The Kentucky SimSmoke Tobacco Control Policy Model of Smokeless Tobacco and Cigarette Use

Background: Smokeless tobacco (SLT) prevalence was decreasing in Kentucky before 2007, but has since increased. This study examines the impact of policies on cigarette and SLT use by applying the SimSmoke tobacco control policy simulation model. Methods: Using data from the large-scale Tobacco Use Supplement of the Current Population Survey (TUS-CPS) and information on state-specific tobacco policies, Kentucky SimSmoke is updated and extended to incorporate exclusive SLT and dual cigarette and SLT use. The model is validated using survey data through 2017. The model was used to estimate the impact on smoking and SLT prevalence and attributable deaths of policies implemented between 1993 and 2018 and the impact of stronger future policies implemented in 2018 and maintained through 2060. Results: SimSmoke generally reflects trends in exclusive cigarette use from the TUS-CPS and the Behavioral Risk Factor Surveillance System (BRFSS), but underestimated the increase in SLT prevalence in recent years. SimSmoke projects that policies implemented between 1993 and 2018 reduced male and female cigarette use by 23.7% and 23.0%, and male and female SLT use by 4.9% by 2018, averting 9018 tobacco-attributable deaths by 2018, increasing to 89 547 by 2060. The largest reductions in cigarette and SLT use were attributed to cigarette price increases. Strengthening tobacco control policies could reduce smoking prevalence by 41% and 40%, and reduce SLT prevalence by 33% and 25% for males and females by 2060. Conclusion: Our results suggest that cigarette-oriented policies were effective in reducing SLT use but have been less successful in recent years. Future use rates can be further reduced through more restrictive statewide policies, which also target non-combustible nicotine products.

Tobacco Use Individuals evolve from never tobacco users to current exclusive smokers, exclusive SLT, or dual users through cigarette, SLT, and dual initiation. Before age 14, the population is considered to consist entirely of never smokers. From age 15, a percentage of each age group transitions to current tobacco users through smoking and SLT initiation rates. Current exclusive smokers, exclusive SLT users, or dual users may quit and thereby become former users. Former exclusive smokers through cessation may return to smoking through relapse and similarly for former exclusive SLT users and former dual users. Former dual users do not relapse to exclusive tobacco users. A discrete time, first-order Markov process is employed to project smoking, SLT use, and dual use rates through initiation, cessation, and relapse.
Baseline estimates of exclusive smoking, exclusive SLT, and dual use status by age and gender were obtained from the nationally-representative 1992/1993 Tobacco Use Supplement of the Current Population Survey (TUS-CPS). 21 Current smoker is defined as having smoked more than 100 cigarettes in his/her lifetime and currently smoking either daily or some days. A question on SLT "use on a regular basis" is used to distinguish SLT users. Dual use is defined as those meeting the respective definitions of current smokers and SLT users. Due to the small female dual use prevalence (<0.1%), we do not consider this group. Former use is defined as those meeting their respective definitions of tobacco-use, but reporting no current use, and they are stratified into quit-years groups (<1, 1-2, 3-4, 5-9, 10-15, 16+ years). Since former smokers were not distinguished by former exclusive users and former dual users, they were estimated by multiplying former smoking prevalence by years quit by the ratio of current exclusive smokers or current dual users to total current smokers.
The TUS-CPS does not provide sufficient information to distinguish initiation, cessation, and switching rates between SLT use and cigarettes, and the evidence on initiation and early transitions to SLT use is mixed. [22][23][24][25][26] To measure initiation while incorporating cessation and switching, we calculate net initiation as the difference between 1993 prevalence at the current age and previous age. Since studies indicate limited switching between SLT use and cigarettes except at younger ages, [27][28][29] we allow switching only via net initiation. Based on cigarette and SLT use peak rates, we allowed cigarette and SLT net initiation through age 27 for males and 34 for females.
