Jproject

Journal of Adolescent Health 54 (2014) 26e32

www.jahonline.org

Original article

Incidence and Determinants of Cigarette Smoking Initiation in Young Adults

Jennifer L. O’Loughlin, Ph.D. a,b,c,*, Erika N. Dugas, M.Sc. a, Erin K. O’Loughlin, M.A. a, Igor Karp, Ph.D. a,b, and Marie-Pierre Sylvestre, Ph.D. a,b a Centre de Recherche du Centre Hospitalier de l’Université de Montréal, Montreal, Quebec, Canada b Department of Social and Preventive Medicine, University of Montréal, Montreal, Quebec, Canada c Institut National de Santé Publique du Québec, Montreal, Quebec, Canada

Article history: Received November 21, 2012; Accepted July 10, 2013 Keywords: Smoking initiation; Longitudinal; Adolescents; Young adults; Incidence

A B S T R A C T IMPLICATIONS AND

Purpose: To describe the incidence and identify predictors of smoking initiation in young adults. Methods: Data were collected in self-report questionnaires in 22 cycles over 13 years in a prospective cohort investigation of 1,293 students recruited in 1999e2000 from all grade 7 classes in a convenience sample of 10 high schools in Montreal, Canada. Participants were 12.7 years of age on average at cohort inception and 24.0 years of age in cycle 22. Independent predictors of smoking initiation in young adulthood (postehigh school) were identified in multivariable logistic regression analysis using generalized estimating equations. Results: Of 1,293 participants, 75% initiated smoking by cycle 22. Of these, 44%, 43%, and 14% initiated before high school, during high school, and in the 6 years after high school, respectively. The incidence density rate of initiation was .33, .13, .14, .11, and .12 initiation events per person-year in grade 7, 8, 9, 10, and 11, respectively, and .05 postehigh school. Independent predictors of smoking initiation in young adults included alcohol use, higher impulsivity, and poor academic performance. Conclusions: A total of 14% of smokers who initiated smoking before age 24 years did so after high school. The predictors of initiation in young adults may provide direction for relevant preventive interventions.

� 2014 Society for Adolescent Health and Medicine. All rights reserved.

* Address correspondence to: Jennifer L. O’Loughlin, Ph.D., Centre de Recherche du Centre Hospitalier de l’Université de Montréal, 3875 St. Urbain, Montreal, Quebec H2W 1V1, Canada.

E-mail address: [email protected] (J.L. O’Loughlin).

1054-139X/$ e see front matter � 2014 Society for Adolescent Health and Medicine. All rights reserved. http://dx.doi.org/10.1016/j.jadohealth.2013.07.009

CONTRIBUTION

The incidence of smoking initiation declines through- out adolescence and young adulthood, but 14% of youth whoinitiatesmokingbefore age 24 years do so after high school. Predictors of smokinginitiationinyoung adulthood include alcohol use, impulsivity, and poor academic performance. Pre- ventive intervention tar- geted specifically to young adults may be warranted.

Transition from adolescence to young adulthood represents a critical life period during which young people graduate from high school and leave home to attend college or university or join the workforce. These changes are typically characterized by decreasing parental control and changing social networks, which may increase susceptibility to smoking [1]. With the marked declines in smoking rates in the past 3 decades [2,3], there is growing concern that the tobacco industry is specifically tar- geting young adults to take advantage of this susceptibility [3], and that the incidence of smoking initiation in young adults may

be increasing. In fact, several reports suggest that the incidence of smoking initiation in young adulthood has increased [1,4,5]. The 2008 National Survey on Drug Use and Health survey, for example, reported that 1 million American adults initiated smoking as young adults, an increase of almost 50% over the 600,000 who initiated in 2002 [4]. If the incidence of smoking initiation in young adulthood is indeed high or increasing, tobacco prevention programs targeting this specific population may be warranted.

Whereas many studies have documented the rates and determinants of smoking initiation in adolescents [6e8] or in populations that include both youth and adults [9], little is known about smoking initiation in young adults [5]. More specifically, it is not known whether the incidence of smoking initiation in young adults is lower, equivalent to, or higher than in

J.L. O’Loughlin et al. / Journal of Adolescent Health 54 (2014) 26e32 27

adolescence, and whether the determinants of later initiation differ from those during adolescence. The objectives of this study were to describe the incidence of cigarette smoking initiation in young adulthood, and to identify predictors of initiation in young adults. To contextualize the importance of initiation in young adults relative to younger persons, the incidence of onset throughout adolescence and emerging adulthood is described.

Methods

Data were drawn from the Nicotine Dependence in Teens (NDIT) Study, a prospective cohort investigation of 1,293 students recruited in 1999e2000 from all grade 7 classes in a convenience sample of 10 high schools in or near Montreal, Canada [10]. Self-report questionnaires were administered at school every 3 months during the 10-month school year from grade 7 to 11, for a total of 20 survey cycles during the 5 years of secondary school. School-specific data on tobacco control poli- cies and activities within schools to promote nonsmoking were collected in self-report questionnaires completed by school administrators in spring 2003. In addition, students and teachers were asked to identify commercial establishments (i.e., conve- nience stores, gas stations, pharmacies, restaurants, fast food chains, grocery stores, and dollar stores) within a 1-mile radius of schools where students gather before school, during recess and lunch, and after school. Each establishment was visited by two trained observers who collected data through direct observation on the availability of and access to tobacco products, the visibility of no-smoking signs, and cigarette promotions, using an assessment tool adapted from previous work [11].

In 2007e2008, when the mean (standard deviation [SD]) age of participants was 20.4 (.8) years, and again in 2010e12 when the mean (SD) age of participants was 24.0 (.7) years, data were collected from 880 (68% of 1,293) and 858 (66% of 1,293) participants, respectively in mailed self-report questionnaires, in survey cycles 21 and 22. Survey cycle 21 covered a median of 3.1 years postehigh school, and survey cycle 22 covered an additional 3.1 years after survey 21. Anthropometric measure- ments (i.e., height and weight) were collected at baseline and in survey cycles 12, 19, and 22. Finally, in addition to participant questionnaires, 654 parents (65% of those eligible) completed mailed self-report questionnaires in 2009e2010.

