Participants
Data on conventional cigarette smoking and the use of alternative tobacco products were available for two cohorts of Dutch adolescents. Cohort I consists of 6819 adolescents aged 11–17 years [mean age = 13.8 (SD = 1.1), 48.2% female] who were enrolled in a study that investigated the impact of school smoking policy on changes in adolescents’ smoking behaviour. Data were collected in 2014–2015 from 19 secondary schools randomly selected across the Netherlands [18]. A comprehensive description of this study is available in the supplementary material. Of the total of 6819 adolescents, 2100 had longitudinal data available on smoking and alternative tobacco use; at time point 0 (T0) and time point 1 (T1) with 6 months in between. At each time point, adolescents were asked to complete a survey containing questions on their smoking behaviour, personality and use of alternative tobacco products.
Cohort II consists of 2758 adolescent participants of the Tr&nds study (Traditional and Novel Substance use among Adolescents) aged 14 to 21 years [mean age = 17.3 (SD = 1.8), 61.3% female]. Tr&nds aims to assess addictive behaviour in a representative group of Dutch adolescents and young adults, with a particular focus on ‘novel’ types of addictive behaviour, including the use of alternative tobacco products [19]. Data were collected in 2016–2017 from 14 educational institutions located mostly in the West of the Netherlands. A small subset of the participants was recruited via a Facebook advertisement (3.8% of the total sample). More details on Tr&nds and the survey data collection can be found in the supplementary material.
Measures
Cigarettes and alternative tobacco products
For conventional cigarettes, electronic (e-)cigarettes with nicotine, e-cigarettes without nicotine (‘shisha-pen’) and waterpipe, there was a question asking ‘How old were you when you used this substance/device for the first time?’. Answer categories were ‘I never used this substance/device’, ‘11 years or younger’, ‘12 years’, ‘13 years’, ‘14 years’, ‘15 years’, ‘16 years’, ‘17 years’, ‘18 years or older’ for Cohort I, while for the slightly older Cohort II the highest two categories were ‘19 years’ and ‘20 years or older’. Next, adolescents were asked how often they had used each of the alternative tobacco products in the past 4 weeks, with answer categories ‘0’, ‘1’, ‘2’, ‘3’,…, ‘9’, ‘10–19’ and ‘40 +’. For conventional smoking there was an additional question asking ‘Have you ever smoked, even if this was only one cigarette or a few puffs?’ with answer categories ‘I have never smoked’, ‘I have smoked once or twice to try’, ‘I smoke once in a while, but not every day’, ‘I used to smoke but I quit’ and ‘I smoke every day’.
With the above information, variables reflecting ever use (0 = no, 1 = yes) of conventional cigarettes and each of the alternative tobacco products were created. Those saying ‘I never used this substance’ to the first question were classified as never users while those who provided an age at which they used the substance for the first time were classified as ever users. For conventional cigarettes, this variable was cross-checked with the additional question on smoking behaviour (participants who were classified as never users based on the first question but answered they (used to) smoke to the second question, or the other way around, were set to missing). Variables reflecting past month use (0 = no, 1 = yes) of conventional cigarettes and each of the alternative tobacco products were created with a similar approach, contrasting no use in the past 4 weeks (0 times) to use in the past 4 weeks (1 time or more). Finally, a measure of smoking status was created. Those who stated to have never smoked cigarettes or only tried once or twice were classified as never smoker, those who smoked but quit were classified as former smoker and those who smoked once in a while or daily were current smokers. For Cohort I, all variables described here were available at both time points (T0 and T1).
When exploring the cross-use of different alternative tobacco products we found clustering such that adolescents who had used one alternative tobacco product, more often than not also used one of the other alternative tobacco products. There were, however, differences in this clustering, depending on the type of alternative tobacco both within and between cohorts (Supplemental Tables 1 and 2). We therefore analyze e-cigarettes with nicotine, e-cigarettes without nicotine and waterpipe separately instead of taking one measure of overall alternative tobacco use.
