Substance use and predictor variables within our data set contained between .2% and 9.5% missing data across all study subjects. No missing value pattern was observed, and data imputation was performed to increase the power of our analysis.

The overall measures of substance use in our analyses showed that 13.8% of adolescents had smoked, on an average, between one and five cigarettes or more each day in the past month. A total of 46.2% of adolescents reported alcohol use problems over the past 12 months and 16.8% had used marijuana at least once in their lives. Summary measures for all studied variables according to substance use can be found in Table 1. Gender-specific sample characteristics are given in supplement Table S1 (available online). Correlations between substance uses and their predictors can be found in supplement Table S2 (available online) (descriptive analysis only).

This study investigated interaction effects between conduct problems, refusal assertiveness, and close friends' substance use on adolescent substance use. In line with our hypothesis, our findings suggested that the relationship between conduct problems and adolescent alcohol use problems is moderated by their friends' alcohol use.

Substance use is common during adolescence and early adulthood ^[4]. In the present study, approximately one-seventh of the adolescents had smoked at least one cigarette per day in the past month. Nearly half of all adolescents reported problematic alcohol use during the last year, and approximately one-sixth had taken marijuana at least once in their life. These rates are comparable with published reports of adolescent substance use in the United Kingdom ^[1,7].

Consistent with previous research, conduct problems and substance use among close friends were strong risk factors for all three substances ^{[6,10–14,22]}. In extending these findings, our analysis showed that moderation of their risk effect however is substance-specific. Both cross-sectional and prospective analyses indicated that conduct problems were only related to an increased risk for alcohol use problems in the presence of substance-using friends whereas no such interaction was observed for tobacco and marijuana use. This suggests that the relationship between conduct problems and alcohol use problems is moderated by the peer group, although a greater density of substance-using friends may be needed for the moderation of the link between adolescent alcohol problems and conduct problems in childhood.

Peer influences have recently been highlighted as an important factor in substance use risk moderation, although research to date has mainly focused on interactions with adolescents' personality, including decision-making, self-reinforcement, or refusal assertiveness skills ^[20,21]. Our results provide evidence for the moderating effect of substance-using close friends that support the peer cluster theory according to which peer affiliation represents an important risk factor for adolescent substance use ^[14]. We extended this notion by finding that the influences of substance-using close friends may be particularly problematic among adolescents with conduct problems as neither conduct problems in childhood nor in adolescence were risk factors for later alcohol use problems in the absence of substance-using peers.

As conduct problems are related to antisocial behavior ^[38], our results may reflect a common, transient and adolescent-limited form of antisocial behavior ^[38]. It has been shown that the adolescence-limited path of antisocial behavior is more strongly associated with delinquent peer affiliation compared to the life-course persistent and pathological path of antisocial behavior ^[38], and peers may be an important moderating influence on delinquency ^[26].

Consistent with previous findings ^[20,21], we also observed a buffering role for high refusal assertiveness on adolescent substance use. The strongest effect was found for cigarette use, followed by marijuana use and alcohol use problems. Previous studies, utilizing a broader definition of friends ^[20,21], also reported that refusal assertiveness mitigates the influence of friends on adolescents' alcohol use ^[20], polydrug use ^[21], and their future smoking intentions ^[21]. The absence of similar moderating effects within our work may reflect differences in study methodology, in particular the definition of friends as substance-using close friends ^[15], and the focus on a white European instead of a multiethnic sample.

Alcohol use ^[1,4], particularly during adolescence, represents common and socially accepted behavior, leaving adolescents particularly susceptible to the influence of alcohol-using close friends. This may also explain why recent research found, in contrast to later tobacco or illicit drug use, no link between early behavior problems and later alcohol use ^[22], which is consistent with our findings from prospective analysis. However, relationships were identified between childhood conduct problems and early alcohol abuse and dependence ^[22].

The absence of an effect moderation of conduct problems through friends' influences as observed for tobacco and marijuana use may suggest that these risk factors act independently. This might support the peer selection model ^[18,28] positing that substance-use initiation precedes the selection of a substance-using peer-group ^[17,18]. For example, marijuana-using adolescents appear to coordinate their friendships toward congruent values and behaviors ^[19]. However, given the lower prevalences of tobacco and marijuana use in comparison to alcohol problem use, the absence of interaction effects may also relate to insufficient study power.

The exploration of moderating effects represents an important area of study among theorized risk and protective factors for adolescent substance use, as it has implications for prevention and intervention efforts. The findings of the present study underscore the importance of structured social interventions preventing youth from early exposure to social groups with extensive substance use and/or decreased levels of adult supervision. In addition, our work underlines the importance of training programs to support social competence and refusal skills, as not only peer influences but also individual decisions will determine the outcome of affiliations with deviant peer groups during adolescence ^[39].

Our study adopted an epidemiological and not a twin study approach as our aims were to assess the moderating effect of the social environment, in particular the interaction with substance-using peers. Although the twin approach enables tests of gene x environment interaction, it is less suitable to address research questions on contextual moderation because of lack of power at the extremes of the moderator ^[40].

This study provides further insight into the complex relationship between conduct problems, close friends' substance use, and refusal assertiveness. Cross-sectional and prospective analyses identified a significant moderating effect of close friends on the risk exerted by conduct problems on young people's alcohol problem use. The findings underline the need to consider not only main but also moderating effects, because risk factors appear to act in combination to increase risk of substance use. In addition, we provided further support for the buffering effect of refusal assertiveness and the risk effect of conduct problems and close friends' substance use across all three substances, a finding that has implications for educational and intervention programs.

Web supplement
Click here for additional data file (mmc1.pdf)

Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.jadohealth.2009.12.022

The authors thank all twin families for their participation in the CaStANET study. This work was supported by the European Research Advisory Board (ERAB) and the Wellcome Trust (GR073063).