Tobacco cessation is explicitly modeled after the last age of net initiation. Smoker quit rates were obtained from the 1992/1993 TUS-CPS, measured as those who quit in the last year, but not the last three months. 30 Because data were not available to estimate quit rates for exclusive SLT and dual users, we assumed that dual and exclusive SLT users quit and relapse rates by years quit are the same as for smokers [31][32][33][34] based on findings that quit rates are at least as high among SLT and dual users as for cigarette users. [27][28][29]35,36 Age, gender, and quit year-specific relapse rates are obtained from the US SimSmoke. [31][32][33][34] Tobacco-Attributable Deaths Mortality rates were developed for each age, gender, and tobacco use category using the relative risks and prevalence rates of the categories and overall mortality by age and gender. Data on mortality relative risk for current and former smokers by age and gender were from the Cancer Prevention Study II. 33,37,38 We assigned the same risks to exclusive cigarette and dual users, with risks declining for both former user categories with years quit. 33,37,38 SLT users were assigned an overall mortality relative risk of 1.15 based on results from a large-scale US study. 13 Tobacco-attributable deaths were calculated as the number of users in each current and former smoking category multiplied by their excess mortality risk (ie, mortality rate of each current and former tobacco use group minus never smoker mortality rate) and then summed.

Policies
SimSmoke begins with policies at their 1993 levels and then incorporates the effect of policy changes between 1993 and 2018. SimSmoke includes tobacco prices (taxes), smoke-free air laws, mass media campaigns, marketing restrictions, health warnings, cessation treatment policies, and youth access. Table 1 shows the policy descriptions and effect sizes. Most policies are modeled as having direct reductions in prevalence rates (immediate impact) with the effect growing over time from changes in the initiation and cessation rates. Each policy effect size is estimated in terms of the percentage change relative to initial rates. All policy effect sizes are multiplicatively applied. Policy levels are based on the specific characteristics of each policy (eg, size of the tax, expenditures on media campaigns, coverage of smoke-free-air laws). Policy levels defined in Table 1, are based on those in effect in January of the corresponding year.
Cigarette and SLT price effects depend on price elasticities, prices through the tracking period, and tax changes in future years. 39 Price elasticities are based on demand studies, which report similar elasticities for SLT and cigarettes. [40][41][42] Cigarette prices are based on average Kentucky cigarette retail prices (including generics) 43 from 1993 to 2018. SLT prices, the weighted average of the prices of chewing tobacco, snuff, and snus, are estimated by manufacturer prices, state and federal taxes, and wholesale and retail mark-ups. 44 We adjusted post-2015 cigarettes and SLT prices by their respective state tax increases and inflation using the consumer price index. 45 Price effects were weighted (75% cigarette, 25% SLT) for dual users. 17 Federal and Kentucky cigarette excise taxes were increased in 2000, 2002, 2005, and 2009. Federal and Kentucky SLT taxes were increased in 2000,2005,2009, and 2010, then slightly decreased in 2013. 17 Smoke-free air laws consider the existence and enforcement of worksite, restaurant, pubs and bars, and other public place bans. 46 The policy effect sizes (shown in Table 1) on exclusive SLT and dual use are assigned 25% of cigarette use. 42 Before 2003, Kentucky had no significant state smoke-free air laws, but since then, local restaurants, bars, and indoor workplace bans were implemented to cover about one-third of the state population. 47 Based on estimates from the US model, 16 we estimated 80% compliance for all users in 1993-2018.
SimSmoke considers tobacco control expenditures, which are usually for media campaigns. 48 Based on evaluation studies of SLT-oriented media campaigns, 42 the policy effect size for exclusive SLT use is 50% less than for exclusive and dual cigarette users. The levels are distinguished as high, medium, or low. 49 Based on state expenditures, 50 this policy level is set at low in 1993-1999, medium in 2000-2002, and low since 2003.
Marketing restrictions are distinguished as high, medium, or low levels. In Kentucky, the policy was assigned a low level in all years with 80% exposure. The effect sizes for smokers and SLT users are the same.
Health warnings depend on the size, location and whether they are graphic. 51 Warnings on cigarette packages at a minimal level (<30% of the package) have not changed since first implemented in 1966. SLT package warnings were also at a minimum level until 2009. Since 2010, SLT packages have been required to display large text warnings covering at least 30% of two principal sides, raising the level to moderate. 42,52 Cessation treatment policies include the implementation and enforcement of four sub-policies: pharmacotherapy (PT) availability, financial coverage of treatments, quitlines, and brief interventions. 53 Reviews of SLT cessation trials find mixed effects for pharmacotherapies, although slightly stronger effects for behavioral interventions. 42 Nonetheless, SLT users utilize these resources at low rates. 54 We set a 50% and 25% reduction of this policy effect for SLT and dual users, respectively, from the effect for smokers. Based on data from private and other healthcare program coverage in Kentucky, we estimate that 30% of the population is covered for PT and behavioral therapy from 1993 to 2009, increasing to 35% in 2010 with Medicaid coverage, and to 50% in 2014 with the Affordable Care Act. About 40% of smokers in all years were receiving brief interventions (asking about smoking, advising to quit and recommending effective cessation treatments) at appropriate levels from their healthcare providers. 55 We also classify Kentucky as having no effective quitline until 2005, when the Kentucky Tobacco Quitline was implemented. 56 Strongly enforced and publicized youth access laws yield a larger reduction in youth smoking initiation for 10-15 yearolds than for [16][17] year-olds, and is enhanced by vending machines and self-service bans. 57 Studies report lower compliance rates for SLT use than for cigarettes sales, 58,59 and Results were also validated using gender-specific estimates from the BRFSS for overall smokers (exclusive cigarette and dual users) from 1996 to 2017. We performed a separate validation for total SLT (exclusive SLT and dual users) from 2011-2016. We also considered confidence intervals from each of the surveys.
