This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on https://publichealth.jmir.org, as well as this copyright and license information must be included.
On January 2, 2020, the US Food and Drug Administration (FDA) released the electronic cigarette (e-cigarette) flavor enforcement policy to prohibit the sale of all flavored cartridge–based e-cigarettes, except for menthol and tobacco flavors.
This research aimed to examine the public perception of this FDA flavor enforcement policy and its impact on the public perception of e-cigarettes on Twitter.
A total of 2,341,660 e-cigarette–related tweets and 190,490 FDA flavor enforcement policy–related tweets in the United States were collected from Twitter before (between June 13 and August 22, 2019) and after (between January 2 and March 30, 2020) the announcement of the FDA flavor enforcement policy. Sentiment analysis was conducted to detect the changes in the public perceptions of the policy and e-cigarettes on Twitter. Topic modeling was used for finding frequently discussed topics about e-cigarettes.
The proportion of negative sentiment tweets about e-cigarettes significantly increased after the announcement of the FDA flavor enforcement policy compared with before the announcement of the policy. In contrast, the overall sentiment toward the FDA flavor enforcement policy became less negative. The FDA flavor enforcement policy was the most popular topic associated with e-cigarettes after the announcement of the FDA flavor enforcement policy. Twitter users who discussed about e-cigarettes started to talk about other alternative ways of getting e-cigarettes after the FDA flavor enforcement policy.
Twitter users’ perceptions of e-cigarettes became more negative after the announcement of the FDA flavor enforcement policy.
An electronic cigarette, also known as an e-cigarette or e-cig, is a battery-powered product that typically delivers nicotine in the form of an aerosol [
The e-cigarette flavor choices in the market have rapidly increased in recent years. One study showed that there were more than 460 brands and 7700 unique e-cigarette flavors as of January 2014 [
In order to prevent youth access to flavored e-cigarettes, in November 2018, the Food and Drug Administration (FDA) announced several policies to protect youth, including restricting the sale of flavored e-cigarettes to physical and online stores, with customer age restriction and verification [
To examine the impact of the FDA flavor enforcement policy, we proposed to investigate how the FDA flavor enforcement policy affects the public perception of e-cigarettes and, subsequently, the potential changes in e-cigarette user behavior using Twitter data. Twitter had around 48.35 million active users in the United States in 2019 [
In this study, we compared the changes in sentiment toward the FDA flavor enforcement policy and e-cigarettes before and after the FDA flavor enforcement policy. In addition, we tried to examine if there was an intention for potential behavior changes in e-cigarette use with the FDA flavor enforcement policy. The findings of this study provide important insights about the potential effects of the FDA flavor enforcement policy, which could be useful for further policy decision making about the regulation of flavored e-cigarettes to protect public health.
E-cigarette–related tweets were downloaded through a Twitter streaming application programming interface (API) using keyword searching based on a list of e-cigarette–related keywords, including “e-cig,” “e-cigs,” “ecig,” “ecigs,” “electroniccigarette,” “ecigarette,” “ecigarettes,” “vape,” “vapers,” “vaping,” “vapes,” “e-liquid,” “ejuice,” “eliquid,” “e-juice,” “vapercon,” “vapeon,” “vapefam,” “vapenation,” and “juul” [
In order to investigate Twitter users’ perceptions of e-cigarettes within the United States only, another layer of geographic filtering was applied to users’ geolocations. US geolocation keywords that contained both the full names and abbreviations of 50 states in the United States, such as “California,” “Illinois,” and “Florida,” as well as big cities, such as “Los Angeles,” “Chicago,” and “Miami,” were used for filtering. As a result, 2,341,660 e-cigarette–related tweets within the United States were obtained, with 644,686 tweets before the announcement of the FDA flavor enforcement policy, 702,488 tweets between the announcement and implementation of the FDA flavor enforcement policy, and 994,486 tweets after the implementation of the FDA flavor enforcement policy.
