcohol disorder later in life than adolescents who started drinking at a later age (King & Chassin, Reference King and Chassin2007), it may be that comorbidity of alcohol dependence and depression develops at an accelerated pace in AA carriers. Neurobiological studies have shown that “liking” processes are associated with opioid neurotransmission in the brain (Berridge, Reference Berridge2003). Stimulation of opioid receptors due to alcohol consumption results in dopaminergic firing in “reward areas.” Whereas dopaminergic effects are thought to be especially important in the later stages of alcohol use, after sensitization has taken place, opioid neurotransmission may be more associated with the early stages of alcohol use. Particularly in AA carriers, who are thought to have less strong binding potential and therefore a more hyposensitive endogenous mu-opioid receptor system, low levels of alcohol use may already produce some form of compensation for the intrinsic opioid response deficit, adding to the risk of developing early signs of withdrawal. These early signs of withdrawal may trigger the experience of unpleasant feelings, which may ultimately be reflected in both increased alcohol use and depressive symptoms over time. In addition, the risk of unpleasant feelings might already be higher in AA carriers, seeing the previous established links between depression and lower binding potential of the mu-opioid receptor (Kennedy et al., Reference Kennedy, Koeppe, Young and Zubieta2006). To avoid these perceived unpleasant consequences, AA carriers may become increasingly motivated to use alcohol more regularly, leading to more severe levels of depressive feelings over time (Robinson & Berridge, Reference Robinson and Berridge2000).
When interpreting the findings of this study, some caveats should be kept in mind. First, we tested one SNP, whereas multiple loci are likely to be involved in the path from initial alcohol use to more problematic patterns of use (Van der Zwaluw & Engels, Reference Van der Zwaluw and Engels2009). Even though we found an effect of a single locus, that does not answer the question of whether this particular SNP is the functional variant responsible for the effect. Because of nonrandom association between alleles (i.e., linkage disequilibrium), genotyping several SNPs within the gene and adjacent genes can provide important insights into other variants that are associated with alcohol use and depression. In addition, it should be noted that our results should be interpreted in a broader framework. Previous studies already identified multiple shared underlying factors for the co-occurrence of alcohol use and depressive symptoms besides genetic factors (i.e., low academic achievement, family dysfunction, externalizing traits, and poor health status; Diaz et al., Reference Diaz, Simantov and Rickert2002; Gau et al., Reference Gau, Chong, Yang, Yen, Liang and Cheng2007; Lewinsohn et al., Reference Lewinsohn, Roberts, Seeley, Rohde, Gotlib and Hops1994; Miller-Johnson et al., Reference Miller-Johnson, Lochman, Coie, Terry and Hyman1998; Pardini et al., Reference Pardini, White and Stouthamer-Loeber2007). A growing body of evidence suggests a common neurobiological basis underlying both addiction and certain psychiatric disorders (Brady & Sinha, Reference Brady and Sinha2005). It is very well possible that multilevel mechanisms might be in play that link increases in alcohol use with depressive feelings for the AA carriers. Future research is needed to discern more comprehensive risk profiles for the co-occurrence of alcohol use and depressive feelings by taking to account a combination of genetic, (neuro)biological, and psychosocial features known to be associated with both phenotypes.
Second, we only used self-reports of alcohol use and depressive feelings of adolescents, and adolescents also reported on the alcohol use of their parents and friends themselves. However, it has been shown that adolescents can validly report their alcohol use (Brener, Billy, & Grady, Reference Brener, Billy and Grady2003) and that of parents (Engels, Van der Vorst, Deković, & Meeus, Reference Engels, Van der Vorst, Deković and Meeus2007) and friends (Belendiuk, Molina, & Donovan, Reference Belendiuk, Molina and Donovan2010) as well as depressive feelings (Kazdin, Reference Kazdin, Reynolds and Johnston1994). It should be kept in mind, though, that self-reports of one's own and other's drinking patterns may be susceptible to over- or underreporting.
Third, we chose to focus on the frequency of alcohol use in the past month and did not include measures on the quantity of alcohol. We did assess quantity of alcohol use in the past week, but seeing that in early adolescents alcohol use is incidental, our 1-week time frame to measure the quantity of alcohol use did not show much variance, and the use of this measure might have resulted in an underestimation of alcohol use.
Fourth, no information was available on depressive feelings of parents and friends, so we could not control for these variables in our analyses. However, previous studies indicated that the relation between parental and child depression in adolescence seems to be moderate (e.g., Van Roekel, Engels, Verhagen, Goossens, & Scholte, Reference Van Roekel, Engels, Verhagen, Goossens and Scholte2011).
Fifth, we used a community sample with adolescents not previously diagnosed with alcohol (mis)use or (sub)clinical depression. It is possible that findings would be different for adolescents who had already initiated alcohol use or had strongly elevated levels of alcohol use and depressive feelings, or for clinical samples.
Our results may also provide some implications for practice. The finding of co-development of alcohol with depressive symptoms can be used to inform and further specify prevention efforts. Recent studies investigated the effects of tailor-made interventions for the at-risk personality populations (Conrod, Castellanos, & Mackie, Reference Conrod, Castellanos and Mackie2008; Conrod, Castellanos-Ryan, & Strang, Reference Conrod, Castellanos-Ryan and Strang2010; Conrod, Stewart, Comeau, & Maclean, Reference Conrod, Stewart, Comeau and Maclean2006). In these studies, tailor-made interventions were provided for the diverse risk groups based on prescreening with the Substance Use Risk-Profile Scale. Depression-prone adolescents, for instance, were provided an intervention program aimed at their specific personality traits or skill deficits (e.g., Conrod et al., Reference Conrod, Stewart, Comeau and Maclean2006). This program targets at-risk groups by selecting adolescents who already had their first drink of alcohol and scored one standard deviation above the school mean on depressive feelings, anxiety sensitivity, sensation seeking, or impulsivity. Studies showed that participants in the intervention conditions show less substance use behaviors compared to a control group (Conrod et al., Reference Conrod, Stewart, Comeau and Maclean2006, Reference Conrod, Castellanos and Mackie2008, Reference Conrod, Castellanos-Ryan and Strang2010). Knowledge about the OPRM1 genetic marker could eventually lead to an even more personalized approach of such an intervention, by selecting those adolescents who are known to be at risk for the co-development of depressive feelings and alcohol use.
To conclude, the results of our longitudinal and age-specific study provide evidence for the notion that the development of alcohol use and depressive mood are already interrelated in early adolescence. In addition, our results suggest that the OPRM1 genotype seems to be a shared underlying factor in the development of both alcohol use and depressive symptoms during this period.