Assessing the Effect of Tobacco Control Policies Upon validating the model, we estimate the effect of three policy scenarios (status-quo, counterfactual, and future) for their impacts on exclusive cigarette, exclusive SLT, and dual user prevalence and tobacco-attributable deaths. The status quo scenario incorporates policies implemented from 1993 to 2018. The counterfactual scenario represents trends in the absence of any policy change and is programmed with all policies at their 1993 levels. We estimate the impact of policies through 2018 by calculating the relative difference between the counterfactual and status-quo scenarios. The contribution of an individual policy is based on programming SimSmoke changes through 2018 on the selected policy with other policies at the 1993 level. The effect of individual policies is measured relative to the summed effects of all individual policies.
We estimate the potential impact of future policies on tobacco prevalence by increasing all policies to their highest levels as shown in Table 1. The effects are presented relative to the status quo level in the same year for smoking and SLT prevalence and deaths averted.
For the effect of past and future policies, we also estimate lower and upper credible ranges of the policy effect sizes. Credible ranges are based on the potential minimum and maximum effect of the policy based on reported effects from the literature review. 62 For cigarettes, they are modeling as +/-50% of the original policy estimated effect, except +/-25% for taxes. For SLT estimates, we use +/-75% and +/-50% for taxes.

Model Validation
Validation comparing SimSmoke estimates to the TUS-CPS are in Table 2. For the total adult population, SimSmoke predicts a decrease in exclusive male (female) cigarette prevalence from 34.2% (29.5%) in 1993 to 21 For male dual users, SimSmoke predicts a decrease in prevalence from 2.1% in 1993 to 1.6% in 2015 (22% relative reduction), while TUS-CPS show a decrease from 2.0% in 1993 to 1.1% in 2015 (45% relative reduction), but dual use increases between 2007 (0.9%) and 2010 (1.3%). Dual user predictions for all years fell within the 95% confidence intervals of the reported data.
SimSmoke predicts male exclusive SLT prevalence to continuously decline by 23% from 7.1% to 5.4% from 1993 to 2015, while TUS-CPS reports a reduction from 6.8% to 3.5% from 1993 to 2007, but a 35% increase from 3.5% to 4.7% between 2007 and 2015. Validation by age and gender also perform relatively well (see Supplementary Report 17 ). Abbreviations: SLT, smokeless tobacco; NRT, nicotine replacement therapy; PT, pharmacotherapy. * Unless otherwise indicated, the effects are on prevalence in the first year, and on initiation and cessation during the years that the policy is in effect. The effect sizes are based on articles referenced in the text.  In a validation against the BRFSS (see Table 3c of Supplementary Report 17 ), male (female) smokers from the BRFSS shows a 20.6% (24.3%) relative reduction between 1996 and 2017 compared to a 38.2% (31.0%) relative reduction predicted by SimSmoke. SimSmoke predictions fell within the BRFSS 95% CIs before 2011and lower than the intervals since 2011 for both genders. SLT use from the BRFSS shows a 7.1% (33.3%) relative increase between 2011 and 2016 compared to a 3.4% (13.0%) relative reduction predicted by SimSmoke, with all the SimSmoke predictions below the BRFSS 95% CI lower bounds. The Effect of Policies Implemented Through 2018

Policy Description Cigarette Effect Size* SLT Effect Size
Results comparing the status-quo scenario (ie, policies implemented between 1993 and 2018) to the counterfactual scenario (ie, policies set at 1993 level) are in Table 3 for  smoking and SLT prevalence and Table 4 for averted tobaccoattributable deaths.
