Cigarette pack size and consumption: a randomized cross‐over trial

Abstract Background and Aims Smoking fewer cigarettes per day may increase the chances of stopping smoking. Capping the number of cigarettes per pack is a promising policy option, but the causal impact of such a change is unknown. This study aimed to test the hypothesis that lowering cigarette pack sizes from 25 to 20 reduces the number of cigarettes smoked. Design This randomized controlled cross‐over trial had two 14‐day intervention periods with an intervening 7‐day period of usual behaviour. Participants purchased their own cigarettes. They were instructed to smoke their usual brand from either one of two sizes of pack in each of two 14‐day intervention periods: (a) 25 cigarettes and (b) 20 cigarettes. Participants were randomized to the order in which they smoked from the two pack sizes (a–b; b–a). Setting Canada. Participants Participants were adult smokers who smoked from pack sizes of 25, recruited between July 2020 and June 2021. Of 252 randomized, 240 (95%) completed the study and 236 (94%) provided sufficient data for the primary analysis. Measurements Cigarettes smoked per participant per day. Findings Participants smoked fewer cigarettes per day from packs of 20 cigarettes [n = 234, mean = 15.7 standard deviation (SD) = 7.1] than from packs of 25 (n = 235, mean = 16.9, SD = 7.1). After adjusting for pre‐specified covariates (baseline consumption and heaviness of smoking), modelling estimated that participants smoked 1.3 fewer cigarettes per day [95% confidence interval (CI) = −1.7 to −0.9], equivalent to 7.6% fewer (95% CI = −10.1 to −5.2%) from packs of 20 cigarettes. Conclusions Smoking from packs of 20 compared with 25 cigarettes reduced the number of cigarettes smoked per day.


INTRODUCTION
Despite the global fall in prevalence of tobacco smoking in the last three decades, the growth in population has increased the number of smokers from 0.99 billion in 1990 to an all-time high of 1.14 billion in 2019 [1]. It remains one of the largest risk factors for disease globally, rated in 2019 as the second most important risk factor for all ages [2] and a major cause of the gap in life expectancy and years in good health between the richest and poorest groups [2].
The size of cigarette packs-the number of cigarettes in a single pack-is a potentially important but neglected tobacco control target [3,4]. However, the impact of cigarette pack size on smoking is uncertain. There are two key uncertainties. First, does smoking from a smaller pack size reduce the number of cigarettes smoked? Secondly, does smoking fewer cigarettes per day increase the chances of subsequently stopping smoking? The aim of the current study is to address the first uncertainty.
An increasing number of jurisdictions have set a minimum size of 20 cigarettes per pack to make cigarettes less affordable to young people, in accordance with the international recommendations under the WHO Framework Convention on Tobacco Control [3][4][5][6][7][8]. Packs of 20 cigarettes are standard in most countries, although larger pack sizes are common in some countries such as Canada and Australia [3].
In an earlier parallel group study using a two-stage adaptive design we randomized smokers in Australia, who usually purchased cigarettes in packs of at least 25, to smoke for 4 weeks either from their usual pack size or from packs of 20 cigarettes. At the interim assessment stage (when 124 participants had been randomized) this adaptive trial was halted as the estimated required total sample of more than 1000 exceeded pre-specified criteria for feasible recruitment [9]. A crossover design, making within-group comparisons, was deemed more efficient, as these typically require smaller numbers of participants.
Based on robust evidence from studies of food consumption [10], it is plausible that smaller pack sizes reduce cigarette consumption.
Limited, non-experimental evidence suggests an association between cigarette pack size and consumption. For example, those smoking more heavily tend to purchase cigarettes in larger packs [11]. In a more recent study based on a hypothetical purchase task, smokers wanting to self-regulate their consumption reported a preference for smaller packs [12].
The causal nature of the association between pack size and cigarettes smoked per day remains uncertain. The aim of the current randomized controlled cross-over trial is to assess the impact on cigarette consumption of using packs of 20 cigarettes compared to packs of 25 cigarettes.

METHODS
The study was approved by the University of Cambridge Psychology Research Ethics Committee (PRE.2019.068) and the University of Waterloo Research Ethics Board (#41353). Participants provided written informed consent to participate.
The study protocol was prospectively registered with ISRCTN on 6 March 2020 (ISRCTN16013277) and the Open Science Framework (OSF) on 21 November 2019, updated on 9 July 2020 (https://osf.io/ zby94). Recruitment was planned to start in March 2020, but was delayed until July 2020 due to the COVID-19 pandemic. The statistical analysis plan was uploaded to both OSF and ISRCTN on 20 July 2021 prior to data analysis.

