Article Text
Abstract
Background and aim In 2016, England initiated the implementation of standardised tobacco packaging, introduced in conjunction with minimum pack sizes and other measures included in the 2014 European Tobacco Products Directive, over the course of a 1-year sell-off period ending in May 2017. These measures have been shown to have been associated with increases in tobacco prices and product diversity. We now investigate the association between implementation of the new legislation and smoking status in England.
Design Segmented regression analysis of repeated cross-sectional surveys using a generalised linear model with individual-level data to test for a change in trend and immediate step change.
Setting England.
Participants Participants in the Smoking Toolkit Study, which involves repeated, cross‐sectional household surveys of individuals aged 16 years and older in England. The sample included 278 219 individual observations collected between November 2006 and December 2019.
Intervention Implementation of standardised packaging legislation (May 2016 and May 2017).
Measurements Individual-level current smoking status adjusted for implementation of tobacco control policies, cigarette price, seasonality and autocorrelation.
Findings The implementation of standardised packaging was associated with a significant step reduction in the odds of being a smoker after May 2017 (OR: 0.93; 95% CI 0.87 to 0.99). The magnitude of the association was similar when modelling the step change in May 2016 at the start of the 1-year policy implementation period (OR: 0.90; 95% CI: 0.83 to 0.97).
Conclusions This is the first independent study demonstrating that implementation of standardised packaging was associated with a reduction in smoking in England which occurred in anticipation of, rather than after, full policy implementation. It appears that the odds of being a smoker were affected by the prospect of the move to standardised packs and accompanying legislation.
- public policy
- packaging and labelling
- advocacy
Data availability statement
Data used for this study are available upon reasonable request from the authors.
Statistics from Altmetric.com
Introduction
Over the past decade, the prevalence of current tobacco smoking in England has been in sustained decline, falling from 19.8% in 2011 to 13.9%, or approximately 5.7 million smokers, in 2019.1 This reduction in prevalence has been particularly marked among children and young adults1 2 and those in higher socioeconomic groups,1 and has been achieved by a range of tobacco control policies implemented by the UK government over the past two decades, the most recent of which was the introduction of standardised packaging legislation in May 2016.3 This legislation determined that after a 1-year transition period to May 2017, manufactured cigarettes and roll-your-own tobacco products could be sold in England only if packaged in generic drab dark brown packs with brand names and a single descriptor presented in a standard font. These requirements were implemented alongside the 2014 European Tobacco Products Directive (TPD), which among other measures mandated minimum pack sizes and larger pictorial health warnings.4
The primary aim of the standardised packaging legislation was to make smoking less appealing to and discourage smoking uptake among young people,5 but there is evidence that standardised packaging legislation might also reduce the prevalence of smoking among adults.6 7 Evidence from Australia, which in 2012 became the first country to introduce standardised packaging, suggests that implementation led to an increase in quitline calls8 and increased the rate of decline in smoking prevalence.9 Research on the introduction of standardised packaging in England has demonstrated that implementation has been associated with considerable increases in the price of tobacco products, switching to less expensive tobacco products and increased use of e-cigarettes among smokers.10–13 However, the effect of standardised packaging on smoking prevalence in England has not yet been explored by researchers independent of the tobacco industry’s funding. In 2018, Tobacco Manufacturers’ Association published analysis using Smoking Toolkit Study (STS) data suggesting that implementation of standardised packaging was associated with an increase in smoking prevalence.14 However, the data analysis was based on a very basic comparison of 3-month rolling average with the data from the same time period in the previous year. The analysis was considerably underpowered and did not consider any potential confounders. The aims of this study were therefore (1) to investigate the effect of the introduction of standardised packaging on smoking prevalence by estimating step and trend changes in individual current smoking status after the policy was implemented in England, and aggregating these results to estimate changes in smoking prevalence; (2) to explore whether differences in step and trend changes were observed in different population subgroups defined by age, sex and socioeconomic status.
