If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
In the fall of 2019, several US states passed short-term bans on the sale of electronic nicotine delivery systems (ENDS) in response to an outbreak of illnesses strongly linked to tetrahydrocannabinol vaping products that received national news coverage. This study assessed how such state-level ENDS bans in 3 US states may have affected cigarette sales.
•
Cigarette sales in states banning ENDS were significantly higher than would have been observed otherwise. A full ban on ENDS was associated with increased cigarette sales of 7.5% in Massachusetts. Banning non-tobacco flavored ENDS was associated with a 4.6% increase in cigarette sales.
•
This study provides new evidence that banning ENDS was associated with increased cigarette sales using commercial sales data. Our results highlight and quantify potential unintended consequences of ENDS sale restrictions, which should be considered in the future as part of public health impact analyses of such policies.
Abstract
Objectives
In the fall of 2019, several states in the United States passed emergency bans on the sale of electronic nicotine delivery systems (ENDS), in response to an outbreak of illnesses strongly linked to tetrahydrocannabinol vaping products that received national news coverage. Given that ENDS are potential alternative nicotine products for adult smokers, banning ENDS may have unintended consequences. This study provides evidence of an association between state-level ENDS bans and cigarette sales.
Methods
We used difference-in-differences and generalized synthetic control methods to estimate the impacts of the emergency ENDS bans on cigarette sales by comparing treatment states that passed ENDS bans in fall 2019 (Massachusetts, Washington, and Rhode Island), halted states that revoked the announced ENDS bans, and control states.
Results
Our results show that cigarette sales in ban states were higher than would have been observed otherwise in the post-ban period. A full ban on ENDS was associated with increased cigarette sales of 7.5% in Massachusetts (P < .01); banning non-tobacco flavored ENDS was associated with 4.6% (P < .1) higher-than-expected cigarette sales. We did not detect statistically significant impacts in halted states, and placebo tests, which randomly assigned control states as treatments, showed no difference in observed cigarette sales in the same period.
Conclusions
This study provides evidence that banning ENDS is associated with increased cigarette sales. Future research is needed to determine the long-term impact of these policies.
U.S. Department of Health Human Services The Health Consequences of Smoking—50 Years of Progress: A Report of the Surgeon General.
Department of Health Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health,
Atlanta, GA2014
Electronic nicotine delivery systems (ENDS), also known as electronic cigarettes (or e-cigarettes), are alternative products for adult smokers that deliver nicotine through heated aerosols. ENDS have the potential to reduce smokers’ exposure to known toxic and cancer-causing chemicals from combustible tobacco, although they do contain nicotine, are addictive, and can have adverse health effects.
Exposure to nicotine and selected toxicants in cigarette smokers who switched to electronic cigarettes: a longitudinal within-subjects observational study.
A recent Cochrane review of the published literature to date, including randomized-controlled trials, found moderate-certainty evidence that rates of quitting cigarettes were higher with ENDS than nicotine replacement therapy or nicotine-free e-cigarettes.
In the fall of 2019, an outbreak of e-cigarette or vaping product use-associated lung injury (EVALI) raised public health concerns and focused media attention on ENDS.
Although it was eventually determined in February 2020 that EVALI cases were strongly linked to vitamin E acetate additive in primarily illicit tetrahydrocannabinol-containing vaping products,
Update: characteristics of a nationwide outbreak of e-cigarette, or vaping, product use-associated lung injury - United States, August 2019-January 2020.
Risk factors for e-cigarette, or vaping, product use-associated lung injury (EVALI) among adults who use e-cigarette, or vaping, products - Illinois, July-October 2019.
several states announced emergency bans on the sales of ENDS in the fall of 2019. Massachusetts (MA) implemented an emergency ban on all ENDS on September 24, 2019, and Rhode Island (RI) and Washington (WA) instituted similar short-term bans on non-tobacco flavored ENDS in early October. These short-term bans on ENDS provided an opportunity to evaluate the effect of restricting ENDS sales and, in particular, the potential for unintended effects on cigarette sales. The analyses on ENDS restrictions have important policy implications, given that ENDS are potential alternatives for adult smokers.