Parents and guardians provided written informed consent for their adolescent to participate at baseline. Participants them- selves (who had attained legal age in survey cycles 21 and 22) provided consent in the postehigh school survey cycles. The study was approved by the Direction de Santé Publique de MontrealeCentre and McGill University Institutional Review Boards and the Ethics Research Committee of the Centre de Recherche du Centre Hospitalier de l’Université de Montréal.

Study variables

Data on cigarette smoking were collected in three indicators in survey cycles 1e22. (1) Lifetime smoking history was measured in one item: Have you ever in your life smoked a ciga- rette, even just a puff (drag, hit, or haul)? [10]. Response choices included “no,” “yes, one or two times,” “yes, three or four times,” “yes, five to 10 times,” and “yes, >10 times.” (2) Current smoking status was measured in one item: “Check the one box that describes you best: “I have never smoked a cigarette, even just a puff”; “I have smoked cigarettes, even just a puff, but not at all

in the past 12 months”; “I smoked cigarettes once or a couple of times in the past 12 months”; “I smoke cigarettes once or a couple of times each month”; “I smoke cigarettes once or a couple of times each week”; or “I smoke cigarettes every day”[10]. (3) To collect data on recent cigarette smoking, participants completed a past 3-month recall in each survey cycle that collected data on cigarette smoking in each of the 3 months preceding each questionnaire. The recall included one item for each month that measured the number of days on which the participant had smoked during the month, and one item for each month that measured the number of cigarettes smoked per day on average during that month [10].

Participants were categorized into one of four categories based on these three indicators, to represent their smoking history during the 11-year follow-up. Never-smokers included participants who self-reported in survey cycle 20 that they had never smoked, even just a puff. To confirm their never-smoked status, each questionnaire completed during high school was checked for any indication of smoking in any of the smoking indicators. A total of 104 participants who reported at the end of high school that they had never smoked had actually smoked, and were therefore reclassified into one of the other three categories. Baseline smokers included participants who reported that they had smoked, even just a puff in survey cycle 1. High school initiators included those who indicated at baseline that they had never smoked, but who indicated that they had smoked, even just a puff, in at least one of the 19 follow-up questionnaires during high school. Finally, young adult initia- tors included participants who indicated that they had never smoked during or before high school, but reported cigarette smoking in either survey cycle 21 or 22 (i.e., he or she initiated smoking between the end of high school and survey cycle 22).

Potential predictor variables included sociodemographic indicators (sex, age, language spoken at home (French, English, or other), single-parent family status, mother’s education, and doing well in school); indicators of smoking in the social envi- ronment (parent(s) smoke, sibling(s) smoke, friends smoke, or teachers/school staff smoke); psychological indicators (family stress, other stress symptoms, depression symptoms, impul- sivity, novelty seeking, self-esteem, worry about weight); susceptibility to cigarette package warning, susceptibility to tobacco advertising; feeling as if one really needed a cigarette, overweight; asthma; lifestyle-related indicators (alcohol use, use of other tobacco products, level of light, moderate, and vigorous physical activity, participation in team sports, television viewing); and neighborhood context (tolerance of smoking at school, and in neighborhood corner stores and restaurants) [10].

Data were collected on most variables in every survey cycle. However, because they were not expected to vary substantially over time, data on several variables were collected in selected surveys only, depending on space available in the questionnaire. Appendix 1, which can be found in the online version of this article, describes potential predictor variables in detail including the reference, survey cycle(s) in which data on the variable were collected, the specific measurement items, response choices, and how response choices were recoded for analysis.

Data analysis

The incidence density of smoking initiation was computed in each survey cycle beginning with survey cycle 2 (because this analysis required participants to be nonsmokers in survey cycle

J.L. O’Loughlin et al. / Journal of Adolescent Health 54 (2014) 26e3228

1) as the sum of new smokers during the survey cycle divided by the sum of person-years at risk during that cycle, calculated as the time difference between the date of the survey and the date of the previous survey. In survey cycles 4, 7, and 8, questionnaires were not administered in 7 schools and 1 and 1 school, respec- tively, and hence no date was documented for that survey. To calculate person-years at risk in this case, a date was assumed at the midpoint between the surveys preceding and after the missing survey. A smooth time trend for the survey cycle-specific incidence density values was fitted using the locally weighted scatterplot smoothing method [12]. The grade-specific incidence density rate was also computed as the sum of new smokers during the grade (i.e., grade 7e11) divided by the sum of person- years at risk during that grade.

Predictors of initiation in young adults were studied in the subset of never-smokers at the end of high school with data on smoking in either survey cycle 21 or 22. For time- varying potential predictor variables (i.e., parent[s], sibling[s], and friends smoke; stress symptoms; depression symptoms; susceptibility to smoking [felt as if you really needed a cigarette], worry about weight; alcohol use; use of other tobacco products; level of physical activity [light, moderate, and vigorous physical activity], participation in team sports; and hours of television viewing per week), the last observation available in grade 11 was retained. For variables measured one to three times during the 20 survey cycles in high school (i.e., self-esteem, asthma, suscepti- bility to tobacco advertising, tolerance of smoking in school, corner stores and restaurants, parental education, teachers or school staff smoke, impulsivity, novelty seeking, academic performance, and overweight), the value of the last observation available was used. A missing value was assigned when no data were available in any survey cycle for a specific variable, so that for each predictor variable, a complete case analysis was performed.

The 6-year cumulative incidence of smoking initiation (i.e., the proportion of never-smokers at the end of high school who initiated smoking during the 6-year postehigh school period) was computed for each response category of each of 32 potential predictor variables. Separate multivariable logistic regression models were then run to assess the association between each potential predictor variable and smoking initiation in young adulthood separately in a model with its own distinct set of covariates (i.e., variables correlated with the predictor variable with a Pearson correlation coefficient at jrj � .20). All models were adjusted for sex and age regardless of their correlation with the potential predictor variable. The procedure that was used to address confounding (i.e., assessment of correlations with exposure and adjustment for the risk factors that were moderately or strongly correlated with exposure) should have simultaneously served to adjust for selection bias (assuming there were no other unmeasured risk factors that would be associated with loss to follow-up in a differential manner between the exposed and unexposed) [13]. Generalized esti- mating equations models were used with an exchangeable correlation structure to take the correlation induced by sampling by school into account. Analyses were conducted using SPSS software, version 16.0 (SPSS, Inc., Chicago, IL) and R (R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. http:// www.R-project.org/; 2012).