Sociodemographic variables
Sociodemographic variables were sex (0 = boy, 1 = girl), age (continuous variable, categorized into age categories appropriate for each respective cohort namely 11–13, 14–15 and 16–17 years for cohort I and 14–15, 16–17 and 18–21 years for cohort II), ethnicity (including the most common ethnic groups in the Netherlands and based on birth country of the parents; 0 = Netherlands, 1 = Surinam/Aruba/Netherlands Antilles, 2 = Morocco, 3 = Turkey, 4 = Other) and educational attainment (0 = low, 1 = average, 2 = middle and 3 = high for Cohort I and 0 = low/average, 1 = middle and 2 = high for Cohort II). The category ‘low’ refers to schooling for students with learning difficulties and the lowest level of pre-vocational secondary education, ‘average’ refers to the higher levels of pre-vocational secondary education or vocational education, ‘middle’ refers to higher general secondary education or higher professional education and ‘high’ refers to pre-university education or university. Given the low numbers of students classified as ‘low’ in Cohort II, ‘low’ and ‘average’ were merged into one category.
Propensity to smoke
In Cohort I only, a composite score of propensity to smoke was computed based on three risk factors for smoking at T0. The first factor, personality, was assessed with the validated ‘Substance Use Risk Profile Scale’ (SURPS) [20]. The SURPS provides sum scores for anxiety sensitivity, hopelessness, sensation seeking and impulsivity. The other two factors, susceptibility to peer pressure and intention to smoke, have also been consistently shown to predict onset of smoking [21]. Intention to smoke was measured by asking adolescents ‘Are you planning to smoke in the coming 6 months?’, with answer categories ranging from 1 ‘Definitely not’ to 7 ‘Definitely yes’, and susceptibility to peer pressure was measured by asking adolescents ‘Imagine that you are with a group of friends who all smoke. They offer you a cigarette, would you take the cigarette and smoke with them?’, with answer categories ranging from 1 ‘Definitely not’ to 7 ‘Definitely yes’. As was done in a study similar to ours [16], we created a composite smoking propensity score by performing a logistic regression analysis and saving the predicted values. In this logistic regression, smoking of conventional cigarettes at T1 (0 = no, 1 = yes) was the dependent variable and the SURPS personality traits, susceptibility to peer pressure and intention to smoke at T0 were the independent variables.
Statistical analysis
Descriptives and cross-sectional associations
Prevalence rates were assessed in each cohort separately. We report ever use and past month use of conventional cigarettes, e-cigarettes with nicotine, e-cigarettes without nicotine and waterpipe in both cohorts and across sociodemographic variables (sex, age, ethnicity, educational level). For alternative tobacco products we also report the mean number of times used in the past month.
Next, we tested cross-sectional associations between conventional smoking and alternative tobacco use. In a GEE (Generalized Estimation Equation) analysis, correcting for clustering within schools, the dependent variable was ever use (0 = no, 1 = yes) of either e-cigarettes with nicotine, e-cigarettes without nicotine or waterpipe while the independent variable was ever use of conventional cigarettes (0 = no, 1 = yes). Covariates were age, sex and educational attainment. Ethnicity was not added as a covariate given the low numbers of adolescents within the different ethnic groups. To check whether ethnicity affected our results, all GEE analyses were repeated in adolescents of Dutch ethnicity only. All analyses were conducted in SPSS Statistical Software.
Longitudinal associations
To investigate whether or not the use of alternative tobacco products predicts the use of conventional cigarettes, longitudinal data (T0 and T1) from Cohort I were analyzed. We first selected adolescents who stated to have never smoked conventional cigarettes at T0. Next, we carried out GEE analysis with ever use of conventional cigarettes at T1 (0 = no, 1 = yes) as the dependent variable, and ever use of either e-cigarettes with nicotine, e-cigarettes without nicotine or waterpipe (0 = no, 1 = yes) at T0 as the independent variable. Besides age, sex and educational attainment, a composite score of smoking propensity at T0 was added as covariate as well as an interaction term between propensity to smoke and ever use of e-cigarettes with nicotine/e-cigarettes without nicotine/waterpipe. Intervention status (0 = no school policy intervention, 1 = school policy intervention) was corrected for but not reported here (for results on effects of the intervention see [18]).
Correction for multiple testing
Given that we perform analyses for three different alternative tobacco products, Bonferonni correction for multiple testing was applied. For Cohort I, three separate cross-sectional regression analyses resulted in a threshold of statistical significance of < 0.017 (0.05/3). For Cohort II the same threshold was adopted given that there were three separate regression analyses in the cross-sectional sample and three in the longitudinal (sub)sample.