E-cigarettes and smoking cessation in the United States according to frequency of e-cigarette use and quitting duration: analysis of the 2016 and 2017 National Health Interview Surveys.
The association between e-cigarette use characteristics and combustible cigarette consumption and dependence symptoms: results from a national longitudinal study.
Using state-level cigarette sales data from a third-party commercial database, we explored the impact of state-level ENDS bans on cigarette sales using both difference-in-differences and synthetic control methods (SCMs). The results will provide further information for policymakers regarding the potential unintended consequences of banning ENDS from the market.
Methods
State-Level Policies
In this study, the treatment group includes 3 states: MA, RI, and WA. In response to the EVALI outbreak, MA instituted an emergency ban on all ENDS starting on September 24, 2019. This full ENDS ban was followed by a new tobacco control policy that only allows sales of tobacco-flavored ENDS that have a nicotine concentration of < 35 mg/mL effective from December 11, 2019. Both RI and WA issued 4-month emergency bans on non-tobacco flavored ENDS (henceforth “flavor ban”), which went into effect on October 4, 2019, and October 10, 2019, respectively. In light of the abovementioned differences in ENDS regulations, we treated MA as the full ban state, and RI and WA as the flavor ban states throughout the rest of the article.
In addition to the full ban and flavor ban states, a few states instituted bans that were subsequently reversed. Michigan banned non-tobacco flavored nicotine vaping products effective September 18, 2019, giving retailers until October 2, 2019, to comply, but the ban was halted by a court on October 15, 2019. Similarly, Oregon banned non-tobacco flavored ENDS effective October 15, 2019, and this temporary ban was halted by a court on October 17, 2019. New York, where the tobacco control policies varied by locality, announced a state-level non-tobacco flavor ban on September 17, 2019, with an enforcement date on October 4, 2019, but it was halted by a court on October 3, 2019. Because either the ENDS bans in these states went into effect for short periods or retailers could have reacted to these announced policy changes before they were overturned, we defined these 3 states as the halted group. We expect the impact of the halted ENDS bans on cigarette sales to be of short duration and possibly too small to be detected.
Data
Information Resources, Inc (IRI), collected weekly marketing data by Universal Product Code sales from a sample of stores in 50 markets and 30 categories, covering approximately 25% to 30% of consumer-packaged goods sales, and the data have been widely used in various academic studies.
This study used state-level syndicated sales data collected between January 1, 2018, and January 5, 2020, a period with 105 weekly data points. IRI projected state-level overall sales of convenience stores in 39 states (due to the limited number of stores in some states, IRI did not provide state-level sales data of tobacco products in Alaska, Delaware, Hawaii, Idaho, Kansas, Minnesota, Mississippi, Montana, Nebraska, New Jersey, and New Mexico). Besides the 3 states in the treatment group and 3 states in the halted group, the control group contained all remaining 33 states for which IRI data were available.
We excluded data from 2020, because the temporary bans on nontobacco flavored ENDS in WA and RI lasted for only 4 months and the temporary full ENDS ban in MA was replaced by a new tobacco control law in December 2019. Thus, the study period of our sample covers 2 complete years of weekly data, allowing us to exploit time-series and across-state variations to identify the impact of the ENDS bans.
Methods
We began with a straightforward difference-in-differences model to estimate the impact of the emergency ENDS bans on cigarette sales with the following equation:
(1)
where is our primary outcome variable, log-transformed cigarette pack sales per capita, in each state in week . In the case of the full ban and flavor ban states, the policy indicator equals 1 if the state had been subject to the ENDS ban in week and 0 otherwise. In the case of the halted states, equals 1 after the announced effective date of an ENDS ban and 0 before that date. The coefficient captures heterogeneous treatment (ie, policy) effects for each state (or group). We included state fixed effects as and year-week fixed effects . The covariate matrix contains state-level control variables, including (1) cigarette taxes, (2) indicators for e-cigarette tax adoption, (3) indicators for cigarette and e-cigarette indoor air laws in bars, restaurants, private workplaces, and government workplaces separately, (4) indicators for state-level Tobacco 21 laws, (5) indicators for marijuana medical use, recreational use, and recreational sales legalization separately, (6) monthly unemployment rate, (7) monthly consumer price index, (8) quarterly log-transformed per-capita gross domestic product, (9) monthly temperature, and (10) weekly retail gasoline price.
