Outline of the approach
The approach involves three stages.
The first stage defines the population in the baseline year, 2000. A hypothetical population of 100,000 US individuals is subdivided by sex and age, and also into never, current and former cigarette smokers, with former smokers subdivided by years quit. The new product, ZYN, is assumed not to have been available before 2000.
In the second stage, the population is followed up annually until 2050. Every year the population is updated to take into account births, net immigrations, deaths and changes in nicotine product use, with the distribution of sex, age and cigarette smoking in the immigrant population assumed to be equal to that used for the initial population.
Follow-up is carried out in two scenarios: the “Base Case” where the new product, ZYN, is never introduced; and the “Modified Case” where it is introduced immediately after baseline. Note that the use of nicotine-containing products other than cigarettes or ZYN is not considered in either scenario.
Table 1 details the nicotine use groups considered. In the Base Case, which concerns only groups 1 (never smokers), 2 (current cigarette smokers) and 3 (former cigarette smokers), our application of the model allows individuals to initiate smoking (change from group 1 to 2) or to quit smoking (change from group 2 to 3), but not to re-initiate smoking (change from group 3 to 2). Transitions to never smokers (group 1) are not possible, while transitioning from never to former smokers (change from group 1 to 3) is assumed not to occur in a single period.
| || |
In the Modified Case nine groups are involved:
1-3: never ZYN by never, current and former cigarettes;
4-6: current ZYN by never, current and former cigarettes and
7-9: former ZYN by never, current and former cigarettes.
Twelve transitions between the groups are considered. Three relate to initiation (changes from group 1 to each of groups 2, 4 and 5), three to quitting (changes from group 2 to 8, 4 to 6 and 5 to 9) and six to switching (changes from group 2 to 5, 2 to 6, 4 to 5, 4 to 8, 5 to 6 and 5 to 8). Other possible transitions are not considered, for comparability with the Base Case.
Based on the initial population distribution, the transition probabilities between groups, and changes in the simulated population due to births, net immigrations and deaths, data are available at each year from 2000 to 2050 on the distribution of the population by age, sex and nicotine use for both the populations in the Base Case and the Modified Case.
Finally, in the third stage, the data on the distributions by age, sex and nicotine use are processed to estimate mortality associated with nicotine use from all causes, the difference between the two measuring the population health impact of introducing ZYN.
Additional details of the methods and data used are described below.
Population at baseline at year 2000
The distribution of age and sex for the US population in 2000 was derived from the Human Mortality Database , which in turn used the 2000-2010 intercensal datasets published by the US Census Bureau . 49.1% of the population was male and 50.9% female. The distribution by age group within sex is given in Additional File 1 Table A1.1.
The initial cigarette smoking distribution by sex and age group is as used by Vugrin et al. , based on data from the National Health Interview Survey (NHIS), and is given in Additional File 1 Table A1.2. The NHIS data only apply to age 18 or over. For ages below 18, the Cancer Intervention and Surveillance Modelling Network (CISNET) provide the current smoking data by age shown in Additional File Table A1.3. CISNET gives no information on former smoking, and the modelling assumes that there are no former smokers age 17 or less in the initial population.
Additional File 1 Table A1.4 gives, by sex and age group, the distribution of years quit in former smokers aged 18 or over, again as used by Vugrin et al. (2015)  based on NHIS data.
Net migration and births
At every year during follow-up, the population is scaled up to allow for the inclusion of immigrants and new births. The sex distribution in newborns and the sex, age and cigarette use distributions for the immigrants are both assumed to be the same as at baseline. The estimates of net migrations and births used were the US Census Bureau 2008 National Population Projections, and are shown in Additional File 2 Tables A2.1 (migrations) and A2.2 (births).
Estimation of histories of cigarette smoking for the Base Case
Additional File 3 Table A3.1 gives the smoking initiation and cessation rates by sex and age used in the Base Case. Using methodology similar to that of Vugrin et al. (2015) , these rates were derived from smoking histories for birth cohorts reconstructed from NHIS data. Initiation rates for ages 0-19 were derived from the 1980-1984 cohorts, for ages 20-24 from the 1975-1979 cohorts and so on. Initiation rates for individuals above age 30 are set to zero, as nearly all initiation occurs at younger ages . Derivation of cessation rates was similar, with rates for ages 0-18 from the 1980-84 cohorts, for ages 19-24 from the 1975-79 cohorts, and so on. CISNET rates were available up to age 84, older individuals being assigned the rate for age 84. As cessation rates in the CISNET data represent successful smoking cessation rates for at least two years, re-initiation rates are set to zero. Both the initiation and cessation rates are assumed to be time independent.
Estimation of histories of product use for the Modified Case
In the pre-approval market setting, transition probabilities in the Modified Case are necessarily based on assumptions that cannot be validated. The transition probabilities in the Modified Case were derived from those in the Base Case as follows:
Initiation. Overall initiation rates were set equal to that for initiating smoking in the Base Case, but split in the proportion 70:25:5 for initiation to exclusive cigarette smoking, exclusive ZYN use, and dual use.
Quitting. Each of the three quitting rates, from exclusive use of either product or of dual use were set equal to that for quitting smoking in the Base Case.
Re-initiation. All re-initiation rates were set to be zero.
Switching from current exclusive smoking to current exclusive ZYN use. Switching rates were taken as 0.18%, 0.24%, 0.36% and 0.54% for the age groups 10-14, 15-19, 20-24 and 25+ respectively.
