The role of alcohol packaging as a health communications tool: An online cross-sectional survey and experiment with young adult drinkers in the United Kingdom

Introduction. Alcohol packaging is a potentially valuable means of communicating product and health-related information, with growing academic and political interest in its role as a health communications vehicle. Methods. An online cross-sectional survey and experiment were conducted with a non-probability sample of 18 – 35-year-old drinkers in the United Kingdom ( n = 1360). The survey assessed exposure to, and engagement with, current messaging on packs, and support for displaying product and health-related information. For the randomised experiment, participants were shown, and asked questions about, a vodka bottle with either no warnings (control), small text warnings, large text warnings or pictorial (image-and-text) warnings; the main binary outcome measures were negative product appeal and social acceptability, and positive cognitive and behavioural impact. Results. Two-ﬁ fths of the sample rarely or never saw on-pack health-related information, with almost three-quarters rarely or never reading or looking closely at this. There was strong support for displaying a range of product and health-related information (e.g. units, ingredients) on packs. Relative to the control, products with warnings were more likely to be perceived as unappealing and socially unacceptable, and to positively impact alcohol-related cognitions and behaviours. For example, pictorial warnings were 10 times as likely to positively in ﬂ uence cognitions and behaviours ( AOR = 10.01, 95% CI: 8.09, 17.46). Discussion and Conclusions. Alcohol packaging could have an important role in delivering health messaging. Large pictorial or text warnings may help counteract the appeal and social acceptability of alcohol products and increase awareness of risks, potentially supporting a reduction in consumption and related harms. [Jones D, Moodie C, Purves RI, Fitzgerald N, Crockett R. The role of alcohol packaging as a health communications tool: An online cross-sectional survey and experiment with young adult drinkers in the United Kingdom. Drug Alcohol Rev 2022;41:1206 – 1215]


Introduction
Alcohol packaging can be used to communicate the harms associated with consumption.There is clearly a need to do so given that alcohol misuse is a key risk factor for illness, disability and death, implicated in more than 200 diseases and responsible for approximately 3.3 million deaths annually [1].Although academic interest in warnings on alcohol packaging is in its infancy when compared to tobacco, evidence suggests that well-designed warnings on alcohol products can capture attention, inform consumers of possible risks, increase awareness and knowledge of harms, decrease speed of drinking and support a reduction in consumption, with moderate consumer support for warnings [2][3][4][5][6][7].Research in the United Kingdom (UK) with young adult drinkers, for instance, found that large, pictorial (image-and-text), front-of-pack warnings with specific health conditions could help reduce individual and social appeal of alcohol products [8].
In the UK, alcohol harm is an important public health issue, with 8974 alcohol-specific deaths in 2020 [9,10].The appearance of alcohol packs in the UK does not reflect this risk, with the inclusion of warnings voluntary; where used, these are typically restricted to Daniel Jones MSc, PhD Student, Crawford Moodie PhD, Senior Research Fellow, Richard I. Purves PhD, Research Fellow, Niamh Fitzgerald PhD, Professor, Rachel Crockett PhD, Lecturer.Correspondence to: Mr Daniel Jones, Institute for Social Marketing and Health, Faculty of Health Sciences and Sport, University of Stirling, Stirling, FK9 4LA, UK.Tel: +44 (0)1786 467390; E-mail: daniel.jones@stir.ac.uk small symbols warning not to drink if pregnant or driving [11].A review of 424 alcohol packs randomly selected in UK shops found that only one (0.24%) included a factual health statement ('Alcohol consumption is injurious to health') on the back label, while 423 (99.76%) packs either had no statement or only a responsibility statement (e.g.'drink responsibly') [12].The only product and health-related information legally required on alcohol packaging in the UK is volume, strength/alcohol by volume (ABV) and presence of common allergens [13], which is not considered to meaningfully inform consumers let alone influence drinking behaviours [8,12].Research suggests that consumers in the UK are not engaging with the information currently provided on alcohol packaging and generally support the inclusion of product and health-related information (e.g.number of alcohol units, calories) [8,14,15].
Academic and political interest in alcohol packaging in the UK is growing [13], as it is elsewhere [7,16,17].This study aimed to assess a sample of young adult drinkers' (i) exposure to, and engagement with, current messaging on packs; (ii) support for displaying product and health-related information; and (iii) reactions to randomly-allocated frontof-pack warnings in terms of product appeal, social acceptability and cognitive and behavioural impact.Based on previous alcohol and tobacco research [18,19], we hypothesised that more salient warnings (i.e.large text, pictorial) would be more effective than small-text warnings and the absence of warnings in reducing product appeal and social acceptability, and positively impacting alcohol-related cognitions and behaviours.

