Article Text

Download PDFPDF

Impact of removing point-of-sale tobacco displays on smoking behaviour among adolescents in Europe: a quasi-experimental study
  1. Maureen M Van Hurck1,2,
  2. Paulien A W Nuyts1,
  3. Karin Monshouwer3,4,
  4. Anton E Kunst1,
  5. Mirte A G Kuipers1
  1. 1 Department of Public Health, Amsterdam Public Health research institute, Academic Medical Center - University of Amsterdam, Amsterdam, The Netherlands
  2. 2 Department of Health Sciences, Faculty of Sciences, VU University Amsterdam, Amsterdam, The Netherlands
  3. 3 Department of Interdisciplinary Social Science, Utrecht University, Utrecht, The Netherlands
  4. 4 Trimbos Institute (Netherlands Institute of Mental Health and Addiction), Utrecht, The Netherlands
  1. Correspondence to Mirte A G Kuipers, Department of Public Health, Academic Medical Center, University of Amsterdam, Amsterdam 22660, The Netherlands; m.a.kuipers{at}amc.uva.nl

Abstract

Background An increasing number of European countries implemented a point-of-sale (PoS) display ban on tobacco products. This study assessed the association between PoS display bans in Europe and adolescent smoking and perceived accessibility of tobacco, 2–6 years after PoS display ban implementation.

Methods In a quasi-experimental design, we compared individuals in countries that did and countries that did not implement a PoS display ban, before and after implementation. We used repeated cross-sectional data of 174 878 15-year-old and 16-year-old adolescents from 25 countries from the 2007, 2011 and 2015 European Survey Project on Alcohol and Other Drugs surveys. Multilevel logistic regression models examined smoking behaviour (regular smoking vs other) and perceived access (easy vs difficult) as a function of display ban implementation and controlled for survey year, gender, parental education and implementation of other tobacco control policies. Interaction with gender was tested.

Results The implementation of a PoS display ban was associated with a 15% larger drop in the odds of regular smoking (OR 0.85, 95% CI 0.79 to 0.91), but was not significantly associated with perceived accessibility of tobacco (OR 0.97, 95% CI 0.892 to 1.03). Associations were similar in males and females (cross-level interactions of gender with display ban were not statistically significant for either outcome).

Conclusion The implementation of PoS display bans in Europe was associated with a stronger decrease in regular smoking among adolescents. This decrease does not appear to be driven by a decreasing accessibility of tobacco, but might be caused by further de-normalisation of tobacco as a result of PoS display bans.

  • advertising and promotion
  • prevention
  • public policy
View Full Text

Statistics from Altmetric.com

Introduction

Bans on the display of tobacco at the point of sale (PoS display bans) are among the more novel tobacco control policies to prohibit all tobacco promotion, as urged in Article 13 of the WHO’s Framework Convention on Tobacco Control.1 Some European countries that are at the forefront of tobacco control, including the UK, Ireland and Iceland, implemented PoS display bans.2 PoS display bans gain popularity because removing this form of tobacco advertisement would reduce smoking cues for smokers and recent ex-smokers and decrease the attractiveness of tobacco branding among young people.3

Young people may be especially susceptible to tobacco displays and advertisement4 as they remain an important target group for tobacco industry marketing efforts.5 6 The removal of tobacco displays has the potential to decrease adolescents’ susceptibility to smoking and de-normalise tobacco use as a result of decreased brand recognition and a decreased perception of tobacco being a normal marketable product.3 7–9 The removal of PoS displays may therefore be essential in protecting young people against tobacco industry marketing and consequential smoking uptake.

