Thread by @ArielleSelya

1/15: My new paper in Harm Reduction Journal simulates what the US situation might look like if we promote e-cigarettes as a harm reduction policy: rdcu.be/cg94K . Hint: #VapingSavesLives

2/15: In system dynamics modeling, it’s important to focus on a specific, measurable problem to solve. Here, it’s the combustible tobacco-related death burden (~480,000 lives/year).

3/15: We also need a specific goal for the tobacco-related death rate. Since it's unreasonable to expect zero tobacco use, I've set the goal to that of accidental deaths (1% of the population) – below that attributed to poor diet/physical inactivity or alcohol.

4/15: The first step is to (approximately) replicate smoking rates so far. In this type of modeling, only the shape of the curve (and not the exact values) matter. The model is able to replicate (slow) exponential decline in smoking, and exponential growth in e-cig use:

5/15: Next, we add a policy to try and achieve our goal! The policy has 2 parts: promoting e-cigs as a less harmful alternative (as is done in the UK, and to correct dangerous misperceptions that e-cigs are at least as dangerous as smoking), and removing existing flavor bans.

6/15: I model the policy as having 3 effects on smoking: 1) diversion from ever starting smoking; 2) reducing progression to established smoking among those who still initiate; and 3) increasing smoking cessation. All in all, the model looks roughly like this:

7/15: The model also conservatively includes an important #UnintendedConsequence: that e-cig use contributes to the total tobacco-related deaths (by PHE's estimate of e-cigs being 5% as harmful as smoking). A "gateway" effect is also modeled, for the sake of argument.

8/15: Under ideal policy roll-out (pink dotted line), we can achieve the target smoking prevalence (blue solid line) by ~2060, versus the status quo (red dot-dashed line) – a fraction of which accumulates into unacceptably high deaths under the status quo projections.

9/15: 2060 may seem like a long time to achieve the policy. This is because 2 of the policy's effects act on newer, incoming generations of smokers, and the 3rd effect is set to a conservatively low value. True numbers may be higher, especially from “accidental quitting.”

10/15: However, policy implementation never goes as planned. So, it's important to account for logistical challenges, including time delays, uncertain likelihood of approval (esp. with allowing flavors), and constrained budgets (for an informational campaign).

11/15: Not surprisingly, logistical barriers to policy implementation delay and weaken its effects, by up to 62%. However, these effects diminish over time.

12/15: Overall simulation results are similar, but exact values differ when using other parameter values (e.g. e-cigs' diversion effect away from smoking, quantified in my other paper onlinelibrary.wiley.com/doi/10.1111/ad…)

13/15: Disagree with my assumptions, in any direction? Run the model yourself, in a user-friendly web format: exchange.iseesystems.com/public/arielle… or download the full model in Stella Architect at: exchange.iseesystems.com/models/player/…

14/15: In summary, promoting e-cigarettes as a less harmful alternative to smoking could reduce tobacco-related deaths to the rate of deaths from accidental causes, by 2060, using current best estimates for e-cigs’ effects on smoking, and e-cigs’ relative harm.

15/15: Disclosures: I now work for PinneyAssociates which consults for JUUL, but this work was completed prior to starting at Pinney, in my NIH-funded academic position. In fact, this line of independent research led me to want to consult on tobacco harm reduction!