19

Modeling of E-Cigarette Use

In Section III, the committee presented a conceptual framework of smoking transitions. The framework captures multiple hypothesized pathways by which e-cigarette use could affect combustible tobacco cigarette use. The hypothesized pathways can be used to understand both individual tobacco use trajectories and population-level effects, and include

• Youth and young adults could begin using e-cigarettes and subsequently start using combustible tobacco cigarettes, either completely switching or using both products concurrently (increasing combustible tobacco cigarette initiation).
• Youth and young adults who otherwise would have begun combustible tobacco cigarette smoking could begin using e-cigarettes instead (reducing or delaying combustible tobacco cigarette initiation).
• Adults who smoke combustible tobacco cigarettes could switch to using e-cigarettes alone or quit both products (cessation).
• Adult combustible tobacco cigarette smokers could start using e-cigarettes in addition to combustible tobacco cigarettes (dual use). Some portion of these adult smokers may subsequently transition to e-cigarette use alone (cessation).
• Former smokers could start using e-cigarettes and subsequently transition to combustible tobacco cigarettes (relapse) either alone or concurrently with e-cigarettes (dual use).

Some of these pathways will result in harms while others may confer benefits. In any population, each of these pathways may occur among both individuals and subpopulations. Thus, e-cigarette use could produce harms for some individuals while conferring benefits to others. These benefits and harms can offset each other, making it difficult to draw inferences about the net effect of e-cigarettes at the population level. Thus, to facilitate an assessment of the overall population health effect of e-cigarettes in the U.S. population as a whole, this chapter uses modeling to apply a common metric (years of life lost or gained as a measure of mortality) to these pathways as they occur simultaneously among different subgroups.

Models of population dynamics have been used in tobacco control for more than two decades. The 2014 Surgeon General’s report, The Health Consequences of Tobacco—50 Years of Progress, presents a summary of those models (HHS, 2014). More recently, several modeling studies have addressed the potential future impact of e-cigarettes under various assumptions, reaching different conclusions (Cherng et al., 2016; Hill and Camacho, 2017; Kalkhoran and Glantz, 2015; Levy et al., 2017; Vugrin et al., 2015). To inform their view of the likely effects of e-cigarette use at a population level, the committee employed a dynamic model of combustible tobacco cigarette smoking prevalence and health effects to examine the potential impact of e-cigarettes on mortality in the U.S. population over the next few decades under various assumptions. Specifically, the committee used a well-established dynamic model of tobacco control (Mendez and Warner, 2004; Mendez et al., 1998) to estimate the cumulative number of life-years lost (or gained) due to e-cigarettes during 2015–2050 and 2015–2070 under different assumptions regarding the harm of e-cigarettes compared with combustible tobacco cigarettes, and their potential effects on the initiation and cessation rates of combustible tobacco cigarettes. The results of the model are not precise forecasts, but rather simulation outputs that inform a qualitative assessment about the potential population health impacts of e-cigarettes.

MODEL

The Mendez-Warner model (Mendez and Warner, 2004; Mendez et al., 1998) tracks individuals in the population from age 0 to a maximum age of 110, additionally differentiated by gender and smoking status. The number of people of age a in year t is computed by multiplying the number of people of age a – 1 in year t – 1 by the appropriate survival rate (1 − death rate). Birth cohort sizes are supplied exogenously to the model. Death rates are differentiated by year, gender, age, and smoking status. The model tracks the adult population smoking status. At age 18, indi-

viduals are characterized as current, former, or never smokers. The definition of an adult current smoker is consistent with that of the National Health Interview Survey (NHIS)—those who have smoked at least 100 cigarettes in their lifetime and are smoking now every day or some days. In this model, adult initiation is measured by the proportion of the population who are current smokers at age 18. Youth smoking history before age 18 is subsumed in the adult initiation measure. Subsequently, current smokers in any given year are estimated as the number of current smokers in the previous year who survived to the current year and did not quit smoking. Former smokers are those who were former smokers the previous year and did not die, plus those who were current smokers the previous year and did not die, but quit. The model differentiates former smokers up to 30 years abstinent, and years-since-quitting−specific death rates are applied accordingly to those individuals.