Evidence for the effectiveness of PoS display bans on smoking among adolescents is still limited. Studies from Ireland and England did not find a significant decline in smoking prevalence among adolescents.10 11 These studies might have missed potential impacts due to the short post-ban observation period and other potential data limitations. Studies from New Zealand12 and Australia13 found significant declines in adolescent smoking prevalence 2 years after the implementation of a PoS display ban, especially in those who visited tobacco retailers most frequently.13

The aim of this study was to test whether the implementation of a PoS display ban was followed by a decline in regular smoking behaviour among 15-year-old and 16-year-old adolescents in Europe. A secondary aim was to study a potential decline in the perceived accessibility of tobacco. A quasi-experimental design was applied, in which countries that implemented a PoS display ban were compared with countries that did not implement a ban by 2015. This is the first study to compare changes in smoking behaviour and perceived tobacco accessibility over time in an international natural experiment.

Methods

Data and study population

Data were obtained from the 2007, 2011 and 2015 European Survey Project on Alcohol and Other Drugs (ESPAD) surveys. ESPAD collects data on substance use among those aged 15–16 years in Europe. ESPAD data have been previously used to evaluate tobacco control in a similar study design.14 The comparability of the data between countries and over the years is important for the purpose of this study. Comparability was achieved by using common questionnaires between years and countries and standardisation of the data collection procedure, as prescribed and coordinated by the ESPAD project leaders.15–17 Data were collected during the same time of the year in each country and each survey year. Students in all countries were from a single year of birth, minimising age variations. In all countries and years, data are collected by group-administered questionnaires. The students answer the questionnaires anonymously in the classroom under supervision of teachers or research assistants. Samples of classes are nationally representative for each year. The average response rate in the three ESPAD surveys of students who were enrolled in the participating classes was 87% and was similar between years (see table 1).15–17 More information on the sampling strategy, validity and reliability of the data collection is available in the ESPAD reports.15–17

Table 1

Response information, and smoking and accessibility prevalence in the study population, derived from the European Survey Project on Alcohol and Other Drugs survey (n=174 848), presented by country

Data from 2007, 2011 and 2015 was available for 41, 40 and 35 countries, respectively. For the current study, 19 countries were excluded. First, Armenia, Bosnia and Herzegovina, Georgia, Kosovo, Serbia and Switzerland were excluded because they did not have data for at least two survey years. Second, Faroes, Isle of Man and Liechtenstein were excluded because they were not established as WHO European region countries. Third, Hungary, France and Italy were excluded because the sale of tobacco differed from the other included countries, in the sense that tobacco products can only be purchased in specialised tobacco shops. Fourth, Germany and Russia were excluded because the sampling frames were inconsistent across survey years; Germany included a different number of federal states in different survey years, and Russia had a national survey in 2007, but covered only Moscow in 2011. Finally, Albania, Macedonia, Moldova, Monaco and Montenegro were excluded because they were not included in the Tobacco Control Scale (TCS).18–20 The TCS was used to control for confounding by other tobacco control policies and is described below in the ‘Country level’ variables section. In total, 25 countries were included in this study (see table 1).

The 25 countries provided repeated cross-sectional data from a total of 214 691 individuals. Individuals with missing data or ‘don’t know’ responses on either of the outcome variables were excluded from this study (total n=39 813), resulting in a total study population of 174 878 individuals. There were no missing values for gender, and parental education was imputed as described in the ‘Study measures’ section.

Study measures

Individual level

Two main outcome variables were derived: perceived accessibility of tobacco and whether adolescents were regular smokers, defined as smoking at least one cigarette per day. Respondents were asked how difficult it was for them to get cigarettes if they wanted and answered with: ‘impossible’, ‘very difficult’, ‘fairly difficult’, ‘fairly easy’, ‘very easy’ or ‘don’t know’. This question was dichotomised into easy (‘very easy’, coded 1) versus difficult (all other options, coded 0). Other studies, examining the perceived accessibility of tobacco on adolescents, categorised this variable in a similar way.21–23

All adolescents were asked how frequently they smoked cigarettes in the last 30 days, answering: ‘none’, ‘<1 cigarette per week’, ‘<1 cigarette per day’, ‘1–5 cigarettes per day’, ‘6–10 per day’ or ‘11+ per day’. We distinguished regular smokers (‘1–5 cigarettes per day’ or more, coded 1) and non-(regular)smokers (‘<1 cigarette per day’ or less, coded 0). In a sensitivity analysis (see ‘Statistical analysis’ section), we separately studied non-regular smoking, defined as smoking ‘<1 cigarette per week’ or ‘<1 cigarette per day’.

