The Role of Affective Reactivity Induced by Cigarette Packaging including Graphic Warning Labels: The CASA Study

Objective: To identify whether three types of cigarette pack designs, including three versions of Graphic Warning Label (GWL) plain packs, one GWL absent and branding absent pack (Blank) and the smoker’s own GWL absent and branding present pack (US), elicit different valence, type and levels of affect. Design: US daily smokers (n=324) were asked to handle each of the five pack types and “Think Aloud” their reactions. To avoid a muted familiarity response, exposure to their own US pack followed exposure to at least one GWL plain pack. Reactions were scored on a reactivity scale (−3 to +3) and the text was coded for speech polarity (−1 to +1) and emotive word frequency. Results: Reactivity scores had excellent inter-rater reliability (agreement≥86%; ICC≥.89) and were correlated with speech polarity (rho’s=.21-.37, p-values<.001). When considering their US pack, approximately two-thirds of smokers had a low (31.5%) to medium (34.6%) positive response (reactivity=1.29; polarity=0.14) with expressed feelings of joy and trust. Blank packaging prompted a largely (65.4%) neutral response (reactivity=0.03; polarity=0.00). The gangrenous foot GWL provoked mostly medium (46.9%) to high (48.1%) negative responses (reactivity=−2.44; polarity=−0.20), followed by neonatal baby (reactivity=−1.85; polarity = −0.10) and throat cancer (reactivity=−1.76; polarity=−0.08) warnings. GWLs varied in their elicitation of disgust, anger, fear, and sadness. Conclusion: Initial reactions to GWL packs, a blank pack, and smokers’ current US pack reflected negative, neutral, and positive affect, respectively. Different versions of the GWL pack elicited different levels and types of immediate negative affect.


INTRODUCTION
Cigarette packaging offers a point-of-use marketing opportunity to influence both a smoker's behaviour and the perceptions of observers, particularly young people. [1][2][3] Branded marketing on cigarette packages is associated with positive affect that supports the decision to smoke another cigarette. [4][5][6] Completely removing industry marketing from the packaging may not be sufficient to counteract positive affect 7 and inhibit incentive salience attribution. 8 9 Graphic warning labels (GWLs) of the health consequences of smoking aim to introduce negative affect with the goal of having the smoker reconsider the decision to smoke. As of January 2021, 127 countries have mandated GWLs on cigarette packaging, 10 and 17 countries have mandated plain packaging pioneered by Australia, 11 which removes industry branding and adds GWLs on 75% of the pack. 12 13 The USA is the only high-income country that has not yet mandated GWLs on cigarette packs.
Multiple studies have shown that GWL packaging is associated with negative affect. [14][15][16][17][18][19] Yet, the measurement of affective response to emotionally evocative stimuli is complicated, with self-report measures requiring recalled responses to the packaging without the presence of the stimulus. 20 The GWL literature mainly uses brief self-report paper and pencil measures of affect, resulting in a simple quantitative scale. Such a measure may be complemented by additional research using observational methods that add rich context. 21 The type of affect that cigarette packaging might induce is thought to be a minor 'emotional episode'. 22 Viewing a GWL package may elicit a minor valenced reaction that would not be strong enough to elicit any major physiological activation (such as fight or flight response), but is enough to have individuals think about their decision to smoke. 23 People are known to use emotive words to express the affect they feel when reacting to such an episode and the act of describing their response often helps them regulate their emotions. 24 A 'think aloud' technique 25 asks smokers to explore cigarette packaging and express their thoughts and feelings as they undertake the exploration. 26 This approach elicits verbalised spontaneous thoughts 27 about the pack presented, that is often influenced by cognition and emotions from previous experiences with the product. 28 This approach is most fruitful when different packaging options are compared, particularly when a very familiar pack is explored after they have been challenged with a pack featuring negative emotive stimuli. 29 This observational methodology when paired with multi-method measurement 30 of responses can capture immediate reactivity, opposed to paper-pencil measures 31 which may promote evaluative reactions.
An examination of real-world emotional responses to various cigarette packaging designs is needed to help inform how GWLs may influence smoking cognition and behaviour, should they be introduced in the USA in July 2022. 32 This paper Original research aims to establish and validate an immediate reactivity measure and identify whether the three types of pack designs elicit distinct affective valence. In this study, we use a structured pack handling task with a 'think aloud' cognitive interview to identify the individual variability in how US cigarette smokers react to five different cigarette packaging options: three plain packs with different GWLs, one blank pack devoid of branding and GWLs, and their usual US cigarette packs (branded without GWLs). We hypothesise reactions to US pack will fall along a range of positive affect, the variability for the blank pack will be in the neutral range, and reactions to all GWL packs will fall along a range of negative affect. All participants in this study were enrolled in a randomised trial where they received 3 months of real-world experience with their cigarettes repackaged into plain packs with GWLs, blank packs or maintained their usual US pack. We expect that the cognitive and behavioural responses in the trial will be determined by the immediate reactivity that the participants had to each of the study packs.