Design and sample
An online cross-sectional survey was conducted in September-October 2020 with a non-probability sample (n = 1360) of 18-35-year-old current drinkers residing in the UK.To explore the impact of the presence and type of warning on product appeal, social acceptability, and cognitive and behavioural impact (aim 3), we incorporated a between-groups experiment into the survey.Surveys and experiments are regularly used in alcohol packaging studies [2,14,[19][20][21], with young adults frequently targeted by alcohol producers when (re)designing packaging [22] and a key population for public health given high levels of hazardous drinking [23].

Recruitment
River sampling is common in non-probability online surveys [24].Participants were recruited via targeted Facebook adverts, a means of accessing specific demographic profiles across a wide geographic area [25], as with previous research involving young adult drinkers [26].The adverts stated that a University of Stirling study about alcohol use and packaging was recruiting 18-35-year-olds living in the UK who drink alcohol (i.e.eligibility criteria), with the chance to win one of 10 £50 Love2shop e-gift cards.Potential participants were invited to complete the survey, which was hosted by Jisc and accessible on any device with an internet connection, via a link in the adverts.

Procedure
Those clicking on the link were informed about the study aims, content, duration (10 min), data privacy and that participation was voluntary and that they could exit the survey at any time, before being asked to provide consent.Participants indicated their country of residence and age, with those living outside the UK or not aged 18-35 screened out.To be considered a current drinker and therefore eligible to participate, participants had to answer anything other than 'Never' to the screening question 'How often do you have a drink containing alcohol?', the first item of the Alcohol Use Disorders Identification Test for Consumption (AUDIT-C) [14].Participants were asked about sociodemographics and drinking behaviours, exposure to, and engagement with, current messaging on packs, support for displaying product and health-related information and response to novel warnings.Most items were compulsory but included 'Neither agree nor disagree' or 'Don't know/ Prefer not to say' options where applicable.
For the between-groups experiment assessing the impact of warning design, participants selected one of four words ('Cord', 'Lock', 'Plug', 'Shed') and, similar to a previous online experiment [20], were randomly allocated to one of four conditions (Figure 1): control (no warnings, n = 383), small-text warnings (n = 307), large-text warnings (n = 350) or pictorial warnings (n = 320).Participants were unaware of the purpose of the word selection and were informed that it did not matter which random word they chose as there was no right or wrong answer, thus, eliminating bias in treatment assignments [27].
All participants were able to enter a prize draw at the end of the survey for their time and provided contact details for alcohol support.Ethical approval was granted by the General University Ethics Panel (GUEP 945) at the University of Stirling.The survey protocol was not pre-registered.

Materials
Each warning set included one general ('Alcohol damages your health') and two specific ('Alcohol causes liver disease', 'Alcohol causes mouth cancer') warnings.The specific warnings were selected as more than three-quarters of alcohol-specific deaths in the UK in 2020 were caused by alcoholic liver disease [9] and past research suggests that it is more effective to specify the type of cancer [8], with alcohol-related mouth cancer prevalent in the UK [28].For those in the pictorial warning condition, an appropriate image was chosen to reflect each warning: 'Alcohol damages your health' (image of blood pressure test); 'Alcohol causes liver disease' (image of person clutching their liver); 'Alcohol causes mouth cancer' (image of CT scanner in a hospital).For consistency, in each condition participants were shown an image of a bottle of Smirnoff Red Label No. 21 vodka.Smirnoff was chosen because it was the most popular alcohol brand in the UK and the highest-selling vodka, with positivity ratings similar among males and females and highest among millennials, who comprised most of the sample [29,30].

