Cigarette smoking is one of the major causes of morbidity and premature death in countries with advanced economies [1, 2]. Research among adult smokers reveals that between 80 and 90% of them began to smoke before 18 years of age [3, 4]. Understanding the social context of adolescent smoking is an important step in designing effective smoking prevention efforts. This has recently been emphasized in writings discussing epidemiological methods to assess risk factors for health risk behaviors such as smoking in adolescence [5–7].

Several studies have revealed that multiple social factors influence the likelihood of smoking during adolescence, the period during which the onset of smoking usually occurs [8–10]. These include family structure [11], socio-economic status [12], perceived parental expectations [13], parental support [8], parental monitoring [13], parental practices [14], parental smoking behavior [2], sibling smoking behavior [15], peer smoking [4], perceived acceptance of friends [13], use of other legal and illegal substances [16], and participation in physical activity [17]. Many of these factors have furthermore been illuminated in previous Icelandic studies [8, 17]. Recently, researchers have called for a more comprehensive modeling of adolescent smoking behavior by including more variables in analyses in order to more clearly understand the complex nature of the social influences on adolescent smoking [18].

In the present study, we analyzed the relationship and relative importance of a broad spectrum of social variables in adolescent smoking in Iceland, a Nordic country with high per-capita income. We included in our analyses several background factors, relationship with parents, family conflict, smoking of significant others, and personal attributes such as participation in physical activity, academic achievement, and substance use. Thus, the aim of our analysis was to investigate how multiple social factors are related to adolescent cigarette smoking, to assess important social features that increase or decrease the likelihood of adolescent smoking, and to demonstrate the importance of the complex nexus of adolescent smoking.

Of all student included in the study base, essentially the entire population of Icelandic 9^th and 10^th graders, 81% responded to the questionnaire. A validity check on the remaining 19% of the population that did not participate in the study (did not attend schools this particular day) revealed no particular difference between non-participants and participants. Population characteristics of the study sample are presented in Table 1. For discrete responses the number of items are shown and the relevant Cronbach's Alpha (CA) calculation of internal consistency.

The family conflict variables, "interparental serious verbal conflict," and "parent-adolescent violence" were only related to smoking in the crude analysis but not in the adjusted models for both daily and occasional smoking but "interparental violence" was borderline significant in the adjusted analyses for occasional smoking. However, the "parent-adolescent serious verbal conflict" variable was significant in the crude- and adjusted analysis for both daily smoking (adjusted OR = 2.83, 95% CI = 1.98–4.04) and occasional smoking (adjusted OR = 1.72, 95% CI = 1.27–2.33). In addition, father's, mother's, and sibling's smoking all increase the risk of adolescent daily smoking in the crude analysis but father's smoking was not significant in the adjusted analysis whereas mother's and sibling's smoking continues to be of importance. Of all the family factors in the study "perceived parental reactions if one would smoke" is the most important risk factor both in the crude and adjusted analyses (adjusted OR for daily smoking = 8.92, 95% CI = 6.04–13.17). Expecting more respect from peers if one would smoke increases the risk in the adjusted analysis for daily smoking (OR = 3.27, 95% CI = 2.09–5.13), and occasional smoking (OR = 2.25, 95% CI = 1.43–3.53).

Moreover, the risk of being a daily smoker increases to over 17-fold when peers smoke when adjusted for other factors in the analysis (OR = 17.17, 95% CI = 6.18–47.71). This is indeed the strongest risk factor for daily smoking in the study. Finding healthy nutrition important was not a significant factor for daily smoking after adjustment. To the contrary, engaging in more physical activity (adjusted OR for daily smoking = 2.15, 95% CI = 1.48–3.13) and achieving better in school (adjusted OR for daily smoking = 2.12, 95% CI = 1.49–3.02) were negatively related to smoking. However, academic achievement was not significantly related to occasional smoking in the adjusted analysis. Drinking alcohol during last 30 days was strongly related to daily smoking (OR = 7.64, 95% CI = 5.11–11.41) and occasional smoking (OR = 8.81, 95% CI = 6.13–12.64), and smoke-free tobacco use was strongly related to both daily (OR = 5.74, 95% CI = 3.93–8.37) and occasional smoking (OR = 3.57, 95% CI = 2.56–4.98) after adjustment.