Results

Of the 1,293 NDIT participants, 971 initiated cigarette smoking during the 11-year follow-up (75%). Of the 971, 424 began smoking before baseline (44%), 415 began smoking in survey cycle 2e20 during high school (43%), and 132 began smoking after high school (14%). The mean (SD) age of initiation reported in survey cycle 22 among young adult initiators was 18.2 (2.7) years, compared with 14.5 (2.1) years among high school initiators and 12.5 (2.6) years among baseline smokers.

Table 1 describes selected baseline characteristics of the 869 participants who had never smoked (not even a puff) at baseline (and were therefore included in the incidence density analyses), and compares these variables in the 547 (63% of 869) who initiated cigarette smoking during one of the follow-up survey cycles (i.e., survey cycles 2e22) with those in the 322 (37% of 869) participants who did not initiate smoking.

Figure 1 documents the incidence density of smoking initiation during the 11-year follow-up among the 869 NDIT participants who had never smoked at baseline. The incidence was highest in survey cycles 2 and 3 (i.e., in grade 7). There- after, there was a steady decline in the incidence over time, as shown by the decreasing smooth trend estimate (test for non- zero trend, p < .0001). The incidence density of smoking initiation was .33, .13, .14, .11, and .12 initiation events per person-year in grade 7, 8, 9, 10, and 11, respectively, and .05 postehigh school.

Predictors of initiation in young adults

A total of 894 NDIT participants had data available in grade 11. Of these, 148 participants were missing data on smoking initia- tion in survey cycle 21 or 22. Selected baseline characteristics of these 148 participants were compared with those of the 746 participants who had both grade 11 and postehigh school data. With the exception of a lower proportion of boys in the subset of participants with data in grade 11 and survey cycle 21 or 22 (46% vs. 57%; p ¼ .009), there were no statistically significant differ- ences in the other characteristics examined (i.e., age, born in Canada, French spoken at home, single-parent family, mother university-educated, parent(s) some, sibling(s) smoke, friends smoke).

Of the 746 participants with data in both grade 11 and postehigh school, 427 had already initiated smoking (either at baseline or during follow-up in high school) and were therefore excluded from the analysis of the predictors of late-onset initi- ation. A total of 283 NDIT participants had never smoked, not even a puff either before cohort inception or during high school, and were therefore eligible for the analysis on predictors of smoking initiation in young adulthood. Of the 283, 108 initiated smoking in either survey cycle 21 or 22 (38%), and 175 remained never-smokers.

Of the 32 potential predictor variables investigated, only four including mother university-educated, not doing well at school, higher impulsivity, and use of alcohol were associated with smoking initiation in young adults univariately. These findings were robust when covariates were taken into account, with the exception that the confidence interval for the odds ratio for mother university-educated included one once covariates were included in the model (Table 2).

Table 1 Baseline characteristics of participants who initiated and did not initiate cigarette smoking during follow-up, Nicotine Dependence in Teens Study 1999e2012

Total (N ¼ 869) Did not initiate smoking during follow-up (n ¼ 322)

Initiated smoking during follow-up (n ¼ 547)

p value

Male, % 50.5 58.1 46.1 .001 Age, years (mean [standard deviation]) 12.7 (.5) 12.7 (.5) 12.6 (.4) .017 Born in Canada, % 90.9 86.6 93.4 .001 French spoken at home, % 21.6 21.4 21.8 .910 Single-parent family, % 7.2 5.9 8.0 .239 Mother university-educated, % 48.1 43.5 50.3 .091 Parent(s) smoke(s), % 29.4 26.5 31.1 .146 Sibling(s) smoke(s), % 10.0 6.8 11.9 .017 Friends smoke, % 21.5 15.2 25.2 .001

J.L. O’Loughlin et al. / Journal of Adolescent Health 54 (2014) 26e32 29

Discussion

Because cigarette smoking usually begins at age 12e13 years [14] and most smokers begin smoking by age 18 years [1,4,15], tobacco prevention programs typically target children and adolescents [1]. There is growing concern that smoking initiation may be increasing in young adults, especially because this subgroup is often overlooked by tobacco control programs [1,5,16]. However, if smoking initiation can be prevented throughout young adulthood, it is highly likely that a person will never become a regular smoker [15].

The current study agrees with others [4] that most youth (75% in NDIT) try smoking. Most individuals who tried did so during high school, and the incidence of smoking initiation declined over time. In this study, 14% of smokers began smoking in the 6 years after high school, which is lower than other reports in which 25% [17e19] and 22% [20] of smokers reported late-onset smoking. This estimate is likely lower than others because it was established that many self-reported never-smokers at the end of high school had actually reported smoking in an earlier survey

Figure 1. Incidence density of cigarette smoking initiation in each survey cycle (beginning in survey cycle 2) among participants at risk of initiating smoking in the given survey cycle. Nicotine Dependence in Teens Study 1999e2012. Width of bars reflects the mean length of time between survey cycles. Incidence density ¼ number of participants who initiated smoking in a given survey divided by the sum of person-years at risk during that survey cycle.

cycle. In a previous report using NDIT data, the authors concluded that there are multiple diverse determinants of adolescent smoking initiation at both the individual and contextual levels [10]. However, young adult initiators seemingly have a unique constellation of risk factors and/or did not have sufficient exposure to risk factors until after high school. Predictors of later initiation reported to date include younger age, male gender, non-white ethnicity, educational attainment, parental education, low socioeconomic status, alcohol use, illicit drug use, marijuana use, exposure to smoking, boredom or stress while serving in military, attending tobacco-sponsored social events while in college, academic difficulties or poor grades, truancy, and exposure to social norms or perceptions that encourage smoking [5,15,17,21].

In the current study, only impulsivity, alcohol consumption, and not doing well in school predicted initiation after high school. This is consistent with a depletion of susceptibles premise such that susceptible youth are exposed during high school to the constellation and intensity of risk factors that cause them to smoke, they begin to smoke, and therefore they drop out of the pool of individuals at risk for later initiation Because the early initiators are no longer in the pool of susceptibles, the profile of risk factors for later initiation is altered [22].

