Predictors of Initiation of Alcohol Use Among US Adolescents: Title and subTitle BreakFindings From a Prospective Cohort Study FREE

Despite recent signs of a decline in prevalence,¹ alcohol use and abuse continue to be pervasive among US adolescents. Approximately 74% of high school students report having had a whole drink of alcohol, 43% report drinking within the past 30 days, and 26% report binge drinking within the past month.² These numbers are disturbing, as alcohol use is associated with the 4 leading causes of death among young people: motor vehicle crashes, unintentional injury, suicide, and homicide.^{3 - 7} In addition, alcohol use increases the risk of violence victimization and perpetration among adolescents.^{8 - 10} Beyond these immediate threats, early alcohol use can lead to the development of major depressive disorders, substance use disorders, and alcohol dependence in adulthood.^{11 - 14} Moreover, adolescents who initiate use at younger ages may be at greater risk for many of these alcohol-related problems.^{12 - 13 ,15 - 20} Identifying precursors of adolescent alcohol use can guide the design of innovative strategies to prevent or postpone initiation.

Previous studies^{6 ,21 - 22} have identified older age, physical maturity, white race, and residing in the northeast region of the United States as individual characteristics that correlate with adolescent alcohol use. Prior studies^{23 - 25} have also highlighted the inverse association between alcohol initiation and use and higher levels of scholastic achievement, school engagement, and global self-esteem; the roles of social and athletic self-esteem are not known.

Family factors such as living in a single-parent household, parental alcohol use, and sibling alcohol use have also been found to be associated with alcohol use among adolescents.^{25 - 29} Moreover, 2 recent cross-sectional studies^{30 - 31} found that eating family dinner at home has an inverse association with adolescent drinking.

Several cross-sectional studies^{32 - 35} have shown that alcohol advertising reaches young audiences and influences their alcohol behaviors. Collins et al³⁴ found that awareness of televised beer advertisements was associated with alcohol use among eighth-grade boys. Fleming et al³² found that increased exposure to alcohol advertising among 15- to 20-year-olds correlated with positive attitudes about alcohol, which in turn correlated with intention to drink. Using prospective data, Snyder et al³⁶ found that past-month exposure to alcohol advertising among 15- to 20-year-olds was associated with past-month drinking. In another longitudinal study, Ellickson et al³⁷ found that among ninth graders, exposure to magazine advertisements predicted alcohol use but exposure to television advertisements did not. Cross-sectional studies have shown a positive association between tobacco promotional items and teen smoking,^{38 - 39} and a similar relationship may exist between alcohol promotional items (APIs) and alcohol uptake. Cross-sectional analyses in a recent study⁴⁰ of 2406 middle school students found that students who owned APIs were 50% more likely than their peers to initiate alcohol use.

Power et al⁴¹ found that among adolescents, predictors of movement from abstinence to non-high-risk drinking were different from predictors of movement into high-risk and problem drinking. It is also plausible that predictors vary depending on stage of behavioral adoption. This information would be useful in targeting intervention strategies.

We prospectively examined predictors of alcohol initiation and binge drinking in the Growing Up Today Study (GUTS), a large, nationwide cohort of adolescents. We compared the factors influencing alcohol uptake among younger and older adolescents and examined predictors among adolescents at varying levels of behavioral progression using a framework based on the transtheoretical model by Prochaska and Velicer,⁴² which posits multiple stages of change, including precontemplation and contemplation, before initiating an action.

The GUTS participants were recruited in 1996 by identifying mothers from the ongoing Nurses' Health Study II who had children aged 9 to 14 years. Further details of the recruitment and establishment of the cohort are described in detail elsewhere.⁴³ The GUTS comprises 16 882 boys and girls. Follow-up questionnaires are mailed approximately annually. This study was approved by the human subjects committees at Harvard School of Public Health and Brigham and Women's Hospital, Boston, Massachusetts.

