Predictors of smoking among Swedish adolescents

Tobacco consumption is a global pandemic, killing half of all lifetime users. In 2011, six million people died as a result of tobacco use, and by the year 2030 eight million people is expected to die annually [1]. According to the WHO study Health Behaviour in School-aged Children (HBSC), approximately 18% of 15 year old adolescents smoke cigarettes at least once a week, corresponding figures in Sweden are 15% among girls and 13% among boys [2]. Nicotine addiction is usually developed during adolescence and has been referred to as a ‘paediatric disease’ [3]. After the age of twenty, the risk of starting to smoke on a regular basis decreases [4], suggesting that if adolescents can be kept tobacco free, most of them will never start [5].

At baseline (7th grade) the mean age of participants was 12.8 years. The prevalence of smoking was 3.3%, increasing to 25.1% at follow-up five years later. Further, 8.1% reported ever using snus, increasing to 37.5% at follow-up. Boys reported higher snus use than girls, and also higher use compared to their own smoking. Figure 2 presents the development of smoking and snus use by gender. At baseline, 2.8% of the boys and 1.4% of the girls reported combined smoking and ever using snus, increasing to 17.2% of the boys and 19.1% of the girls at follow-up. Univariable logistic regression analysis (Table 2) showed several possible significant predictors of smoking at follow-up: female sex, lower parental education, poorer family mood, poorer self-rated health, poorer self-esteem, less negative attitude towards smoking, binge drinking, snus use and smoking. In a multivariable logistic regression analysis (Table 2) four predictors remained independently related to smoking: female sex, poorer self-esteem, less negative attitude towards smoking and ever using snus.

The study confirms previous findings of a multi-factorial explanation to initiation and persistence of smoking. Health attitudes, family and intrapersonal factors, binge drinking, smoking and ever using snus all predicted future smoking.

The results showed that poor self-esteem predicted future smoking. As being concluded by McGee and Williams (2000) longitudinal studies addressing this topic are few and the findings are inconclusive. Self-esteem is a concept closely related to self-efficacy, which according to Bandura (1995) is central in models attempting to explain addictive behaviours [34]. In a short-term longitudinal perspective, Engels and colleagues (2005) confirms that self-esteem in the early years of adolescence is linked to smoking in subsequent years among girls [15].

The Theory of Planned Behaviour suggests that behaviours can be predicted by a person’s intention to enact the behaviour. Since attitudes and intentions are closely related, our finding that less strong anti-smoking attitude increases the risk of smoking initiation is in line with this theory. The follow-up time in our study was relatively long; anti-smoking attitudes had a strong association with smoking five years later. Thus, our findings suggest that attitudes towards smoking in early adolescence are related to behaviour later in life.

Initial health attitudes were measured when the adolescents were entering teenage, and only a month earlier had moved to a school level that among adolescents in Sweden is associated with teenage values and growing independence. Therefore, at the time of the first questionnaire, the study cohort may not have fully adopted ‘teenage norms’, i.e. they were probably still much influenced by norms of significant adults, e.g. their parents. We were unable to study the impact of parental smoking since this information was unavailable. It is, however, well documented that parental smoking is a risk factor for adolescent smoking. In a Swedish study with focus group interviews of adolescents, one central theme was that concerned adults make a difference [12].

In this cohort, snus use in early adolescence was a strong predictor for smoking in late adolescence. However, a large drop-out of snus users between baseline and follow-up adds some uncertainty to this finding. The results are consistent with previous findings that use of smokeless tobacco facilitated smoking at follow-up among high school students in California [20], and partly consistent with findings from a Norwegian cohort study of young men in which snus use at baseline elevated the risk of a combined use of snus and smoking at follow-up [22]. Thus, in both studies snus facilitated smoking. The Norwegian study also found that a combined use at baseline implied a higher risk for combined use (daily smoking and daily snus use) at follow-up. As all boys who reported smoking at baseline in our study also reported ever using snus, this indicates a widespread combined use. However, due to small numbers of smokers and snus users at baseline, it is not feasible to divide tobacco use into smokers, snus users and combined users as the groups get too small for proper analyses.