Measures
Exposure to, and engagement with, current messaging and support for displaying product and health-related information.To address aim 1, participants were asked how often they see, and read or look closely at, the health-related information, messages or warnings on alcohol packaging [8,14], with responses coded as 'Sometimes', 'Often' or 'Always' versus 'Rarely' or 'Never'.To address aim 2, we explored whether participants supported displaying the following information on alcohol packaging: strength/ABV; number of units; ingredients; number of servings equal to recommended weekly guidelines; weekly unit guidelines for men and women; calories; health conditions which can result from drinking alcohol; health warnings on the front of packs [31].Responses were 'Strongly Disagree' (1) to 'Strongly Agree' (5).
Response to packaging with/without warnings.For aim 3, participants were shown images of a Smirnoff bottle with warnings (experimental conditions) or without warnings (control condition) and asked a series of questions (Table S1) informed by previous research [8,20,32], e.g.'These alcohol products would make me aware of the health risks of drinking'.For each question, responses were 'Strongly Disagree' (1) to 'Strongly Agree' (5).The responses to the item 'I find these alcohol products off-putting' were reverse-coded.Sample characteristics.Information on country of residence, age, gender, ethnicity, religiosity, occupation, education and social grade was captured; social grade was categorised according to the occupation of the person in the household with the greatest income [33], with grades A, B and C1 indicating higher and middle-class groups and C2, D and E working-class groups.
Alcohol consumption.The three-item AUDIT-C scale was used to measure consumption [14]: (i) frequency of consumption (0 = 'Never' to 4 = 'Four or more times a week'); (ii) number of units drunk in a typical drinking occasion (0 = 'One or two' to 4 = 'Ten or more'); and (iii) how often they drunk six or more units if female, or eight or more if male, on a single occasion in the last year (0 = 'Never' to 4 = 'Daily or almost daily').
Awareness and perceptions of Smirnoff vodka.Participants were asked about awareness of the Smirnoff vodka brand and how they rated it on a scale from 1 = 'Very Unappealing' to 5 = 'Very Appealing'.Participants were asked these questions as a Smirnoff bottle was used to display the novel warnings.

Analysis
Data were analysed using SPSS (Version 28).Analyses were unweighted and were not pre-registered.A total of 3987 people clicked on the survey link, with 1511 completing it and the remainder screened out or choosing to exit.To ensure meaningful engagement with the survey content, we excluded participants in the bottom five percentile (range: 04:01 to 05:53, n = 76) and top five percentile of completion time (range: 24:05 to 10:02:14, n = 75).Identifying implausible response times is a common means of removing careless respondents and improving data quality in alcohol research [34], with a median completion time of 09:07 for the final sample (n = 1360).Frequencies examined exposure to, and engagement with, current messaging (aim 1), see Table 2, and support for displaying product and healthrelated information (aim 2), see Table 3. Principal components analysis using varimax rotation was conducted to determine the factorability of warning response items (Table S1), with two components extracted, defined and used as composites: 'appeal and social acceptability' and 'cognitive and behavioural impact'.Binary outcomes are commonly measured in global health research [35], and have been used in alcohol and tobacco packaging studies [21,36].Comparable to previous tobacco research [37], composite scores were recoded into binary variables based on the mid-point of possible scores to mutually exclude, and enable comparison of, positive and negative reactions to the stimuli.The main outcome measures were product appeal and social acceptability scores (midpoint = 21), coded as negative (≤20, n = 611) or neutral/positive (≥21, n = 749) (Table S2), and cognitive and behavioural impact scores (mid-point = 15), coded as positive (≥16, n = 612) or neutral/negative (≤15, n = 748) (Table S3).
For aim 3, hierarchical binary logistic regression models with simple contrasts examined the main effect of warning condition on negative product appeal and social acceptability (Table 4), and positive cognitive and behavioural impact (Table 5), respectively.In both models, block one controlled for sociodemographic and drinking-related factors identified in past research as influencing responses to warnings [21,[38][39][40][41]: age, gender, social grade, higher education qualifications or professional/vocational equivalents, occupational status, AUDIT-C category and Smirnoff appeal, with warning condition entered in block two.The control condition was chosen as the reference category to The role of alcohol packaging determine the effect of including warnings on packaging, with subsequent analyses using the large-text condition as the reference category to assess the relative effect of warning design.
Exposure to, and engagement with, current messaging and support for displaying product and health-related information While most participants (59.9%) reported seeing healthrelated information, messages or warnings on alcohol packaging sometimes, often or always, two-fifths (40.1%) rarely or never do, with most (69.9%)rarely or never reading or looking closely at it (Table 2).Participants largely supported displaying a range of product and health-related information on packaging (Table 3).For example, 94.7% agreed or strongly agreed that units should be displayed on packaging, with 88.8%% agreeing or strongly agreeing that ingredients should be included.