In 1998 and 1999, an expanded section on alcohol use was included in the questionnaire, allowing for a more detailed examination of these behaviors. Approximately 75% of the girls (n = 6807) and 64% of the boys (n = 5027) who completed the baseline questionnaire in 1996 completed follow-up questionnaires in both 1998 and 1999, yielding an overall response rate of 70%. We excluded 1257 boys and 1787 girls who had initiated alcohol use by 1998, as well as those who did not respond to the 2 primary alcohol questions in both 1998 and 1999 and therefore lacked sufficient data to determine alcohol initiation status. We further excluded participants who did not respond to questions pertaining to variables of interest including Tanner stage, self-esteem, cigarette smoking, family composition, frequency of family dinners, adult alcohol use in the home, sibling alcohol use, peer alcohol use, owning or being willing to use APIs, alcohol advertising, and attitudes toward alcohol. Rates of alcohol initiation were similar among those missing data on any individual variable and those who responded, and among all of those excluded (18%) vs all of those included (18%) in the study. The final sample included 3283 girls and 2228 boys aged 11 to 18 years in 1998.

The Tanner stage of pubic hair development was assessed by a series of 5 sex-specific sketches, with stage 1 being the least developed and stage 5 being fully developed.^{44 - 45} Race was dichotomized into white and nonwhite groups. We used a modified version of the Harter Self-perception Profile for Children⁴⁶ to measure self-esteem in 4 domains: social, athletic, global, and scholastic. Self-esteem scores ranged from 6 to 18 for each domain, with higher scores indicating higher levels of self-esteem. Because the 1998 question included only global and scholastic domains, we used 1997 values for social and athletic self-esteem in these analyses. Cigarette smoking was assessed with the question, “In the past year, have you smoked a cigarette, even a few puffs?”

Participants reported which adults they live with most of the time. Those who reported living with 2 parents (any combination of mother, father, stepmother, or stepfather) were compared with those who reported living with only 1 parent (mother or father). Participants who lived with other relatives or adults but not their mother or father were not included in the family composition variable so that a direct comparison between single- and 2-parent families could be made. To assess the frequency of family dinner at home, participants were asked, “How often do you sit down with other members of your family to eat dinner or supper?” Those who responded never or some days were compared with those who responded most days and with those who responded every day. Participants were asked whether any adults who live in their household drink alcohol and whether they have a sibling younger than 21 years who drinks alcohol.

Peer alcohol use was assessed with the question, “How many of your friends drink alcohol?” Participants who responded none were compared with those who responded 1 or a few and with those who responded most or all. Awareness of alcohol advertising was assessed with the question, “In the last month, have you talked with your friends about any advertisement or TV commercial involving alcohol?” Participants were asked whether they had ever bought or been given an API such as a hat, T-shirt, or bag with the name of an alcohol beverage on it and whether they would be willing to use such an item. Those who responded yes to either question were defined as owning or being willing to use APIs.

The Alcohol Expectancy Questionnaire–Adolescent version (AEQ-A) is a validated measure of propensity to initiate alcohol use.^{47 - 48} Shortened adaptations of this scale have been shown to be a valid indicator of the susceptibility of young people to drink alcohol.^{49 - 50} We used a shortened version of the AEQ-A to assess participants' attitudes and beliefs about alcohol. Participants indicated whether they agreed with each of 17 statements about alcohol. Total scores ranged from 0 to 17, with higher scores representing more favorable attitudes and beliefs about alcohol consumption.

The main outcome of interest was initiation of alcohol use. Participants were asked, “Have you ever tried drinking alcohol (beer, wine, or liquor), even a few sips?” Those who responded yes were further asked, “Have you ever had a whole ‘drink’ of alcohol? (One drink means a whole glass, can, or bottle of beer; a whole glass of wine; or a whole ‘mixed drink’ or shot of liquor.)” Those who responded no to the first or second question were classified as never whole drinkers, and those who responded yes to both questions were classified as initiators of alcohol use.

To assess binge drinking, participants who had consumed a whole drink of alcohol were further asked, “Over the past year, how many times did you drink 5 or more alcohol drinks over a few hours?” Those who reported consuming 5 or more drinks on at least 1 occasion were considered to have engaged in binge drinking.

Stage of alcohol uptake⁴² in 1998 was assessed among those who reported never trying alcohol. These participants were further asked, “Do you think you will try drinking alcohol in the next year?” Those who responded definitely not were defined as precontemplators. Those who responded maybe or probably were categorized as contemplators. Those who reported that they had tried alcohol but never consumed a whole drink were defined as experimenters.

