Public perceptions of synthetic cooling agents in electronic cigarettes on Twitter

Amid a potential menthol ban, electronic cigarette (e-cigarette) companies are incorporating synthetic cooling agents like WS-3 and WS-23 to replicate menthol/mint sensations. This study examines public views on synthetic cooling agents in e-cigarettes via Twitter data. From May 2021 to March 2023, we used Twitter Streaming API, to collect tweets related to synthetic cooling agents with keywords such as ‘WS-23,’ ‘ice,’ and ‘frozen.’ Two deep learning roBERTa models, classified attitudes expressed in tweets about synthetic cooling agents and identified e-cigarette users. The BERTopic deep-learning model identified major topics of positive and negative tweets. Two proportion Z-tests were used to compare the proportions of positive and negative attitudes between e-cigarette users (vapers) and non-e-cigarette-users (non-vapers). Of 6,940,065 e-cigarettes related tweets, 5,788 non-commercial tweets related to synthetic cooling agents. The longitudinal trend analysis showed a clear upwards trend in discussions. Vapers posted most of the tweets (73.05%, 4,228/5,788). Nearly half (47.87%, 2,771/5,788) held a positive attitude toward synthetic cooling agents, which is significantly higher than those with a negative attitude (19.92%,1,153/5,788) with a P-value < 0.0001. The likelihood of Vapers expressing positive attitudes (60.17%, 2,544/4,228) was significantly higher (P < 0.0001) than that from non-vapers (14.55%, 227/1,560). Conversely, negative attitudes from non-vapers (30%, 468/1,560) was significantly (P < 0.0001) higher than vapers (16.2%, 685/4,228). Prevalent topics from positive tweets included “enjoyment of specific vape flavors,” “preference for lush ice vapes,” and “liking of minty/icy feelings.” Major topics from negative tweets included “disliking certain vape flavors” and “dislike of others vaping around them.” On Twitter, vapers are more likely to have a positive attitude toward synthetic cooling agents than non-vapers. Our study provides important insights into how the public perceives synthetic cooling agents in e-cigarettes, insights that are crucial for shaping future FDA regulations aimed at safeguarding public health.


Introduction
Electronic cigarettes (e-cigarettes) are devices used to aerosolize a solution typically containing propylene glycol, vegetable glycerin, flavorings, and in most cases, nicotine for inhalation by consumers [1 -2].The prevalence of e-cigarette use has risen exponentially in the last decade, especially in youth and young adults [3,4].E-cigarettes have become the most frequently used tobacco product among high school students with an estimated 3.6 million middle and high school students reporting current e-cigarette use in 2022 [5].Toxicity and health impact studies of e-cigarettes have shown that components in e-cigarette aerosols could lead to severe adverse health problems, including respiratory disorders, cardiovascular diseases, mental health issues, and possibly cancer [6][7][8][9].In February 2020, the FDA issued a policy to limit the sale of cartridge-based e-cigarette products to only tobacco and menthol flavored e-cigarettes [10].Following the announcement and implementation of the FDA's flavor enforcement policy, there was a significant increase in the sales of menthol-flavored e-cigarettes [11].As mentholflavored e-cigarettes gain popularity, menthol-flavored cigarettes are also becoming increasingly popular among youth and young adults, particularly in the African American community.In April 2022, the FDA announced the proposed standard to ban menthol-the last allowable flavorin cigarettes in hopes of saving lives and reducing health disparities [11 -13].
In response to the proposed menthol cigarette ban, the market has seen a significant increase in 'ice' flavors that attempt to circumvent these regulations by incorporating synthetic cooling agents like WS-3 and WS-23 into e-cigarette refill liquids to mimic the cooling sensation like menthol or mint [14].Ice-flavored e-cigarette use correlates with higher vaping frequency and dependence, and 44.8% of e-cigarette users primarily use ice flavors [15].Synthetic cooling agents in e-cigarettes imitate menthol's cooling sensation but induce a unique 'cold' feeling during inhalation, unlike the 'hot' sensation from capsaicin or menthol's 'cool' feeling [16].These agents are now prevalent in 'ice' flavored e-cigarettes, despite many consumers' intake of the agents exceeding established safety thresholds [17 -18].This increasing popularity of synthetic cooling agents creates new concerns about the potential health risks posed by industry attempts to bypass regulations [17].While the long-term health effects of synthetic cooling agents are largely unknown due to their relative newness, one recent study showed that synthetic cooling agents may disrupt respiratory airway epithelial cell (AEC) responses, a vital defense against inhaled toxins and allergens [19].This disruption could impact the physiology of the airway and impact susceptibility to health-threatening respiratory diseases [19].
Understanding the public perceptions and discussions of synthetic cooling agents is crucial for shaping future regulation.Social media platforms, notably Twitter (now rebranded as "X"), have been widely used in capturing public perceptions and discussions of tobacco products [20].As of mid-2023, Twitter has over 235 million daily active users, with an average of 500 million tweets posted daily [21].This massive volume of interactions has made it an important resource for investigating public perceptions and discussions on various topics, including health and lifestyle behaviors like vaping.For instance, researchers analyzed thousands of tweets to investigate attitudes toward e-cigarette flavors and found that preferences for certain flavors, such as fruit/fruit beverage, dessert/pastry, and candy, chocolate, or sweets, might be associated with continued e-cigarette use [22].Twitter has also been employed to study the perceived health impacts of vaping [23].A study analyzing Twitter posts found that many e-cigarette users reported negative health effects, ranging from minor complaints to more serious conditions [24].
A cross-sectional study on marketing of electronic cigarettes on Twitter highlighted Twitter's value in tracking real-time public sentiment [25].Overall, Twitter served as an ideal resource for understanding public perceptions and discussions of synthetic cooling agents.
Our study aims to examine the public perceptions and discussions of 'ice' flavored ecigarettes and associated synthetic cooling agents using Twitter data.Using natural language processing and deep-learning algorithms, our study results will provide a snapshot of public discourse and attitudes towards 'ice' flavored e-cigarettes and synthetic cooling agents.Our findings will provide vital insights into the dynamic e-cigarette landscape, specifically in relation to 'ice' flavors and synthetic cooling agents.Understanding these dynamics can inform future regulations and strategies to protect public health.This aim was successfully achieved, as evidenced by our detailed analysis and significant results, which provide a comprehensive understanding of public sentiment regarding synthetic cooling agents in e-cigarettes.