Switching from current exclusive smoking to current dual use. Switching rates were taken as 0.04%, 0.05%, 0.07% and 0.11% for the age groups 10-14, 15-19, 20-24 and 25+ respectively.
Other switching rates. These were taken as 0.48% for age 10+. In the absence of other information, these switching rates were as used in a previous publication .
Estimating mortality attributable to cigarette smoking in the Base Case
US death rates for never smokers aged 0-34 in 2000, given in Additional File 4 Table A4.1, were extracted from the Human Mortality Database . As tobacco-attributable mortality is considered minimal before age 35 , the risk of death for current smokers was assumed to be equal to that for never smokers, an assumption also made by Vugrin et al. (2015) .
Death rates for never and current smokers for ages 35 and above are taken from Vugrin et al. (2015) , the never smoker death rates being estimated from NHIS Linked Mortality Files. Similarly to Vugrin et al. (2015) , the never smoker death rates are adjusted for changes over the period 2000-2050 using mortality scaling factors estimated using the Lee-Carter method , and are available in Additional File Table A4.2 (males) and Table A4.3 (females). Individuals aged 85 years or older were assumed to face the same risk of death, regardless of age, given the same sex and smoking history. Individuals were tracked until age 101, being then counted as dead. The never smoker death rates for ages 35 and above were then converted into annual probabilities of death by sex and age groups using standard demographic methods , and are shown in Additional File 4 Table A4.4.
Current smoker death rates were calculated by multiplying the never smoker death rates by the estimates of relative risk by age and sex given by Vugrin et al. (2015) , which were derived from NHIS Linked Mortality Files, and are shown in Additional File 4 Table A4.5.
Former smoker probabilities were estimated, following Hill and Camacho (2017) , by assuming that the decline in excess risk (compared to never smokers) follows a negative exponential function, with a half-life of 9.08 years. Thus a current smoker with an excess risk of E at the time of quitting, would have an excess risk of E/2 at 9.08 years after quitting, and of E/4 at 18.16 years after.
Estimating mortality attributable to cigarette smoking and ZYN use in the Modified Case
The risk of death for those who have never used cigarettes or ZYN was taken to be that of never smokers in the Base Case. The excess risk of exclusive current ZYN users was taken to be 3.5% of that of exclusive current smokers of the same sex and age, while the excess risk of current dual users was taken to be that of current smokers. The estimate of 3.5% was based on the average of estimates made of 5% for snus use and 2% for nicotine replacement products made by a panel of experts at the Independent Scientific Committee on Drugs . For current ZYN users who were former smokers, the excess risk was taken to be the maximum of that for current ZYN use and former smoking. For former ZYN users who were current smokers, the excess risk was taken to be that of current smokers. For former ZYN users who had never smoked, the excess risk for current ZYN users was scaled down using the same half-life as used for former smokers. For former users of both products, the excess risk was taken to be the maximum of that for former ZYN use and former smoking.
Note that, in the following text, the term “product-related deaths” is used generally to describe either deaths attributable to cigarette smoking in the Base Case or deaths attributable to both cigarettes and ZYN in the Modified Case.
Sensitivity analyses investigated the effect of varying the assumed values of each of a number of different parameters:
Excess risk factor for exclusive ZYN use. Instead of 3.5%, alternatives of 0%, 7% and 20% were tested.
Excess risk factor for dual use. Instead of taking the maximum of the individual excess risks from smoking and from ZYN use, alternatives of the mean and the sum were tested.
Excess risks for current smoking. Instead of using the excess risks implied by the relative risk values shown in Additional File 4 Table A4.5, the excess risks were multiplied either by 0.8 or by 1.2.
Half-life of the negative exponential function. Instead of using 9.08 years, values of 4.54 or 18.16 years were used,
Transition probabilities. For each of the quitting rates and the switching rates, alternatives of half or double the rate used in the Modified Case were used. For initiation, the Modified Case rates were constrained to add to the initiation rate used in the Base Case. One set of sensitivity analyses for initiation retained this constraint, so that if one of the rates were halved or doubled, the three rates were then multiplied by a scaling factor so that their sum remained the same. The other set of sensitivity analyses, which varied only the initiation rates for exclusive ZYN use and for dual use, did not keep this constraint, but kept the initiation rate to cigarettes unchanged, so allowing study of the possibility that use of ZYN might affect overall initiation rates.
Note that most of the sensitivity analyses only affect the Modified Case. However, the sensitivity analyses which vary the excess risks for current smoking, the half-life of the negative exponential function, or the quitting rate for cigarette smoking, affect both the Base and the Modified Case.
Each of the sensitivity analyses described above varies only one parameter at a time. A further analysis was carried out based on a Pessimistic Scenario in which 12 assumptions relating to the introduction of ZYN were varied simultaneously, each of which were found to reduce the advantage to ZYN when considered individually. Thus the excess risk for exclusive ZYN was taken to be 20% of that from smoking, and the excess risk for dual use was taken to be the sum of those for smoking and ZYN individually. Also the transition probabilities were halved for initiation rate for ZYN, quitting rates for ZYN and for dual use, and switching rates from smoking to ZYN use, from smoking to dual use, and from dual use to ZYN, while the transition probabilities were doubled for initiation rate for dual use, and switching rates from ZYN to smoking, from dual use to smoking, and from dual use to ZYN. The changes to the initiation rates described were made while keeping the overall initiation rate fixed.