Tobacco smoking, the main cause of preventable morbidity and death worldwide, is a behavior that starts in adolescence for 90% of the adults who self-report smoking. Serious consequences of smoking, such as cancer and cardiovascular disease, are known to occur later in life [1]. Studies have indicated to various smoking-associated factors in adolescence such as depression, suicidal ideation [2] and the abuse of other substances [3].

In Europe, the prevalence of past month use of cigarettes among high school students was remarkably different between genders, varying from 15% of boys in Iceland to 48% of girls in Austria, with a large number of countries showing higher cigarette consumption among girls [4]. Although there has been a recent decrease in tobacco consumption among high school students in the United States (from 28.3% to 12.6%), boys in older age categories did have slightly higher tobacco consumption [5]. In Brazil, studies conducted among middle and high school students revealed a 10-16% prevalence of cigarette consumption in the last month, with gender differences in different cities. Like the U.S., there has been a decrease in regular tobacco use over the last ten years [6, 7].

Diverse psychosocial factors have been associated with the use of cigarettes by adolescents; including: peer pressure, poor parental supervision or familial relationships [8–10], smoking among parents, siblings or partners [11], leisure activities, especially those that involve the company of friends such as nightlife and aspects of religiosity [12].

Evidence from North America suggests that there are gender differences in factors associated with smoking. Boys with strong religious involvement showed slower progression from experimentation to regular smoking. Meanwhile, a positive familial relationship, including care and family activities, protected girls from the first use of tobacco [13]. Studies based on developmental theories suggest that boys and girls may perceive and react to familial influence in different ways and thereby are exposed to different risk factors for drug consumption [14]. These latter studies also suggest the need for gender-specific prevention programs.

Few studies have investigated the relationship between these factors. Recent studies suggest the need to concurrently evaluate the impact of religious and familial themes on drug consumption by adolescents because these are integrated issues with mutual influence [15–17]. Other studies also suggest that familial factors are modified by leisure activities or partner influence when evaluated with regression models [18, 19].

This study aims to verify the association among familial factors, religious factors and leisure time activities with smoking by gender among adolescent private school students in Sao Paulo, Brazil, based on the Social Development Model [20]. The Social Development Model presupposes theories of control and social learning, constituting a picture of pro and antisocial behaviors [21, 22]. According to this model, adolescents tend to engage in antisocial behaviors when they develop bonds with groups with antisocial beliefs (e.g., drug-using peers) and vice-versa. In addition, external factors such as family relationships and parental monitoring may interfere in the process of bonding with bad behavior peers [21, 22].

The relationship between psychosocial factors and smoking may vary by culture [13, 14]. The majority of studies have been conducted in North American and European populations [23–25], and studies focusing on Latin American, Asian and African populations are rare. This knowledge gap calls for studies in these cultures.

This study aims to examine gender differences in familial, religious and leisure factors in a sample of adolescents in Sao Paulo, the largest city in Latin America.

The results of this study confirmed the importance of familial, religious and leisure factors on smoking behavior among adolescents of both genders. For both genders, older age, not having a religion, living with someone who smoked and the frequency of nightlife with friends were associated with smoking behavior. However, the familial and religious factors identified for girls were different from the ones observed for boys. For girls, receiving care and attention from parents and attending religious meetings were also inversely associated with CS. In contrast, for boys, the death of one or both parents was associated with CS.

In this study, we did not observe gender differences in the one-month CS prevalence. This result differed from studies performed at public schools in São Paulo, which suggested that the CS prevalence was higher among girls [6]. This also differs from findings in other countries, which show a higher prevalence of CS among boys [19, 29]. This suggests that the CS prevalence even within the same city may vary between genders according to socio-cultural contexts.

Despite the similar prevalence of smoking in both genders, there were gender differences in factors associated with CS. Familial relationships and parental monitoring were unevenly associated with smoking by gender. Despite the finding that second-hand smoke exposure at home increased by 50-70% the likelihood of recent tobacco consumption in both genders, which is in agreement with the majority of the literature [1, 11], the type of familial interaction that may influence smoking differed by gender. Perhaps, due to the characteristics of the female gender, girls found it important to be able to recognize their parents' attention and care. In this context, O'Byrne et al. [10] suggested that tobacco prevention programs aimed at adolescents must be focused on improving the relationship between parents and children. In their study with 7^th to 12^th-grade adolescents, a healthy familial relationship was the main protective factor for the first use of tobacco. However, it is still not clear whether this association between familial relationship and adolescent smoking is modified by parents' use of tobacco [30] or not [31]. Nevertheless, in studies on tobacco experimentation, this association has a stronger influence on girls than boys [13].

For boys, the death of one or both parents was associated with tobacco use. This finding suggested that more important than the parent-son relationship is the need for both parents to be alive, even if they may be divorced. A study performed in 11 European countries showed that the parents marital status only influenced the initiation of smoking when divorce was rare in the country [24]. With that in mind, the only rare family structure status in our sample was the death of one or both parents (about 4% of the total sample).