To control for trends in local markets, state-specific linear time trends were included. The robust errors were clustered at the state level.
The common trends assumption on which the difference-in-differences approach relies is not directly testable, but we conducted a formal event-study analysis to check whether cigarette sales in treatment states had similar trends compared with control states during the period before emergency ENDS bans went into effect.
Because there were only 3 treatment states, the point estimates of treatment effects deriving from the difference-in-differences methodology with a small number of policy change groups are not consistent. Although these estimates are unbiased if the error term has mean zero conditional on the regressors, there remains a concern about time-series autocorrelation in this case.
The GSC method first obtained a fixed number of time-varying coefficients (latent factors) from the IFE model using only the control states and then estimated the state-specific intercepts (factor loadings) for each treated state. This enables the prediction of counterfactuals for the treated states based on estimated latent factors and factor loadings. GSC is able to addresses multiple treatments with different intervention dates and heterogeneous treatment effects and allows weak serial dependence of the error terms.
To ensure that the results were robust, multiple specifications were tested with selected covariates and controls. We conducted a specification of difference-in-differences model with only 2-way fixed effects and state-specific linear time trends and a specification of the GSC model with only latent effects from the IFE model without any other control variables, to avoid confounding effects from any potentially endogenous control variables. To address the concern of potential spillover effects of the ENDS bans in treatment states within close geographic proximity, another specification was conducted by excluding states sharing borders with treatment states from the control group. In addition to the GSC specifications, we applied the traditional SCM, which allowed inclusion of time-invariant covariates such as the median number of hospitalized EVALI cases in each state as reported by the Centers for Disease Control and Prevention. To further validate that the changes in cigarette sales were due to the ENDS bans and not coincidental, we also performed a placebo test by assigning control states the treatment status in the framework of a GSC model.
Results
Descriptive Statistics
Table 1 provides summary statistics of the selected observable covariates, and Figure 1 shows overall cigarette sales per capita for full ban state, flavor ban states, halted states, and control states, during the study period. (Appendix Figure 1, in Supplemental Materials found at https://doi.org/10.1016/j.jval.2021.12.006, shows overall and certain flavored ENDS and cigarette sales.) Although the states were heterogeneous as shown in Table 1 and the mean cigarette sales per capita in treatment states were lower than the means in the control or halted states (Fig. 1), the overall trends in cigarette sales were very similar across these 4 groups before the imposition of the emergency ENDS bans (indicated by vertical dotted lines in Fig. 1). Note that cigarette sales have strong seasonal patterns with relatively high sales in summer and low sales in winter.
Nevertheless, after the policy interventions, the overall cigarette sales trends of the control and halted groups continued to decrease, whereas the downward trend in these treatment states became less steep and, in the case of the flavor ban states, even reversed, suggesting the presence of other factors that countered the seasonal decline in cigarette sales.
Table 1Summary statistics for treatment, halted, and control states.
Covariates
Control states
Flavor ban states
Full ban state
Halted states
Weekly per-capita cig. pack sales
0.666
0.350
0.258
0.450
Weekly per-capita ENDS volume sales
0.050
0.061
0.062
0.046
% smoking prevalence
17.94
14.24
13.70
16.49
Cigarettes tax $
1.45
3.64
3.51
2.56
Population (million)
7.49
4.31
6.89
11.23
% population aged 15-24
13.34
13.36
13.86
12.99
Consumer price index
250.21
267.26
267.26
266.81
% unemployment rate
3.6
4.1
3.1
4.0
GDP per capita
57 620.1
67 157.9
84 560.3
64 870.7
Temperature (F)
54.14
49.19
48.98
46.03
Gasoline price ($ per gallon)
2.69
3.07
2.74
2.94
Note. All variables were averaged from January 2018 to December 2019.