The education level of the most highly educated parent was used to control for socioeconomic status (SES). Studies showed that low parental SES is a predictor for offspring to start smoking.24 25 Parental education level was assessed with: ‘What is the highest level of schooling your [mother/father] completed?’, with options: ‘completed primary school or less’, ‘some secondary school’, ‘completed secondary school’, ‘some college or university’, ‘completed college or university’, ‘don’t know’ or ‘does not apply’. ‘Don’t know’ and ‘does not apply’ responses were defined as missing. To account for differences in the distribution of parental education between the countries, a country-specific parental education measure was constructed.26 Individuals were assigned the mean rank for each answer category, which was divided by the number of respondents per country. This resulted in a continuous variable, ranging between 0 and 1, where 0 represents the lowest parental education in the country and 1 represents the highest. Missing values were imputed with the median (0.5 by definition). In the regression analysis, this country-specific measure resulted in the Relative Index of Inequality (RII).26

Country level

PoS display ban implementation was defined by whether the ban was introduced or not. Display bans were defined as national legislation that prohibits the display of tobacco products in shops. Products may be kept behind shutters, under a counter, or elsewhere in the shop where they are out of sight for customers. We collected information on whether and when countries had implemented PoS display bans from reports and scientific papers,18–20 27–30 as well as information from the American Nonsmokers’ Rights Foundation and the international legal consortium of the Campaign for Tobacco-Free Kids. For each country–year combination, the ‘PoS display ban implemented’ variable was assigned a value of 1 if a PoS display ban had been implemented and 0 if it had not (see table 1). For example, the value was 0 for all three years in countries that did not implement a PoS display ban at any point until 2015 and in Iceland, where the display ban was implemented in 2001, the value was 1 in all three years.

In Denmark, Greece, the Netherlands, Slovenia and the UK, the minimum age of tobacco sales increased from 16 to 18 years during the study period. The variable ‘age restriction’ defined having implemented a minimum age of sales to minors of 18 years before the survey year (coded 1) or having a minimum age of 16 years (coded 0). See online supplementary table I for details.

The TCS is a scale that was developed to track tobacco control progress across Europe and is updated every three years.18–20TCS quantifies the strength of tobacco control policies in each country, by awarding points for price of cigarettes, smoke-free work and other places, spending on public information campaigns, bans on advertising and promotion, health warning labels and treatment services to help smokers stop. TCS scores from 2007, 2010 and 2013 for each country were assigned to individuals in the surveys of 2007, 2011 and 2015, respectively. As the allocation of points was different in 2007 than in 2010 and 2013, the 2007 scores were converted to be directly comparable over time, according to Bosdriesz et al.31 For the purpose of this paper, two points that TCS awards for ‘Ban on display of tobacco products at the point of sale’ were excluded from the scale. The maximum TCS score was 98, and ranged from 27 to 85 in our sample. The TCS scores of 2007 and 2010 were missing for Croatia and Ukraine because these countries were included in the TCS from 2013 onwards. The TCS scores for Ukraine were calculated for the SILNE project based on information from local sources (M Leinsalu, personal communication, 2014; see http://silne.ensp.org/ for more information on SILNE). The TCS scores for Croatia were determined with information on their tobacco control policies between 2007 and 2013.32 33

Statistical analysis

Analyses were performed in Stata V.14.2. Multilevel logistic regression analyses (first level: individual adolescents; second level: countries) included a random intercept for countries to control for country-level clustering. Outcome variables were regular smoking and perceived easy access to tobacco, respectively. Data of all available survey years were used for all countries. In all regression models, the ‘PoS display ban implemented’ variable compared country–year combinations in which the display ban had been implemented with country–year combinations in which there was no display ban. In this design, we controlled for survey year in order to account for confounding by trends in smoking among all countries combined. Moreover, we took clustering at the country level into account in order to control for cross-country variations in smoking among all years combined. By using this design, we isolated the effect of the PoS display ban as much as possible.