Regarding religiosity, studies that sought to comprehend the role of religiosity in drug use suggested that there was a distinction among internal factors (religion membership, believing in God and the importance given to religion), as well as external factors (frequency of attendance at religious groups) [32–35]. However, the majority of the studies that sought to assess the role of this external expression of religiosity in drug use have not distinguished attendance frequency at conventional religious activities, such as services and masses, from the attendance frequency at youth religious meetings as we did [32]. In our study, the only religious factor that was associated with CS was the latter, which occurred only among girls. For both genders, not having a religion increased the odds of CS. Yet, among boys none of the internal or external religiosity variables were associated with recent CS. A possible explanation for this finding is the fact that in Brazil it is rare to openly discuss smoking during religious services, and therefore, these interactions would have added little to the adolescents' decision to smoke cigarette [36]. Nevertheless, the finding among boys is in agreement with results from a longitudinal study of 12-18 year-old British adolescents [13]. Among girls, merely having a religion did not seem to protect them from CS; instead, the network of friends in the religious group had a more protective effect because a statistical significance was found only for the religious factor that represented the public expression of religiosity in a group. This points to the group role once again as a mediator for the use or nonuse of tobacco specifically in girls. Perhaps, girls who attended adolescent religious groups had friends who did not use tobacco and who disapproved of this behavior [15].

In this context, we should highlight the important role of religion in the psychosocial development of adolescents, a domain of the developmental process that is not always included in the health literature [37]. In this developmental perspective, religiosity, family and friends merge to form a foundation for decision making: to experiment or not to experiment tobacco. Religion seems to develop as a "social controller" through its moral standards. This role places drugs, including cigarette smoking, in a category of reprehensible actions [38], in other words, actions that are not approved by "God". Gorsuch [39] posited that the Church prevents drug use by encouraging parents to supervise their children and establish anti-drug rules in their home. As in our sample we also found that parents smoking status is associated with adolescents CS, it suggests that, religiosity and family may simultaneously influence the decision of not smoking: family religiosity may be influencing adolescent religiosity and parents' religiosity may be the reason of parents not smoking. Stylianou [40] proposed a theory suggesting that the perception of immorality and personal responsibility on physical self-destruction that religions bring to their members controls these individuals' attitudes when faced with opportunities to use drugs, including tobacco.

This protective role of religion was also observed in an investigation that collected data on the education, religiosity, and moral attitudes of 16,604 individuals in 15 countries [41]. According to the authors, individual behavioral patterns and moral attitudes were more strongly oriented by religion in countries where religiosity was generally more important in social life. For example, in a cross-sectional survey among almost 13,000 adolescents in seven Latin American countries, all mainly Catholic by history, researchers identified that higher levels of religious practice behaviors were significantly associated with lower odds of initiation of tobacco smoking and lower odds of opportunities to use tobacco [12].

In both genders, hanging out with friends at night for parties was associated with CS, and this association was stronger among girls. Although participants were not questioned about the use of tobacco by their friends, our results indirectly pointed to the relevance of the frequency of hanging out with friends without parental supervision. Similarly, a North American study of African-American adolescents showed that the factor most strongly associated with smoking was the relationships with friends. When the variables describing the activities with friends were inserted in the model, all the familial factors that had previously demonstrated an association with CS were no longer significant for both genders [18]. In contrast, in Barcelona, a cross-sectional survey among high school students showed that the influence of a group on using tobacco was stronger among boys [29]. Thus, the findings corroborate the Social Developmental Theory [20] because tobacco use was associated with social influences both in terms of family members and leisure activities with friends. Even the religiosity aspects that were characterized by social activities seemed to have reinforced the anti-tobacco attitude of the adolescents [9].

Therefore, the present study suggests that preventive measures against CS among adolescents should emphasize gender peculiarities. Prevention programs that focus on family and school need to be designed separately and tailored for boys and girls in order to take into account the possible adverse risk factors and protective factors that are gender-specific. Moreover, this study highlighted the importance of reducing the adverse risk factors associated with tobacco use during adolescence to reduce the likelihood of other substances consumption [3, 42].

One strategy would be to encourage parents to have a clear anti-tobacco attitude to pass on to their children from their adolescence. A recent Swedish study showed that the majority of adolescents in Sweden were receptive to their parents' anti-tobacco attitude because the practice was not based on punishments but on explanations and limits [43]. Although our analysis suggested that the use of tobacco by someone at home was a possible risk factor for tobacco use by adolescents, the literature has also shown that this factor may be reversed by communication on the topic. A study among Dutch adolescents suggested that parents who used tobacco were a risk factor for their children only when the frequency and quality of the communication between parents and children were poor [8].

In the present study, the frequency of hanging out with friends was the factor most strongly associated with CS in both genders, with evidence of a "dose-response" association. Although hanging out with friends is part of a healthy process of adolescent socialization, these results suggested that parents should be more prudent when allowing their sons and daughters to frequently hang out. And especially for girls, the additional attention and care may help to lower the likelihood of CS. In addition, parental attitudes may help to shorten the communication distance between parents and their children; parents may monitor their adolescents more closely and also become more aware of the behavior of their children's friends. Similarly, encouraging a religious connection may also play a positive role in reducing smoking among adolescents.

Since this was a cross-sectional survey, the data analyses have some limitations. For instance, the analyzed factors may have had associations with CS, but it was not possible for us to establish causal relationships. Additionally, due to the fact that a self-reported questionnaire was used, the questions were subject to interpretation by the participants. The adolescent respondents may have underreported their tobacco use [44], and smoking behaviors may have been underestimated because the study asked questions only about cigarettes. Further, we did not have data from public high schools students in São Paulo, so the data is not generalizable to the overall adolescent population in the city. The strength of this study included the inclusion of individuals rarely studied, wealthy adolescents from the largest and richest city of Brazil. Carlini-Cotrim et al. [45] have pointed out that is difficult to access private schools in Sao Paulo; despite this difficulty, were able to overcome barriers through our various telephone contacts and visits to the school principals.

The results of this study may be substantially important for the programs that could influence determinants connected to gender and for building preventive models based on factors that could be more significant for either boys or girls in order to reduce CS among adolescents.