Methods
Data sources and research design
We used data from STS, a monthly interview-based household survey of smoking status in representative samples of ~1700 adults aged 16 years and over in England that has collected data since 2006.15 The survey used a random location sampling design to select grouped output areas (~300 households) stratified by sociodemographic characteristics, while interviewers selected households within areas based on quotas targeted to the characteristics of the output area. Face-to-face computer-assisted interviews were then carried out with one household member. More details about the methods and the data can be found elsewhere.16–19 Comparisons with other national surveys suggest the design produces a sample of the population in England representative of key demographic variables, smoking prevalence and cigarette consumption.15
We used individual-level data from November 2006 to December 2019 to carry out a segmented regression analysis to study level and post-slope changes20 21 in current smoking status after the implementation of the standardised packaging policy in England, using a before and after approach.21 22 Our analysis first explored effects before and after May 2017, the end of the 1-year implementation period, as the main implementation point, and then before and after May 2016 and explored effects after the start of that implementation period. To check the robustness of our results, we performed a sensitivity analysis using implementation dates based on the proportion of tobacco sales in standardised packs, derived from Nielsen sales data, as a marker of the extent to which standardised packaging had been implemented during the implementation period.11 These indicated that standardised packs began to appear on the UK market in July 2016,11 and that sales of branded packs probably did not end until July 2017.11 13
Measures
Outcome variable
The outcome variable was current smoking status. All participants from STS were classified as current smokers (with the value ‘1’) if they responded affirmatively to any of the statements ‘I smoke cigarettes (including hand-rolled) every day’, ‘I smoke cigarettes (including hand-rolled), but not every day’, or ‘I do not smoke cigarettes at all, but I do smoke tobacco of some kind (eg, pipe, cigar or shisha)’; and as a non-smoker (with the value ‘0’) if they responded affirmatively to any of the statements ‘I stopped smoking completely in the last year’, ‘I stopped smoking completely more than a year ago’, or ‘I have never been a smoker’. This question differs from the question used in other nationally representative surveys. Hence, two actions were taken as a precautionary measure. First, we compared quarterly trends in smoking prevalence from STS data to Annual Population Survey (APS); and second, we compared general monthly smoking prevalence in STS with that of different tobacco product users, such as manufactured cigarettes only, hand-rolled tobacco only, combined manufactured cigarettes and hand-rolled tobacco users, and other tobacco product users, which would include those using pipe cigar or cigarillos among others. These two comparisons can be found in online supplemental figures S1 and S2 and show that trends in STS were fairly similar for most quarters when compared with APS data and that most of the smoking prevalence figures refer to only manufactured cigarette and only hand-rolled tobacco users, which were the most affected by standardised packaging legislation.
Supplemental material
Segmented regression analysis variables
For our analysis after full implementation of the policy, we created a level variable that took the value ‘0’ for all observations from months up to and including May 2017 (before), and the value ‘1’ for all observations after May 2017 (after). For our analysis after the policy start date, we added a before and after variable for May 2016, while our sensitivity analysis studied level changes using July 2016 and July 2017 as the start and full implementation dates. We also created a slope variable (post-intervention) with values between 0 and 1 that increases in equal amounts each month after full implementation up to 18 months to study changes in smoking status trend after full policy implementation (post-slope after May 2017 and post-slope after July 2017 in our sensitivity analysis). An equivalent slope variable was created for the implementation period (between May 2016 and May 2017) as well as for the period between July 2016 and July 2017. In addition to the level and post-intervention slope variables, the model included a linear time-trend variable with equally increasing values starting in the first month of the data until the last month of our study period.23
Subgroup analysis by age and socioeconomic status
We studied changes in the likelihood of being a current smoker among six population subgroups defined by sex (male and female), age (individuals aged 16–25 years and those above 25 years of age) and socioeconomic status (routine and manual occupations vs non-routine occupations).
Control for the effects of other tobacco control policies
During our study data period, the following tobacco control policies were implemented: smoke-free public places legislation in July 2007, an increase in the minimum age of sale from 16 to 18 years in October 2007, and a ban on point of sale displays which applied to large shops from April 2012 and small shops from April 2015. We created a dummy variable for each of these policies which assigned the value ‘0’ for all months up to and including the implementation month, and the value ‘1’ for all subsequent months. To adjust for the effects of tobacco tax and other price rises,24–28 we used the average monthly price for a 20-cigarette pack,29 adjusted for inflation using the Consumer Price Index.30
Seasonality and autocorrelation
Two additional variables were included to adjust our models: seasonality and autocorrelation. Regarding the first, evidence suggests that smoking has a seasonal pattern.31 32 Hence, we used a categorical ‘month of the year variable’ to account for possible differences in smoking status specific to the month of the year in which the survey took place. For autocorrelation, as we were using individual-level data to estimate grouped policy effects, we used robust SEs and created a variable with lagged values (one lag) of smoking prevalence (general population smoking prevalence and subgroup smoking prevalence to use accordingly).33
Statistical analysis
We initially plotted aggregated monthly trends in current smoking prevalence in the general population and in the subgroups defined by sex, age and socioeconomic status using weighted STS data from November 2006 to December 2019 to illustrate overall prevalence trends for the population in England.15 We compared smoking prevalence during the year the policy was implemented with the year before using a t-test to have a simple estimate of difference in prevalence before and after policy implementation.
