Factors associated with different smoking status in European adolescents: results of the SEYLE study

The multi-center study “Saving and Empowering Young Lives in Europe” (SEYLE) was initiated to evaluate school-based prevention measures for risk behaviors and suicidality in youths. Study design and characteristics of the sample have already been published [57, 58]. In total 168 schools, comprising a sample of 12,395 adolescents from 11 countries (i.e., Austria, Estonia, France, Germany, Hungary, Ireland, Israel, Italy, Romania, Slovenia and Spain), with Sweden as the coordination center, were included. In each country eligible schools were randomly selected and ethical permissions were obtained from local ethical committees. Sixty-seven of the 12,395 participating adolescents were excluded, based on missing relevant data and thus a total of 12,328 adolescents [mean age: 14.9 years, range 13–17; number of female/male participants: 6799 (55.2%)/5529 (44.8%)] were included in further analyses.

Baseline evaluations were conducted between October 2009 and December 2010. The baseline structured self-report questionnaire was submitted to adolescents aged between 13 and 17. It contained socio-demographic items (e.g., sex, age, religious affiliation, etc.), risk behaviors and mental symptoms. Family setting was assessed by the question: “We would like to know about your family. Please answer this question about your home, where you live permanently or most of the time and put down the people who live with you at your home”. Questions from the Global school-based student health survey (GSHS; [59]) were applied to determine risk behaviors, such as alcohol or drug consumption. The following item was used to assess alcohol consumption in youths: “How often do you have a drink containing alcohol? For example, 0.33 l beer or cider; glass of wine or 4 cl of strong alcohol” and drug use was investigated by the question: “During your life, how many times have you ever used drugs”. Alcohol consumption of family members was assessed by: “Have you ever seen a family member when they are drunk?”. GSHS-Items were also used to identify adolescent and parental smoking behavior. Juvenile smoking behavior was assessed by the questions “Have you ever smoked cigarettes?” and “How many cigarettes did you smoke per day, during the last 6 months?”, whereas parental smoking behavior was gathered by the question “Which of your parents or guardians use any form of tobacco?”. Emotional symptoms, conduct problems and hyperactivity were investigated by questions of the Strengths and Difficulties Questionnaire (SDQ, [60]). Anxiety was assessed using the Zung Self-Rating Anxiety Scale (SAS; [61]) and suicidality (i.e., previous suicide attempts) was ascertained employing the Paykel Suicide Scale (PSS; [62]).

There are various terms to define self-injurious behaviors such as non-suicidal self-injury (NSSI), deliberate self-harm (DSH) and direct self-injurious behavior (D-SIB). NSSI refers to deliberately inflicted damage to one’s body without suicidal intent [63]. The term deliberate self-harm is a more comprehensive term for self-injurious behaviors [64]. DSH is defined as intentional self-injury or self-poisoning irrespective of with or without suicidal intent [65]. DSH also includes indirect damage to one’s body such as severe substance abuse or taking overdose [54]. The term DSH so far was used mainly within Europe and Australia, whereas many researchers within Canada and the United States prefer the term NSSI [64].

The present paper focuses on D-SIB, which is defined as self-inflicted damage to an individual’s body on purpose, regardless of with or without suicidal intent. It includes direct damage such as self-cutting, -burning, -biting, -hitting, and skin damage by other methods [54]. To investigate adolescents engaging in D-SIB, a 6-item questionnaire was used. This 6-item questionnaire is based on the 9-item Deliberate Self-harm Inventory (DSHI) questionnaire from Bjärehed and Lundh [66], which is a shortened version of the 16-item DSHI by Lundh et al. [67] that originated from the original 17-item DSHI by Gratz [68]. This modified version tends to assess direct self-injurious behavior (D-SIB) to one’s body only and self-injurious acts were combined to shorten and simplify the measure. However, it contains the same facets on frequency, severity and duration [54].

Statistical analysis was performed using IBM SPSS and GraphPad Prism.

Dichotomous variables were constructed for alcohol consumption (at least twice a week vs. no to low use), drug consumption (at least three times in life vs. no to low use), family drinking (seen family member drunk at least sometimes vs. never seen family drunk), parental smoking (no parent smokes vs. one parent or both parents smoke), family setting (broken homes—one birth parent with or without step-parents vs. both parent households—both birth parents), and previous suicide attempts (lifetime history of suicide attempts vs. no previous suicide attempts).