Several longitudinal studies [17,18,20] support an association between alcohol consumption and smoking initiation in young adults. With decreased parental monitoring after high school, young persons may increasingly frequent bars and other places where young people gather to drink, thereby increasing expo- sure to peer smoking [17,23]. Alternatively, the Alcohol Myopia Model [24] posits that alcohol impairs cognitive processing dependent on intact attentional capacity. This impairment creates a myopic effect on attention that restricts the range of internal and external cues that can be perceived and processed. As a result, remaining attentional resources are allocated to the most salient and easy-to-process cues. Thus, under the influence of alcohol, nonsmokers may be less inhibited and feel stronger urges to smoke cigarettes [23]. Although studies supporting this theory were conducted in laboratory settings, it is possible that vulnerability to smoking in later initiators is enhanced by alcohol use [23].

Higher impulsivity was associated with later initiation in the current study. It is possible that during high school, young people with higher impulsivity were more closely monitored by their parents because of problems related to impulsivity [25]. There is indeed evidence that adolescents whose parents use effective monitoring practices are less likely to smoke cigarettes [26]. However once these young people leave the home environment, the protective influence of parental monitoring may no longer be

Table 2 Six-year cumulative incidencea of young adult cigarette smoking initiation according to potential predictor variables, and crude and adjusted odds ratios (ORs) and 95% confidence intervals (95% CIs) for the association between potential predictor variables and young adult initiation, Nicotine Dependence in Teens Study 1999e2012

Potential predictor variable n 6-year cumulative incidence (n ¼ 283), %

ORcrude (95% CI) ORadj (95% CI) Covariates included in model

Sociodemographic Sex Age, other stress, depression, worry about weight,

vigorous physical activityMaleb 150 38.0 Reference Reference Female 133 38.3 1.0 (.6e1.7) .9 (.5e1.7)

Age at baseline, yearsc .6 (.3e1.1) .5 (.3e1.0) Sex, school tolerance �12.6 (median) 141 40.4 >12.6 142 35.9

Language Sex, age, school tolerance, corner store intolerance, restaurant intoleranceOtherb 224 40.6 Reference Reference

French 59 28.8 .6 (.3e1.1) .6 (.2e1.6) Single-parent family Sex, age Nob 255 38.0 Reference Reference Yes 28 39.3 1.1 (.4e2.5) 1.0 (.4e2.5)

Mother university-educated Sex, age, team sports, corner store intolerance Nob 138 32.6 Reference Reference Yes 136 43.4 1.6 (1.0e2.6) 1.5 (.9e2.5)

Doing well at school Sex, age A bit/very trueb 258 35.7 Reference Reference Not at all true 25 64.0 3.2 (1.4e7.4) 3.5 (1.5e8.2)

Smoking in social environment Parent(s) smoke Sex, age Nob 236 38.1 Reference Reference Yes 47 38.3 1.0 (.5e1.9) 1.1 (.5e2.1)

Sibling(s) smoke Sex, age Nob 247 37.2 Reference Reference Yes 36 44.4 1.4 (.6e2.9) 1.4 (.6e2.9)

Friends smoke Sex, age Noneb 85 31.8 Reference Reference Few or more 198 40.9 1.5 (.9e2.6) 1.5 (.9e2.6)

Teachers/school staff smoke Sex, age, restaurant intolerance Not at all trueb 81 43.2 Reference Reference A bit/very true 202 36.1 .7 (.4e1.2) .8 (.4e1.3)

Psychosocial indicators Family stressc 1.1 (.6e1.9) 1.1 (.6e2.0) Sex, age, other stress, depression symptoms,

worry about weight�1 (median) 181 38.7 >1 102 37.3

Other stressc .9 (.6e1.6) .8 (.4e1.6) Sex, age, family stress, depression symptoms, worry about weight, asthma�1.4 (median) 144 40.3

>1.4 139 36.0 Depression symptomsc 1.1 (.8e1.5) 1.2 (.8e1.7) Sex, age, family stress, other stress, self-esteem,

worry about weight�2 (median) 145 37.2 >2 138 39.1

Impulsivityc 1.4 (1.0e1.9) 1.5 (1.0e2.3) Sex, age, novelty seeking <2 (median) 137 32.1 �2 146 43.8

Novelty seekingc 1.2 (.9e1.7) .9 (.6e1.4) Sex, age, impulsivity <2.7 (median) 146 35.6 �2.7 137 40.9

Self-esteemc .6 (.3e1.2) .6 (.3e1.2) Sex, age, depression symptoms <2.8 (median) 132 41.7 �2.8 146 34.9

Worry about weight Sex, age, family stress, other stress, depression symptomsNob 178 38.8 Reference Reference

Yes 105 37.1 .9 (.6e1.5) .9 (.5e1.6) Psychosocial exposures Cigarette package warnings

make me afraid to smoke Sex, age

Not at all trueb 120 41.7 Reference Reference A bit/very true 163 35.6 .8 (.5e1.2) .8 (.5e1.2)

Cigarette advertisements make me want to smoke

Sex, age

Not at all trueb 270 38.5 Reference Reference A bit/very true 4 50.0 1.6 (.2e11.6) 1.7 (.3e12.0)

Felt as if one really needed a cigarette Sex, age Neverb 277 37.5 Reference Reference Rarely/sometimes/often 6 66.7 3.3 (.6e8.7) 3.5 (.6e20.1)

J.L. O’Loughlin et al. / Journal of Adolescent Health 54 (2014) 26e3230

Table 2 Continued

Potential predictor variable n 6-year cumulative incidence (n ¼ 283), %

ORcrude (95% CI) ORadj (95% CI) Covariates included in model

Overweight Sex, age Nob 227 39.2 Reference Reference Yes 51 33.3 .8 (.4e1.5) .8 (.4e1.5)

Asthma Sex, age, other stress Nob 247 37.2 Reference Reference Yes 36 44.4 1.4 (.7e2.7) 1.4 (.7e2.8)

Lifestyle-related indicators Alcohol use 126 27.0 Sex, age Nob 157 47.1 Reference Reference Yes 2.4 (1.4e4.1) 2.3 (1.4e4.0)

Use other tobacco products Sex, age Nob 270 37.8 Reference Reference Yes 13 46.2 1.4 (.5e4.0) 1.3 (.4e3.6)