Smoking prevalence for any specific age group in a specific year is computed by taking the ratio of current smokers to the total number of people within the group that year. Baseline cessation rates were estimated within the model using the NHIS and the National Survey on Drug Use and Health data for the period 1990–2014. The model uses permanent quit rates, that is, quitting net of relapse, so that these rates are smaller than those used in models that include quits that eventually result in relapse. The model is calibrated periodically, and is tracking with excellent accuracy the overall adult smoking prevalence in the United States. At the model baseline (2014), e-cigarette use among working adults was 3.8 percent (Syamlal et al., 2016). However, the model considers that e-cigarette use may increase combustible tobacco cigarette initiation among non-smokers and may also promote cessation among dual users. Thus, the effects of increased e-cigarette prevalence are subsumed in the assumptions of increases in both cessation and initiation of combustible tobacco cigarettes. The model does consider gender differences (e.g., relative risks and death rates), although the committee applied the same values to men and women for some parameters (e.g., background cessation rates). The model uses age, gender, and smoking-status−specific death rates, derived from data from the Cancer Prevention Study II. The model assumes that no smoking-related deaths occur before age 35.

MODELING ASSUMPTIONS

In the model, the committee assumes that the introduction of e-cigarettes has the potential to increase smoking initiation among young adults, and smoking cessation among adults. The committee also assumes that e-cigarettes are not harmless, and that e-cigarette use increases the risk of mortality over that of a non-vaper, non-smoker individual. At the

same time, the risk of mortality among e-cigarette users is lower than that among combustible tobacco cigarette smokers. Dual users are treated as current smokers in terms of risk but also as having a different cessation rate than non-vaper smokers. To be conservative, dual users who quit are assumed to continue using e-cigarettes for the rest of their lives and are given a reduction of risk consistent with the direct harm effect assumed for e-cigarettes. For example, if we assume that e-cigarettes are 10 percent as harmful as cigarettes, a dual user who quits smoking will be given 90 percent of the reduction in risk that a non-vaper quitter would attain as a former smoker.

The committee’s assumptions about possible effects on smoking initiation among young adults, smoking cessation among adults, and the harm of e-cigarettes in relation to combustible tobacco cigarettes are informed by the committee’s review of the literature presented in the preceding chapters. Some of the parameters were also chosen to provide an extreme upper limit for the harmful effects of e-cigarettes and to illustrate the level of such negative effects necessary to counterbalance the potential benefits of e-cigarettes at the population level. In particular, the simulations contain scenarios where e-cigarettes are 50 percent as harmful as cigarettes and/or increase initiation by 50 percent. The committee considers those scenarios to be extreme and highly unlikely.

The committee considered the following specific effect levels:

• E-cigarettes increase the smoking initiation rate by 0 percent, 5 percent, 10 percent, 25 percent, or 50 percent;
• E-cigarettes increase the net smoking cessation rate by −5 percent, 0 percent, 5 percent, 10 percent, or 15 percent; and
• E-cigarettes are 0 percent, 10 percent, 25 percent, or 50 percent as harmful as combustible tobacco cigarettes.

The range of parameter values were selected according to the criteria described in the following sections: e-cigarette effect on initiation, e-cigarette effect on cessation, and e-cigarette harm.

E-Cigarette Effect on Initiation

As concluded in Chapter 16, e-cigarette use likely increases the risk of ever using combustible tobacco cigarettes among youth. However, it is unclear whether this increase in ever use results in an increased adult initiation rate. The committee decided to examine a wide range of effect levels, from no impact on initiation to a 50 percent increase in initiation. The upper limit of 50 percent implies that e-cigarettes will not only stop the currently observed downward trend in the adult initiation rate, but

that they will increase initiation from its 2015 value of 13 percent (Jamal et al., 2016) to 19.5 percent, a level not observed since 2011 (CDC, 2012). The committee considers this level extreme and very unlikely, as discussed above.

E-Cigarette Effect on Cessation

Recent meta-analyses including randomized controlled trials and cohort studies report an adjusted odds ratio for cessation around 0.7 (with versus without e-cigarettes); on the other hand, a recent population study (Zhu et al., 2017) reports an adjusted odds ratio of 1.65 (1.40–1.93). Taking the 2014 prevalence of dual users as 16.2 percent (Syamlal et al., 2016), a 0.7 odds ratio translates into 16.2 percent × (0.70 – 1) = −4.86 percent increase (or 4.86 percent decrease) in the overall cessation rate, while 1.65 odds ratio implies a 16.2 percent × (1.65 – 1) = 10.53 percent increase in cessation, with an upper bound of 16.2 percent × (1.93 – 1) = 15 percent. Based on these values, the committee chose to model values between −5 percent and 15 percent for the effect of e-cigarettes on the overall population cessation rate.