Note: States were divided into 4 groups. The flavor ban group includes RI and WA, which imposed nontobacco flavored ENDS bans on October 4, 2019, and October 10, 2019, respectively. The full ban group only has MA, which initiated a full ENDS ban on September 24, 2019. The vertical dotted lines show when these ENDS bans came into place in RI, WA, and MA. The halted group covers Michigan, Oregon, and New York. We do not call them the placebo group because Michigan and Oregon’s flavor bans did go into effect for a short period and retailers might have reacted to these announced policy changes in all 3 states although the bans were halted later. Eventually, the control group contains all remaining 33 states that are available from the IRI data set.
The main results from the difference-in-differences approach described in Eq. (1) earlier are presented in Table 2. To estimate the treatment effects of imposing full or flavor ENDS bans on cigarette sales, we took the natural logarithm of the weekly per-capita cigarette pack sales in each state to generate effect estimates in percent terms. The specification in column (1) controlled for state and week fixed effects along with state-specific time trends. The model in column (2) further included all control covariates mentioned in the Methods section earlier. The 1-week lag of weekly ENDS sales per capita was included to control for the uptake of ENDS in the specification shown in column (3), although the lag term of ENDS sales could potentially be endogenous. The specification in column (4) excluded states neighboring the treatment states to eliminate possible confounding spatial spillover effects resulting from cross-border purchasing behaviors in response to the emergency ENDS bans. Nonetheless, we can observe that the estimated effects of both full and flavor ENDS bans are consistent across all these different specifications.
Table 2Estimated effects of emergency ENDS bans on cigarette sales using difference-in-differences method.
Note. Robust standard errors are clustered at the state level; with 95% confidence intervals reported in parentheses. Column (1) only controlled for state and week fixed effects along with state-specific time trends; columns (2) and (3) further included all control covariates mentioned in the Methods section, whereas column (3) involved additional 1-week lagged ENDS sales. Column (4) excluded states sharing the same borders with MA, RI, and WA from the control group. Column (5) estimated the state-specific effects for RI and WA, rather than a pooled effect of flavor ban states.
In particular, the results based on column (2), as our preferred estimates, suggest that per-capita cigarette sales increased by approximately 7.5% (P < .01) after MA imposed a full ENDS ban. A smaller average impact of a 4.6% (P < .1) increase in per-capita cigarette sales can be observed when states imposed flavor bans on ENDS sales.
Column (5) further shows the results of a model with state-specific estimates. In this case, cigarette sales increased in both RI and WA after they imposed flavor bans on ENDS. Given the extremely small size of the treatment group with only 3 states, the difference-in-differences approach could have struggled with the precision of estimates, and it is not surprising to see relatively large p-values.
In all the specifications shown in Table 2, we also estimated the effects of the halted ENDS bans on cigarette sales in New York, Oregon, and Michigan, failing to reject the null hypothesis of no effect on cigarette sales in states with halted bans. Appendix Table 1, in Supplemental Materials found at https://doi.org/10.1016/j.jval.2021.12.006, reports the estimated effects of these emergency ENDS bans on ENDS sales with the same difference-in-differences approach and shows that ENDS bans eliminated corresponding ENDS sales.
An essential identifying assumption underlying the above analyses is that the per-capita cigarette sales in the states that imposed emergency ENDS bans would have similar trends to those in control states before the ENDS bans. We conducted a formal event study to examine evidence for the validity of this assumption based on pre-trends before the ENDS bans. To gain statistical power, we pooled all 3 treatment states. The x-axis of Figure 2 indicates the number of months after (with positive values) or before (with negative values) the imposition of ENDS bans in the treatment states. Although the coefficients to the left of the vertical line at (the months when treatment states imposed ENDS bans) were all small and insignificant (with event indicators at one month before the ENDS bans, at , dropped as the benchmark), we observed large and significant, positive impacts on cigarette sales after the ENDS bans came into effect. The results suggest that the parallel pretrends assumption underlying the difference-in-differences is likely to hold, although the 95% confidence intervals of the pretrends coefficients are quite large because of the imprecision stemming from the small sample size.
