DEPICT: study design and recruitment

DePICT recruited French-speaking children aged 12–17 years, residing in metropolitan areas of France. Participants were recruited by trained interviewers from randomly generated telephone lists. Landline and mobile phone numbers were called up to 30 times by trained professionals using a computer-assisted telephone interviewing system. The oral consent of at least one parent was required.

DePICT recruited both adolescents (12–17 years) and adults (18–64 years, data not shown), using somewhat different methodologies in order to over-represent adolescents in the final sample, because changes in tobacco-related perceptions and behaviours in that age group were of paramount interest. We describe the way in which we recruited both populations (adolescents and adults), even though the present study focuses exclusively on adolescents, because a large fraction of them were recruited via one of their parents.

Adults and adolescents recruitment

Adults were randomly recruited (up to 30 recalls) among all members of eligible households reached via a landline or mobile phone, using the Kish method. The participation rate among adults was 63% for the first wave and 50% for the second wave. To recruit adolescents using the Kish method was not feasible, because this subgroup represents a small proportion of the population (7.4%). Moreover, oral consent from at least one parent was necessary. Therefore we constituted a convenience sample, recruiting adolescents: (1) In each household where an adult was reached or participated. (2) Through randomly generated phone number files created exclusively to recruit adolescents. Only one adolescent was recruited in each household; in households where more than one adolescent was eligible, we interviewed the youngest one whenever possible and after obtaining parental permission, because we anticipated that parents would rather we surveyed the oldest. Adolescents’ participation rates were 58% in 2016 and 56% in 2017. After weighting the sample on age, sex, educational setting, parental socioeconomic status and family situation, DePICT adolescents were comparable to the adolescents in the general population of France.

Data collection procedures were independently monitored to ensure data quality, and the same methodology was used in both study waves. No weighting procedure taking into account the sampling procedure was used. The first wave of the DePICT study took place between 15 August and 15 November 2016, (prior to PP implementation), and the second wave between 15 August and 20 November 2017, (post PP implementation).

Smoking-related perceptions

Participants rated: (1) Their fear of the health consequences of smoking. (2) The dangerousness of smoking. (3) Their friends’ and family’s acceptance of smoking. (4) If smokers are less socially accepted than non-smokers on a Likert Scale (strongly agree, somewhat agree, neither agree nor disagree, tend to disagree, totally disagree). Responses were then coded dichotomously (agree vs disagree). The exact phrasing of these questions, which are also routinely used in the French National Health Survey,19 and the way in which they were recoded are presented as online supplementary materials.

Smoking experimentation and current smoking

Participants were asked about their lifetime and current (daily and occasional) use of tobacco, smoking experimentation (having tried smoking at least once, even a single puff) and current daily smoking (≥1 cigarette/day, yes/no). Respondents indicating they had ever smoked but were not current smokers were classified as experimenters (ever smokers who are not current smokers) (yes/no). We distinguished two definitions of ‘smoking experimenters’: (1) All experimenters, comprised of current, and former smokers as well as experimenters (those who tried smoking at least once but never smoked neither regularly nor occasionally). (2) ‘current experimenters’ combining current smokers and experimenters (vs former-smokers and non-experimenters).

Occasional or daily smokers were classified as current smokers (yes/no).

Perceptions of tobacco brands

Current smokers (daily and occasional) who reported smoking at least one tobacco brand regularly, were asked about their attachment to their main brand (‘I am very attached to this brand’), their opinion of its name (‘I like its name’) and the perceived harmfulness of this brand (‘My brand is less bad than other brands’). All answers were rated on a Likert Scale (strongly agree, somewhat agree, neither agree nor disagree, tend to disagree, totally disagree) and dichotomised into yes/no depending on the distribution (‘Yes’ corresponded to the answer ‘Strongly agree’ for attachment to the brand; ‘Strongly agree’ and ‘Somewhat agree’ for opinion regarding the brand’s name and its perceived harmfulness). The coding and distribution of these variables are shown in online supplementary materials.

Sociodemographic characteristics and other covariates

We also assessed participants’ age, sex and educational setting (general vs technical/vocational training). Parental characteristics previously shown to be linked to adolescents smoking20–22 were also measured: family situation (parents living together: yes/no), occupational grade (at least one parent with a high occupational grade: yes (manager, executive,…)/no (middle-level manager, technician, clerk, manual worker,…), and parental smoking (at least one parent smokes regularly: yes/no). Since participants were interviewed by telephone, we also asked them whether one of their parents was present next to them during the interview.

