The public health impact of NVP use among smokers and non-smokers is estimated by comparing two scenarios: a) the No-NVP Scenario which projects future cigarette use and associated mortality outcomes for each birth cohort in the absence of NVPs, and b) the NVP Scenario which incorporates NVP use patterns into each cohort’s cigarette use trajectory. To simplify the model and ensure that health outcomes reflect regular (i.e. stable) use patterns over time (Jha et al. 2013), we focus on the regular (rather than experimental) use of NVPs and cigarettes and the transitions between those uses.
The No-NVP Scenario
The SAVM begins with separate cohorts of males and females by individual age. Within each cohort, the population evolves with age. The No-NVP Scenario projects the prevalence of current and former smokers over time using age- and gender-specific initiation and cessation rates for each cohort that were previously developed using an age-period-cohort statistical smoking model (Holford et al. 2014a, Holford, Levy and Meza 2016, Jeon et al. 2018, Tam et al. 2018). The initiation and cessation rates are projected forward based on data from the US National Health Interview Survey (NHIS) through the year 2013, before NVPs were in more widespread use (Levy et al. 2019c) and thus reflects smoking patterns in the absence of NVPs. The transitions are shown in Figure 1.a.
Individuals are born as never users, which in the context of the No-NVP model is just never smokers. For a given cohort, never users at age a and year t includes surviving never users in the previous year (age a-1, year t-1) who did not initiate smoking, where surviving never users are modeled as never users at age a-1 in year t-1 multiplied by (1- mortality rate of never usersa-1,t-1).Smoking initiation among never users can occur through age 40, and reflects the transition to becoming an established current smoker, defined as having smoked at least 100 cigarettes during one’s lifetime and currently smoking every day or some days. Smokers at age a and year t include surviving never users at age a-1 and year t-1 who initiated smoking and surviving smokers (using current smoker mortality rates) who did not quit since the previous year (age a-1, year t-1). Smoking cessation among current smokers reflects the permanent transition to becoming a former smoker and requires that the smoker quit at least two year; this avoids the need to model relapse and reflects the underlying inputs used for this and other tobacco models (Holford et al. 2014b, Holford, Levy and Meza 2016, Tam et al. 2018). Former smokers include surviving former smokers (based on age- and former smoker-specific mortality rates) and surviving smokers who quit smoking for at least two years.
The NVP Scenario
Starting from the same initial current, former and never smoking prevalence as in the No-NVP Scenario, the NVP Scenario is expanded to include current and former NVP use. The transitions are shown in Figure 1.b.
Because SAVM focuses on measures of regular use, the NVP Scenario does not explicitly model individual youth transitions directly from short-term vaping to regular smoking. Any such transitions are however incorporated indirectly through a smoking initiation multiplier reflecting the net effect of vaping on smoking initiation rates. Similarly, the NVP Scenario does not explicitly model short-term NVP use leading to quitting both cigarettes and NVPs, but is indirectly incorporated through a smoking cessation multiplier reflecting the net effect of vaping on smoking cessation rates. In addition, the NVP Scenario does not distinguish dual cigarette and NVP users from exclusive smokers, because dual users have generally been found to either remain dual users or soon transition to either exclusive cigarette or exclusive NVP use (Azagba, Shan and Latham 2019, Borland et al. 2019, Coleman et al. 2018, Robertson et al. 2019, Stanton et al. 2020, Taylor et al. 2020) and the health risks of dual users are similar to those of exclusive smokers (Goniewicz et al. 2017, Shahab et al. 2017a, Shahab et al. 2017b). However, transitions from dual use to exclusive NVP use or neither cigarette nor NVP use are reflected in the switching and smoking cessation parameters.