Light physical activityc 1.0 (.9e1.1) 1.0 (.9e1.1) Sex, age, moderate physical activity <1 (median) 141 35.5 �1 142 40.8

Moderate physical activityc 1.0 (1.0e1.0) 1.0 (1.0e1.0) Sex, age, light physical activity, vigorous physical activity�6 (median) 145 38.6

>6 138 37.7 Vigorous physical activitya 1.0 (.9e1.0) 1.0 (.9e1.0) Sex, age, team sports, moderate physical activity �1 (median) 145 41.4 >1 138 34.8

Team sports Sex, age, mother university-educated, vigorous physical activityNob 131 37.4 Reference Reference

Yes 152 38.8 1.1 (.7e1.7) 1.0 (.6e1.7) Hours of television/weekc 1.0 (.9e1.1) 1.0 (1.0e1.1) Sex, age �3.5 (median) 144 37.5 >3.5 139 38.8

Neighborhood context School tolerancec Sex, age, language, corner store intolerance �.9b 198 38.9 Reference Reference >.9 85 36.5 .9 (.5e.5) 1.0 (.5e2.0)

Corner store tolerancec .8 (.3e1.8) 1.9 (.5e8.2) Sex, age, language, school tolerance, restaurant intolerance, mother university-educated�.25 (median) 157 38.2

>.25 126 38.1 Restaurant tolerancec 2.1 (.5e8.2) 1.4 (.1e13.5) Sex, age, language, corner store intolerance,

teachers smoke�.40 (median) 151 36.4 >.40 132 40.2

a From the end of high school to survey cycle 22. b Reference category. c Potential predictor variable was included in the model as a continuous variable, with higher scores indicating higher levels of exposure. The OR indicates the increase

in the probability of smoking initiation per one-unit increase in the potential predictor variable.

J.L. O’Loughlin et al. / Journal of Adolescent Health 54 (2014) 26e32 31

present, such that smoking initiation is simply delayed in this subgroup [21].

Consistent with previous reports [15,20,21], later initiators were more likely to report academic difficulties in high school. This subgroup is more likely to seek employment immediately after high school, many in blue collar occupations, rather than continue with their education [27]. It is well established that the prevalence of smoking is higher in blue collar workers [28]. It is possible that increased exposure to work-related smoking in this subgroup may underpin the association between doing poorly at school and later initiation. Ellickson et al. [15] suggested that school performance programs in conjunction with smoking prevention programs may be warranted.

Whereas smoking initiation remains an issue in young adulthood, it is not clear whether the intensity, frequency, and/or duration of smoking differ between adolescent and young adult initiators. The current data suggest that smoking patterns differ in early versus later initiators (only 10% of young adult initiators were daily smokers, compared with 24% of high school initiators

and 28% of baseline smokers), although elapsed time since initiation may not have been sufficient for young adult initiators to develop daily smoking. Longer follow-up is needed to assess whether smoking patterns converge over time between adoles- cent and young adult initiators.

These results need to be interpreted with caution and the associations observed need to be replicated in other studies. However, this study has several potential program and policy implications. Smoking prevention programs tailored specifically to young adults may be needed. In particular, education about the dangers of smoking may be effective when young adults are starting new households and new families. Educating young adult parents (and parents-to-be) about the dangers of second- hand smoke may not only benefit the child, but may also prompt cessation. Practitioners and policy makers may need to be made aware that smoking initiation not only occurs among adolescents, but remains an issue into early adulthood. Finally, there is a need to continue to promote smoke-free environments in places where young adults work and play (i.e., at work, on

J.L. O’Loughlin et al. / Journal of Adolescent Health 54 (2014) 26e3232

campuses, in restaurants, bars and nightclubs, or at home). Because alcohol use is a determinant of young adult initiation, tobacco control campaigns conducted in places where alcohol is sold could be particularly effective in this age group.

Limitations of this analysis include the problem that self- report data may be subject to recall bias. Loss to follow-up may have resulted in selection bias in the estimates of incidence density and/or cumulative incidence. However, the authors believe that their analytic approach ensured that it is unlikely that the estimated measures of association between cigarette smoking initiation and its potential determinants were subject to selection bias. Use of a convenience sample may limit the generalizability of the findings. Finally, the denominators used to compute incidence density may be inaccurate when a date was assumed at the midpoint between the surveys preceding and after a missing survey.

The incidence of smoking initiation declined steadily through high school into young adulthood. However,14% of young people who initiated smoking did so after high school, which indicates that tobacco control interventions tailored to young adults are warranted to reduce the public health burden of tobacco. Risk factors for young adult initiation including alcohol use, impul- sivity, and poor academic performance may provide direction for the development of relevant tobacco control interventions tar- geted to this subgroup.

Acknowledgments

This work was supported by the Canadian Cancer Society (Grants 010271 and 017435). J.O.L. holds a Canada Research Chair in the Early Determinants of Adult Chronic Disease. The authors thank the NDIT participants and their parents.

References

[1] Hammond D. Smoking behaviour among young adults: Beyond youth prevention. Tob Control 2005;14:181e5.

[2] GilmoreJ.ReportonsmokinginCanada,1985to2001(StatisticsCanada,Catalog 82F0077XIE). October 15, 2012. Available at: http://publications.gc.ca/ collections/Collection/Statcan/82-0077-X/82F0077XIE2001001.pdf; 2002.

[3] United States Department of Health and Human Services. Preventing tobacco use among youth and young adults: A report of the Surgeon General. Atlanta (GA): United States Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2012.

[4] Lantz PM. Smoking on the rise among young adults: Implications for research and policy. Tob Control 2003;12(Suppl 1):i60e70.

[5] Freedman KS, Nelson NM, Feldman LL. Smoking initiation among young adults in the United States and Canada, 1998-2010: A systematic review. Prev Chronic Dis 2012;9:E05.

[6] Paynter J, Edwards R. The impact of tobacco promotion at the point of sale: A systematic review. Nicotine Tob Res 2009;11:25e35.

[7] Derzon JH, Lipsey MW. Predicting tobacco use to age 18: A synthesis of longitudinal research. Addiction 1999;94:995e1006.

[8] Lovato C, Watts A, Stead LF. Impact of tobacco advertising and promotion on increasing adolescent smoking behaviours. Cochrane Database Syst Rev 2011;10:CD0003439.