Predictors of negative appeal and social acceptability, and positive cognitive and behavioural impacts
Both hierarchical logistic regression models were statistically significant compared to respective null  4 and 5) and multicollinearity not present (variance inflation factors ≈ 1).After controlling for covariates, participants who viewed products with warnings were significantly more likely to perceive the products as unappealing and socially unacceptable than the control condition ( (P = 0.011), and having higher education qualifications or professional vocational equivalents (P < 0.001) and a lower Smirnoff rating (P < 0.001) were associated with negative appeal and social acceptability in the final model.After controlling for covariates, participants who viewed products with warnings were significantly more likely to report positive cognitive and behavioural impacts than the control condition (Table 5): smalltext warnings, AOR = 8.87, P < 0.001, 95% CI [6.00, 13.10]; large-text warnings, AOR = 11.88,P < 0.001, 95% CI [8.09, 17.46]; pictorial warnings, AOR = 10.01,P < 0.001, 95% CI [6.78, 14.77].Relative to the large-text condition, the likelihood of positive cognitive and behavioural impacts was not significantly different in the small-text condition, AOR = 0.75, P = 0.075, 95% CI [0.54, 1.03], or pictorial condition, AOR = 0.84 P = 0.296, 95% CI [0.61, 1.16].Of the other covariates, only being female (P = 0.003) and a lower-risk drinker (P < 0.001) were associated with positive cognitive and behavioural impact in the final model.

Discussion
This sample of young adult drinkers were influenced by warnings on alcohol packaging.We found that including a front-of-pack health warning (small-text, large-text or pictorial) on a market-leading brand of vodka reduced product appeal and social acceptability, and could positively influence alcohol-related cognitions and behaviours.As expected, participants who viewed the product with no warnings had the highest social appeal and acceptability ratings, and the lowest reported impact on cognition and behaviour.That those in the control condition were not informed about alcohol-related harms via the packaging is reflective of products on the market, but needs to be considered when interpreting the findings.
Larger warnings (with or without an image) were particularly effective in reducing product appeal and social acceptability, with previous research indicating that prominent warnings on alcohol packaging may help capture attention, counteract positive product perceptions, and reduce consumption intentions [19,43,44].Surprisingly, we did not find clear differences between warning designs in terms of positively impacting cognitions and behaviours.This is in contrast to the considerable evidence in the tobacco packaging field, where larger warnings with images are more capable of positively influencing smoking-related thoughts and behaviours than smaller, text-only warnings [18,45].Further research is needed to determine optimal alcohol warning designs.For instance, one study suggests that pictorial warnings may communicate messages more effectively than having just text as it requires less cognitive effort from consumers [8], while others report inconclusive differences in effectiveness (e.g.reducing speed of consumption, increasing risk perception) between pictorial and text-only conditions [6,46,47].
Two-fifths of participants reported rarely or never seeing health-related information, messages or warnings on alcohol packaging, with almost three-quarters, rarely or never reading or looking closely at it.As most of the sample were classified as higher-risk drinkers, it is reasonable to assume they are exposed to alcohol packaging when drinking at home or in licensed premises, yet most did not interact with the information currently provided.This is unsurprising considering alcohol packaging in the UK is not designed to meaningfully engage or inform consumers, with ambiguous messaging in small fonts usually positioned on the back of packs and over 70% of labels not including upto-date low-risk drinking guidelines [8,11,12].Salient warnings with specific health-related messages, as used in our study, can positively influence consumer attention, comprehension, recall, judgement and behavioural compliance [3].The lack of exposure to warnings on alcohol packaging may help to explain why we failed to find significant differences between pictorial and text warnings in terms of perceived cognitive and behavioural response, in direct contrast to the vast majority of studies in the tobacco field [45].
Warnings have been displayed on cigarette packs for several decades in many countries, so even prior to the inclusion of pictorial images on packs from this century (they were first required in Canada in 2000) people would have been exposed to, and familiar with, warnings on packs.For alcohol, the lack of exposure to any meaningful warning messages on packs may have meant that the novelty of seeing these, irrespective of warning size and type, was sufficient to influence their perceptions.Qualitative research exploring possible reasons for these findings would be of value.
There was strong support for displaying strength/ ABV, number of units, ingredients, weekly guidelines (by serving and gender) and calories on alcohol packaging, with moderate support for displaying warnings, consistent with international findings [7,48,49].Limited consumer engagement with current provisions and support for more information on packaging may be partially explained by legal requirements.While alcohol packaging requirements vary globally [50], minimal information is legally required in the UK (volume, strength/ABV, presence of common allergens) [13], with other product and health-related information (e.g.ingredients, calories, warnings) self-regulated by alcohol companies.Tinawi et al. [51] argue that voluntary warnings on alcohol packaging in New Zealand, like the UK, are inadequately designed to inform consumers and that evidence-based, standardised requirements outlining alcohol-related risks are necessary.As over 740 000 of all new cancer cases in 2020 globally were attributable to alcohol consumption [52], with alcoholic liver deaths in the UK increasing [53], our findings could help inform packaging regulations to meaningfully inform consumers about alcoholic products and potential harms.
This study has several limitations to consider.While the warnings used aligned with World Health Organization [54] recommendations (e.g.evidence-based content, different messages in bold and using capital letters, placed in a standard location with clear separation of text from other information, and inclusion of images), this study is unable to provide insight into optimal message content (e.g.specific vs. general text, message framing, image criteria) and real-world reactions to, or longer-term impacts of, warnings on alcohol packaging.The images used were not pre-tested, it should be noted, as would be the case if pictorial warnings were required on packs, and as such the images may not have resonated with or been clear to participants.The images were also somewhat benign, with some arguing that the goal of deterring images is simply to inform consumers and questioning the need to evoke fear or an elevated perception of risk to prevent people from drinking [47]; nonetheless, more severe images (e.g.diseased organs) could elicit stronger responses from drinkers, with alcohol and tobacco research suggesting that severe images are more likely to elicit negative emotional reactions and avoidance behaviour, reduce product appeal, and increase thinking about harms and motivation to drink less [55][56][57][58].
Non-probability online surveys are cost-effective and efficient yet lack representativeness and generalisability, partially due to topical self-selection bias and potential differences between those who choose to participate and those who do not [24,59,60].It has been argued that river samples, like ours, are not formally generalisable beyond the population of users accessing the service (e.g.Facebook) within the sampling period [24].While careless responding is no more prevalent in alcohol research than other fields [34], we excluded participants with implausible completion times to improve data quality; including an attention check (e.g.careful-read item) could have helped to identify additional careless respondents, potentially further increasing data quality and motivating participants to exert more cognitive effort when answering survey items [61].
Although controlling for Smirnoff vodka brand appeal helped overcome the limitation of showing one product, which ensured brand awareness, consistency and removed potential bias or confounding effects from exposure to multiple products [37], a broader range of alcoholic drinks, brands and pack formats would better represent the choices and preferences consumers have, which may elicit different reactions.This study had a small budget and while this enabled a modest sample size and we achieved relatively good spread across the warning conditions, our sample was not sufficiently large to allow for either more warnings or warnings on a variety of alcohol products.As over 80% of the sample were classified as higher-risk drinkers, future research with other populations that may benefit from exposure to warnings, such as non-drinkers, susceptible drinkers, adolescents and older adult drinkers, is needed.This sample of young adult drinkers reported limited engagement with the health information, messages and warnings on current alcohol packaging and largely supported the inclusion of product and health-related information.Prominent warnings may help to counteract the appeal and social acceptability of alcohol products, encourage consumers to think about their drinking and, potentially, support a reduction in alcohol consumption and related harms.