We found that the smoking behavior of friends and their attitude toward smoking were social factors strongly associated with adolescent cigarette smoking. Another strong force protecting adolescents from smoking was parents attitudes towards smoking, the quantity of time spent with parents, and absence of "serious" verbal conflict between parents and child. Alcohol consumption during the previous 30 days greatly increased the risk of smoking both in the unadjusted and adjusted analysis for both daily and occasional smoking.

In this study demographic factors seem to be of little importance for adolescent smoking, probably reflecting the homogeneity of the Icelandic population and relatively well developed social security system. Parental support, parental control and time spent with parents are all protective factors of adolescent smoking in the crude analysis but only time spent with parents continues to be of importance in the adjustment models. This is consistent with prior knowledge from Iceland and is presently one of the cornerstones in the Icelandic Model of substance use prevention (a detailed paper describing this prevention approach is currently in submission). Moreover, this study supports the notion that the quantity of time parents spend with their children may be a major protective factor for adolescent smoking. But the adolescents' perception that parents are not against adolescent smoking greatly increases the risk of being a daily smoker, independent of other factors (OR = 8.92, CI = 6.04–13.17). However, since this study was carried out with a cross-sectional design, this finding may reflect that parents are perceived to tolerate the smoking of adolescents rather than being a risk factor for smoking initiation. Yet, data from other studies indicate that parental expectations do function as a protective factor for the initiation of adolescent smoking [13], which is consistent with our preferred interpretation of these findings.

Family conflict has been reported to be strongly related with increased risk of adolescent smoking [22, 23]. The social context of this relationship is not fully understood. Findings in this study indicate that it may particularly be serious verbal conflicts between the adolescent and his/her parents that is of importance, being the only family-conflict variable of the four (with interparental serious verbal arguments, interparental violence, and parent-adolescent violence) that remains significantly related to daily smoking after adjustments for other variables in the analysis. This issue will need further study in the future.

Although mothers' and sibling smoking remained significantly related to adolescent smoking after adjustments, and this relationship has shown to be of importance in previous studies [24], the impact was relatively small in comparison to other factors with adjusted odds ratio of 1.69 (CI = 1.20–2.40) for mother's smoking and 1.55 (CI = 1.10–2.18) for sibling's smoking. Reporting having friends that smoke increased the risk of being a daily smoker 17-fold when adjusted for other variables in the analysis. This was by far the strongest risk factor in the study after adjusting for other factors. The reason for this high odds ratio is the group division of our dependent variable. Most public health studies on adolescent smoking divide participants into "never" and "ever" smokers. In this study, those who smoke daily are distinguished from those who smoke occasionally and both groups are analyzed against those who do not smoke at all. The present finding nevertheless supports previous research that the smoking of peers is of central importance for adolescent smoking [8, 17, 25]. Recently, however, this notion has been doubted by Arnett [26] who argues that we need to look more closely into the peer-selection process to fully understand the important role of peer smoking in adolescent smoking.

One problem with interpreting the role of peer influences is that we have no data on the order of initiation to smoking in the peer group and consequently do not know who influenced whom. Notwithstanding, what we can safely say is that birds of a feather tend to flock together when it comes to adolescent tobacco smoking. Also, by including in the model a measure of perceived respect from peers, if one would smoke cigarettes, in addition to the smoking of peers, we may be able to analyze further the influence of the peer group. In this study, if smoking cigarettes was perceived to increase peer respect, the odds of daily smoking was 3.27 in the adjusted analyses, down from an OR = 9.96 in the crude analysis which is probably because of other protective factors in the model.

Finally, higher levels of physical activity and academic achievement were protective factors to the risk of adolescent daily smoking after adjustment. This is consistent with the findings of previous studies [8]. In contrast, alcohol consumption during the last 30 days and having used smoke-free tobacco greatly increase the risk of daily and occasional smoking, underlining the fact that it is often the same individuals who engage in risk behaviors of more than one kind.

Due to the study's cross-sectional design we were not able to draw any firm conclusions regarding causality between the dependent and independent variables. Also, small adjusted odds ratio values in a sample of this size (7,430 respondents) should be interpreted with caution, particularly when the 95% confidence intervals are close to, but do not include, 1.0 [27].

The present findings indicate that multiple social factors are related to adolescent smoking. Important factors protecting adolescents from smoking included higher academic achievement and physical activity. Abstinence from other tobacco products and alcohol were also protective factors; however, the strongest protective factors comprised peers abstinence from smoking and their perceived negative attitudes to smoking. Also, parent's negative attitudes to smoking, absence of serious verbal arguments with parents and the quantity of time spent with parents, were highly protective against adolescents smoking in this study.