Never users include surviving never users from the previous age and year who do not initiate into regular smoking or NVP use. The model directly relates the NVP Scenario to the No-NVP Scenario through separate linear multipliers applied to the smoking initiation rate in the No-NVP Scenario. Using a linear multiplier assures that NVP and smoking initiation follows the same age patterns for smoking initiation as in the No-NVP case, e.g., smoking initiation mostly occurs before age 21 and is minimal after age 30. NVP initiation is similarly tied to smoking initiation in the No-NVP Scenario, whereby a multiplier value less than (or greater than) 100% implies that NVP initiation rates are less than (or greater than) smoking initiation rates in the No-NVP Scenario.
In the NVP Scenario, smokers become former smokers in one of two ways: 1) they may switch to regular exclusive NVP use, or 2) they may quit both regular smoking and regular NVP use (e.g., dual users would quit both) and thereby become former smokers. In the latter case, the smoker may temporarily use NVPs but quit both smoking and vaping. The overall reduction in smokers in the NVP Scenario is the reduction in smokers from switching and complete cessation, and this sum may be less (e.g. harm reduction) or more (harm increasing ) than the number of smokers in the No-NVP Scenario. Smokers age a in year t include surviving smokers at age a-1 and year t-1 who did not switch to exclusive vaping and who did not quit smoking, and surviving never users at age a-1 and year t-1 who initiated smoking.
Similar to the initiation process, smoking and vaping cessation are modelled as separate linear multipliers of smoking cessation in the No-NVP Scenario, so that smoking cessation follows the same age pattern in the No-NVP Scenario and tends to increase with age. Since studies to date do not indicate different age patterns, the smoking and vaping initiation and cessation multipliers are simply modeled as independent of age. In addition, these multipliers are assumed to remain constant over time, so that smoking and vaping initiation and cessation in the NVP Scenario follow the same temporal patterns as in the No-NVP Scenario.
Former smokers who either switch to or quit smoking before age 35 are distinguished from those who quit or switch at age 35 or above. Those who quit smoking (by switching to NVPs or cessation) before age 35 are classified the same as never smokers who vape rather than as former smokers, because mortality risks of smoking are minimal when quitting before age 35 (Jha et al. 2013). Those who quit at age 35 or above maintain the former smoker status and its resulting mortality risks. Thus former smokers include surviving former smokers, surviving smokers (age> 35) who quit smoking and do not vape, surviving smokers who switched to vaping (age> 35) and surviving former smokers who at one point regularly vaped but quit vaping.
Exclusive NVP users include surviving never users who initiate vaping, surviving vapers who do not quit, and those switching to vaping from surviving current smokers before age 35. Similar to smokers, those who quit smoking before age 35 are treated the same as never smokers who vape rather than former smokers. Since former smokers using NVPs may quit vaping, former smokers using NVPs includes surviving former smokers using NVPs who do not quit vaping and smokers who switch to vaping (after age 35).
A detailed discussion of the parameters and equations used in both scenarios can be found in the Supplement 1.
Public Health Outcomes
The SAVM considers two public health outcomes: 1) Smoking- and vaping-attributable deaths, and 2) smoking- and vaping-attributable life-years lost (LYLs). Both are based on the excess risks of smoking or vaping and the number of current and former smokers and vapers.
In the No-NVP Scenario, smoking-attributable deaths (SADs) by age and gender for current smokers are calculated by applying the excess risks of current smokers relative to never smokers (current smoker mortality ratea,t – never smoker mortality ratea.t) to the smoking population and for former smokers are calculated by applying the excess risks of former smokers relative to never smokers (former smoker mortality ratea,t – never smoker mortality ratea.t) to the former smoker population. SADs for current and former smokers are summed over all ages in a particular year to obtain total SADs in that year. LYLs are estimated as the number of premature deaths multiplied by the remaining life expectancy of a never smoker at age a in year t. The sum over all ages in a particular year obtains the LYLs in that year.
