Estimating the effect of differing assumptions on the population health impact of introducing a Reduced Risk Tobacco Product in the USA

Abstract

We use Population Health Impact Modelling to assess effects on tobacco prevalence and mortality of introducing a Reduced Risk Tobacco Product (RRP). Simulated samples start in 1990 with a US-representative smoking prevalence. Individual tobacco histories are updated annually until 2010 using estimated probabilities of switching between never/current/former smoking where the RRP is not introduced, with current users subdivided into cigarette/RRP/dual users where it is. RRP-related mortality reductions from lung cancer, IHD, stroke and COPD are derived from the histories and the assumed relative risks of the RRP. A basic analysis assumes a hypothetical RRP reduces effective dose 80% in users and 40% in dual users, with an uptake rate generating ∼10% RRP and ∼6% dual users among current users after 10 years. Sensitivity study changes in tobacco prevalence and mortality from varying effective doses, current smoking risks, quitting half-lives and rates of initiation, switching, re-initiation and cessation. They also study extreme situations (e.g. everyone using RRP), and investigate assumptions which might eliminate the RRP-related mortality reduction. The mortality reduction is proportional to the dose reduction, increasing rapidly with time of follow-up. Plausible increases in re-initiation or dual users' consumption, or decreased quitting by smokers would not eliminate the drop.

Modeling the Population Health Impact of Introducing a Modified Risk Tobacco Product into the U.S. Market

“…The closeness of the null scenario predictions to actual epidemiological and authoritative statistics from the U.S. population across the 20-year study period provides a solid basis for assessing the potential population benefit of a cMRTP [Lee et al 2017]. As shown in that paper, the introduction of a cMRTP would result in fewer SADs in all but the most unlikely situations.…”

“…Separately for the null and alternative scenarios, the PHIM then derives estimates of the relative risk (compared to never users) for lung cancer, COPD, ischemic heart disease, and stroke for every individual of each sex at each year of follow up. The estimation uses a negative exponential model that has been described elsewhere [Lee et al 2017]. It requires not only knowledge of each individual's tobacco use at each year but also estimates of the "effective dose" corresponding to the tobacco use pattern.…”

“…The population health impact is then derived by subtracting the numbers under the alternative scenario from those under the null scenario. As equal population size is assumed under both scenarios, an adjustment can be made for the decrease in death rates in the alternative scenario [Weitkunat et al 2015;Lee et al 2017]. However, the simulations presented in this paper do not use this adjustment, as it had a small effect and did not affect the overall picture.…”

“…The derivation of the smoking histories under the null validation scenario uses the set of transition probabilities described in Table 2 of our earlier publication [Lee et al 2017]. That publication demonstrates that there was a reasonably good fit between the smoking distributions generated by PHIM across all these ages in 1990 and 1995, though the estimates for the older age groups do not correspond as strongly with the International Smoking Statistics (ISS) data.…”

“…PMI has developed the PHIM to quantify, in the absence of epidemiological data, the effect that marketing a cMRTP may have on the health of the population as a whole [Weitkunat et al 2015; Lee et al 2017]. This activity is accompanied by an increasing number of publications on alternative modeling approaches for estimating the population health impact following changes in the nicotine and tobacco product landscape [Bachand and Sulsky 2013;Vugrin et al 2015; Hill and Camacho 2017; Poland and Teischinger 2017].…”