Compared to the counterfactual, SimSmoke projects that exclusive cigarette prevalence was reduced in relative terms by 23.7% (with credible range of 16.7%, 30.4%) for males and 23.0% (16.2%, 29.4%) for females by 2018. The 2018 reduction for male dual users was 16.4% (11.6%, 21.2%) and for exclusive male SLT users was 4.9% (1.2%, 8.6%). By 2060, the model projects a relative reduction of 32.7% Range Health warning alone Total Youth access alone  By individual policy, increased prices (largely due to cigarette taxes) reduce exclusive cigarette use rates by an average of 17.5% for males and females by 2018 and avert 65,826 total deaths by 2060. Smoke-free air laws yield a 2.5% reduction in exclusive cigarette use by 2018 and avert 9598 deaths by 2060. Cessation treatment policies show a relative reduction of 2.2% in 2018, with 12 464 averted deaths by 2060. Youth access restrictions show a 2.6% reduction in exclusive cigarette use and avert 6530 deaths by 2060.
SimSmoke estimated the highest reduction in 2018 exclusive smokers' prevalence for taxation (68%), followed by youth access policy (11%), smoke-free air laws (10%), cessation treatments (9%), and tobacco control campaigns (2%). For male dual prevalence, impacts were higher for taxation (76%), cessation treatment (10%), and youth access enforcement (9%) but generally lower for other policies. For SLT use, the order is taxes (44%), cessation programs (24%), health warnings (24%), and youth access policy (5%). Tables 5 and 6 show the projected smoking prevalence and exclusive SLT use from strengthening tobacco control policies. Increasing the excise tax per pack of cigarettes by $2.00 and SLT by $2.00 reduces the average male and female smoking prevalence by 9.6%, reduces male and female SLT use by 10.3% and 9.6% initially increasing to 18.5% and 12.7% by 2060, averting 11,072 deaths by 2060. An SLT tax increase of $2.00 yields a 10.3% (5.2%, 15.2%) and a 9.6% (4.8%, 14.1%) immediate relative reduction for male and female exclusive SLT users, increasing to 18.5% and 12.7% by 2060 and averting 615 deaths. Increasing smoke-free air laws to their highest level, as shown in Table 1, reduces
With the above policies in combination, SimSmoke predicts that by 2060, male and female prevalence are reduced by 41%, male and female exclusive SLT use are reduced by 33% and 25% respectively. Premature deaths are reduced by 61 626 for smoking-attributable deaths and 1276 from exclusive SLTattributable deaths.

Discussion
Our projection of trends in cigarette and dual use from Kentucky SimSmoke are relatively well-supported by two large scale nationally representative surveys, TUS-CPS and BRFSS. However, the analysis of SLT does not fit the data as well. The estimated effect sizes of cigarette-oriented policies on SLT use in SimSmoke reflects studies of use patterns prior to 2007. With major changes in tobacco control policies between 1993 and 2007, these findings suggest that cigarette-oriented policies may have additional impacts on SLT use. Since reductions in cigarette use were associated with reductions in SLT use, these patterns suggest a complementary relationship between the use of these two products. However, since dual use did not increase, these patterns may reflect changes in norms towards single tobacco use rather the dual use of the products.
Although not predicted by SimSmoke, SLT use increased in recent years according to the TUS-CPS and BRFSS estimates and other recent studies. [65][66][67][68][69][70] These increases may reflect the growth in SLT marketing by cigarette manufacturers. 71 The increase in unregulated sales of flavored SLT products also could contribute to smoking initiation at younger ages. 72,73 74 During that period, cigarette companies increased the marketing of SLT, lowered prices and, in particular, marketed SLT as a product that could be used in places (eg, worksites) where smoking was not allowed. 75,76 Similar and even stronger patterns of increased SLT use were observed in Minnesota, 77 which implemented a statewide smoke-free air ban. The increase in unregulated sales of flavored SLT products also could have contributed to SLT initiation at younger ages, 72,73 as indicated by the particularly large increase in exclusive and dual SLT use of young adults (especially ages [18][19][20][21][22][23][24]. These patterns suggest that SLT use has acted as a substitute for cigarettes in recent years. The findings in this study have implications for the recent growth in e-cigarette and heated tobacco product use. From 1993-2002, SLT use fell during a period of active cigarette-oriented policies. This finding suggests that more restrictive cigarette-oriented policies may also serve in a complementary fashion to reduce e-cigarette use. At the same time, e-cigarettes may be a better substitute for cigarettes than SLTs, since they appear to more efficiently deliver nicotine and provide sensorimotor effects closer to cigarettes than other non-combustible tobacco products. 15 In that case, stronger policies may cause increased substitution of e-cigarettes for cigarettes, especially if e-cigarette policies are weak. In recent years (since 2006), cigarette and SLT companies have been more active in promoting new forms of oral tobacco products. 10,11,78 The recent growth in SLT use by young adults also suggests that cigarette companies may actively promote e-cigarettes to be used with cigarettes (eg, dual use), again suggesting a complementary relationship. In particular, Altria recently purchased oral tobacco firm "On" 10 and introduced heated tobacco products (ie, IQOS). With this increasing diversification of products, the landscape for nicotine delivery products is likely to become increasingly complex. 79,80 It will be important to monitor the behavior of cigarette companies and how they promote products that may serve as an alternative to cigarettes.