Figure 1 .
Figure 1.Alcohol packaging displaying no warning (control) and warnings differing by text-size (small; large), form (text-only; pictorial) and image content.

Table 1 .
Sample characteristics bAs the third AUDIT-C item measures binge drinking by gender ('How often have you had 6 or more units if female, or 8 or more if male, on a single occasion in the last year?'),cumulative AUDIT-C scores were not computable for those who did not indicate their gender as either male or female (n = 30), with two participants preferring not to answer the third item.AUDIT-C, Alcohol Use Disorders Identification Test for Consumption.

Table 3 .
Support for displaying product and health-related information on alcohol packaging a How many servings of the product are equal to the recommended weekly guidelines.b Guidelines on how many units men and women should drink each week.c Information on health conditions which can result from drinking alcohol.d Health warnings on the front of alcohol packaging.ABV, alcohol by volume.

Table 2 .
Exposure to, and engagement with, health-related information, messages or warnings on alcohol packaging a To measure frequency of exposure, participants were asked 'How often do you see health-related information, messages or warnings on alcohol packaging?'.b To measure frequency of engagement, participants were asked 'How often do you read or look closely at the health-related information, messages or warnings on alcohol packaging?'.1210 D. Jones et al. models, with acceptable goodness-of-fit as per Hosmer-Lemeshow tests (Tables

Table 5 .
Logistic regression exploring association between warning condition and positive cognitive and behavioural impact