Study design
The study was a randomized controlled cross-over trial with two conditions-pack size of 20 cigarettes versus pack size of 25 cigarettes-separated by a period of 'usual behaviour'.

Participants
A total of 252 smokers in Canada participated in the study from July 2020 until June 2021, recruited by a research agency (https:// leger360.com/). In Canada cigarettes are sold in two pack sizes, 20 and 25. Most sales (71%) are for the larger pack sizes [13]. Eligible participants were those with the following characteristics: • aged 19 years and over • did not live in the same household as someone who has enrolled in the study.

Intervention
Participants were instructed to smoke their usual brand variant and length (king size or regular) of cigarettes from a single size of cigarette pack in each of the two intervention periods: (a) 20 cigarettes and (b) 25 cigarettes. Each intervention period lasted 14 days. Participants could smoke cigarettes of any brand variant and from either pack size in the 'usual behaviour' period, which lasted at least 7 days between the intervention periods. 2 This is classified as a size × product intervention within the typology of interventions in proximal physical micro-environments (TIPPME) [14].

Procedure
Participants were recruited and screened for eligibility by a research agency (https://leger360.com/), which sent a screening questionnaire based on the inclusion and exclusion criteria to 84 873 members of their research panel known to be smokers or on whom there was no information regarding their smoking status. Those who passed the screen were sent the study information sheet and invited to participate. The research agency referred 538 potential participants to the research team in total.
The study was presented as investigating how cigarette pack size affects the effectiveness of health warnings. This was to reduce the chance of participants focusing on their cigarette consumption in relation to pack size. Potential participants provided information on their demographic characteristics and smoking behaviour. Potential participants were asked to purchase one pack of 20 and one pack of 25 cigarettes in their usual brand variant. They sent photographs of these packs and their receipts to the research team. This was to check the eligibility criterion that they used a brand or brand variant available to buy in both pack sizes from a shop convenient to them.

Randomization
Allocation of participants to the order in which they completed the conditions was determined using a computer-generated random number sequence prepared by one of the project statisticians (R.M.), using Stata version 15 (StataCorp LLC, TX, USA). Block randomization was used to generate an equal number of participants allocated to each treatment order and to reduce the potential of selection bias compared to simple randomization. The blocks were in sizes of 2, 4 and 6. The random number sequence, with IDs for the sequence of potential participants, was concealed from the research team, research agency and participants until the participant had consented to take part in the study and shown that they were able to purchase their usual brand variant of cigarettes in packs of 20 and packs of 25.
When a participant was deemed eligible for randomization the research team accessed the next random allocation in the sequence, and this participant was assigned the corresponding ID. Instructions (see Supporting information, Material 1) were e-mailed and mailed to participants, together with a set of stickers to attach to all the cigarette packs from which they would smoke during the intervention periods.
The stickers had space to record the following information:

Sample size estimation
The within-person standard deviation (SD) of cigarettes smoked per day obtained in our previous study [9] was used to calculate that, with 80% power, we would detect a difference of 1.5 cigarettes smoked per day (which is a size of importance to detect, and consistent with previous research, Lee et al. 2021) as significant at a two-sided significance level of 5%, with 210 participants available for analysis (105 per sequence group). To account for potential attrition, 252 participants were recruited for randomization.

Statistical analysis
Intention-to-treat analyses of the primary and secondary outcome included all randomized participants who completed at least one intervention period. A mixed-effects normal regression model was used to estimate the mean difference according to pack size condition, with a 95% confidence interval (CI) and P-value. This involved a repeated-measures analysis with terms included for the treatment effect, period effect (intervention periods 1 and 2) and order effect [(a-b) or (b-a)]. Evidence for a treatment × order interaction was examined. As pre-specified in the statistical analysis plan, the following variables were examined as potential analysis model covariates: (1) self-reported number of cigarettes smoked per day at enrolment; (2) HSI, as measured at enrolment; (3) MTSS, as measured at enrolment; (4) price per cigarette for packs of 20 and packs of 25 cigarettes; (5) duration of 'usual behaviour' period; and (6) number of non-study cigarettes smoked.