We then used a generalised linear model (GLM) to estimate changes in level and slope of the likelihood of being a smoker after implementation of the standardised packaging policy using individual-level data on a binary smoking status variable. We did not use survey weights for this analysis, but we performed the same analysis using quarterly data from APS (2010–2019), and ran the general population analysis for different types of tobacco users in order to check the robustness of our results. The results from these two analyses can be found in online supplemental tables S1 and S2. Our GLMs were defined using binomial family and logit link to estimate OR and 95% CIs for each of the variables included in the regression. Only results for level (before/after) and slope (implementation period and 18 months after implementation) were reported here, while full list of results for the main analyses can be found in the online supplemental files 4–10. We used robust SEs, and adjustments by seasonality and autocorrelation since we estimated aggregate before/after and slope effects.
We estimated unadjusted models, which only included seasonality and autocorrelation, and adjusted models, which added other tobacco control policies implemented during the period 2006–2019 (smoking ban, change in minimum age of sale, and tobacco display ban in small and large shops), as well as our monthly average retail price variable. We first estimated the model exploring changes after May 2017, the full implementation date (level and post-intervention slope). We then added May 2016, the start of policy implementation period, and explored level changes for before/after May 2016, the slope for the implementation period (slope May 2016–May 2017), level changes before/after May 2017, and post-intervention slope 18 months after May 2017.
We estimated changes in level and slope among population subgroups by running the same models described for each population subgroup (the four model specifications six times), in order to study each group’s smoking status separately. Interaction effects by subgroup were also investigated for the main analysis. These results can be found in the online supplemental table S3.
We performed a sensitivity analysis using the same models described above but exploring level and slope changes before/after July 2017, and before/after July 2016. Using July 2017 instead of May 2017 also allowed us to disentangle policy effects from any tax effect that were not captured in our price variable, since in 2017 there were changes in tobacco taxes in March, May and November.
Finally, we plotted the linear predictions of our model for the whole sample of England against a counterfactual prediction reflecting the hypothetical situation ‘if the policy was not in place’ to visually compare smoking prevalence trends with and without (counterfactual) standardised packaging policy. To obtain the standardised packaging policy predicted trends, we ran our unadjusted models, removing adjustments by seasonality to compute linear trends. To obtain the counterfactual predicted trend, we estimated our GLM only including a time-trend variable and limited the regression to the period before full implementation (May 2017), and to the period before implementation start date (May 2016). Then, we aggregated individual-level predicted values from the unadjusted models, and from the counterfactual model to generate scatter plots of smoking prevalence combined with line graphs for the linear predictions. We performed all analyses in Stata V.16.0, and the confidence level was set to 95%.
Results
Our sample included 278 219 individual observations collected between November 2006 and December 2019, of which 48.6% were from men and 51.5% from women, 15.8% from persons aged up to 25 years and 84.2% above 25 years old, while 40.3% were classified as manual workers and 59.7% as non-manual workers.
There was a secular downward trend in smoking prevalence throughout the study period for the general population of England (figure 1A). During the standardised packaging policy implementation period (May 2016–May 2017), the prevalence of smoking was on average 17.9% (95% CI: 17.2% to 18.6%), while it was 19.2% (95% CI: 18.6% to 19.9%) in the year before the policy was implemented. Trends within age, sex and socioeconomic subgroups were similar to those in the total population, though among those aged 16–25 years, the prevalence of smoking declined rapidly from November 2006 until May 2012, then remained at around 24.9% (95% CI: 24.2% to 25.7%) until May 2016, declined to an average of 22.5% (95% CI: 21.1% to 23.8%) between May 2016 and May 2017, and then after a brief increase again assumed a decreasing trend (figure 1C).