An important note is that, as described above, the model uses permanent quit rates (i.e., quitting net of relapse). Thus, cessation in the modeling refers to net cessation rates. In other words, a positive value indicates that more people in a population have quit smoking than nonsmokers who have started/relapsed, and a negative value indicates the opposite. This net cessation statistic is not a common measure in the literature, which generally reports only a cessation rate based on the percentage of smokers who successfully quit smoking, without regard to non-active smokers at baseline who started or relapsed within a specified time period. For clarity, the committee uses the term “net cessation” when discussing the modeling.

E-Cigarette Harm

As concluded in previous chapters, e-cigarettes are likely to be less harmful than combustible tobacco cigarettes. Estimates of how harmful they are relative to combustible tobacco cigarettes range from 5 percent estimated by the UK Royal College of Physicians (RCP, 2016)¹ to 30–50 percent estimated by Glantz (2016), with most agreement concentrated around the lower figure. The committee examined a wide range of values for the relative harm of e-cigarettes compared with combustible tobacco cigarettes, from 0 to 50 percent as harmful as combustible tobacco ciga-

___________________

¹ This reference was added since the final release of the publication.

rettes. The upper limit of 50 percent was selected as an extreme and improbable value, used to set an upper limit to the potential harm of e-cigarettes. The likelihood that e-cigarettes have none of the harm of combustible tobacco cigarettes is equally extreme and improbable.

SIMULATION SCENARIOS

The model runs were designed as follows: First, to establish a base case, the model was used to estimate the number of life-years lost due to smoking over the periods 2015–2050 and 2015–2070, under the assumption that the annual initiation and net cessation rates observed in 2015 (13 percent among young adults age 18–24 and 4.35 percent among adult smokers, respectively [Jamal et al., 2016; Mendez et al., 2017]) would remain constant in the future. Then, starting in 2015, the committee increased the base initiation and net cessation rates by different percentages that reflect the impact of e-cigarettes on those rates and again calculated the cumulative life-years lost or gained over the same periods as in the base-case scenario. As described in the model assumptions, the committee also assumed that individuals who quit smoking because of e-cigarettes will continue to use e-cigarettes for the remainder of their lives, and so they only achieve a fraction of the health benefits due to quitting combustible tobacco smoking. Finally, the committee compared the different scenarios with the base case to calculate the extra number of life-years gained or lost due to the effects of e-cigarettes.

Overall, the committee considered 85 different simulation scenarios. They reflect a range of likely real-world scenarios as well as scenarios that the committee views as extreme and unlikely, for heuristic purposes. The committee only considered five cases in which e-cigarette use reduces net cessation because it chose not to increase the relative risk of death for anyone beyond that of a current smoker. That is, the differential effect of reducing the net cessation rate would be to increase the number of smokers, who would then be subject to the mortality risk of a current smoker, regardless of the harm associated with e-cigarettes.

The entirety of the simulation runs is summarized in Table 19-1.

RESULTS

All scenarios show a decrease in combustible tobacco cigarette smoking prevalence, which reflect effects from past tobacco policies. Table 19-2 presents the model-estimated life-years lost during 2015–2050 due to e-cigarettes, under the assumption that e-cigarettes cause no harm directly, but their health consequences stem from their effects on initiation and net cessation of combustible tobacco cigarettes. The first section

TABLE 19-1 Summary of Simulation Runs Considered by the Committee

of the table (upper part) shows the life-years lost due to combustible tobacco cigarettes and e-cigarettes combined; the second section shows the life-years lost attributable to e-cigarettes; and the third section shows the same figure as in the second section, as a fraction of the total toll of combustible cigarettes over 2015–2050. In this table, as in all subsequent tables, results shown in red text in parentheses indicate life-years saved compared with a baseline scenario of no effect of e-cigarettes on initiation or net cessation of combustibles.

For example, under the scenario that e-cigarette use causes a decrease of 5 percent (from 4.35 percent to 4.13 percent) on the net cessation rate, and an increase of 5 percent on the initiation rate (from 13 percent to 13.65 percent), the estimated total life-years lost due to smoking (counting the extra smokers because of e-cigarettes) would be 296,067,599. Given

TABLE 19-2 Model-Estimated Life-Years Lost During 2015–2050 Due to E-Cigarettes^a

^a Under the assumptions that e-cigarettes cause no harm directly and a range of assumptions about smoking initiation and net cessation. E-cigarettes = 0% × risk of combustibles.