Study population
This study uses cross-sectional data collected during the initial in-person visit (V1) for the CASA randomised trial of the effects of cigarette packaging on smoking cognition and behaviour. 33 Volunteer daily smokers, aged 21-65 years from San Diego County, California, were enrolled using community advertising. All participants signed an informed consent (overseen by Institutional Review Boards at University of California San Diego and Cal State San Marcos), completed questionnaires and followed a protocol to think aloud their reactions as they explored study cigarette packaging.

Pack handling task
Participants were handed one pack at a time and asked to verbalise what thoughts came to their mind as they explored each side of each pack. For each pack, verbalisations were timed, recorded and transcribed. There were five study packs (online supplemental figure 1) each labelled with the participant's brand and variant: three GWL plain packs; one blank pack (devoid of marketing with GWLs absent) and their current US pack (branding present but GWL absent). In a pretest, 33 we selected three of eight GWL plain packs licensed from the Commonwealth of Australia using negative affect scores from the Positive and Negative Affect Scale. 34 We were concerned that familiarity with their own pack might lead to only cursory attention if it was presented first in the pack handling task. Accordingly, we required at least one GWL pack to be presented as the first pack which focused the individual's attention on the packaging elements that were being changed. GWL packs were thus randomised to the first, third or fifth presentation, and the blank pack and US pack to the second or fourth presentation.

Coding reactivity to study packaging
Using a multi-method qualitative approach, 30 35 two coders in consort with an anthropologist (SH) developed a coding manual 36 for a 7-point affect scale (high, medium, low for both negative and positive reactivity as well as a central neutral category; table 1) using a training set of 30 transcriptions. Four additional coders were trained using this set until group concordance (±1) was reached on 80% of transcriptions. In total, six coders used the coding manual to independently rate each transcription for each pack. The coders met weekly to discuss their scores and resolve instances of coding discordance. High reactivity was indicated by use of highly emotional words or amplified moderately emotional words that suggested a somewhat visceral reaction to the packaging. If moderately emotional words or highly emotional words were used and de-amplified (eg, 'somewhat disgusting') or emotional statements accompanied by qualifications (eg, 'that's disgusting but it would not stop me from smoking'), that indicated medium reactivity. A low level was a mild reaction followed by a rationalisation. Neutral reactivity was when no emotional or reactive language was uttered. For each pack, reactivity scores were averaged, and categorical reactivity scores generated by rounding mean scores to their nearest integer.

Natural language processing of initial reactivity
Using R V.4.0.3 with the 'SentimentR' package, 37 we conducted natural language processing of the transcribed speech from the pack handling task to quantify the number of words uttered and polarity of word choice. Using the Jockers-Rinker sentiment lexicon of 11 710 polarised words, 38 sentences were classified Table 1 Coding system for reactivity to each study pack

Rating
Participant reactions and descriptions of study packs include: High negative Score: −3 Highly emotional words or amplified moderately emotional words that are negatively valanced to describe pack aversion. Visceral reaction and repeated exclamations of aversion; might repeat emotional words. Language that indicates they do not want to handle the pack.

Score: −2
Moderately emotional words or de-amplified highly emotional words that are negatively valanced to describe pack aversion.
No visceral reaction and a lower emotional response than high aversion. Strong initial negative reaction followed by rationalisation (eg, pack design would not modify behaviour).