Statistical analysis

Data were weighted to match the structure of the French population aged 12–17 years with respect to sex, age, educational setting, the parents’ socioeconomic status (at least one parent working as a manager or equivalent) and family characteristics (parents living together or not). Weighting was based on data published by the National Institute of Statistics and Economic Studies23 24 and on the same structure of the population (2016). Log-binomial regression models were used to estimate prevalence ratios (PRs) between the two study waves, adjusting for the listed covariates. All statistical analyses were conducted using SAS V.9.4 (SAS Institute); statistical significance was set at p=0.05.

Interpretation

Our findings need to be interpreted in a context where the introduction of PP and larger graphic health warnings occurred in parallel to other tobacco control policies (eg, the adaptation of the Stoptober campaign, announced price increases). Moreover, recent data suggest a declining trend in the prevalence of smoking between 2014 and 2016 among adolescents living in France, which changes in packaging may have strengthened.26

In such a complex setting, it is difficult to identify the exact role of each type of tobacco control policy on smoking-related perceptions and tobacco experimentation among adolescents, especially that different policies can interact and influence potentially mediating factors such as parental smoking, or images and identities related to smoking.27 28 Nevertheless, PP probably plays a role in the decrease in young smokers’ attachment to their tobacco brand, since unattractive packaging has been repeatedly linked with less positive pack and product perceptions and a decrease in smoking pleasure.29 30 Adolescents appear especially likely to be influenced by tobacco branding and marketing, and it is quite plausible that their perceptions are affected by changes in packaging.31

We observed a considerable drop in adolescents’ perception of the social acceptability of smoking; this may be due to increase in antitobacco mass media campaigns as well as the introduction of other tobacco control policies which presented smoking more negatively and may have contributed to the subject of smoking being discussed between family members and friends.

The increase in adolescents’ perceptions of the harmfulness of smoking is another mechanism by which public health interventions can influence smoking initiation and uptake.32 In our study, perceptions of the harmfulness and social acceptance of smoking changed more among girls than among boys. In parallel, decreases in smoking experimentation were also more pronounced in girls than in boys. Girls might be more sensitive to social norms surrounding smoking,33 and more motivated to try tobacco because of the ‘glamourous’ images of smoking covertly promoted by the tobacco industry.34 35 PP and larger graphical health warnings aim to interrupt these associations. In France, adolescent girls are as likely to smoke as boys of the same age.26 Since women seem to be especially vulnerable to the negative effects of smoking,36 37 our finding that the implementation of PP is associated with a decrease in smoking initiation among girls is important. Nevertheless, as we did not find comparable decreases in smoking experimentation in boys, it may be necessary to identify other health promotion and tobacco control policies that are effective in both boys and girls.38

Limitations

Several limitations of our study need to be noted. First, since PP was introduced at the same time as larger graphic health warnings and other antitobacco measures, we cannot assert that the changes we observed are exclusively related to this policy. These changes probably result from a combination of different interventions. In order to be effective, tobacco control policies should be comprehensive and complementary, yet this makes the identification of the specific effects of each policy component difficult. Moreover, our prepost study design cannot establish a causal effect. Nonetheless, changes in adolescents’ attachment to tobacco brands, which we observed, are likely to be specifically related to the appearance of tobacco products, indicating that PP is effective. Second, we examined changes in smoking-related perceptions and behaviours over the course of 1 year, which is a relatively short period. In Australia, the big dip in smoking levels in adolescents and adults was observed 3 years after the introduction of PP,15 and we expect that in France the pattern will be similar. PP will likely contribute to long-term changes in the norms surrounding tobacco use, and changes in perceptions of smoking as well as smoking-related behaviours need to be monitored over the mid term to the long term. A third possible limitation is the possibility of selective non-response to our survey. It is possible that non-smokers were more likely to answer our survey than smokers; however this bias should have been the same in the two study waves. To mitigate this potential selection bias, we statistically weighted the data to render our sample representative of adolescents living in France. The weighted prevalence of smoking we observed in 2016 is comparable to other nationally representative sources of data in France, which is reassuring.26

Implications

Our study provides early and positive results regarding changes in smoking-related perceptions and behaviours among adolescents after the implementation of PP of tobacco products in France. While further studies with more robust designs are needed to gain a more thorough understanding of the effects of PP on smoking initiation and daily use rates over time, these encouraging results add to already existing data suggesting the importance of PP as a component of tobacco control policies.