At a time that the FDA is deciding whether to authorize continued sale of Juul e-cigs a group of Juul scientists have published “Modeling the population health impact of E.N.D.S in the U.S.” that concludes that “the availability of ENDS [electronic nicotine delivery systems, which includes e-cigs] is projected to reduce smoking and prevent 2.5 million premature deaths by 2100.”
Like other similar efforts to estimate the population health effect of e-cigs, Juul draws this conclusion based on a mathematical model of changes in e-cig and smoking behavior over time that considers people starting and stopping smoking and e-cigs, including switching between products and becoming a dual user (both cigs and e-cigs at the same time). The overall structure of the model is reasonable and similar to models that others have proposed.
The important thing in such efforts, however, is not usually the structure of the model but rather the numbers that are in the model to quantify changes between using different products and the risk of different behaviors (smoking, e-cigs, dual use, quitting and switching between various products).
It is in this area that the Juul model is seriously flawed.
First, and most notably, the model is based on the discredited claim that e-cigs are only 5% as dangerous as cigarettes. Virtually any model that assumes e-cigs are nearly harmless will show population benefits.
The authors do present a sensitivity analysis that allows this risk to be as high as 40% as dangerous as cigarettes, but the top line conclusions are based on the 5% number. In addition, as we learn more about the cardiovascular and pulmonary dangers of e-cigs, this 40% number is almost certainly too low.
The authors also assume that dual use is no more dangerous than smoking despite the consistent evidence that dual use is more dangerous than smoking.
They also assume that e-cigarettes help smokers quit smoking, despite the fact that the population-based studies show that e-cigarettes as consumer products do not increase cigarette smoking cessation.
The presentation of e-cigarette recruitment of youth to nicotine addiction is hard to follow, but it does not seem to adequately account for the well-documented gateway effect to cigarette smoking. Rather than relying on the high quality meta-analyses of this effect, Juul does their own estimate based on the FDA/NIH PATH dataset. Reading how they did this illustrates the convoluted approach that Juul takes:
For youth use, transition rates were derived from PATH Waves 3 to 4, which coincided with a significant increase in ENDS use among youth. However, because the PATH-derived ENDS initiation rates were much lower than what was needed to reproduce ENDS use prevalence as re-ported in NYTS, they were scaled up significantly to calibrate youth prevalence of ENDS use against NYTS data from 2011–2019, while scaling down by half as much the subsequent transitions to not artificially inflate youth smoking rates. Analyses of PATH were also used to model the share of past 30-day youth ENDS users that go on to become established adult ENDS users once they turn 18. [citations dropped]
It is not clear why Juul needed to follow this convoluted procedure when there are direct estimates of both e-cig use and of the gateway effect.
The authors say that the important parameters quantifying how adult behavior changes (known as “transition rates” or “transition probabilities”) come from the PATH study and Juul’s ADJUUST study, but do not provide details on how they actually came up with the numbers.
One thing I have learned when assessing modeling efforts like this is to examine the reference list. Looking at Juul’s references shows heavy reliance on older literature (before many e-cig risks were documented) with a strong bias toward papers written by investigators and groups sympathetic to e-cigarettes while ignore the more recent critical studies. For a model like this to be meaningful, it needs to be based on all the literature.
Given all these problems I was surprised that they managed to get the paper published in a peer reviewed journal until I realized that the entire May 2021 edition of AJHB ( V. 45:3 ) is a Special Issue on Juul. The authors of all the papers are primarily either employees of JUUL Labs itself, part of the ENDS industry and 35% owned by Altria Group, or of Pinney Associates, which is “working exclusively with JUUL Labs, Inc. to advance relative risk-based regulation of nicotine and tobacco products.” Such special issues are usually paid for by the sponsoring agency.
Neither the FDA nor anyone else should rely on this paper when assessing the population health effect of e-cigarettes, including Juul.
Here is the abstract:
Objectives: Our objective was to improve understanding of the population health impact of electronic nicotine delivery systems (ENDS) availability in the US via computational modeling. Methods: We present an agent-based population health model (PHM) that simulates annual smoking, ENDS use, and associated mortality for individual agents representing the US population, both adults and youth, between 2000 and 2100. Model transitions were derived from key population surveys and a large longitudinal study of JUUL purchasers. The mortality impact of ENDS is modeled as excess risk relative to smoking. Outcomes are compared between a cigarettes-only Base Case and a Modified Case where ENDS are introduced in 2010. Model validation demonstrates that the PHM simulates population-level behavior and outcomes realistically. Results: The availability of ENDS in the US is projected to reduce smoking and prevent 2.5 million premature deaths by 2100 in the Modified Case. Sensitivity analyses show that a significant population health benefit occurs under all plausible scenarios. Conclusions: Our results suggest the availability of ENDS is likely to result in a significant health benefit to the US population as a whole, after accounting for both beneficial and harmful uses.
The full citation is: Wissmann R, Zhan C, D’Amica K, Prakash S, Xu Y. Modeling the Population Health Impact of ENDS in the U.S. Am J Health Behav. 2021 May 1;45(3):588-610. doi: 10.5993/AJHB.45.3.12. PMID: 33894802.. It is available here.