All of the analyses were stratified by sex to detect differences in factors influencing alcohol initiation between boys and girls. The association between each potential predictor and 1999 alcohol initiation status was tested, and χ² tests for proportional distributions and t tests for mean values were calculated. Predictors that retained significance in univariate logistic regressions were included in multivariate logistic regressions. For factors that appeared significant among girls but not boys or vice versa, we tested for an interaction between the variable and sex.

We examined effect modification by age to compare the influence of various factors on alcohol uptake among younger (aged < 15 years) and older (aged ≥ 15 years) adolescents. Age 15 years was selected as the age cut point for these analyses, as it is the approximate age when adolescents enter high school. We also examined effect modification by stage of alcohol uptake in 1998. Because findings for contemplators and experimenters were similar in preliminary analyses, these 2 groups were combined and compared with precontemplators. For factors that showed statistical significance among one age group or among one stage of uptake but not the other, we tested for an interaction. The categorical variables of peer drinking and family dinner were transformed into continuous variables to test for interaction.

Logistic regressions were used to compute odds ratios (ORs) for binge drinking among those adolescents who initiated alcohol use in 1999. The 90 participants who initiated alcohol use but did not respond to the question on binge drinking were excluded from this subanalysis.

Overall, 611 girls (19%) and 384 boys (17%) who were never drinkers in 1998 had a first whole drink of alcohol in 1999 (Table 1).

In multivariate logistic regression models, age, Tanner stage, cigarette smoking, adult drinking in the home, sibling drinking, peer drinking, owning or being willing to use APIs, and higher AEQ-A score were predictive of alcohol initiation among both girls and boys (Table 2). Higher social self-esteem among girls and higher athletic self-esteem among boys were positively associated with alcohol initiation. Among girls, living in a single-parent household was positively associated with alcohol initiation, whereas eating family dinner at home every day was inversely associated. Because social and athletic self-esteem, single-parent households, and family dinner had an association with alcohol uptake for one sex but not the other, we tested for an interaction between sex and these variables. No statistically significant interactions were found (all P > .10). As neither race nor US geographic region were statistically significant variables in univariate models (data not shown), these covariates were not included in the multivariate models.

In logistic regression models stratified by age group, we found evidence of an interaction between age group and both sibling drinking (P < .001) and peer drinking (P = .003) among girls, with stronger associations seen among those younger than 15 years compared with those aged 15 years and older (sibling drinking: OR, 2.68, 95% confidence interval (CI), 1.78-4.01 vs OR, 0.75, 95% CI, 0.42-1.34, respectively; peer drinking among 1 or a few friends: OR, 2.16, 95% CI, 1.63-2.86 vs OR, 1.30, 95% CI, 0.83-2.06, respectively; peer drinking among most or all friends: OR, 3.37, 95% CI, 1.45-7.83 vs OR, 1.32, 95% CI, 0.66-2.63, respectively). Among boys, we found an interaction between age group and Tanner stage (P = .006), with stronger effects seen among younger teens than among older teens (OR, 1.38, 95% CI, 1.15-1.67 vs OR, 0.75, 95% CI, 0.49-1.13, respectively). There was evidence of interaction with age group and both adult drinking in the home (P = .08) and owning or being willing to use APIs (P = .08) among boys, with strong influences on older adolescents as compared with younger adolescents (adult drinking in the home: OR, 1.95, 95% CI, 1.19-3.22 vs OR, 1.26, 95% CI, 0.89-1.76, respectively; APIs: OR, 2.43, 95%, CI 1.51-3.91 vs OR, 1.50, 95% CI, 1.08-2.09, respectively).