[9] Choi WS, Harris KJ, Okuyemi K, et al. Predictors of smoking initiation among college-bound high school students. Ann Behav Med 2003;26:69e74.

[10] O’Loughlin J, Karp I, Koulis T, et al. Determinants of first puff and daily cigarette smoking in adolescents. Am J Epidemiol 2009;170:585e97.

[11] Centers for Disease Control and Prevention. Point-of-purchase tobacco environments and variation by store typeeUnited States, 1999. MMWR Morb Mortal Wkly Rep 2002;51:184e7.

[12] Cleveland WS. Robust locally weighted regression and smoothing scat- terplots. J Am Statist Assoc 1979;74:829e36.

[13] Hernan M, Hernandez-Dias S, Robins JM. A structural approach to selection bias. Epidemiology 2004;15:615e25.

[14] Youth smoking in Canada. Health Canada. Available at: http://www.hc-sc. gc.ca/hc-ps/tobac-tabac/research-recherche/stat/_ctums-esutc_fs-if/2000- youth-eng.php. October 15, 2012.

[15] Ellickson PL, McGuigan KA, Klein DJ. Predictors of late-onset smoking and cessation over 10 years. J Adolesc Health 2001;29:101e8.

[16] Tjora T, Hetland J, Aarø LE, et al. Late-onset smokers: How many, and associations with health behaviours and socioeconomic status. Scand J Public Health 2012;40:537e43.

[17] Tercyak KP, Rodriguez D, Audrain-McGovern J. High school seniors’ smoking initiation and progression 1 year after graduation. Am J Public Health 2007;97:1397e8.

[18] Myers MG, Doran NM, Trinidad DR, et al. A prospective study of cigarette smoking initiation during college: Chinese and Korean American students. Health Psychol 2009;28:448e56.

[19] Bernat DH, Klein EG, Forster JL. Smoking initiation during young adult- hood: A longitudinal study of a population-based cohort. J Adolesc Health 2012;51:497e502.

[20] Costa FM, Jessor R, Turbin MS. College student involvement in cigarette smoking: The role of psychosocial and behavioral protection and risk. Nicotine Tob Res 2007;9:213e24.

[21] White HR, Violette NM, Metzger L, Stouthamer-Loeber M. Adolescent risk factors for late-onset smoking among African American young men. Nicotine Tob Res 2007;9:153e61.

[22] Caro JJ, Möller J, Getsios D. Discrete event simulation: The preferred technique for health economic evaluations? Value Health 2010;13: 1056e60.

[23] Reed MB, McCabe C, Lange JE, et al. The relationship between alcohol consumption and past-year smoking initiation in a sample of undergrad- uates. Am J Drug Alcohol Abuse 2010;36:202e7.

[24] Steele CM, Josephs RA. Alcohol myopia: Its prized and dangerous effects. Am Psychol 1990;45:921e33.

[25] Patock-Peckham JA, King KM, Morgan-Lopez AA, et al. Gender-specific mediational links between parenting styles, parental monitoring, impul- siveness, drinking control, and alcohol-related problems. J Stud Alcohol Drugs 2011;72:247e58.

[26] Choquet M, Hassler C, Morin D, et al. Perceived parenting styles and tobacco, alcohol and cannabis use among French adolescents: Gender and family structure differentials. Alcohol Alcohol 2008;43:73e80.

[27] United States Department of Commerce, Economics, and Statistics Administration, United States Census Bureau. http://www.census.gov/ prod/2004pubs/p20-550.pdf; 2003. October 15, 2012.

[28] Lee DJ, Fleming LE, Arheart KL, et al. Smoking rate trends in U.S. occupa- tional groups: The 1987 to 2004 National Health Interview Survey. Occup Environ Med 2007;49:75e81.

Appendix 1 Potential predictor variables used in the analysis to identify predictors of young adult cigarette smoking initiation

Potential predictor variable and reference

Data collected in survey cycle(s)

Item(s) Response choices/creation of score Recoded for use in analysis

Sex 1e20 Are you a boy or a girl? Male, female e Age 1e20 Date of birth, date of survey �12.6 (median),

>12.6 for univariate analysis

Language 1e20 What language do you speak most often at home?

English, French, English and French, other French, other

Single-parent family

1e20 Do you live with your: biologic mother, biologic father, stepmother, stepfather, aunt(s), uncle(s), grandmother(s), grandfather(s)?

No, yes (for each person) No, yes

Mother university- educated

13, 17 How much education has your mother had? Did not finish high school, high school graduate, vocational, technical school, CEGEP, university, do not know, other

No, yes

Doing well in school

16, 20 How true is the following for you? I’m doing well at school this year.

Not at all, a bit true, very true A bit/very true, not at all true

Parent(s) smoke 1e20 Does your father currently smoke cigarettes? Does your mother currently smoke cigarettes?

No, yes (for each parent) No, yes (1e2 parents smoke)

Sibling(s) smoke 1e20 You have ____ sisters who smoke cigarettes. You have ____ brothers who smoke cigarettes.

0, 1, 2, 3, 4, 5, �6 No, yes (�1 siblings smoke)

Friends smoke 1e20 Now, think about your friends. How many of the people whom you usually hang out with smoke cigarettes?

None, a few, about half, more than half, most or all

None, a few or more

Teachers/school staff smoke

12, 16, 20 How true is the following for you? I often see teachers or staff smoking near this school.

Very true, a bit true, not at all true A bit/very true, not at all true

Family stress symptoms [1,2]

1e20 During the past 3 months, have you been worried or stressed by any of the following? (1) Your parents separating or divorcing; (2) your relationship with your father; (3) your relationship with your mother; (4) your relationship with your brother(s)/sister(s); (5) your new family (parents remarried)

Not at all, a little bit, quite a bit, a whole lot. Scored 1e4. Responses were summed, then divided by the number of items responded to, to create a continuous score. Higher scores indicate higher stress.

�1 (median), >1 for univariate analysis

Other stress symptoms [1,2]

1e20 During the past 3 months, have you been worried or stressed by any of the following? (1) Breaking up with your boyfriend or girlfriend; (2) your relationship with your friends; (3) a health problem (such as acne or asthma); (4) sex; (5) financial problems in your family; (6) schoolwork

Not at all, a little bit, quite a bit, a whole lot. Scored 1e4. Responses were summed, then divided by the number of items responded to, to create a continuous score. Higher scores indicate higher stress.