NOTE: Results shown in red text in parentheses indicate life-years saved compared with a baseline scenario of no effect of e-cigarettes on initiation or net cessation of combustible tobacco cigarettes.

that approximately 50 percent of life-long smokers die prematurely of smoking-related causes, losing an average of 20 years of life, this figure translates into 14,803,380 total premature deaths over a span of 35 years, or an average of 422,954 premature deaths per year. Out of this figure, the extra toll imposed by e-cigarettes is 1,461,811 life-years lost (or 2,088 premature deaths per year), representing 0.5 percent of the total toll of smoking over the 35-year span.

If, on the other hand, e-cigarettes increased net smoking cessation rates by 5 percent (4.57 percent) while still increasing initiation by 5 percent, 1,228,751 life-years would be saved in 2015–2050, representing approximately 1,755 premature deaths averted per year. Under the scenario of a 5 percent increase in the net smoking cessation rate due to e-cigarettes, even assuming that e-cigarettes increase the initiation rate by 50 percent (to 19.5 percent), there would still be 650,927 life-years saved by 2050.

Scenarios extending outcomes through 2070 under the same assumptions indicate worse outcomes in all scenarios compared with those through 2050. This is because, under any scenario that increases adult initiation, the benefits of increased net cessation are felt much sooner than the negative effects of increased initiation. Of note, the committee chose to keep the background rates on smoking initiation and cessation constant, to avoid forecasting future values of those parameters. In reality, current trends indicate that the adult smoking initiation rate is decreasing while the cessation rate is increasing. If those trends continue into the future, the negative effects of e-cigarettes on smoking initiation will be smaller while the positive effects on cessation will be larger.

Table 19-3 illustrates this fact. It shows cumulative life-years lost (or saved) over 2015–2070. For example, assuming that e-cigarettes increase the initiation rate by one-quarter (to 16.3 percent), and the net cessation rate by 5 percent (to 4.57 percent) in 2015, around 577,000 life-years would be lost by 2070 due to e-cigarettes. However, under the same conditions, there would be 971,940 extra life-years by 2050; by 2070, this gain would be offset by the excess mortality brought by the increased initiation.

The rest of the scenarios show the same results as Tables 19-2 and 19-3, considering different levels of harm associated with e-cigarettes.

Tables 19-4 and 19-5 show the life-years lost by 2050 and 2070, respectively, under the assumption that e-cigarettes cause 10 percent of the harm of (i.e., are 90 percent less harmful than) combustible tobacco cigarettes. These tables show that, if net smoking cessation increases by 5 percent, by 2050, there would be life-years gained. The gains would range from 467,228 life-years if smoking initiation increases by 50 percent to 1,110,728 life-years if there is no increase in smoking initiation. By 2070, if net smoking cessation increases by 5 percent, there would be life-years gained

TABLE 19-3 Model-Estimated Life-Years Lost During 2015–2070 Due to E-Cigarettes^a

^a Under the assumptions that e-cigarettes cause no harm directly and a range of assumptions about smoking initiation and net cessation. E-cigarettes = 0% × risk of combustibles.

NOTE: Results shown in red text in parentheses indicate life-years saved compared with a baseline scenario of no effect of e-cigarettes on initiation or net cessation of combustible tobacco cigarettes.

TABLE 19-4 Model-Estimated Life-Years Lost During 2015–2050 Due to E-Cigarettes^a

^a Under the assumptions that e-cigarettes are 90 percent less harmful than combustible tobacco cigarettes and a range of assumptions about smoking initiation and net cessation. E-cigarettes = 10% × risk of combustibles.

NOTE: Results shown in red text in parentheses indicate life-years saved compared with a baseline scenario of no effect of e-cigarettes on initiation or net cessation of combustible tobacco cigarettes.

TABLE 19-5 Model-Estimated Life-Years Lost During 2015–2070 Due to E-Cigarettes^a

^a Under the assumptions that e-cigarettes are 90 percent less harmful than combustible tobacco cigarettes and a range of assumptions about smoking initiation and net cessation. E-cigarettes = 10% × risk of combustibles.