In the NVP Scenario, current and former smoker SADs are calculated in the same manner in the No-NVP Scenario as above except using their respective current and former smoker population sizes from the NVP scenario. For current and former exclusive NVP users, attributable deaths are a product of the number of NVP users and the excess risks of smoking adjusted by NVP relative risk, denoted by RiskNVP, i.e., RiskNVP*(current smoker mortality ratea,t – never smoker mortality ratea.t) for current users and RiskNVP*(former smoker mortality ratea,t – never smoker mortality ratea.t) for former users. As a special case in the NVP Scenario, the mortality rate of NVP users who previously smoked is determined by the mortality rate of former smokers plus the portion of excess NVP risk of current relative to former smokers. Smoking and vaping-attributable deaths (SVADs) for former smokers who currently use NVPs is measured as a product of the number of former smokers who currently use NVPs and the excess mortality risk of these users to never smokers, calculated as (former smoker using NVPs mortality ratea,t – never smoker mortality ratea.t). Total attributable deaths are calculated by summing SVADs over all ages in each year. LYLs in the NVP Scenario at each age are calculated as the product of SVADs and life-years remaining of never smokers. Summing over all ages in a particular year obtains the total LYLs in that year.
The public health impact of NVP use each year is evaluated as the difference in attributable deaths between the No-NVP and NVP Scenarios, and similarly for LYLs.
Data and Parameter Specification
In SAVM, the analysis of current, former and never smoking and vaping is in terms of their prevalence. These prevalence rates are translated into population numbers using actual and projected US population size estimates (Centers for Disease Control and Prevention 2019a). While the model itself does not explicitly incorporate projected births, mortality and immigration, the CDC projections incorporate projected births, mortality and immigration.
A detailed description of model parameters is provided in Table 1.
The No-NVP Scenario
The initial level of current, former and never smoking prevalence, the smoking initiation and cessation rates used to project these future prevalence rates, mortality rates by smoking status, and the life expectancy of never smokers used in the No-NVP model were previously developed (Holford et al. 2014a, Holford et al. 2014b, Holford, Levy and Meza 2016, Rosenberg et al. 2012) available on the National Cancer Institute: Cancer Intervention and Surveillance Modeling Network (CISNET) website (CISNET 2016). These measures apply data only through 2013 to incorporate trends prior to the time when NVPs became more widely used.
The NVP Scenario
The NVP Scenario requires six input parameters: NVP mortality risks, NVP switching rate, smoking initiation, vaping initiation, smoking cessation, and vaping cessation.
The NVP relative risk multiplier, RiskNVP, represents the relative risk of death associated with current NVP use as a percentage of the excess mortality risk experienced by current or former smokers as defined above. NVP relative mortality risk is designated at 5% of cigarette excess risks based on a multi-criteria decision analysis (Nutt et al. 2016) and a Public Health England review (McNeill et al. March 2020). Since others have suggested higher risks (de Groot et al. 2018, Nelluri et al. 2016, Tegin et al. 2018, Unger and Unger 2018), an NVP relative risk multiplier of 40% is also considered.
The NVP switching rate is the annual rate at which current smokers switch from smoking to NVP use, leading to a direct reduction in smoking prevalence. Baseline male (female) NVP yearly switching rates are based on weighted data by age group that we generated from the 2013/2014, 2014/2015, 2015/2016, and 2016/2017 Population Assessment of Tobacco and Health (PATH) surveys (U.S. Food and Drug Administration, National Institutes of Health and Abuse. 2019) averaged over years, as described in Supplement 2. These are: 8% (5%) age 10-17; 4.0% (2.5%) for ages 18-24, 2.5% (2.0%) for ages 25-34, 2.5% (1.6%) for age 35-44, 1.3% (1.4%) for ages 45-54, 1.2% (1.4%) for ages 55-64, and 0.6% (1.0%) for ages 65. We also consider switching rates that are 50% lower and 100% higher than the baseline estimates. These rates are initially assumed constant over time (i.e., 0% decay), but we also consider a 10% decay (i.e., annual rate of decline) to reflect the possibility of reduced innovation in NVPs over time and the tendency for those who are most amenable to switching to or quitting NVPs to have already switched, leaving a population less amenable to switching and quitting. However, because innovation in NVP design may increase their substitutability for cigarettes and potentially increase switching, we also consider an annual 5% increase rate for the first five years (2018-2022).
