Moderators of peer influence effects for adolescents' smoking and vaping norms and outcomes in high and middle-income settings.

Background: Peer influence is central to adolescent smoking initiation, yet its impact varies depending on individual and contextual factors. Understanding which moderators (personality, contextual, cultural, and environmental traits) shape these processes can inform more effective prevention strategies. We investigated hypothesized moderators of peer influence for adolescent smoking/vaping norms and other smoking-related outcomes in high and low-middle-income countries (LMICs): Northern Ireland and Bogotá. Methods: Across 12 schools ( Results: Peer influence was moderated by study setting, intervention, gender, school SES, personality characteristics (pro-sociality, fear of negative evaluation, extraversion), and social network structure. Effects were stronger among girls and in schools with lower SES. ASSIST schools showed greater peer influence effects than Dead Cool, reflecting the programs' distinct mechanisms, as ASSIST operates primarily through network diffusion and Dead Cool through teacher-led instruction and skills-building. Network measures highlighted that peer influence was stronger amongst more central individuals and more homogenous networks. Conclusion: Susceptibility to peer influence depends on contextual, individual, and network factors. Future social norms interventions should provide information on both injunctive and descriptive norms and highlight the social consequences of smoking, particularly in LMICs. Gender-tailored approaches are needed to address heightened susceptibility among girls. Future intervention research should combine peer-led diffusion approaches with teacher-led instruction to maximize reach and sustainability in different contexts. Social influence-based interventions may be particularly beneficial for schools with lower SES or in LMICs without tobacco control legislation, where smoking remains largely normalized. Network-based interventions like ASSIST could benefit from careful consideration of which network metrics are used to select peer leaders (e.g., eigenvector or closeness centralities) and exploring alternative approaches for more heterogeneous networks (e.g., 'segmentation', which targets clusters of individuals within social networks).