NOTE: Results shown in red text in parentheses indicate life-years saved compared with a baseline scenario of no effect of e-cigarettes on initiation or net cessation of combustible tobacco cigarettes.

TABLE 19-6 Model-Estimated Life-Years Lost During 2015–2050 Due to E-Cigarettes^a

^a Under the assumptions that e-cigarettes are 75 percent less harmful than combustible tobacco cigarettes and a range of assumptions about smoking initiation and net cessation. E-cigarettes = 25% × risk of combustibles.

NOTE: Results shown in red text in parentheses indicate life-years saved compared with a baseline scenario of no effect of e-cigarettes on initiation or net cessation of combustible tobacco cigarettes.

TABLE 19-7 Model-Estimated Life-Years Lost During 2015–2070 Due to E-Cigarettes^a

^a Under the assumptions that e-cigarettes are 75 percent less harmful than combustible tobacco cigarettes and a range of assumptions about smoking initiation and net cessation. E-cigarettes = 25% × risk of combustibles.

NOTE: Results shown in red text in parentheses indicate life-years saved compared with a baseline scenario of no effect of e-cigarettes on initiation or net cessation of combustible tobacco cigarettes.

TABLE 19-8 Model-Estimated Life-Years Lost During 2015–2050 Due to E-Cigarettes^a

^a Under the assumptions that e-cigarettes are half as harmful as combustible tobacco cigarettes and a range of assumptions about smoking initiation and net cessation. E-cigarettes = 50% × risk of combustibles.

NOTE: Results shown in red text in parentheses indicate life-years saved compared with a baseline scenario of no effect of e-cigarettes on initiation or net cessation of combustible tobacco cigarettes.

TABLE 19-9 Model-Estimated Life-Years Lost During 2015–2070 Due to E-Cigarettes^a

^a Under the assumptions that e-cigarettes are half as harmful as combustible tobacco cigarettes and a range of assumptions about smoking initiation and net cessation. E-cigarettes = 50% × risk of combustibles.

NOTE: Results shown in red text in parentheses indicate life-years saved compared with a baseline scenario of no effect of e-cigarettes on initiation or net cessation of combustible tobacco cigarettes.

under scenarios with 0 percent, 5 percent, and 10 percent increases in smoking initiation, and life-years lost if smoking increased by 25 or 50 percent.

Tables 19-6 and 19-7 show the life-years lost by 2050 and 2070, respectively, under the assumption that e-cigarettes cause 25 percent of the mortality harm of combustible tobacco cigarettes. If e-cigarettes increase the net cessation rate by 5 percent, by 2050 there would be life-year gains, ranging from 198,061 if e-cigarettes increase the initiation rate by 50 percent to 843,764 if e-cigarettes have no impact on the initiation rate. Extending the same scenarios to 2070, the results show that, with an increase of 5 percent in net cessation, there would be cumulative life-year gains under the 0, 5, and 10 percent increase in initiation scenarios, but life-year losses below 25 percent and 50 percent increase in initiation assumptions.

Tables 19-8 and 19-9 show the life-years lost by 2050 and 2070, respectively, under the assumption that e-cigarettes are half as harmful as combustible tobacco cigarettes. If e-cigarettes increase the net cessation rate by 5 percent, by 2050 there would be life-year gains if net smoking initiation increases by 0, 5, 10, or 25 percent, but life-year losses if initiation increases by 50 percent. By 2070, if e-cigarettes increase the net cessation rate by 5 percent, there are life-year gains only if there are no increases in smoking initiation.

SUMMARY

The specific time frame and magnitude of population health effects of e-cigarettes will depend on their impact on the rates of initiation and net cessation of combustible tobacco cigarettes and their intrinsic harm. Any population health effect includes the possibility of some groups incurring harm (e.g., youth who initiate combustible tobacco cigarettes), while others benefit (e.g., adult combustible tobacco cigarette users who completely quit or reduce smoking). As with other models of population health effects of tobacco use, the effects of changing net cessation rates are seen earlier than effects of changing initiation rates, due to the lag in time for serious chronic health effects of combustible tobacco cigarettes to manifest.