The basic model included implementation of the PoS display ban, time and gender. In model 1, we additionally controlled for regular smoking and parental education. Regular smoking was controlled for in the analysis of accessibility of tobacco as smokers perceive tobacco to be more easy to obtain than non-smokers.22 23 In model 2, TCS and age restriction were added. Finally, an interaction analysis with gender was conducted as studies showed that males are more likely to depend on commercial sources of tobacco compared with females,34 and may therefore be more heavily exposed to tobacco retailers. A model that included interaction between all independent variables and gender tested differential associations by gender.

In five sensitivity analyses, we evaluated whether the results from the main analysis were robust:

  1. We used a more lenient cut-off point for accessibility, with   ‘fairly easy’ included in the ‘easy’ category.

  2. We explored the role of non-regular smoking by (a) using non-regular smoking versus non-smoking as the outcome (excluding regular smokers; n=1 43 315), (b) using regular smoking versus non-smoking as the outcome (excluding non-regular smokers; n=154 416) and (c) using a three-category smoking variable (regular smoking, non-regular smoking, non-smoking) as a covariate in the analysis of accessibility.

  3. Missing values on parental education were excluded instead of imputed (total population n=151 568).

  4. We only included countries with data of all three surveys, leading to exclusion of Austria and the UK.

  5. ‘Don’t know’ responses on tobacco access were considered to imply ‘difficult’, meaning that we distinguished especially those who were confident that tobacco was easy to obtain (total population n=212 699).

Results

Characteristics of the included countries are presented in table 1, and additional individual-level characteristics can be found in online supplementary table II. Regular smoking decreased across years, except for Poland and Portugal, where an increase from 2007 to 2011 took place. Perceived accessibility of tobacco also decreased across years, except for Romania, where accessibility remained the same and Finland and the Netherlands, where accessibility slightly increased from 2007 to 2011.

Table 2 presents the results for regular smoking. The implementation of a PoS display ban was associated with lower odds of regular smoking (OR 0.85, 95% CI 0.79 to 0.91). This association was very similar across the three models with increasing adjustment for socio-demographics and country characteristics. The association also showed to be robust in all sensitivity analyses (see online supplementary table III), although it was weaker and non-significant when non-regular smoking was used as the outcome variable (online supplementary table III, sensitivity analysis 2a: OR 0.94, 95% CI 0.87 to 1.03). Regular smoking did not change between 2007 and 2011, but decreased after 2011. Regular smoking was more prevalent in males than in females, and in adolescents with poorly educated parents than in adolescents with highly educated parents.

Table 2

Associations between point-of-sale (PoS) tobacco display ban implementation and regular smoking

Table 3 presents the results for perceived tobacco access. We found a weak inverse association between PoS display ban implementation and perceived easy access to tobacco (OR most adjusted model: 0.97, 95% CI 0.92 to 1.03). The sensitivity analyses resulted in very similar findings (see online supplementary table III). Adolescents perceived tobacco to have become significantly more difficult to obtain over the years. Males and adolescents with more highly educated parents perceived tobacco to be more easily accessible than females and adolescents with poorly educated parents. The implementation of a ban on sales to minors below the age of 18 was associated with lower odds of perceived easy access (OR 0.74, 95% CI 0.70 to 0.78).

Table 3

Associations between point-of-sale (PoS) tobacco display ban implementation and perceived easy accessibility of tobacco

Stratified analyses by gender are presented in table 4. Both for regular smoking and perceived accessibility, the association with the PoS display ban was comparable between both genders (p for interaction 0.895 and 0.253, respectively).