Regression results
Our model before and after full implementation demonstrated a statistically significant level decrease in the odds of being a smoker after May 2017 (adjusted OR: 0.93; 95% CI: 0.87 to 0.99) with no statistically significant change in post-intervention slope. However, when May 2016 was included in the model (before/after policy start date), the observed level decrease in the odds of being a smoker was similar to the analysis after full implementation in both the unadjusted (OR: 0.91; 95% CI: 0.85 to 0.98) and adjusted models (OR: 0.90; 95% CI: 0.83 to 0.97), again with no significant change in post-intervention slope. Hence, our results indicate that the level decrease in the odds of being a smoker was associated with the onset of standardised packaging in May 2016, and not the full implementation of, standardised packaging and other TPD measures (table 1).
Our subgroup analyses explored each population subgroup’s smoking status and showed statistically significant step changes for women, men, those aged over 25 years old and manual occupations. For women, there was a significant level decrease after May 2016 (adjusted OR: 0.89; 95% CI: 0.79 to 1.00), which was also observed among men (unadjusted OR 0.89; 95% CI: 0.80 to 0.98). Men also showed a statistically significant level decrease in our model exploring effects before/after May 2017 (unadjusted OR: 0.92; 95% CI: 0.85 to 1.00; adjusted OR: 0.90; 95% CI: 0.83 to 0.98). For population aged over 25 years old, there was a significant level decrease after May 2017 in our model exploring effects before/after full implementation (unadjusted OR: 0.92; 95% CI: 0.87 to 0.98; and adjusted OR: 0.90; 95% CI: 0.84 to 0.96) and after May 2016 (unadjusted OR: 0.90; 95% CI: 0.83 to 0.98; and adjusted OR: 0.90; 95% CI: 0.82 to 0.98) in our model exploring effects before/after the policy start date. Finally, there was a significant decrease after May 2016 for manual occupations in our model exploring effects after the policy start date (adjusted OR: 0.85; 95% CI: 0.75 to 0.96). The only increase in the odds of being a smoker was observed among population aged 25 years old or younger after May 2017 in our model exploring effects before/after the policy start date (adjusted OR: 1.29; 95% CI: 1.03 to 1.62).
Sensitivity analysis
Our sensitivity analysis of step and trend changes between July 2016 and July 2017 (table 2) was consistent with our two models for the general population of England with similar step changes observed at both the beginning and end of the policy implementation period, though only statistically significant after May 2016 (unadjusted OR: 0.93; 95% CI: 0.86 to 1.00; adjusted OR: 0.92; 95% CI: 0.85 to 0.99).
Subgroup results in our sensitivity analyses differed in that the only step decreases observed were among men after May 2016 in our model exploring effects after the policy start date (unadjusted OR: 0.89; 95% CI: 0.80 to 0.98), among population aged over 25 years old after May 2016 in our model exploring effects after full implementation (adjusted OR: 0.93; 95% CI: 0.87 to 0.99) and among manual occupations after May 2016 in our model exploring effects after the policy start date (adjusted OR: 0.85; 95% CI: 0.75 to 0.96), and in that there was a decreasing trend observed between July 2016 and July 2017 among non-manual occupations (adjusted OR: 0.79; 95% CI: 0.64 to 0.97).
Model predictions
Figure 2 shows predictions of smoking prevalence obtained from the unadjusted model, excluding autocorrelation and seasonality, allowing us to explore linear changes before and after the implementation of standardised packaging. The two models show that there was a step decrease in prevalence and no significant change in trends 18 months after full implementation—in line with our regression results. Moreover, both the predictions from our models after full implementation and after the policy start date show no complete return to the pre-policy level of smoking prevalence, and a larger difference between the counterfactual trend and the predicted trend when the model accounts for the start of the implementation date (May 2016) in figure 2B.
Discussion
To our knowledge, this is the first study evaluating the possible impact of standardised packaging on smoking prevalence in England. Our individual-level findings found that the implementation of standardised packaging legislation was associated with a step decrease in the odds of being a smoker which was associated with the onset of standardised packaging after May 2016, when the transition to the new policy officially began, rather than May 2017 when the policy was fully implemented. Insofar that the association reflected a causal impact, the suggestion is that smokers were influenced more by the prospect of standardised packs, and possibly also of minimum pack sizes and other TPD measures, or of changes in the tobacco market introduced by the tobacco industry in advance of standardised packaging, than the actual adoption of standardised packaging. Within this study, we were unable to investigate what the underlying reasons were though we hypothesise that smokers might have been aware of the policy due to media coverage and had consider what effect this specific policy might have on their smoking behaviour (costs, no brand loyalty, lack of appealing packaging). We found no strong evidence in our subgroup analyses that this effect differed markedly between the sexes, between older and younger smokers, or those of high or low occupational socioeconomic status.