Lastly, a third layer of filtering was applied to obtain tweets related to the FDA flavor enforcement policy. The filtering keywords included “FDA ban,” “flavor ban,” “ban,” and any combination of “FDA,” “flavor,” and “ban.” A total of 190,490 FDA flavor enforcement policy–related tweets were collected, with 29,120 tweets before the announcement of the FDA flavor enforcement policy, 89,539 tweets between the announcement and the implementation of the FDA flavor enforcement policy, and 71,831 tweets after the implementation of the FDA flavor enforcement policy. The complete data collection and filtering process is showed in a flowchart in
VADER (Valence Aware Dictionary and Sentiment Reasoner) was used as the sentiment analyzer to analyze Twitter users’ thoughts and perceptions of e-cigarettes, and analyze Twitter users’ thoughts and perceptions of the FDA flavor enforcement policy [
Topic modeling, specifically latent Dirichlet allocation (LDA) modeling, was used to determine the popular topics among e-cigarette–related tweets. LDA is a generative text model for analyzing and clustering words and terms in the given document and generating topics with keywords and their corresponding weights, which indicated the possibility of appearance in the document [
According to a research survey published in 2018, the LDA method is one of the most powerful and popular methods used for topic modeling of social network data for knowledge discovery and behavior analysis [
We applied topic modeling to the e-cigarette tweets in the 3 time periods. First, in order to ensure consistency in the process of the training model, all characters were in lower case, and all words were in the same tense by using the spaCy lemmatization function. In addition, stop words, such as personal pronouns and prepositions, were removed by using Natural Language ToolKit (NLTK) packages. Furthermore, in order to get precise and meaningful results, frequent bigrams (eg, flavor ban) and trigrams (eg, food drug administration) were identified by using the Gensim package, which were considered as a single term rather than separated words during model training. The number of topics was chosen from 3 to 10, and the optimal number of topics was determined by the coherence score of each LDA model result. Lastly, the keywords of the fitted LDA topic model and the percentage distribution of each topic were obtained using the pyLDAvis package.
To examine the impact of the FDA flavor enforcement policy on the discussion about e-cigarettes on Twitter, the total amount of tweets was normalized by the number of days before and after the FDA flavor enforcement policy. As shown in
Average number of daily tweets about e-cigarettes and the Food and Drug Administration flavor enforcement policy.
In order to investigate the perceptions of Twitter users toward e-cigarettes and the FDA flavor enforcement policy, sentiment analysis was conducted, and the proportions of tweets with positive, neutral, and negative sentiments were calculated before and after the FDA flavor enforcement policy. For the better understanding of the sentiment results,
As shown in
Changes in the public perception of e-cigarettes on Twitter with the announcement and implementation of the Food and Drug Administration flavor enforcement policy.
Different from the changes in sentiment toward e-cigarettes, for the FDA flavor enforcement policy–related tweets, the proportion of tweets with positive sentiment increased significantly (
Changes in the public perception of the Food and Drug Administration flavor enforcement policy on Twitter with the announcement and implementation of the policy.
The LDA topic modeling was applied to e-cigarette–related tweets in order to determine any change in the e-cigarette–related topics discussed by the Twitter users over time. An optimal number of topics was selected by the highest coherence score. Across the 3 time periods, the topics discussing
Top topics related to e-cigarettes before and after the Food and Drug Administration flavor enforcement policy.
Time frame and topic | Tokens, n (%) | Keywords | |||
|
|
|
|||
|
Stop vaping and smoking to protect health | 201,142 (31.20) | vape, vaping, lung, smoke, get, go, people, stop, cancer, health | ||
|
New e-cigarette flavor use among friends | 158,593 (24.60) | vape, new, ude, level, link, case, dear_ncan, space_nasking, use, friend | ||
|
Single Juul pod equals a pack of cigarettes | 153,435 (23.80) | juul, hit, be, pod, say, still, stare, go, iterally, single | ||
|
Vaping leads to nicotine addiction for those who unlikely smoke | 131,516 (20.40) | smoking, cigarette, generation, whole, create, first, start, addiction, unlikely, statistically | ||
|
|
|
|||
|
Ban on flavored tobacco products due to lung disease | 231,119 (32.90) | vaping, vape, cigarette, smoking, product, flavor, ban, quit, lung, people | ||
|
Ways to buy vaping products | 174,920 (24.90) | vape, would, buy, get, think, shop, go, need, take, look | ||
|
Time to stop vaping and smoking | 167,192 (23.80) | smoke, time, vape, juul, stop, early, good, hit, read, drink | ||
|
Epidemic of teenager vaping | 129,258 (18.40) | vape, kid, school, new, vapefam, high, top, epidemic, give, vaper | ||
|
|
|
|||
|
Vaping and smoking have risks to get COVID-19 | 354,037 (35.60) | vape, vaping, smoking, smoke, want, could, know, risk, covid, people | ||
|
Buy Juul products through shipping | 225,748 (22.70) | vape, juul, get, shop, keep, still, flavor, product, sure, ship | ||
|
Intention to stop vaping | 210,831 (21.20) | take, vape, stop, late, note, dah, photo, see, guy, friend | ||
|
Vaping and corona virus can cause respiratory disease | 202,875 (20.40) | go, lung, vape, virus, people, young, bro, respiratory, disease, affect |
Before the announcement of the FDA flavor enforcement policy, the e-cigarette–related topics included “Stop vaping and smoking to protect health,” “New e-cigarette flavor use among friends,” etc. After the announcement of the FDA flavor enforcement policy, the topic about “Flavor ban” became popular. At the same time, discussions about “Ways to buy vaping products” or “Buy products through shipping” were getting popular. In addition, after the implementation of the FDA flavor enforcement policy (after February 6, 2020), there were increasing discussions about COVID-19 and e-cigarettes with the appearance of “covid” and “virus” keywords in the topics.