Under the assumption that the use of e-cigarettes increases the net cessation rate of combustible tobacco cigarette smoking among adults (i.e., the increase in permanent quitting offsets the potential relapsing of former smokers because of e-cigarettes), the modeling projects that use of these products will generate a net public health benefit, at least in the short run. The harms from increased initiation by youth will take time to manifest, occurring decades after the benefits of increased cessation are seen. However, for long-range projections (e.g., 50 years out), the net

public health benefit is substantially less, and is negative under some scenarios. With the range of assumptions used, the model projects that there would be net public health harm in the short and long term if the products do not increase net combustible tobacco cessation in adults.

Factors that would maximize potential health benefits associated with these products include determining with more precision whether and under which conditions e-cigarettes could serve as an effective smoking cessation aid; discouraging their use among youth through standard tobacco control strategies, such as education and access restrictions; and increasing their safety through data-driven engineering and design.

REFERENCES

CDC (Centers for Disease Control and Prevention). 2012. Current cigarette smoking among adults—United States, 2011. Morbidity and Mortality Weekly Report 61(44):889–894.

Cherng, S. T., J. Tam, P. J. Christine, and R. Meza. 2016. Modeling the effects of e-cigarettes on smoking behavior: Implications for future adult smoking prevalence. Epidemiology 27(6):819–826.

Glantz, S. 2016. Accumulating evidence suggests e-cigs 1/3 to 1/2 as bad as cigs (maybe higher). https://tobacco.ucsf.edu/accumulating-evidence-suggests-e-cigs-13-12-bad-cigs-maybe-higher (accessed November 30, 2017).

HHS (U.S. Department of Health and Human Services). 2014. The health consequences of smoking—50 years of progress: A report of the Surgeon General. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Disease Prevention and Health Promotion, Office on Smoking and Health.

Hill, A., and O. M. Camacho. 2017. A system dynamics modelling approach to assess the impact of launching a new nicotine product on population health outcomes. Regulatory Toxicology and Pharmacology 86:265–278.

Jamal, A., B. A. King, L. J. Neff, J. Whitmill, S. D. Babb, and C. M. Graffunder. 2016. Current cigarette smoking among adults—United States, 2005–2015. Morbidity and Mortality Weekly Report 65(44):1205–1211.

Kalkhoran, S., and S. A. Glantz. 2015. Modeling the health effects of expanding e-cigarette sales in the United States and United Kingdom: A Monte Carlo analysis. JAMA Internal Medicine 175(10):1671–1680.

Levy, D. T., Z. Yuan, Y. Luo, and D. B. Abrams. 2017. The relationship of e-cigarette use to cigarette quit attempts and cessation: Insights from a large, nationally representative U.S. survey. Nicotine & Tobacco Research. https://doi.org/10.1093/ntr/ntx166 (accessed February 8, 2018).

Mendez, D., and K. E. Warner. 2004. Adult cigarette smoking prevalence: Declining as expected (not as desired). American Journal of Public Health 94(2):251–252.

Mendez, D., K. E. Warner, and P. N. Courant. 1998. Has smoking cessation ceased? Expected trends in the prevalence of smoking in the United States. American Journal of Epidemiology 148(3):249–258.

Mendez, D., J. Tam, G. A. Giovino, A. Tsodikov, and K. E. Warner. 2017. Has smoking cessation increased? An examination of the U.S. adult smoking cessation rate 1990–2014. Nicotine & Tobacco Research 19(12):1418–1424.

RCP² (Royal College of Physicians). 2016. Nicotine without smoke: Tobacco harm Reduction. London, UK: Royal College of Physicians.

Syamlal, G., A. Jamal, B. A. King, and J. M. Mazurek. 2016. Electronic cigarette use among working adults—United States, 2014. Morbidity and Mortality Weekly Report 65(22): 557–561.

Vugrin, E. D., B. L. Rostron, S. J. Verzi, N. S. Brodsky, T. J. Brown, C. J. Choiniere, B. N. Coleman, A. Paredes, and B. J. Apelberg. 2015. Modeling the potential effects of new tobacco products and policies: A dynamic population model for multiple product use and harm. PLoS ONE 10(3):e0121008. https://doi.org/10.1371/journal.pone.0121008 (accessed February 8, 2018).

Zhu, S. H., Y. L. Zhuang, S. Wong, S. E. Cummins, and G. J. Tedeschi. 2017. E-cigarette use and associated changes in population smoking cessation: Evidence from U.S. current population surveys. BMJ 358:j3262. https://doi.org/10.1136/bmj.j3262 (accessed February 8, 2018).

___________________

² This reference was added since the final release of the publication.

This page intentionally left blank.