AIMS: To examine the role of national wealth in the association between family affluence and adolescent weekly smoking, early smoking behavior and weekly smoking among former experimenters.

DESIGN AND PARTICIPANTS: Data were used from the 'Health Behaviour in School-aged Children (HBSC)' study conducted in 2005/2006 in 35 countries from Europe and North America that comprises 60,490 students aged 15 years. Multilevel logistic regression was conducted using Markov Chain Monte Carlo methods (MCMC) to explore whether associations between family affluence and smoking outcomes were dependent on national wealth.

MEASUREMENT: Family Affluence Scale (FAS) as an indicator for the socioeconomic position of students. Current weekly smoking behavior is defined as at least weekly smoking (dichotomous). Early smoking behavior is measured by smoking more than a first puff before age 13 (dichotomous). Weekly smoking among former experimenters is restricted to those who tried a first puff in the past.

FINDINGS: The logistic multilevel models indicated an association of family affluence with current weekly smoking [Odds ratio (OR) = 1.088; 95% credible interval (CrI) = 1.055-1.121, P-value (P) < 0.001], early smoking behavior [OR = 1.066; CrI = 1.028-1.104, P < 0.001] and smoking among former experimenters [OR: 1.100; CrI: 1.071-1.130; P < 0.001]. GDP per capita was positively and significantly associated with the relationship between family affluence and current weekly smoking [OR: 1.005; CrI: 1.003-1.007; P < 0.001], early smoking behavior [OR: 1.003; CrI: 1.000-1.005; P = 0.012] and smoking among former experimenters [OR: 1.004; CrI: 1.002-1.006; P < 0.001]. The association of family affluence and smoking outcomes was significantly stronger for girls.

CONCLUSIONS: The difference in smoking prevalence between rich and poor is greater in more affluent countries.