In logistic regression models stratified by stage of uptake, there was a suggestion of interaction between stage of uptake and adult drinking in the home (P = .07), family dinner (P = .06), and owning or being willing to use APIs (P = .004) among girls, with stronger associations among precontemplators compared with contemplators or experimenters (adult drinking in the home: OR, 1.79, 95% CI, 1.23-2.62 vs OR, 1.12, 95% CI, 0.82-1.54, respectively; family dinner every day: OR, 0.47, 95% CI, 0.29-0.76 vs OR, 0.86, 95% CI, 0.60-1.22, respectively; APIs: OR, 2.27, 95% CI, 1.49-3.47 vs OR, 1.24, 95% CI, 0.93-1.67, respectively). Among boys, ownership of or willingness to use APIs and awareness of alcohol advertising showed evidence of interaction with stage of uptake (P = .003 and P = .03, respectively), with stronger associations among precontemplators than contemplators or experimenters (APIs: OR, 2.63, 95% CI, 1.61-4.30 vs OR, 1.24, 95% CI, 0.88-1.76, respectively; alcohol advertising: OR, 1.70, 95% CI, 1.06-2.72 vs OR, 1.00, 95% CI, 0.71-1.40, respectively).

Among those who were never drinkers in 1998, 149 girls (5%) and 112 boys (5%) engaged in binge drinking in 1999. These 261 binge drinkers comprise 24% of girls and 29% of boys who first initiated alcohol use in 1999. Among both girls and boys, a higher AEQ-A score was positively associated with binge drinking (Table 3). For girls, sibling drinking and owning or being willing to use APIs were predictive of binge drinking. Among boys, older age was associated with binge drinking. There did not appear to be an interaction between sex and sibling drinking, owning or being willing to use APIs, or age (all P > .09).

One of the most striking findings of this study was the protective effect of eating family dinner every day among girls. Although the effect appears to be limited to precontemplators, these girls were more than 50% less likely to initiate alcohol use compared with those who ate family dinner some days or never. Family dinner may provide an opportunity for communication between parents and adolescents that may influence adolescent behavior, or there may be an unmeasured protective factor in families who dine together regularly that was not captured in our data. Consistent with literature that has found a greater risk of alcohol use among children living in a single-parent home,^{7 ,25} we found that living in a single-parent household increased the risk of alcohol initiation among girls. Because we did not see a statistically significant interaction between sex and these family variables, we cannot conclude that these relationships are limited to girls.

The roles that alcohol advertising and marketing play in adolescent alcohol uptake and abuse have been explored in several studies.^{32 - 34 ,40} Owning or being willing to use APIs had a greater impact than advertising on alcohol behaviors, increasing risk especially among precontemplators. Owning or being willing to use APIs was further associated with binge drinking among girls. While APIs take the form of seemingly benign items such as T-shirts and key chains, our research suggests that these products contribute to underage alcohol use and abuse. A positive association between awareness of alcohol advertising and alcohol initiation was limited to boys who were precontemplators. However, our measure did not incorporate frequency and it may be that variation in advertising exposure has a broader effect.

Consistent with other prospective studies,^{21 - 22 ,51 - 53} we found that older age and greater physical maturity were associated with alcohol initiation among boys and girls. Cigarette smoking also increased the likelihood of alcohol initiation, as found in other prospective studies.^{54 - 55} However, among those who initiated alcohol use, smoking did not appear to increase the risk of binge drinking.

Among boys, greater athletic self-esteem was associated with alcohol initiation. Our findings are consistent with those of a prior longitudinal study⁵⁶ of adolescents showing that boys and girls who self-identified as jocks were more likely to engage in heavier drinking and binge drinking. Although the measures of athleticism were different from our own, the consistency of the results among boys suggests that an underlying mechanism contributes to the relationship between athleticism and alcohol use. Higher athletic self-esteem among boys may indicate participation on a school sports team, and social influences associated with school teams may encourage drinking and provide increased opportunities for social activities involving alcohol. At this time, however, we can only speculate as to the reasons behind this association.

For girls, social self-esteem was predictive of alcohol initiation, especially among younger girls. A similar positive association was found in a prior cross-sectional study.⁵⁷ Similar to our hypothesis about boys, girls with higher levels of social self-esteem as compared with their peers may attend more social functions that offer greater exposure and access to alcohol. Because neither social nor athletic self-esteem showed an interaction with sex, we cannot conclude that these associations are limited to either girls or boys.

Several prior studies^{25 ,58 - 59} found that high scholastic self-esteem, or scholastic achievement, was inversely associated with alcohol use. Although our results suggested a similar finding, they were not statistically significant. Alcohol use may be related more directly to quantitative measures of scholastic performance than to perceived achievement.