�1.4 (median), >1.4 for univariate analysis

Depression symptoms [3]

1e20 During the past 3 months, how often have you: (1) felt too tired to do things; (2) had trouble going to sleep or staying asleep; (3) felt unhappy, sad, or depressed; (4) felt hopeless about the future; (5) felt nervous or tense; or (6) worried too much about things?

Never, rarely, sometimes, often. Scored 1e4. Responses were summed, then divided by the number of items responded to, to create a continuous score. Higher scores indicate more depression symptoms.

�2 (median), >2 for univariate analysis

Impulsivity [4,5] 14, 18 How true are each of the following statements for you? (1) I often do things without stopping to think. (2) I am an impulsive person. (3) I often talk quickly, before thinking things out. (4) I often get involved in things I later wish I could get out of. (5) I need to use a lot of self-control to keep out of trouble. (6) I often get into trouble because I do things without thinking. (7) I get carried away be new and exiting ideas, but I don’t think of the possible problems.

Not at all true, a little true, somewhat true, pretty true, very true. Scored 1e5. Responses were summed and then divided by the number of items responded to, to create a continuous score. Higher scores indicate higher impulsivity.

<2 (median), �2 for univariate analysis

(continued on next page)

J.L. O’Loughlin et al. / Journal of Adolescent Health 54 (2014) 26e32 32.e1

Appendix 1 Continued

Potential predictor variable and reference

Data collected in survey cycle(s)

Item(s) Response choices/creation of score Recoded for use in analysis

Novelty seeking [4,5] 14, 18 Here are some things people may say about themselves. How true are each of the following statements for you? (1) I often try new things just for fun or thrills, even if most people think it is a waste of time. (2) When nothing new is happening, I usually start looking for something that is exciting. (3) I can usually get people to believe me, even when what I’m saying isn’t quite true. (4) I often do things based on how I feel at the moment. (5) I sometimes get so excited that I lose control of myself. (6) I like it when people can do whatever they want, without strict rules and regulations. (7) I often follow my instincts, without thinking through all the details. (8) I can do a good job of “stretching the truth” when I’m talking to people. (9) I change my interests a lot, because my attention often shifts.

Not at all true, a little true, somewhat true, pretty true, very true. Scored 1e5. Responses were summed and then divided by the number of items responded to, to create a continuous score. Higher scores indicate higher novelty seeking.

<2.7 (median), �2.7 for univariate analysis

Self-esteem [6] 12 For each of the following statements, indicate the response that best describes your situation. (1) I think I am someone who has something valuable to offer, at least as much as other people do. (2) I think I have a certain number of good qualities. (3) Everything considered, I tend to think I’m a failure. (4) I think I am capable of doing things as well as other people my age. (5) There’s little reason to be proud of myself. (6) I have a positive attitude toward myself. (7) I find it difficult to accept myself as I am. (8) Sometimes I think I’m really useless. (9) I’ve thought of myself as a good-for- nothing on occasion.

Not at all true, a little true, very true. Scored 1 e3. Responses were summed and then divided by the number of items responded to, to create a continuous score. Higher scores indicate higher self- esteem.

<2.8 (median); �2.8 for univariate analysis

Worry about weight [1] 1e20 During the past 3 months, have you been worried or stressed about your weight?

Not at all, a little bit, quite a bit, a whole lot No, yes

Cigarette package warnings make me afraid to smoke

13 How true is the following for you? Warnings on cigarette packages make me afraid to smoke.

Very true, a bit true, not at all true Not true, true

Cigarette advertisements make me want to smoke

13 How true is the following for you? Cigarette advertisements make me want to smoke

Very true, a bit true, not at all true Not at all true, a bit/very true

Felt as if you need a cigarette

1e20 How often have you felt as if you really need a cigarette?

Never, rarely, sometimes, often Never, rarely/ sometimes/often

Asthma 12, 16, 20 Do you have any of the following chronic health problems that have been diagnosed or confirmed by a doctor or other health professional? A chronic health problem means a health problem that has lasted or will probably last for �6 months: Asthma

No, yes e

Overweight [7] 1, 12, 19 Height and weight were measured by technicians trained according to a standardized protocol.Two measures of height to the nearest .1 cm and weight to the nearest .2 kg were obtained for each subject. If discrepancies >.5 cm for height or .2 kg for weight were observed, a third measure was obtained. The average of the two closest measures was recorded.

Body mass index was computed as weight (kg)/height (m)2. Participants whose body mass index was �85th age and sex- specific percentile were categorized as overweight

No, yes

Alcohol use 1e20 During the past 3 months, how often did you drink alcohol (beer, wine, or hard liquor)?

Never, a bit to try, once or a couple of times a month, once or a couple of times a week, every day

No, yes

J.L. O’Loughlin et al. / Journal of Adolescent Health 54 (2014) 26e3232.e2

Appendix 1 Continued

Potential predictor variable and reference

Data collected in survey cycle(s)

Item(s) Response choices/creation of score Recoded for use in analysis

Use other tobacco products

1e20 During the past 3 months, how often did you: (1) smoke a cigar or cigarillo; (2) use chewing tobacco or snuff?

Never, a bit to try, once or a couple of times a month, once or a couple of times a week, every day

No, yes

Light physical activity [8,9]

1e20 Now, think about the physical activities that you did last week from Monday to Sunday outside your regular school gym class. For each activity that you did for �5 minutes at one time, mark an X to show the day(s) on which you did that activity (list of 29 different activities).

No, yes for each activity, Monday to Sunday. Two of the 29 activities were designated light (i.e., <3 METs). Each activity was summed to create a continuous score (possible range: 0e14)

<1 (median), �1 for univariate analysis

Moderate physical activity [8,9]

1e20 Now, think about the physical activities that you did last week from Monday to Sunday outside your regular school gym class. For each activity that you did for �5 minutes at one time, mark an X to show the day(s) on which you did that activity (list of 29 different activities).

No, yes for each activity, Monday to Sunday. Twenty-one of 29 activities were designated moderate (i.e., 3e6 METs). Each activity was summed to create a continuous score (possible range: 0 e147)

�6 (median), >6 for univariate analysis

Vigorous physical activity [8,9]

1e20 Now, think about the physical activities that you did last week from Monday to Sunday outside your regular school gym class. For each activity that you did for �5 minutes at one time, mark an X to show the day(s) on which you did that activity (list of 29 different activities).