Group variables were generated for adolescent smoking status (non-smoker—never have smoked cigarettes; non-daily smoker—have ever smoked or do smoke occasionally but do not smoke daily; daily smoker—smoke daily at least one cigarette) and onset of smoking (younger than 10 years; younger than 14 years; 14 years or older).

The relationship of different adolescent smoking status with internalizing, externalizing and family problems was first analyzed using Chi-square test. Group differences were tested by Kruskal–Wallis H test (i.e., nonparametric analysis of variance) followed by post hoc Mann–Whitney U tests for group comparisons. Multinomial logistic regression analysis was used to verify the association between different adolescent smoking status and internalizing, externalizing and family problems controlling for covariate risks. The level of statistical significance was predefined at p < 0.05.

Another association with adolescent daily and non-daily smoking was found for D-SIB.

Chi-square test revealed a significant difference between defined adolescent smoking groups and D-SIB (χ ² = 640.208; df = 2; p < 0.001). In the non-smoking group, 7.9% reported D-SIB, whereas in the daily smoking group 26.5% of the adolescents showed D-SIB. Further, 56.4% of adolescents with D-SIB and only 26.5% of adolescents without D-SIB reported daily smoking (p < 0.001).

Further group comparisons revealed statistically significant differences between different adolescent smoking status and D-SIB (p < 0.001). The largest differences emerged between the non-smoking group and the daily smoking group (effect sizes; DSHI, r = 0.29). Smaller but still significant differences were found between non-smokers and non-daily smokers (DSHI, r = 0.12) and between non-daily smokers and daily smokers (DSHI, r = 0.14). Non-daily and daily smokers obtained significantly higher mean scores on the DSHI compared to adolescent non-smokers.

Multinomial logistic regression analysis revealed that with each increase on the DSHI the odds of belonging to the adolescent daily or non-daily smoking group than to the non-smoking group rose about 36.5 or 22.8% (p < 0.001) for both genders, all age groups and countries. Even after controlling for influencing factors these findings remained significant (Table 3). With increasing D-SIB scores the likelihood of belonging to the daily or non-daily smoking group rather than to the non-smoking group was now about 19 or 15% and still significant (p < 0.001).

The sooner adolescents started smoking, the more they smoked at a later point in time and the higher were their mean scores for smoking frequencies. Adolescents who started smoking below 10 years reported 7 cigarettes per day on average, whereas adolescents that reported the onset of smoking under the age of 14 smoked on average 6 cigarettes per day. Adolescents who started smoking only after 14 years reported smoking four cigarettes per day (Fig. 1). Further, the earlier adolescents started smoking the higher the scores they reached on the DSHI score. Group comparisons revealed statistically significant differences between different onset of smoking groups and D-SIB (p < 0.001).

As a very high percentage of youths reported the onset of smoking under 14 years, our data revealed significant differences between different onset of smoking groups and DSHI scores. Many studies state that the early onset of regular tobacco use is predictive of psychiatric and psychological health problems like depression, substance consumption as well as personality disorders [4‐7]. Thus, our findings suggest that the early onset of smoking is associated with D-SIB. Our data further verify a coherence of adolescent smoking behavior (especially daily smoking) and D-SIB. Klonsky and Muehlenkamp [49] also argue that self-injurious behavior and substance abuse have to be strongly associated, because both include causing physiological harm to the body. In addition, prior studies found similar correlations between smoking and self-injurious behavior in youths [45‐48, 69], arguing that similar psychological processes underlie these behaviors [49] or that substance use helps individuals to habituate to SIB [50].

Further, our data show a relationship between adolescent smoking and psychological issues like anxiety, emotional symptoms and previous suicide attempts. Several studies found similar associations corroborating, for example the relationship between adolescent smoking and anxiety. Patton et al. [39] already stated that subjects reporting high levels of anxiety were twice as likely to be smokers and that anxiety predicted the initiation of experimental smoking [39]. Further, Johnson et al. [36] suggested that early cigarette smoking may increase the risk of certain anxiety disorders during late adolescence. Recently, Cavazos-Rehg et al. [32] reported that smoking cessation is associated with risk reduction of anxiety even among smokers who have had a pre-existing disorder. This is in contrast to the hypothesis that smoking might function as a sort of compensation or self-medication and suggests the assumption of juvenile smoking as a risk factor for anxiety disorders.