Low negative
Score: −1 Moderately emotional words that are negatively valanced followed by detracting statements or de-amplifiers that over-rule the response.
No visceral reaction or high/moderate negative emotional response. Mild reaction or acknowledgement of pack aversion followed by rationalisation (eg, pack design would not modify smoking behaviour).

Score: 0
No emotional words to describe pack. No or little reaction to the pack and/or appear to be unaffected by the pack. Text on the pack may be read without saying how it makes them feel.
according to their overall polarity (eg, the degree to which the speech and its linguistic modifiers had a positive, neutral or negative valence; online supplemental table 1). To account for extreme negative words occurring more commonly in natural language, 39 polarity scores were scaled from −1 to +1 using a general rescaling function. 37 Linguistic modifiers were accounted for by examining the four words following, and two words preceding, each polarised word and tagged as one of the following: neutral, negators (flip the ± polarity sign of a word, for example, 'I do not like it'), amplifiers or de-amplifiers (increase or decrease the impact of a word by multiplying polarity scores using standard preset weights, 38 for example, 'I really like it. I hardly like it'), or conjunctions (over-rule previous clauses, for example, 'I like it but it's not worth it').
The sentiment lexicon was augmented to neutralise polarised words that had different connotations in our study (eg, baby, child, surgeon). Sentence-level polarity scores were averaged to generate composite polarity scores per participant per pack. The prototypical emotions of fear, disgust, anger, sadness, anticipation, trust, joy and surprise 40 were explored using 'SentimentR's' emotion function and the National Research Council of Canada (NRC) Hashtag Emotion Lexicon look-up of 8265 emotion terms. 41 42 The rate of emotion expressed was evaluated as the number of emotional words uttered relative to the total number of words spoken, with scores ranging between 0 (no emotional utterances) and 1 (all emotional utterances).

Study covariates
Sociodemographics (age, sex, race/ethnicity and educational attainment), 33 tobacco use (daily use frequency and primary brand smoked), 33 the Fagerström Test of Nicotine Dependence Scale, 43 brand loyalty 4 and health anxiety 44 were measured covariates. We assessed brand appeal using a 6-point Likert scale ('The design on the brand of cigarettes I currently smoke is… Stylish, Fashionable, Cool, High quality, Attractive, Appealing'; α=0.92). 45 46

Statistical analysis
Inter-rater reliability of the coded reactivity scores across the five pack conditions was evaluated in two ways 47 : (a) by computing the percentage agreement across the scores while allowing for a tolerance of 1 in ratings, and (b) by modelling the intraclass correlation coefficient (ICC) among the raters. With the goal of constructing composite scores, a two-way random effects (ie, participants within pack type) ICC model was used 47 with raters' scores evaluated for consistency. 48 To examine patterns in highest levels of reactivity, quintile cut-off points were calculated. To examine differences in the time to explore packs, total words uttered, polarity of word choice and verbalised reactivity expressed, we conducted Kruskal-Wallis tests and post-hoc examination of pairwise comparisons using Dunn's tests. Spearman rho correlation coefficients were used to evaluate construct validity between reactivity scores and word polarity. To explore differences in emotion expressed during pack handling, we plotted the average rates of emotional utterances using a radar chart. 49 To explore the associations between sample characteristics and reactivity scores, we fit an intercept only conditional mixed-effects model with bootstrapped CIs using the 'Lme4' package. Reactivity scores were the outcome of interest, with package viewing order, age, gender, race/ethnicity, education, health anxiety, nicotine dependence, brand appeal, brand loyalty and brand smoked included as fixed effects. All two-way interactions between pack condition and covariates were examined using the 'LmerTest' package and significant terms (p<0.05) retained using an omnibus F-test. Estimated marginal means were computed from model terms using the 'effects' package and then plotted.

Original research
The frequency of prototypical emotions expressed in the 'think aloud' is presented in the radar chart (figure 1). The foot gangrene pack elicited more emotions characterised as disgust, fear and, to a lesser extent, anger. A similar distribution of expressed emotions was seen in response to the throat cancer GWL pack, although at a lower frequency. The primary emotion elicited by the neonatal baby GWL pack was sadness. The two main emotions elicited by their own pack were trust and joy.