Figure 2Event study analysis and examination on parallel pre-trends.
To further validate these results and overcome limitations of the difference-in-differences approach using a small treatment group, models with more advanced GSC methods were estimated. Figure 3 displays the actual and estimated weekly cigarette sales per capita in MA, WA, and RI using GSC methods. Although the predicted cigarette sales per capita aligned well with the actual sales in the pre-ban period, the actual and predicted cigarette sales diverged immediately after the implementation of ENDS bans in the 3 treatment states. The alignment in the pre-treatment periods indicates that the covariates and latent factors from the control states provided appropriate approximations of cigarette sales per capita in the absence of the bans. Overall, the actual cigarette sales in the treatment states after the ENDS bans exceeded the estimates of cigarette sales that would have occurred in the absence of these bans.
Figure 3Actual and estimated cigarette sales for states with emergency ENDS bans by generalized synthetic control methods.
Note: The black lines are the actual weekly per-capita cigarette sales in the treatment states; the dashed blue lines are the counterfactual sales estimated by the generalized synthetic control methods based on weighted trends in the control states. The dashed blue lines fit the actual cigarette sales in each treatment state well before the implementation of ENDS bans and represent what sales would have been observed in the same time period if the states had not passed a ban in the grayed period.
Table 3 is a summary of the estimated impacts of the ENDS bans on cigarette sales measured as the average percent difference between actual cigarette sales and estimated cigarette sales using multiple specifications of GSC and SCM. The model in column (1) used GSC with the latent factors from the IFE model only and did not include any control variables. The specification in column (2) used GSC with all control variables as used in the difference-in-differences approach. Note that all time-invariant variables and state-level dummy variables would naturally be absorbed through the IFE process. The model in column (3) excluded states sharing borders with MA, RI, or WA from the control group, given that the ENDS bans could have spatial spillover effects on the neighboring states. The specification in column (4) used the SCM with all control variables and additional time-invariant state-level observations.
Table 3Estimated effects of emergency ENDS bans on cigarette sales using GSC and SCM.
Treatment state
GSC with no covariates (1), %
GSC with full control variables (2), %
GSC excluding neighboring states (3), %
SCM (4), %
MA
6.9 (5.6-8.3)
5.3 (4.0-6.5)
8.0 (6.0-9.9)
3.4 (2.3,4.5)
RI
4.6 (2.7-6.5)
4.0 (1.6-6.3)
5.7 (3.0-8.4)
0.9 (−1.0 to 2.7)
WA
5.0 (3.2-6.7)
5.8 (3.5-8.0)
6.9 (4.9-9.0)
3.6 (1.1-6.1)
Note. 95% confidence intervals are reported in parentheses. Column (1) used GSC without any control variables but with latent factors only. Column (2) used GSC with all control variables as used in the difference-in-differences approach. Note that all state-level dummy variables would naturally be absorbed through the interaction fixed effect scheme. Column (3) used GSC by excluding states sharing the same borders with MA, RI, and WA from the control group. Column (4) used SCM. Overall, the results are significant by all specifications proposed in the work.
Although specific estimates varied across different methods and specifications, overall, the results are consistent and significant. Combining results from Tables 2 and 3, the full ban on ENDS in MA was associated with an increase in overall cigarette sales ranging from 3.4% to 8.8%. Estimates of the effect of the flavor ban on cigarette sales in WA ranged from 3.6% to 8.7%, and the impact of the flavor ban in RI ranged from 0.9% to 5.7%.