To ameliorate the high prevalence of e-cigarette use in the United States, especially among teenagers and young adults, the FDA announced and implemented an enforcement policy on flavored e-cigarettes in 2020. In this study, we investigated the changes in perceptions of e-cigarettes and the FDA flavor enforcement policy before and after the announcement and implementation of this policy using Twitter data. In addition, frequent topics discussed together with e-cigarettes by Twitter users were examined.
The proportion of tweets with positive sentiment toward e-cigarettes decreased while the proportion of tweets with negative sentiment increased after the announcement of the FDA flavor enforcement policy compared with before the announcement of the policy. These results suggested that the perceptions of US Twitter users toward e-cigarettes were significantly affected by the announcement and implementation of the FDA flavor enforcement policy. The Twitter users’ perceptions of e-cigarettes in general became more negative with the announcement and implementation of the FDA flavor enforcement policy. Different from e-cigarette–related tweets, tweets about the FDA flavor enforcement policy had opposite trends. The proportion of tweets with positive sentiment increased while the proportion with negative sentiment decreased with the announcement and implementation of the FDA flavor enforcement policy.
Several topics that were frequently discussed with e-cigarettes were common during the 3 time periods, such as health concerns (lung cancer and respiratory disease) and quit vaping, which might be partially due to the occurrence of e-cigarette or vaping product use–associated lung injury in 2019 [
To prevent the epidemic of e-cigarette use, the FDA announced several tobacco regulation policies. For example, in April 2014, the FDA released proposed regulation on selling and distributing tobacco products and enhancing the requirement for warning notices on the products [
With the FDA flavor enforcement policy, the public perception of e-cigarettes became more negative. Furthermore, with the announcement of the FDA flavor enforcement policy, e-cigarette users started to discuss about what they would do, for example, quit vaping or find an alternative. These results suggest that the FDA policy had some significant effects on the use of flavored e-cigarettes, which might potentially change user behavior. A recent survey study examined the effectiveness of the FDA warning label on e-cigarette–related products among college students, and the results showed that the warning label proposed by the FDA was more effective than that created by companies, which reduced more college students’ intentions to use e-cigarettes with the FDA warning notices [
During the process of exploring e-cigarette–related conversations on Twitter, we identified topics about the ongoing COVID-19 pandemic and vaping at the beginning of 2020, which was consistent with another social media study on Twitter about COVID-19 and vaping [
In this study, Twitter data were used to investigate the changes in the public perception of the FDA flavor enforcement policy. User demographic information, including gender, age, and ethnicity, were not directly available from the Twitter public API, which might limit our further study on the public perceptions of the FDA flavor enforcement policy and e-cigarettes in different demographic groups. Twitter users do not represent the general population, and Twitter users who tagged their geolocation may or may not represent all Twitter users, which might introduce some bias in this study. Moreover, social bots (agents that communicate more or less autonomously on social media) were not identified and removed from the final data, which might bias the results. In this study, we had different numbers of days of Twitter data before and after the FDA flavor enforcement policy to avoid a possible effect of the New York State law on banning all flavored vaper products that was announced on April 3, 2020, and implemented on May 18, 2020, which might cause comparison bias in the results [
Our study showed that the announcement and implementation of the FDA flavor enforcement policy might have influenced Twitter users’ perceptions of e-cigarettes. The findings of this study provide valuable insights into public responses to the FDA flavor enforcement policy, which can be used as an important guideline for future FDA policies on further regulating flavored e-cigarettes.
Flowchart of the data collection and filtering process.
Tweet examples in positive, neutral, and negative sentiment categories.
application programming interface
Food and Drug Administration
latent Dirichlet allocation
Valence Aware Dictionary and Sentiment Reasoner
The research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health (NIH) and the Food and Drug Administration (FDA) Center for Tobacco Products under Award Number U54CA228110. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or the FDA.
XL, ZX, and DL conceived and designed the study. XL analyzed the data. XL wrote the manuscript. XL, LS, ZX, and DL assisted with interpretation of analyses and edited the manuscript.
None declared.