Since the smoking initiation in the NVP Scenario is measured relative to smoking initiation in the No-NVP Scenario, the smoking initiation multiplier is greater than 100% if NVP use increases net smoking initiation beyond the rate at which individuals would have otherwise initiated smoking in the absence of NVPs (i.e., gateway in > gateway out), and is less than 100% if those who would have initiated smoking tend to transition to exclusive NVP use instead of smoking (i.e., gateway out > gateway in). Based on the more rapid downward trend in US youth and young adult smoking as NVP use increased in recent years (Levy et al. 2019b), the age- and year-invariant baseline smoking initiation multiplier is initially set at 75%, i.e. a 25% net decrease in smoking initiation due to vaping. A range of 25% to 125% is also considered.
The NVP initiation rate multiplier reflects the youth and young adult initiation of NVP use relative to smoking initiation rates in the No-NVP Scenario. If less than 100%, NVP initiation rates are lower than smoking initiation rates in the No-NVP Scenario. i.e., fewer individuals become regular NVP users than who would have become smokers in the No-NVP Scenario. Reflecting the increased US youth and young adults regular NVP users in recent years (Cullen et al. 2019, Hammond et al. 2019, Levy, Yuan and Li 2017, Miech et al. 2019b, Miech et al. 2019c), the age- and year-invariant baseline NVP initiation multiplier is initially set at 50%. Sensitivity analysis is conducted over the range of 25% to 75%.
Since smoking cessation in the NVP Scenario is measured relative to smoking cessation in the No-NVP Scenario, the smoking cessation multiplier is greater than 100% if smoking cessation rates are higher in the NVP Scenario than in the No-NVP Scenario, e.g., if the availability of NVPs leads to increased smoking cessation. A parameter less than 100% implies that smokers are less likely to quit smoking in the NVP Scenario compared to the No-NVP Scenario, e.g., NVP use leads to dual use (continued smoking) rather than complete cessation among smokers. The baseline smoking cessation rate multiplier is initially set at 100%, with sensitivity analysis at 50% and 150%.
The NVP cessation multiplier is greater than 100% if NVP cessation rates are higher than smoking cessation rates in the No-NVP Scenario, e.g., NVPs are less addictive than cigarettes. The parameter is less than 100% if NVP cessation rates are lower than smoking cessation rates in the No-NVP Scenario. The baseline NVP cessation multiplier is set at 100%, with sensitivity analysis at 50% and 150%.
Model Validation and Analysis
The model estimates the NVP effects over time for the prevalence of current and former smokers, current and former NVP users, and former smokers using NVPs, and for smoking-attributable and NVP-attributable deaths and LYLs. The model was first validated over the years 2013 to 2018 by comparing model predictions of current smoking prevalence to current smoking prevalence rates from the NHIS (Centers for Disease Control and Prevention 2019b). We focus on relative reductions over the period 2013-2018, because levels of model predictions in the initial year 2013 differ from those in the NHIS.
We also validated NVP use. Although NVP use was already occurring in 2013, the model itself begins with no NVP use that year and is adjusted to reach the 2018 levels (Levy et al. 2019b). Consequently, we validated NVP use prevalence against NHIS estimates for 2018, the latest year for which data was available rather than rely on prior trend data. We defined NVP users as those who used NVPs at least 10 of the past 30 days to reflect more regular use (Amato, Boyle and Levy 2016, Levy et al. 2019c).
Upon validating the model, we consider how NVP use affects smoking prevalence, smoking-attributable and NVP-attributable deaths and LYLs. We conduct sensitivity analyses for the NVP transition and risk parameters over the plausible ranges specified above focusing on their impact on premature attributable deaths and LYLs.