Figure 1
Average rate of emotive words spoken during pack exposure period (n=324). A semantic analysis of transcribed speech that was text mined for emotive utterances using an emotion word lexicon and computing the rate of emotive words expressed per sentence between 0% (no emotional utterances) and 100% (all emotional utterances).   From separate intercept only conditional mixed-effects models with bootstrapped 95% CIs (n=1000) predicting reactivity to cigarette packaging design. *Measured on 7-point affect scale (−3 to +3: high, medium, low for both negative and positive reactivity as well as a central neutral category).

Original research
a reactivity score of 1.45 (95% CI=1.37 to 1.54) which was much higher than the 25th percentile level with a score of 1.16 (95% CI=1.08 to 1.24); figure 2). No relationship was observed between ratings of brand appeal and reactivity scores for GWL or blank packaging. More positive reactivity scores for the US pack were observed when the pack was viewed later in the pack handling task (fourth position=1.43 (95% CI=1.33 to 1.53)) compared with when it was viewed earlier in the task (second position=1.16 (95% CI=1.06 to 1.26)).

DISCUSSION
US daily smokers, with minimal previous exposure to GWLs, demonstrated consistent negative reactions when they were exposed to the GWLs used as part of plain packaging licensed from the Commonwealth of Australia. While reactivity to GWL packaging was negative across the board, the level of reactivity appeared to align with the negative emotional response found in prior work. 33 50-53 Conversely, smokers' current branded US cigarette pack was associated with positive reactivity which was higher when their branded pack occurred after exposure to two different GWL packs in the study's pack handling protocol. This finding supports previous research that found current cigarette packaging in the USA to be associated with positive affect for smokers, which may promote more regular smoking behaviour. 4 Blank packs, devoid of all marketing, drew a neutral response. Thus, this study has established that these three packaging conditions (GWL plain pack, blank pack, US pack) elicit markedly different initial participant reactions to the cigarette packaging. A major objective of Australia's GWL health consequences messaging was to induce thoughts (eg, 'I cannot bear to think of that happening to me') that might promote future quitting behaviour. 54 Notably, this study found a significant negative affect experienced by US smokers in response to the GWL plain packs, most markedly with the foot gangrene image. Both the images of the neonatal baby and throat cancer were associated with negative affect where the emotions appeared to be a mix of fear, disgust, anger and sadness-which appear consistent with the goal of this health consequences messaging. 54 However, the foot gangrene image was associated with much stronger negative emotions that were more likely to be characterised as visceral. The emotions expressed appeared to be disgust, fear and anger much more than sadness. In future work, we aim to explore the transcribed text for the directionality of anger, which could be directed at the tobacco industry, 55 governmental regulations 56 or somewhere else. One of the strengths of this qualitative methodology is that it facilitates such further detailed analyses. In the randomised trial, we aim to use ecological momentary assessment to test whether the high initial reactivity to the GWL pack images is associated with increased cognition when participants reach for a cigarette. 57 With twice daily measurement, we will be able to assess whether and how this reactivity is associated with avoidance and/or pack hiding behaviour, 58 which may 59 or may not 58 be related to cessation behaviours. The detailed and frequent measurement of both cognition and behaviour in the randomised trial will be a major advance on most of the studies completed to date. 23 59 GWL plain packs may disrupt the incentive salience attributed to the cigarette packaging via the removal of industry marketing and inclusion of visceral imagery and aversive design characteristics (eg, fonts and colours). Cue-learning models suggest that appealing design features on packaging capture attention, generate positive affective reactions and motivate behaviour that may facilitate a desire to smoke. 60 61 We found that the more brand appeal smokers reported for their own US-marketed pack (eg, cool, stylish, etc), the more positive their reaction was when asked to express their thoughts and feelings about it. When appealing marketing cues are affixed to tobacco products and perceived immediately prior to use, the cues themselves can acquire similar motivational significance and evoke a desire to smoke. 8 62 63 Yet, levels of brand appeal did not influence the reactivity to the GWL plain packs, despite the packs being matched to the smoker's cigarette preference and clearly labelled with brand and variant name. Thus, plain GWL packaging may have the intended effect of inhibiting incentive salience attribution by quelling the appeal of the product, an effect consistent with prior research suggesting that plain GWL packaging impedes the product's ability to generate appeal. [64][65][66] Nevertheless, reactivity to the blank pack did not vary by levels of brand appeal, indicating that the appeal of the product may be suppressed by simply removing tobacco industry marketing.
There are a number of factors that limit the generalisability of these findings: (a) the study recruited volunteer smokers and the sample was not representative of the US population or smokers in other countries; (b) under-representation of minorities in the study also resulted in a lower proportion of menthol smokers; (c) all participants were from California which has stronger social norms against smoking than the rest of the USA. 67 Other limitations included the loss of <10% (n=33) of the 'think aloud' data due to a computer hardware failure at our storage facility that was unrelated to the trial, indicating that the data are most likely missing-at-random. 68 The GWL packaging proposed for use in the USA is not on the plain packaging used in this study, but a hybrid packaging condition that includes reduced industry marketing with smaller GWLs, a design quite common in many countries. 12 We would expect that such hybrid packaging would be associated with a lower level of initial reactivity to the GWLs than was observed in this assessment.
Despite limitations, the study had numerous strengths. It allowed smokers to openly express their thoughts and feelings about GWL packaging, thus resulting in more rich emotive Figure 2 Relationship between level of brand appeal and reactivity to five cigarette pack designs (N=324). Estimated marginal means and 95% confidence interventions extracted from intercept only conditional mixed-effects model predicting reactivity to cigarette packaging design with age, race/ethnicity, education, health anxiety, nicotine dependence, brand loyalty, and brand smoked included as fixed main effects and package viewing order, gender, and brand appeal as interaction effects.
details than structuring their response through a questionnaire. All study packs were matched to the participants' preferred cigarette brand and variant in an effort to maintain cigarette expectancy and isolate the effects of the reactivity. We used observational measurement of reactions to the various pack designs with high-quality coding, which yielded a full range of valenced reactivity and was concurrently valid with the polarity of speech as identified by natural language processing. We used an exposure to GWL plain packs prior to assessing reactivity to US packs which likely focused the participant's thinking on what they liked about their current pack, resulting in more reliable reactions.