No, yes for each activity, Monday to Sunday. Six of 29 activities were designated vigorous (i.e., >6 METs). Each activity was summed to create a continuous score (possible range: 0e42)

�1 (median), >1 for univariate analysis

Team sports 1e20 Since September of this school year, did you belong to any of the following intramural or extramural school sports teams (teams that were not part of your regular gym class)? (list of 13 different teams). Now think about sports teams and lessons outside school. In the past 3 months, did you belong to a.? (list of 12 different teams).

No, yes (for each team or lesson) No, yes (�1 team)

Hours of television/ week

1e20 On weekdays, I usually watch ____ hour(s) of television a day. On weekends, I usually watch ____ hour(s) of television a day.

Number hours television/weekday, number hours television/weekend. Responses were summed to create a continuous score

�3.5 (median), >3.5 for univariate analysis

School tolerance [10]

16 (1) What is the school policy regarding students smoking on school property? (No smoking anywhere on school property; students in grades 9e11 can smoke outside the school building in designated locations; students of all grades can smoke outside the school building). (2) What is the school policy for teachers regarding smoking on school property? (No smoking anywhere on school property; teachers can smoke in designated location(s) outside the school building; other). (3) Does your school have any problems enforcing the current policy on smoking? (no; yes). (4) Have any outside speakers come to your school this year to talk about smoking with students? (yes; no). (5) A count of the number of cigarette butts lying on the ground in the school periphery (<100; 100e1,000; >1,000 butts).

Responses to each item were summed to create a tolerance score for each school. Lower scores indicated less tolerance of smoking.

�9.0 (median), >9.0 for univariate analysis

(continued on next page)

J.L. O’Loughlin et al. / Journal of Adolescent Health 54 (2014) 26e32 32.e3

Appendix 1 Continued

Potential predictor variable and reference

Data collected in survey cycle(s)

Item(s) Response choices/creation of score Recoded for use in analysis

Corner store tolerance [10]

16 To measure corner store of smoking in each of the 10 school neighborhoods, data on 7 indicators were collected by direct observation in commercial establishments located within a 1-mile radius of study schools. Indicators included: (1) proportion of establishments that sold tobacco products (range: 22%e54% in the 10 school neighborhoods); (2) proportion of establishments that sold tobacco products other than behind the counter only (range:11%e54%); (3) proportion of establishments with vending machines selling cigarettes (range: 0%e8%); (4) proportion of establishments permitting smoking inside the premises (range: 17% e50%); (5) mean number of no-smoking signs in each establishment (mean 1.0 e2.8); (6) proportion of establishments that did not have signs indicating that they do not sell cigarettes to minors (range: 0%e18%); (7) mean number of tobacco advertisements per establishment (range .5e2.0).

Study schools were ranked from 1 to 10 on each of the seven indicators. Tied values were assigned the highest corresponding rank. Ranks for each school were summed to create a corner store tolerance of smoking score, with lower scores indicating lower tolerance of smoking.

�.25 (median), >.25 for univariate analysis

Restaurant tolerance [10]

16 Data on 7 indicators were collected by direct observation in restaurants located within a 1-mile radius of study schools. Indicators included: (1) proportion of restaurants that sold tobacco products (range: 22%e54% in the 10 school neighborhoods); (2) proportion that sold tobacco products other than behind the counter only (range:11%e54%); (3) proportion with vending machines selling cigarettes (range: 0%e8%); (4) proportion permitting smoking inside the premises (range: 17%e50%); (5) mean number of no-smoking signs (mean: 1.0e2.8); (6) proportion that did not have signs indicating that they do not sell cigarettes to minors (range: 0%e18%); and (7) mean number of tobacco advertisements per establishment (range: .5e2.0).

Schools were ranked from 1 to 10 on each of the seven indicators. Tied data values were assigned the highest corresponding rank. Ranks for each school were summed to create a tolerance scores, with lower scores indicating lower tolerance of smoking.

�.4 (median), >.4 for univariate analysis

CEGEP ¼ Collège d’enseignement général et professionnel/General and Vocational College; MET ¼ metabolic equivalent of task.

J.L. O’Loughlin et al. / Journal of Adolescent Health 54 (2014) 26e3232.e4

References

[1] Deschesnes M, Schaefer C. Style de vie des jeunes du secondaire en Out- aouais, Tome 1. Hull, Quebec, Canada: Direction de la Santé Publique. Régie Régionale de la Santé et des Services Sociaux de l’Outaouais et les Centres Jeunesse de l’Outaouais; 1997.

[2] Dugas E, Low NC, Rodriguez D, et al. Early predictors of suicidal ideation in young adults. Can J Psychiatry 2012;57:429e36.

[3] Kandel DB, Davies M. Epidemiology of depressive mood in adolescents: An empirical study. Arch Gen Psychiatry 1982;39:1205e12.

[4] Wills TA, Windle M, Cleary SD. Temperament and novelty seeking in adolescent substance use: Convergence of dimensions of temperament with constructs from Cloninger’s theory. J Pers Soc Psychol 1998;74:387e406.

[5] Eysenck SB, Eysenck HJ. The place of impulsiveness in a dimensional system of personality description. Br J Soc Clin Psychol 1977;16:57e68.

[6] Vallieres EF, Vallerand R. Translation to French (Canadian) and validation of Rosenberg’s self-esteem scale. Int J Psychol 1990;25:305e16 [in French].

[7] Evers SE, Hooper MD. Dietary intake and anthropometric status of 7 to 9 year old children in economically disadvantaged communities in Ontario. J Am Coll Nutr 1995;14:595e603.

[8] Pate RR, Pratt M, Blair SN, et al. Physical activity and public healthdA recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA 1995;273:402e7.

[9] Bélanger M, Gray-Donald K, O’Loughlin J, et al. When adolescents drop the ball: Sustainability of physical activity in youth. Am J Prev Med 2009;37: 41e9.

[10] Pabayo R, O’Loughlin J, Barnett TA, et al. Does intolerance of smoking at school or in restaurants or corner stores decrease cigarette use initiation in adolescents? Nicotine Tob Res 2012;14:1154e60.