Table 4

Associations between point-of-sale (PoS) tobacco display ban implementation with accessibility of tobacco and regular smoking stratified by gender through interaction analysis

Discussion

Key findings

This study was the first to provide evidence on the effect of PoS display bans on adolescent smoking by comparing multiple European countries over time. We found that the implementation of PoS display bans was associated with a 15% decrease in the odds of regular smoking. We observed no significant association between the implementation of PoS display bans and perceived ease of accessing tobacco products. Both associations were comparable for males and females.

Potential limitations

Our study had some limitations. First, all individual-level variables used in this study were self-reported, potentially leading to socially desirable answers, which could potentially underestimate regular smoking.35 The introduction of a PoS display ban, through its potential impact on de-normalisation of smoking, may have been associated with a higher rate of socially desirable answers.13 36 If so, this may have overestimated the association between a PoS display ban and regular smoking. However, studies showed that self-reported answers in questionnaires on substance use among adolescents provide reliable estimates of prevalence rates.37 38 Moreover, the ESPAD surveys were anonymous and all the surveys were conducted in schools, which has been shown to result in limited social desirability bias.35 39

A second potential limitation is that we were unable to control for differences between countries in long-term trends in smoking. With three data points, we had one data point available before or after the implementation of PoS display bans. As a result, we cannot rule out the possibility that pre-existing long-term trends in youth smoking were more favourable in countries that implemented a PoS display ban. However, an English study among adults did take secular trends into account and found that the PoS display ban accelerated the rate of decline in smoking after implementation of the PoS display ban.40 Similar studies on youth smoking should determine whether the declines observed in this study are also independent of pre-existing trends.

A third limitation is potential residual confounding. First, the TCS may inadequately control for all confounding caused by other tobacco control policies. For example, the TCS includes simplified measurements of price (ranking of price of the country’s most popular price category) and anti-tobacco campaigns (spending on tobacco control). If countries with PoS display bans were more likely to have implemented policies that are less well captured by the TCS, this may have resulted in an overestimation of the associations. Second, we did control for bans on sales to minors, but did not take the strength of enforcement of this ban into account. For example, introducing new ways of checking identification or more frequent inspections were not considered. If this had mainly occurred in countries where a PoS display ban was implemented, an association with perceived accessibility may have been overestimated. Third, we were unable to control for e-cigarette use. Recent longitudinal studies show that e-cigarette use in adolescents is associated with higher smoking initiation rates.41–43 During the studied period, the use of e-cigarettes increased and the magnitude of this increase varied by country.44 We do not have reason to believe that the increase was smaller in countries with PoS display bans than in those without, but there is a potential for some residual confounding.

Interpretation of results

Our results are in accordance with two other studies.12 13 These studies, conducted in New Zealand and Australia, used a before-after design and also found a decrease in youth smoking prevalence after the PoS display ban was implemented.12 13 Studies from England and Ireland did not find an effect on smoking prevalence.10 11 McNeill et al 10 suggested that the lack of an effect in the Irish data may be attributable to a small sample size, loss to follow-up and a short post-ban period. The English study also included a relatively short post-ban period of a year, compared with the 2–6 years in the current study.11 The current study design, which included multiple countries and a large sample of individuals, suggests that evidence of the impact of the PoS display in New Zealand and Australia may be generalisable to European settings.

We did not find an effect of PoS display ban implementation on the perceived ease of accessibility of tobacco, while previous experimental studies in lab settings did suggest an effect.7 8 In these studies, adolescents who were shown photographs of shops without PoS tobacco displays reported a significantly lower perceived accessibility of tobacco compared with those who were shown photos of shops with visible displays.7 8 However, these results may not fully translate to real-life settings, for which we suggest the following possible reasons. First, young people may have a shared knowledge of which shops provide potential access to cigarettes, which would be unaltered when tobacco displays are covered. Their ability to identify and access these shops is illustrated by many studies evaluating the ban on sales to minors, which consistently found that young people continued to have access by asking strangers or by directly buying from small local shops.45–47 Second, although precise evidence is lacking, the actual availability of tobacco might remain about the same after implementing the PoS display ban as the ban does not aim to reduce the number of tobacco retailers. Third, in the current study, the definition of accessibility was not limited to commercial sources, but also included social sources (ie, cigarettes provided by friends, older peers, family or strangers). Young people often rely on social sources of cigarettes,34 48 which are likely unaffected by PoS display bans.

