Descriptive statistics at baseline and attrition analysis
gives the descriptive statistics for all interviewed never-smokers at baseline, for those lost to follow-up, and the final analysed sample, allowing comparisons of differences owing to attrition. Never-smokers lost to follow-up were of significantly younger age, more often male, had lower scores on the SES scale, rated their school performance more poorly, had higher scores in sensation seeking/rebelliousness and more often reported at least one parent who smoked. No differences were found with regard to tobacco or non-tobacco advertising contact.
Descriptive sample statistics at baseline and attrition analysis
Smoking initiation during the observational period
Post-30 months of the baseline assessment, 436 never-smokers reported trying cigarette smoking, including a few puffs (33% incidence rate); 138 reported smoking in the past 30 days (10.5% incidence rate), 66 had smoked more than 100 cigarettes and were classified as established smokers (incidence rate 5%), and 58 reported daily smoking (incidence rate 4.4%). Daily smoking incidence was not significantly related to age (p=0.526) or sex (p=0.153), with 33% of the daily smokers at follow-up being 14 years of age or younger and 24% being 16 or older.
Exposure to advertisements at baseline
gives contact frequencies (how often the students had seen the ad) for all advertised products at baseline. The cigarette ad with the highest contact frequency was Lucky Strike, for which about half of the sample reported at least one contact. The lowest tobacco ad contact frequency rate was found for F6, a regional German cigarette brand sold mainly in eastern Germany. Ad contact frequency for non-tobacco products was generally much higher than for tobacco products. For example, almost all students (96%) reported having seen the ad for Kinder Pingui, a chocolate bar. The range of the sum of contacts over all depicted advertisements was 0–55 (mean=7.9) for the tobacco ads, and 0–88 (mean=42.2) for the non-tobacco ads, also reflecting the lower number of tobacco ads (6 vs 8).
Contact frequency for tobacco and non-tobacco advertisings (n=1320 never-smokers at baseline)
Zero order associations
shows pairwise Spearman rank correlations between the study variables, demonstrating significant crude associations between the assessed covariates and smoking behaviour as well as between covariates and advertising contact, justifying their inclusion in the multivariate analyses. The highest correlation with all smoking outcomes was found for peer smoking, followed by tobacco advertising contact. There were some differences in the correlational pattern between tobacco and non-tobacco advertising contact. Compared to the amount of contact with tobacco ads, non-tobacco advertising exposure was stronger related to age, showing no association with gender, and also had a stronger correlation with SES, TV screen time and parental smoking. The zero-order correlation between tobacco and non-tobacco advertising contact indicated a proportion of about 20% shared variance.
Zero-order correlation matrix for all study variables
Association between advertising contact and smoking initiation
A,B shows the adjusted predictions of established smoking and daily smoking based on the amount of tobacco and non-tobacco advertising contact. The curves illustrate an increasing risk for the two smoking outcomes dependent on the amount of tobacco ad contact, but not for non-tobacco advertising contact.
Dotted line represents tobacco advertising; solid line represents non-tobacco advertising. Figures in brackets = 95% CI. IRR, Incidence Rate Ratio for 10 additional advertising contacts. n.s., not significant; *=p<0.05; ***=p<0.001.
The figures also report the adjusted incidence rate ratios associated with an increase in advertising exposure. There was an adjusted IRR for established smoking of 1.38 (95% CI 1.16 to 1.63; p<0.001) for each additional 10 tobacco ad contacts and 1.00 (95% CI 0.84 to 1.19; p=0.996) for each additional 10 non-tobacco ad contacts. For daily smoking, the corresponding IRRs were 1.30 (95% CI 1.03 to 1.64; p=0.029) for 10 tobacco ad contacts and 0.92 (95% CI 0.79 to 1.08; p=0.296) for 10 non-tobacco ad contacts, respectively.
Owing to the skewed distribution of tobacco ad contact frequency (more than half of the never-smoking students had fewer than 10 contacts), we repeated the analysis using contact frequency parsed into tertiles, representing relative low (0–2.5), medium (5–10) and high (11–55) advertising contact. For established smoking, the adjusted IRRs were 1.52 for tobacco ads (95% CI 1.14 to 2.03; p=0.004) and 1.05 for non-tobacco ads (95% CI 0.68 to 1.62; p=0.819). Using daily smoking as an outcome variable, the IRRs were 1.43 (95% CI 1.08 to 1.90; p=0.012) and 0.84 (95% CI 0.58 to 1.22; p=0.363) for each additional tertile of tobacco and non-tobacco advertising contact. These IRRs relate to a 3.1%, 4.8% and 7.3% established smoking attributable incidence rate or a 3.1%, 4.6% and 6.4% daily smoking incidence for low, medium and high tobacco advertising contact, respectively, assuming that the adjusted analysis adequately controlled for the third variable influence.
To address the question if some never-smokers had higher tobacco advertising contact because they were already more susceptible to smoking at baseline, we conducted a sensitivity analysis with only never-smokers with low susceptibility. These students reported at baseline that they would definitely never-smoke in the future and also would definitely not try cigarettes if a friend offered one (n=803). In this restricted subsample, the adjusted IRR for each additional 10 tobacco ad contacts was 1.37 for established smoking (95% CI 1.07 to 1.76; p=0.012) and 1.33 for daily smoking (95% CI 1.02 to 1.75; p=0.038). Again, no significant associations were found for non-tobacco advertisements.