Ethics Statement
This study was reviewed and approved by the Research Subjects Review Board of the Office for Human Subject Protection at the University of Rochester (STUDY00006570).Patient consent for publication is not required as the data were analyzed anonymously.

Data collection and preprocessing
For this study, we leveraged Twitter data from the Twitter Streaming API (Application Programming Interface) collected from May 3, 2021, to March 14, 2023 using keywords related to electronic cigarettes [26].Tweets were initially filtered based on keywords related to ecigarettes, including but not limited to keywords like "e-cig", "vaping", "e-liquid", and "vapenation".After filtering out the retweets from e-cigarette related tweets, we obtained 2,597,206 tweets.We applied another layer of filter to filter out promotional content with promotional keywords within the username, retweets, and content of the tweet from further analysis [26].This resulted in a collection of 2,426,615 non-promotional tweets.We applied a third layer of filter to filter out "ice" flavored e-cigarettes or synthetic cooling agents related tweets using keywords like 'ice', 'frozen', and 'arctic' (S1 Table ) and ultimately identified 5,788 tweets related to synthetic cooling agents and sensations.

Deep learning models for classification
We used an inductive method to manually label 2,000 randomly selected tweets, which served as the training data for our deep learning classification models.We first randomly selected 200 tweets and manually labeled them according to the users' attitudes toward synthetic cooling agents and sensations into negative, neutral, and positive categories.Additionally, we determined whether the Twitter users were e-cigarette users (vapers), indicated from the tweets.
Two authors (AL, and JH) hand-coded the 200 tweets independently and reached a high interrater reliability (Cohen's Kappa coefficient was 0.9219 for attitude and 0.9256 for whether the Twitter user was an e-cigarette user).Then, the remaining 1,800 tweets were manually coded by the same two authors (AL and JH) with each coding 900 tweets.We built two roBERTa (Robustly Optimized BERT Pre-Training Approach) models using the 2,000 manually coded tweets as the training dataset [27], one for sentiment analysis and the other for binary classification to determine whether a Twitter user was an e-cigarette user.RoBERTa is a deep learning model developed by Facebook AI, which builds upon BERT (Bidirectional Encoder Representations from Transformers) [28].This model differs from its predecessor by enhancing the pretraining process, involving increased training data and a more dynamic masking pattern, leading to improved accuracy on various natural language processing tasks.We employed ten-fold cross-validation to ensure the robustness of our models.Our models achieved an F1-score of 0.83 for attitude and 0.92 for e-cigarette user classification, ensuring the validity of our model results.The two-proportion Z-test was used to compare differences in the proportions of various sentiment categories.