The presence of adults drinking in the home was predictive of alcohol initiation, consistent with previous findings.⁶⁰ This may result from increased access and availability of alcohol or may reflect adolescent mimicking of adult behaviors. Sibling drinking was positively associated with alcohol use and was further associated with binge drinking among girls. This highlights the influence of sibling behavior on adolescent drinking perhaps through behavior modeling, increased access to alcohol, or other mechanisms.

Consistent with previous findings, peer drinking was associated with initiation.^{23 ,25 ,28 ,54 ,61 - 65} We did not find peer drinking to be associated with binge drinking among those who had initiated alcohol use. Although we did not measure peer binge drinking, future studies should examine whether peer binge drinking predicts onset of binge drinking.

Finally, adolescents with higher AEQ-A scores, irrespective of sex, age, or stage of uptake, were more likely to initiate alcohol use and binge drink. Previous research also found that nondrinking teens who report more positive views about alcohol are more likely to initiate use during 1 year of follow-up.⁶⁶ Participant AEQ-A score was further predictive of binge drinking among those who initiated alcohol use and may be a good tool for predicting this behavior.

This study was limited to analyzing those measures included on GUTS questionnaires and did not examine other potential predictors of adolescent alcohol initiation, such as family history of alcoholism.^{67 - 70}

These findings are subject to several potential limitations. Although the GUTS cohort includes participants from across the United States, it does not represent a random sample of adolescents. Participants are predominantly white (94%) and their mothers hold nursing degrees. However, these potential limitations in generalizability of results do not negate the internal validity of the study.

The greatest potential source of misclassification of the data is the self-report nature of GUTS data. However, self-reported data from other studies of adolescent health, including the Youth Risk Behavior Surveillance System, the Longitudinal Survey of Adolescent Health, and the National Longitudinal Survey of Youth, have reasonable validity and reliability^{71 - 75} and we believe that potential misclassification in this study should be limited.

The potential limitations are offset by several strengths of this study. First, the longitudinal nature of the data allowed us to draw conclusions about the temporal effect of multiple factors on adolescent alcohol behaviors. In addition, we were able to identify predictors of alcohol use separately for girls and boys, for younger and older adolescents, and for teens at varying stages of alcohol uptake. Evaluating risk factors within these strata provides data for designing tailored interventions for adolescent alcohol initiation and binge drinking.

Parents should be aware that while high social and athletic self-esteem are positive attributes, they do not protect children from initiating alcohol use. Parents who drink alcohol in the home should be aware that their children are at increased risk for alcohol use and binge drinking. Having family dinner together on a regular basis may be a simple way for parents to reduce the chances that their children will initiate alcohol use.

Although federal legislation banned television advertising for tobacco products and the Master Settlement Agreement restricted tobacco promotional item distribution,⁷⁶ advertising and marketing guidelines for the alcohol industry are self-regulated.⁷⁷ Our findings suggest that more formalized restrictions are needed.

Correspondence: Laurie B. Fisher, SM, Channing Laboratory, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, 181 Longwood Ave, Boston, MA 02115 (laurie.fisher@channing.harvard.edu).

Accepted for Publication: March 12, 2007.

Author Contributions:Study concept and design: Fisher, Miles, Austin, Camargo, and Colditz. Acquisition of data: Austin, Camargo, and Colditz. Analysis and interpretation of data: Fisher, Miles, Austin, Camargo, and Colditz. Drafting of the manuscript: Fisher and Miles. Critical revision of the manuscript for important intellectual content: Fisher, Miles, Austin, Camargo, and Colditz. Statistical analysis: Fisher. Administrative, technical, and material support: Fisher and Colditz. Study supervision: Colditz.

Financial Disclosure: None reported.

Funding/Support: Dr Austin is supported by grant 6T71-MC00009-15-01, Leadership Education in Adolescent Health project, Maternal and Child Health Bureau, Health Resources and Services Administration. Dr Colditz is an American Cancer Society Clinical Research Professor.

Additional Contributions: The GUTS Research Team provided input. We acknowledge the contribution of the thousands of adolescents who complete the GUTS surveys.