While the current role of SLT use vis-à-vis e-cigarettes needs further policy consideration, there is a potential for further reductions in tobacco usage in Kentucky. For example, a SimSmoke model for Minnesota, 62 one of the leading tobacco control policy states, estimates that policies implemented since 1993 led to a reduction in smoking prevalence of 35% and SLT use by 23% by 2018. By that same year, SimSmoke results for Kentucky estimate a smoking prevalence reduction of 24% and 5% for male SLT users. The variations are mostly due to the different levels of policies implemented. By 2018, Minnesota had a cigarette excise tax of $3.04, smoke freeair laws in workplaces, schools, and restaurants and bars, an adequate level of cessation coverage, and spent 32% of the CDC's recommended budget for tobacco control. In contrast, Kentucky had an excise tax of $0.60 and a third of the state with smoke-free air laws in schools, government buildings, workplaces, and restaurants, and the state spent 10% of the CDC recommended level for tobacco control expenditures. 5 Since early 2018, Kentucky has started to move towards more aggressive regulations by implementing a CDCrecommended comprehensive cessation program and increasing the cigarette tax in July 2018. An increase in the Kentucky cigarette excise tax of $0.50 to $1.10 81 could achieve a reduction in smoking prevalence of 2.8% in the first few years. If the cigarette tax is increased by an additional $0.90 to $2.00, Kentucky would then be above the mean US state taxation of $1.82, which would result in further reducing smoking prevalence by 5.1% by 2060.
Like all models, our results are only as strong as the assumptions and underlying data. We assumed that projections of cigarette use are based on initiation and cessation rates from 1993, but subject to policy changes over time. Thus, the initiation and cessation rate estimated using smoking prevalence data from 1993, and the policy levels and effect sizes play an important role. Also, we do not explicitly incorporate the role of industry, which may have had a major impact when the major SLT producing firms were acquired by cigarette manufacturers. 74 In addition, cessation and relapse data were not available for SLT use. Also, we treated SLT as a homogeneous category in terms of risks and an ability to substitute for cigarettes, although new forms, such as snus and other oral tobacco products, have come onto the market. 10,11 Another limitation of the model is that it includes only cigarettes and SLT use and does not incorporate the use of other nicotine delivery products, including cigars, water pipes, heated tobacco products, and e-cigarettes. Each may act as substitute or complement to the use of cigarettes and SLT. While use of these products may be a relatively minor contributor to overall tobacco-related harms, policies should be directed at all of these products, particularly small cigars. 82 Finally, we assume that cigarette-oriented and SLT-oriented policies have independent effects. As suggested above, cigarette-oriented policies may have indirect impacts on SLT use through social norms. The literature on the interrelationship between the effects of different types of policies is sparse. 42,62 While some smokeless demand studies incorporate cigarette prices and some cigarette demand studies incorporate smokeless prices, [40][41][42]64 the studies obtain mixed results regarding whether the two products are substitutes or complements. 42 In general, greater information is needed on the inter-related effects of policies targeting different tobacco products.

Conclusion
While the landscape for nicotine delivery products has dramatically changed in the last 10 years, some lessons can be gleaned from our results. First, with cigarettes still the dominant form of nicotine delivery, cigarette-oriented policies appear to be an effective means of reducing the use of nicotine delivery products. However, with SLT use increasing in recent years, policies directed at SLT use may also play a role. With cigarette manufacturers having acquired major SLT firms, it is important to monitor the role of the cigarette industry. With strong incentives to protect the high profits from cigarettes, cigarette firms can be expected to encourage dual use rather than switching to other tobacco products. Well-targeted policies and regulations, such as cigarette and SLT tax increases and media campaigns, will be needed to achieve reductions in SLT use.

Ethical issues
None, the study uses publicly available data.