Topic modeling
To understand the topics discussed in tweets related to synthetic cooling agents and sensations, we employed BERTopic [29].BERTopic is an advanced, context-aware topic modeling technique.BERTopic is based on the BERT language model, which allows us to grasp the larger context of the conversation by capturing semantic relationships between words and topics.We determined the optimal number of topics based on the coherence score and inter-topic distances [30 -31].

Public discussions on synthetic cooling agents, sensations, and Ecigarettes on Twitter
We observed a clear increasing trend in the number of tweets related to ice flavors or synthetic cooling agents in e-cigarettes from May 2021 to March 2023 with some variations (Fig 1

Principal findings
We identified an increasing trend in the discussions of the synthetic cooling agents and sensations in e-cigarettes from May 2021 to March 2023.Interestingly, the public attitudes toward the synthetic cooling agents and sensations in e-cigarettes remains relatively constant over the study period.Furthermore, we found a predominantly positive attitude toward synthetic cooling agents and sensations in e-cigarettes.Further analysis of Twitter users revealed that this positive sentiment was largely driven by e-cigarette users, or vapers.The topics discussed within these positive tweets mainly centered on the enjoyment of ice-flavored vaping.Though, the importance of flavor variety in vaping experiences was also noted.Conversely, the primary concerns arising from tweets with negative attitudes involved second-hand exposure to vaping and potential addiction.

Comparisons with prior work
Consistent with our findings of a mostly positive attitude towards synthetic cooling agents and sensations in e-cigarettes, a previous content analysis of ice-flavored e-cigarette products using Twitter data from January to July 2021 also revealed a majority favorable attitude towards these products [32].
We observed an increasing trend in the discussions of the synthetic cooling agents and sensations during the study period, which might partially be due to the aggressive marketing of ice-flavored e-cigarette products identified from the previous study [32].Another reason for this increasing trend might be due to their increased popularity and conversations related to their regulations.This trend suggests that synthetic cooling agents are becoming more prevalent and widely used, which may be due to their perceived benefits or effectiveness in providing a pleasant experience [33].However, there is also growing concern about their potential health risks, leading to increased discussions about policy and regulation.Studies have already shown that the amount of synthetic cooling agent, WS-3/WS-23, in e-cigarettes often cause levels of exposure higher than the safety threshold [17,33].Elevated exposure to WS-3/WS-23 has been linked to changes in airway epithelial cells, potentially increasing susceptibility to respiratory diseases [19].Although more research is needed on the broader health effects that these synthetic agents may have, preliminary findings suggest a need for immediate regulatory attention.The comparison of attitudes between vapers and non-vapers towards synthetic cooling agents and sensations indicated that a higher percentage of vapers expressed positive sentiments, while non-vapers displayed more negative and neutral sentiments.This underlines the divergent views between these two groups and can be attributed to their direct experiences and exposure to synthetic cooling agents, especially enjoyment of specific vape flavors and preference for certain vape types.For example, tweets like "I don't vape until someone pulls out lush ice" emphasize the allure of specific mixtures like Lush Ice, which offers a mix of watermelon and melon flavors with a minty kick.This contrasts with standard ice vapes, known for a straightforward cooling sensation.Tweets like "Told mum I like vaping for the minty flavor and she goes I'll make you a mint chutney pls stop," underline the minty flavors' appeal.Additionally, mentions of enjoying the vapor smell, such as in the tweet "hi moots I ran out of my disposable vape and now I miss my ice grape-smelling room," show that sensory enjoyment isn't limited to taste.The recurring theme of addictiveness and pleasurable sensations highlights the sensory and innovative appeal of vaping.This trend, reflecting a growing acceptance or normalization of vaping within certain communities, has crucial implications for future policy.Policymakers may need to address these specific attractions when instituting regulations, ensuring that marketing and promotion of vaping products do not exploit these preferences.Tweets with negative attitudes expressed dislike of certain flavors and concerns about second-hand exposure.Examples like "@mentholdiet banana ice OR literally anything but watermelon (watermelon vapes make me vomit in a bad way :/)" and "Minty vape flavors give me major brain freeze vibes.I can't do it," point to specific aversions to flavors, ranging from mint to fruit.These dislikes could be due to individual taste preferences or sensations like "brain freeze," contrasting with a broader trend of users finding non-traditional flavors like fruit or candy more satisfying [34].Concerns about second-hand exposure also resonate strongly.
Tweets such as "you blow peach ice in my face and watch how i shove that vape down your throat," or "Vape smokers will blow banana ice in ya face while u telling em how ya grandma died," spotlight the intrusion of vapor into personal spaces, exacerbating existing concerns about the inhalation of harmful substances such as nicotine and toxic chemicals by bystanders [35 -37].These findings provide insight into the specific factors that influence people's perceptions of synthetic cooling agents, highlighting concerns about health, etiquette, and the social impact of vaping.
Our study has several limitations.Despite demonstrating the impressive performance of the deep-learning models we employed for tweet labeling, they are not flawless and could potentially introduce some misclassification and bias into our findings.Furthermore, the absence of demographic data (such as age, gender, and race/ethnicity) for the Twitter users in our dataset restricts our ability to explore the attitudes and topics discussed based on each demographic subgroup.We also deliberately avoided using the term "cool" as a keyword to minimize the collection of irrelevant tweets.While this decision reduced noise, it also likely omitted some relevant tweets discussing synthetic cooling agents in a relevant context.We also did not account for the potential influence of bot accounts, which could have skewed both the sentiment and the topical focus of the tweets collected.Furthermore, the longitudinal nature of our data collection doesn't consider temporal shifts in public opinion that could be influenced by new research findings or policy changes.Lastly, our study is constrained by its focus on English-language tweets and its reliance on Twitter as the sole platform, which may not fully represent global and/or multilingual perspectives.