Regarding emotional symptoms we found interesting different results. Whereas Chi-square test revealed positive associations between emotional symptoms and adolescents smoking, multinomial logistic regression analysis showed negative associations between these two topics. These findings in turn support the hypothesis that smoking might function as a sort of self-medication. Also Audrain-McGovern and colleagues [70] provide evidence for self-medication processes in the relationship between adolescent smoking and depression.

The association of smoking behavior and suicidality has already been demonstrated in several studies [31, 33‐35, 37, 38, 40, 41]. Our findings partially confirm these previous results for a large sample of European adolescents.

To summarize, our data show a relationship between adolescent smoking and psychological issues like anxiety, emotional symptoms, previous suicide attempts and D-SIB.

The present paper further suggests that hyperactivity as well as conduct problems increase the risk for smoking in adolescents. Previous studies have documented that adolescents with attention-deficit/hyperactivity disorder (ADHD) or conduct problems are at increased risk of substance use problems [25‐28]. Chang et al. [29] found that especially hyperactivity and impulsivity predict early onset of tobacco and alcohol use in youths. Recently, Roberts et al. [30] also stated that specifically hyperactive symptoms were associated with alcohol and nicotine use in young adults, underlining that increased levels of impulsivity are thought to contribute to their increased levels of risk.

Our results further show that there exists a relationship between adolescent smoking and several other risk factors. The strongest associations with juvenile smoking were observed between other substance consumptions like drug and alcohol use. Several studies ascertained strong significant associations between these topics as well and therefore corroborate our results [22‐24].

Thatcher and Clarke [71] proposed that psychological dysregulation can be seen as a predictive phenotype for substance use disorders in adolescents. It is characterized by cognitive, behavioral and emotional difficulties in childhood, connecting heritable predispositions and early environmental influences to later substance use disorders. During adolescence, brain circuits including those for motivation, reward and decision-making are not yet fully developed, thereby making this group more susceptible to addiction [72].

We further found that family behavior and issues like parental smoking, family drunkenness and living in broken homes influences adolescent smoking behavior. Previous studies confirm the influence of family factors on adolescent smoking behavior. Especially the relationship of parental smoking or tobacco use and adolescent smoking is reported to be very high [17, 73]. Bauman et al. [15] already verified that lifetime parental smoking was as strongly correlated with adolescent smoking as peer smoking. Farkas et al. [74] found that parental smoking cessation keeps adolescents off smoking and that the earlier parents quit smoking, the less likely their children are to become smokers. In addition to that, Almutairi [75] showed that parents had been perceived to be the first source for smoking. Further, Nosa et al. [19] found that exposure to smoking at home seems to be an important risk factor for ever-smoking. Compared to prior findings we found strong effects regarding this topic. Adolescents with smoking parents had a higher likelihood for belonging to the daily smoking group. Furthermore, Griesbach et al. [20] found that family structure (like intact vs. stepfamilies) is significantly associated with smoking among 15-year-olds in seven European countries. Recent findings of Du et al. [21] in multicultural students of Hawaii also show that family structure could be a serious risk factor for smoking among young people. These results demonstrate how important and necessary it is to raise awareness in parents and legal guardians of the damaging influence of their behavior on the smoking behavior of their children to ideally cause lifestyle changes and to make their homes smoke-free.

It is important to mention that our data do not imply causality, meaning that they do not give an explanation about cause and effect. This limitation of the study is due to the cross-sectional design, where a direct link between determinants and outcome is missing. Concerning this matter further research will be needed. For example, does smoking itself represent a form of self-injurious behavior or does it rather function as a compensation of existing problems?

Another limitation of this analysis would be the missing cross-cultural comparison. Our study comprised adolescents from 11 different countries from all across Europe. Of course there exist cultural differences between the majority of countries, which might also influence adolescent smoking behavior. Future analysis of differences across countries might therefore help to develop country and culture-specific treatment programs. Nevertheless, controlling for country in our regression analysis did not attenuate the effect of the influencing factors we found.

The major strength of this study is its large sample size together with the random selection of schools across 11 representative study sites in Europe. A huge body of literature corroborates our results, most of which are studies from North America. Therefore, to our knowledge, this study is the first investigating factors associated with adolescent smoking in Europe on such a large scale.