In addition to the multiple specifications shown in Tables 2 and 3, as an additional robustness check on the selection of treatment states, Appendix Figure 2, in Supplemental Materials found at https://doi.org/10.1016/j.jval.2021.12.006, shows the average treatment effect on states without any bans using the placebo test. If a state without a ban were chosen as a treatment state, we would find no difference in cigarette sales during the pre- and post-treatment periods, confirming that the observed changes in cigarette sales were not due to the random assignment of bans, but rather because of the actual implementation of bans on ENDS.
Discussion
This article presents evidence of the impact of state-level, short-term bans of ENDS on cigarette sales using both difference-in-differences and SCMs with careful controls for state and time fixed effects, time-varying state characteristics, and other unobserved latent factors with GSC. Our statistical analyses examined how cigarette sales were affected unintendedly as a consequence of abrupt restrictions on ENDS. Using state-level commercial data, we found that banning ENDS likely increased cigarette sales in states that had a full ban and those that only banned non-tobacco flavored ENDS. Notably, we showed that after a full ban, which was instituted in MA, cigarette sales increased by 7.5%, and the flavor bans, which only allowed sales of tobacco-flavored ENDS, were also associated with higher-than-expected cigarette sales in WA and RI, with an average estimate of 4.6%. The effect of the full ban in MA was almost double that of the corresponding increase associated with the partial, non-tobacco flavored bans in RI and WA. Based on the actual cigarette sales and the estimated counterfactuals in each state, we estimated that approximately 3.4 million additional cigarette packs were sold through the convenience store channel across the 3 treatment states during the 3-month ban period (from the effective dates to the end of the study period), including approximately 1.7 million packs in MA and 1.7 million packs in WA and RI.
This study provides some of the first rigorous evidence of the impact of ENDS restrictions at a market level. Previous studies with discrete choice experiments
Buckell J, Marti J, Sindelar JL. Should flavours be banned in cigarettes and e-cigarettes? Evidence on adult smokers and recent quitters from a discrete choice experiment [published online May 28, 2018]. Tob Control. https://doi.org/10.1136/tobaccocontrol-2017-054165.
have suggested that taking ENDS out of consumer choice sets may result in smoking relapse among adult smokers. Nevertheless, these estimates using experimental settings may be biased because they measure the potential impact of banning ENDS within a hypothetical scenario of unavailability, and these hypothetical choice scenarios are different from those of an equivalent realistic choice context.
For instance, a real-world sales ban cannot prevent people from purchasing ENDS through either cross-border shopping or other unregulated channels. Our findings advance the literature by building on the expanding research and empirical evidence of the population-level consequences of ENDS bans on cigarette sales in a real-world setting.
Additionally, this study contributes to the literature on ENDS and cigarettes as substitute products among the smoking population. Our results are consistent with most of the tobacco market literature, which suggests that greater difficulty in accessing ENDS is associated with increased cigarette sales and smoking prevalence. Previous studies using direct or indirect evidence from taxation on ENDS,
have confirmed the substitution relationship between cigarettes and ENDS. In addition, recent research evaluating the impact of heat-not-burn product market entry in Japan and ENDS in Canada concluded that the introduction of these noncombustible alternatives was associated with reduced cigarette sales.
Nevertheless, knowledge about the direct impact of banning ENDS remains sparse, despite the growing literature suggesting substitution effects. Our findings that restricting ENDS leads to increased cigarette sales indicate that ENDS and cigarettes are alternative choices for adult smokers.
Our findings were robust to multiple specifications and robustness checks under the complex and evolving tobacco control environment during the study period. Additionally, we did not detect significant increases in cigarette sales in halted states, in which the proposed ENDS sales bans were reversed or invalidated by the courts shortly after they were enacted or announced. This suggests that the main reason for the increases in cigarettes sales in treatment states was the actual market unavailability of ENDS, rather than state-level variations in risk perceptions or other unobserved reasons associated with proposed ENDS sales bans. Dave et al
have shown that the media coverage of EVALI significantly changed the risk perceptions of ENDS within the US adult population, which could potentially contribute to ENDS users substituting for cigarette smoking. Both difference-in-differences and GSC approaches were able to address this nationwide shock, including changes in risk perception of ENDS, on cigarette sales with two-way fixed effects or latent factors generated from the IFE model. The results from our placebo test further confirmed this given that we did not observe significant changes of cigarette sales in these control states that were treated as treatment units in the analysis. Nonetheless, the results using different specifications remained consistent with the main conclusion and thus suggested our estimated cigarette sales increase was mainly driven by market unavailability of ENDS.