Participant characteristics
Of the 538 smokers referred by the research agency, 349 (65%) met the eligibility criteria, of whom 252 (72%) consented to participate and were randomized to one of the two study arms (Figure 1).
Of those randomized, 240 (95%) completed the study; 236 (94%) provided sufficient data for the primary analysis of the primary outcome.
Demographic characteristics appeared comparable between participants who completed the study (n = 240, 94%) and those who did not (n = 12; 6%) ( Table 1). Those completing the study were predominantly female (n = 171, 71%) and white (n = 208; 87%), with a mean age of 46 years (SD = 12). Forty-five per cent of participants (n = 104) had household incomes less than the average of approximately Can$60 000, with 21% (n = 51) having one or more university degrees. Participants were deemed to be adherent to study instructions if at least 90% of the cigarette packs from Of the 4952 cigarette packs, 58 had at least one piece of information missing, either from the sticker or information relating to the cigarette type (brand variant, length or size). Of the 236 participants included in the primary analysis of the primary outcome, 12 had missing information imputed for the primary outcome. See the OSF registration for the protocol that was followed for handling missing information (https://osf.io/zby94), which assumes any missing packs were missing at random [17].  to −0.9) or 7.6% fewer (95% CI = −10.1% to −5.2%) from packs of 20 cigarettes (Table 3). The interaction between pack size and allocation order was investigated but was not retained in the final model (P = 0.286). These findings were robust to different approaches to imputation for data missing from 12 participants.
This showed that MTSS tended to increase from intervention period 1 to intervention period 2, regardless of the pack sizes received.

Other outcomes
Responses to the end of study questionnaire are described in Supporting information, Material 6. We report here only responses to the question regarding preferred pack size, following completion of the study.
Eighty-two per cent of participants (n = 234) preferred using packs of 25, 16% preferred using packs of 20 and 2% expressed no preference. For those preferring pack sizes of 25, the most cited reason was that the packs of 25s lasted longer (n = 62). Among those preferring pack sizes of 20, the most cited reason was that they smoked fewer cigarettes (n = 15) (see Box 1).

DISCUSSION
Using packs of 20 compared with 25 cigarettes reduced the mean number of cigarettes smoked per day by 1.3, a reduction of 7.6%.
Motivation to stop smoking was similar when smoking from either pack size. This study therefore provides the strongest evidence to date that smaller pack sizes reduce the number of cigarettes smoked.
This finding was robust to four sets of sensitivity analyses.

Comparison with previous findings
The results of the current study provide the first evidence that at least part of the observed association between cigarette pack size and consumption-those smoking from larger pack sizes smoke more cigarettes [11]-is causal. These results also fitted with the more robust evidence from studies of food and alcohol consumption showing that smaller portion and pack sizes reduce consumption [10,18]. They are also consistent with the growing evidence from studies of alcohol consumption showing similar effects on consumption of reducing glass and bottle sizes [19,20]. While food, alcohol and tobacco vary in the extent to which they are addictive, consumption of even the most addictive-tobacco-is influenced nonetheless by external cues such as constraints on where it can be consumed, with some evidence that such constraints lead to voluntary restrictions on home smoking [21].