Conclusion
Using Twitter data, we investigated public perceptions and discussions of synthetic cooling agents and sensations in e-cigarettes.Our analysis revealed an increasing trend in discussions about these agents and sensations over time, with positive attitudes consistently dominating the study period.Vapers were found to be the major contributors to the overall positive attitudes, expressing their preference for ice-flavored e-cigarette products.Tweets with negative attitudes expressed dislike for certain flavors and concerns about second-hand vapor exposure.Our results provide valuable information for future regulation of synthetic cooling agents.The favorable view of ice-flavored e-cigarette products raises significant public health concerns, given the potential harms of these synthetic cooling agents.Public health campaigns are needed to educate the public about the potential risks associated with synthetic cooling agents.These campaigns could be delivered through popular social media platforms such as Twitter to reach a large and engaged audience.
Administration (FDA).The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or the FDA.

The attitude toward synthetic cooling sensation on Twitter over time. Twitter users' attitudes toward synthetic cooling agents and sensations between vapers and non-vapers
Among 5,788tweets related to synthetic cooling agents and sensations in e-cigarettes, 47.81% (2,771/5,788) of tweets had a positive sentiment, 19.92% (1,153/5,788) had a negative sentiment, and 32.20% (1864/5,788) had a neutral sentiment.The proportion of tweets with a positive attitude was significantly higher than that with a negative attitude (P-value <0.0001).As shown in Fig2, the proportion of tweets with either a positive or negative attitude from May 2021 to April 2023 was relatively consistent.While it is not significant, there is a small increase in the proportion of negative tweets and a slight decrease in the proportion of 645/3,822) expressed negative attitudes.Nearly half of all non-vapers expressed neutral attitudes (49.77%, 652/1,310).More than a third non-vapers showed negative attitudes (34.73%, 455/1,326) and less than one sixth non-vapers showed positive attitudes (15.50%, 203/1,326).
). Fig 1.The number of tweets on synthetic cooling agents and sensations over time.positivetweetsovertime.Among 3,924 tweets related to synthetic cooling agents and sensations in e-cigarettes with positive or negative sentiment (omitting neutral), approximately 70.62%(2,771/3,924) had positive sentiment and 29.38% (1,153/3,924) had negative sentiment.The proportion of tweets with a positive attitude is significantly higher than that with a negative attitude when observing only tweets of positive and negative sentiment (P-value <0.0001).Fig 2.2,243/3,757).In addition, 23.13% (869/3,822) of vapers expressed neutral attitudes, and 17.17% (

Table 1 . Major topics identified from tweets with positive or negative attitudes
*Indicates the presence of an emojiTable1also highlights the main issues brought up in tweets with a negative attitude towards the synthetic cooling agents and sensations in e-cigarettes.The most recurrent topic is disliking certain vape flavors (53.69%, 619/1,153), followed by discomfort from others vaping around them (20.82%,240/1,153), disagreement with the act of vaping itself (6.42%, 74/1,153), and disliking the mintiness of vapes (4.51%, 52/1,153).