The results were subject to certain limitations. First, IRI only provided sales data for 39 states from convenience stores sales channels for both ENDS and combustible cigarettes. Therefore, it is possible that the overall effects of emergency ENDS bans on cigarette sales were underestimated, because the data excluded other channels such as independent retailers, online sales, or untracked sales. A second limitation is that we could not capture tobacco policy changes that were made below the state level, such as any district or municipal policy changes during the study period, given that the primary outcomes were at the state level. Although this is a common approach in studies using aggregated data, it should be recognized that such tobacco-related policy changes may also have confounding impacts on the presented estimates.
Conclusions
This study provides novel evidence that banning ENDS was associated with increased cigarette sales, using state-level commercial sales data. The results highlight and quantify potential unintended consequences of ENDS sale restrictions, which should be considered in future policy deliberations on tobacco products. Additional research is also needed to investigate the impact on spatial spillover effects,
illicit markets, and other scenarios that may arise in response to ENDS restrictions. Furthermore, the long-term impact of ENDS sales bans on ENDS and cigarette sales, as well as the distal public health outcomes, will need to be studied as additional data become available.
Article and Author Information
Author Contributions:Concept and design: Xu, Jiang, Prakash
Analysis and interpretation of data: Xu, Jiang, Chen
Drafting of the manuscript: Xu, Chen
Critical revision of the paper for important intellectual content: Xu, Jiang, Prakash, Chen
Statistical analysis: Jiang, Chen
Supervision: Xu, Prakash
Conflict of Interest Disclosures: Dr. Chen is a full-time employee of JUUL Labs, Inc (JLI); Drs. Xu and Prakash and Ms. Jiang were full-time employees of JLI, during the time that this work was conducted. All authors reported holding restricted stock units in JLI; reported that JLI provided support for presentations and manuscript writing; and reported that JLI provided support for attending meetings and travel.
Funding/Support: Funding for this study was provided by JLI.
Role of the Funder/Sponsor: The sponsor approved the research plan and provided comments on a near-final draft of the article.
Acknowledgment: The authors thank Siddharth Chandra and Saul Shiffman, who consult with JLI on matters of harm reduction through Pinney Associates, for their valuable suggestions and the 4 anonymous reviewers for their insightful comments and careful consideration of the manuscript.
The Health Consequences of Smoking—50 Years of Progress: A Report of the Surgeon General.
Department of Health Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health,
Atlanta, GA2014
Exposure to nicotine and selected toxicants in cigarette smokers who switched to electronic cigarettes: a longitudinal within-subjects observational study.
Update: characteristics of a nationwide outbreak of e-cigarette, or vaping, product use-associated lung injury - United States, August 2019-January 2020.
Risk factors for e-cigarette, or vaping, product use-associated lung injury (EVALI) among adults who use e-cigarette, or vaping, products - Illinois, July-October 2019.
E-cigarettes and smoking cessation in the United States according to frequency of e-cigarette use and quitting duration: analysis of the 2016 and 2017 National Health Interview Surveys.
The association between e-cigarette use characteristics and combustible cigarette consumption and dependence symptoms: results from a national longitudinal study.
Buckell J, Marti J, Sindelar JL. Should flavours be banned in cigarettes and e-cigarettes? Evidence on adult smokers and recent quitters from a discrete choice experiment [published online May 28, 2018]. Tob Control. https://doi.org/10.1136/tobaccocontrol-2017-054165.
00008-0&id=gr1.jpg){kind=link}
00008-0&id=gr2.jpg){kind=link}
00008-0&id=gr3.jpg){kind=link}