Use of smokeless tobacco varies around the world, but in a number of countries it counts for a substantial part of the total tobacco consumption. In Sweden, health authorities regard snus a less serious health hazard than cigarette smoking [35]. Based on our results, the most important health hazard of snus use could be the early introduction of nicotine and by that the risk of nicotine addiction. A recent Swedish study found symptoms of nicotine dependence to be more common among adolescents who used both cigarettes and snus [36]. An implication should be that preventive efforts against snus should be introduced in an early age.

The higher risk for girls to become smokers, prominent in many high-income countries, was confirmed in this study. However, it should be noted that the loss to follow-up was significantly higher for boys, making the interpretation of the finding uncertain.

The study suggests implications for preventive actions of smoking among adolescents. The predictive value of attitudes imply that actions influencing smoking attitudes during the first six school years should be a good investment. This is consistent with the conclusion from Edvardsson and colleagues [37] stating that, in order to influence attitudes, preventive actions need to be established well before tobacco is introduced. Modelling and social reinforcement are core concepts for the understanding of factors promoting smoking [38]. Flay and colleagues [39] investigated the influence of parental and friends’ smoking on adolescents’ initiation and continuation of smoking. They conclude that parental approval of smoking, i.e. having a positive attitude, plays a significant role, more so for girls than for boys.

The strengths of the study are the longitudinal design, structured recruitment of schools with no schools refraining from participation, a moderate drop-out rate, and a relatively long follow-up time. A limitation is the higher drop-out rate among boys, adolescents born outside Sweden and snus users. Another limitation is a low statistical power to examine smoking as a predicting factor, especially in the multivariable analysis, due to the low prevalence of smoking among 12–13 year olds. This could explain why smoking in early adolescence was not significantly associated with smoking later in adolescence although the OR was high. Due to the low number of smokers it was not possible to explore subgroups of smokers (e.g. smokers divided into occasional and daily smokers and combined users). Further, the study was conducted in Sweden, a high-income country scoring high in health and social indexes, thus affecting the generalization of the results to countries with other living conditions. As noted above, a larger sample might have been able to demonstrate further significant results. One example is country of birth, where the comparatively small number of adolescents born in other countries did not allow stratification based on native country. In the multivariable analysis, important factors associated to smoking in previous studies have been included. However, some factors that previously have been found associated with adolescent smoking were not included in the questionnaire, e.g. parents smoking and smoking attitudes, and peer smoking. Furthermore, when discussing the finding regarding attitudes, it should be noted that we cannot know if the intention has varied during the elapsed time between initial assessment of the attitude, nor know what may be the causal link between attitude and subsequent behaviour. Since the beginning of this study there has been some changes in the legislation; in 2005, smoking bans in restaurants, ban of advertising outside stores selling tobacco products and ban of cigarette sales in packets with fewer than 19 cigarettes was introduced. Despite these new legislations, Swedish official data show smoking habits in upper secondary school to have been fairly stable since the end of our study in 2008 [24]. Limitations regarding the wording of the smoking question should also be noted. Adolescents’ perceptions of time, as it was defined in the question (‘these days’) might have changed between 12–13 years of age and 17–18 years of age. Hence, the possibility of time-dependent misclassification must be acknowledged. Furthermore, we lack information regarding previous smoking (information that we have regarding snus), nor was smoking status confirmed by objective measurements. However, in most studies, self-reports have been shown to be a reliable proxy for actual smoking [40].

The results of this study show that in early adolescence, attitudes towards smoking, family and intrapersonal factors, gender, binge drinking, smoking and ever using snus can predict persistent smoking in later adolescence. The study stresses the importance of actions to strengthening adolescents’ self-esteem, of promoting anti-smoking attitudes in early adolescence, as well as avoidance of early initiation of snus use, implying a need of action by parents, schools, youth associations and legislating authorities.