The main limitation of our study was that we were only able to include policy implementation as May 2016 and May 2017 though the policy was gradually implemented over 9 months of the 1-year transition period,11 and our model did not account for that. The standardised packaging policy was actually implemented across the UK but the results of this study were based on individual-level data from a large sample representative to population in England only. Therefore, generalisability of the findings to other UK countries and elsewhere in the world is limited and further research exploring differences in the UK countries and globally is warranted. To our knowledge, studies similar to this have not been carried out elsewhere. Also, time-series analysis using ARIMA models at the aggregated level would be more suitable for assessing the effect of policy on smoking prevalence though at the time of analysis the power was too low to produce such analysis. Therefore, we will use aggregated level data to evaluate longer term effects.
Although UK law did not require tobacco manufacturers to adopt standardised packaging simultaneously with minimum pack sizes, updated health warnings and other measures, in practice the changes were introduced simultaneously within individual cigarette brands. Consequently, we were unable to determine the effect of each of these policy elements separately. However, the findings of our study, which used individual-level data and estimated odds of being a smoker in various population groups instead of measuring the effect of policy on aggregated prevalence data, are consistent with previous research7–9 and provide further evidence that standardised packaging, or at least the prospect of standardised packaging, influences smoking status. In this study, we were unable to determine whether the reduction in odds of smoking occurred due to changes in quitting or smoking uptake though detailed analysis of each of these aspects is necessary in further research.
A possible explanation for this is that prospect of standardised packs proved to be a stimulus for smokers intending to quit smoking to act on that intention, rather than a direct visual effect of the pack itself. We have previously described substantive changes in the diversity of products available on the market both in advance of and after standard packs appeared in the UK11 and described substantial price increases with the adoption of standardised packaging.11 13 These included the introduction of, and widespread consumption of, low-price cigarettes in packs of less than 20 in advance of the change to standardised packaging.11 Whether any of these changes, introduced by the tobacco industry in preparation to standardised packs and the loss of distinctive branding and brand descriptors, contributed to the step change in smoking prevalence when the law mandating the change to standardised packs came into force is not known and will be difficult to determine. The fact remains however that standardised packaging occurred in the UK in conjunction with a wide range of other legislative and market changes, so the precise contribution of standardised packs per se to the reduction in smoking prevalence we observed is impossible to determine.
In the long run, more pronounced effects on uptake of smoking and smoking cessation might be observed related to reduced appeal of packaging and more prominent health warnings. However, our analysis focusing on short-term to medium-term effects provides clear evidence that these marked changes in packaging policy have had an appreciable beneficial effect on smoking prevalence in England.
What this paper adds
What is already known on this subject
Implementation of standardised packaging legislation is associated with changes in tobacco market and pricing.
Legislation was implemented with an aim to reduce smoking uptake and encourage quitting among existing smokers.
What important gaps in knowledge exist on this topic
Are there any changes in smoking behaviour associated with implementation of standardised packaging?
What this paper adds
This is the first paper to explore the effect of standardised packaging on smoking prevalence in England. It shows that implementation of standardised packaging legislation was associated with reduction in smoking in England.
Data availability statement
Data used for this study are available upon reasonable request from the authors.
Ethics statements
Patient consent for publication
Acknowledgments
Department of Behavioural Science and Health, University College London, London, UK.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors MOB designed the study, carried out analysis and was involved in writing up the manuscript. JBritton contributed to designing the study and was involved in writing up the manuscript. JBrown was involved in designing the study and in writing up the manuscript. EB was involved in designing the study and in writing up the manuscript. IB was involved in designing the study, contributed to analysis of the data and was involved in writing up the manuscript.
Funding The work was supported by Cancer Research UK grant number C45256/A20606. Cancer Research UK (C1417/A22962) supported the data collection and EB’s salary. All authors are members of SPECTRUM, a UK Prevention Research Partnership (UKPRP) Consortium (MR/S037519/1). UKPRP is an initiative funded by the UK Research and Innovation Councils, the Department of Health and Social Care (England) and the UK devolved administrations, and leading health research charities.
Competing interests JB and EB have received unrestricted research grants from Pfizer related to smoking cessation. JB, MOB and IB have no competing interests to declare. All authors declare no financial links with tobacco companies or e-cigarette manufacturers or their representatives.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.