If the potential effect of PoS display bans is unlikely to be mediated by a reduction in access, other mechanisms would be implied. Possibly, PoS display bans may not affect the supply side but rather the demand side of tobacco. This mechanism is supported by studies that suggested that PoS display bans may result in reduced susceptibility to smoking, brand recognition and perceptions of tobacco being a normal marketable product, which could contribute to the de-normalisation of tobacco use.10 36 49

Regular smoking decreased in most of the included countries between 2007 and 2011, but especially between 2011 and 2015. This decrease also occurred in countries without a PoS display ban, although to a smaller extent. Between 2007 and 2012, great progress was made with the implementation of smoke-free policies, advertising bans, health warning labels on tobacco products and increased tobacco taxes in Europe.18–20 50 This development, especially banning smoking from most public places, could affect the norms of smoking as studies have shown that smoke-free policies are associated with an increase in social unacceptability of smoking.51–53 The greater decrease after 2011 is suggestive of delayed effects of smoke-free policies, possibly because of its indirect effect through de-normalisation.

Our findings for adolescent smoking are in line with studies that found the introduction of PoS display ban to be associated with adult smoking behaviour.54–59 These studies showed that a PoS ban could reduce the frequency of impulse purchases of cigarettes,54 55 and that removing PoS displays would help adults quit smoking.56–59 As adult smoking rates may decline, adolescents are probably less exposed to adult smoking, and in particular parental smoking, which is strongly associated with adolescent smoking.60–62 Through such a reduction in adult smoking, the impact of PoS display bans on adolescent smoking may be strengthened in the longer term.

Conclusions

The implementation of PoS display bans in spearhead countries in Europe was associated with a decrease in regular smoking among adolescents. PoS display bans were not associated with a decrease in perceived accessibility of tobacco, suggesting that other mechanisms were at play, such as changes in tobacco-related social norms among young people. This study provides important evidence for the many European countries where PoS display bans are yet to be implemented as their implementation may help accelerate the declining trend in adolescent smoking across Europe.

What this paper adds

What is already known on this subject

  • Tobacco displays at the point of sale remain an important marketing tool for the tobacco industry in Europe.

  • Some studies suggest that the removal of point-of-sale tobacco displays contributed to the de-normalisation of smoking among adolescents.

What important gaps in knowledge exist on this topic

  • Evidence on the impact of tobacco display bans on actual adolescent smoking behaviour is lacking.

  • Studies to date have been restricted to single countries, which do not take secular trends into account.

What this paper adds

  • This is the first international quasi-experimental study to provide evidence on the effect of point-of-sale tobacco display bans on adolescent smoking.

  • The implementation of a display ban was associated with a 15% stronger decline in the odds of adolescent smoking than when no display ban had been implemented.

  • This decrease in smoking does not appear to be driven by a decreasing perceived accessibility of tobacco.

  • The implementation of point-of-sale display bans throughout Europe could contribute to a substantial further decline in adolescent smoking.

References

View Abstract

Footnotes

  • Contributors MAGK and AEK conceived and designed the study. KM provided access to and advice on the data. MMVH and MAGK prepared and analysed the data. MMVH drafted the manuscript. PAWN, KM, AEK and MAGK contributed to the interpretation of the data and critically revised the manuscript at various stages for important intellectual content. All authors have read and approved of the final manuscript.

  • Funding This study was funded by H2020 Health (grant no: 635056).

  • Competing interests None declared.

  • Patient consent Not required.

  • Ethics approval National ethics committees in each country, according to national law.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.