Interpretation of findings
A reduction in cigarettes smoked is an appropriate aim for tobacco control policy if it leads either to direct or indirect health benefits.
Therefore, one critical question is whether the magnitude of reduction observed in this study is likely to translate to population health benefit. A decrease of 1.3 cigarettes per day, or fewer than 10% of usual cigarettes smoked, is unlikely to confer any direct reduction in risk due to decreased exposure. First, smoking reduction studies suggest that a minimum reduction of 50% consumption is required to reduce indicators of risk [22,23]. Secondly, smokers typically engage in compensatory smoking, whereby reductions in cigarette consumption at the individual level are offset by compensatory increases in smoking behaviour, such as taking more puffs and deeper inhalation, in an effort to maintain nicotine intake [24]. Compensatory changes to reductions in cigarettes per day have also been observed at the population level: average cigarette consumption has declined in the United States over several years, but nicotine intake among smokers has remained unchanged [25]. Rather, the extent to which a reduction in cigarettes per day might contribute to harm reduction depends, in part, on the extent to which smoking fewer cigarettes a day increases the chances of quitting smoking, indirectly leading to health improvement.
There is evidence that smoking fewer cigarettes per day-which may be a marker of lower dependence-increases the probability of cessation. Smoking one fewer cigarette per day at baseline was estimated in one study to increase the chances of being a former smoker by between approximately 7 and 11% [26]. There is also evidence that deliberate attempts to reduce cigarette consumption increase the likelihood of subsequent cessation [27]. However, the impact of interventions that aim to reduce cigarette consumption that do not require deliberate effort on the part of the cigarette smoker is unknown. It is possible that reduced consumption might, over time, lead to extinction of conditioned associations between the behaviour of smoking (and associated cues) and the reward association with nicotine consumption [28,29]. However, this may not occur for an average reduction of one cigarette per day and, if it does, it is unknown whether it would be of sufficient magnitude to reduce dependence and promote cessation. The current study was not designed to address these questions. Therefore, while we provide evidence that capping pack size might lead to a reduction in cigarette consumption-an important proof of concept-the impact on population smoking rates remains unknown. We judge the potential effect likely to be small or negligible.
Motivation to stop smoking was low in this sample of smokersthat is, below the mid-point of a scale for motivation to stop smoking-reflecting the inclusion criterion that participants did not intend to quit smoking in the next 3 months. An unpredicted interaction between pack size and order was found, such that motivation to stop smoking increased from intervention period 1 to intervention period 2 regardless of which pack size was used first, and participants who were allocated to use packs of 20 cigarettes first had lower motivation to stop smoking at the end of both intervention periods compared to participants who were allocated to use packs of 25 cigarettes first.

Strengths
This is the first experimental study, to our knowledge, to estimate the impact on daily cigarette consumption of smoking from smaller packs.
Retention of randomized participants was very high which, together with the study design and procedures, minimized the risk of bias.

Limitations
The two main limitations of the study concern: first, its generalizability

Implications for research
The current study raises several questions for further research. These concern first, the mechanisms for the observed effect and secondly, the optimal pack size for cigarettes. Understanding the mechanisms by which cigarette pack size affects consumption could provide the basis for optimizing the intervention [30]. Free text responses to the end of study questionnaire were compatible with two linked possible explanations for what is known as 'the portion size effect'-the tendency to consume more the larger the portion or package [31]. First, smaller packs might reduce consumption by making it more effortful to smoke more-that is, to buy or open a new pack [10]. Secondly, it might reflect the tendency to consume a specific number of units in a pre-specified period of time [32]. This might be one glass or bottle of wine with dinner, and one pack of cigarettes in a day or during a 2-day period. When the glass, bottle or pack contain more, more is consumed. Regarding the optimal cigarette pack size, the current study compared just two pack sizes. It is unknown whether the size of effect observed is linear-that is, whether smoking from a pack size of 25 compared with a pack of 30 would reduce the number of cigarettes smoked daily by the same proportion-approximately 7%-as that observed in the current study when smoking from a pack size of 20 compared with a pack of 25.

Implications for policy
The novel findings from this study raise questions regarding whether pack size might be a useful target for intervention in addition to existing tobacco control policies.
There are at least two ways to shift smokers away from buying larger packs, which are not mutually exclusive. The first is to use price-based measures to ensure that purchasing cigarettes and tobacco in larger quantities-in packs or bundles of packs in cartonsdoes not provide value for money. As noted by some participants in the current study (Box 1), their preferences for larger packs reflected better value for money; that is, paying less per cigarette when buying in packs of 25 than in packs of 20.
Proportionate pricing of cigarettes and hand-rolling tobacco is needed so that the price per stick or gram of tobacco is the same, regardless of pack size. This builds upon the robust evidence that tax and price increases are the most impactful tobacco control policies, particularly for children and young adults [33]. The second approach, to complement such price-based measures, is to regulate the maximum size at which cigarettes and tobacco can be sold. An increasing number of countries have regulated a minimum pack size of 20 to reduce the affordability of cigarettes to children [3] resulting, for example, in a virtual disappearance of pack sizes below 20 in the European Union [6]. A few have regulated to cap the maximum size at 20 [3]. Different tobacco control policies may therefore need to be balanced to find an optimum pack size. While the current study cannot be used to determine whether and at what size a maximum cap might be set, it provides the first experimental evidence that people smoke more when smoking from larger pack sizes.

CONCLUSION
Smoking from packs of 20 compared with 25 cigarettes reduced the number of cigarettes smoked per day. The extent to which the observed reduction would translate into reduced population rates of smoking remains uncertain.