CONCLUSION
GWLs are an integral part of the recommended suite of tobacco control strategies for governments to reduce the health costs associated with disease caused by cigarettes. 13 Yet, to date, litigation by the tobacco industry has blocked implementation in the USA by arguing that GWLs are too aversive and are aimed at forcing smokers to quit. 69 In this study, we have demonstrated that US smokers do react with a range of negative emotive reactions to GWL packs, contrasting with the appeal of their regular branded pack. Future studies are needed to demonstrate whether GWL packaging achieves the Food and Drug Administration's (FDA's) goal of encouraging smokers to think about the health consequences of using these products.
Twitter Matthew Stone @MatthewDavStone What this paper adds What is already known on this subject ► Graphic warning labels (GWLs) on cigarette packaging have been implemented in 120+ countries and jurisdictions, but not in the USA. ► GWLs can introduce negative affect when they remind smokers of the health consequences of smoking.
What important gaps in knowledge exist on this topic ► There has been no systematic examination of the range of affect on current smokers produced by cigarette packaging with and without tobacco industry branding and with and without GWLs. ► An examination of real-world emotional responses to various components of packaging elements is needed to help inform how packaging designs may influence cognition and behaviour.
What this paper adds ► Handling current US-branded cigarette packaging without GWLs was associated with moderate positive affect and feelings of trust and joy, an effect that was amplified when viewed immediately after exposure to packs without current branding but with GWLs. ► Handling blank packs (without GWLs and without current branding) was associated with lower positive affect than that associated with exposure to US-branded packs without GWLs, regardless of prior exposure to packs without current branding but with GWLs. ► Handling three examples of packs without current branding but with GWLs induced a range of moderate-to-severe negative affect and provoked feelings of disgust, fear, anger and sadness. ► Exposure to blank, branded and blank+GWLs packaging elicited immediate neutral, positive and negative affect, respectively. Different versions of the blank+GWL pack elicited different levels and types of negative affect. ► The immediate reactivity measure revealed the range of affect needed to explore the role of differentially valenced packaging on smokers' cognition and behaviour in a realworld randomised trial.