Electronic cigarettes and health outcomes: epidemiological and public health challenges

Electronic cigarettes (e-cigarettes) are battery-powered devices that heat and aerosolize an ‘e-liquid’. The aerosol is then inhaled by the user. Since their first introduction to broad global markets in 2006–7, use has become increasingly common, particularly among youth. The population and individual health impacts of e-cigarettes relate to patterns of use, their direct effects on health and their indirect effects, through impacts on smoking behaviour, as well as background disease risks and tobacco control. Based on the current worldwide evidence summarized in our recent systematic review, use of nicotine e-cigarettes increases the risks of addiction, poisoning, toxicity from inhalation (including seizures) and trauma and burns. Applying the National Academies of Science, Engineering and Medicine framework for rating the strength of conclusions based on evidence, the review found strong evidence that, among young non-smokers, uptake of conventional smoking is increased by an average of 3-fold in e-cigarette users versus non-users. Use of e-cigarettes can cause e-cigaretteor vaping-associated lung injury (EVALI), largely attributable to e-liquids containing tetrahydrocannabinol/ vitamin E acetate, although one in eight cases in the largest case series to date were reported to be related to use of nicotine e-cigarettes. The review found that the effects of e-cigarettes on most important clinical outcomes are not known, due to insufficient or absent evidence. This includes outcomes related to cancer, cardiovascular disease, respiratory conditions other than lung injury, mental health, development, reproductive health, sleep, neurological conditions other than seizures, and endocrine, olfactory, optical, allergic and haematological conditions. There was conclusive or substantial evidence that e-cigarettes can cause indoor air pollution, waste and fires. There was less direct evidence that e-cigarette use can adversely affect cardiovascular health markers, such as blood pressure and heart rate, lung function and adolescent brain development and function. The review also found limited evidence that freebase nicotine e-cigarettes used in the clinical setting are efficacious smoking cessation aids. Evidence was also limited that former smokers who use e-cigarettes are around twice as likely to take up smoking again than those not using e-cigarettes. Based on the systematic review findings, there is currently strong evidence that use of e-cigarettes by nonsmokers is harmful to health overall, with multiple health harms and no health benefits identified in this population. Ex-smokers would be likely to reduce e-cigarette-related health effects if they avoid ongoing e-cigarette use. Dual tobacco smoking and e-cigarette use is the commonest pattern of e-cigarette use in most settings where evidence is available. The direct health effects of combining smoking and e-cigarette use are unknown. Dual use may facilitate prolongation of smoking, which would be likely to increase risks. E-cigarette use in smokers may offset the pressure of many effective tobacco control measures,

as evidenced by common reported reasons for use including that, compared with tobacco cigarettes, e-cigarettes are cheaper, more socially acceptable, able to be consumed in settings where tobacco smoking is prohibited and considered less hazardous to health. 3,8 Most smokers who quit successfully do so without specific smoking cessation aids. 12,13 Therapeutic products must demonstrate an acceptable balance of benefits and risks to be approved by regulatory authorities [14][15][16] and a range of approved smoking cessation aids are available. 17 Internationally, at the time of writing, e-cigarettes are not approved as therapeutic goods. The United States Preventive Services Task Force concluded recently that 'the current evidence is insufficient to assess the balance of benefits and harms of electronic cigarettes (e-cigarettes) for tobacco cessation in adults, including pregnant persons' and 'recommends that clinicians direct patients who use tobacco to other tobacco cessation interventions with proven effectiveness and established safety'. 18 In England, e-cigarettes combined with behavioural support are recommended as an option for smoking cessation, particularly for smokers who have tried other methods without success, 19 and in Australia current guidelines list them as an option following use of approved products. 17 Smoking is extremely harmful and our systematic review findings indicate that the balance of probabilities may be that e-cigarettes benefit smokers who have tried other measures unsuccessfully and who use e-cigarettes to quit tobacco smoking completely and promptly. However, the ultimate balance of safety and efficacy of the use of e-cigarettes for smoking cessation remains unclear; most of their effects on health outcomes are unknown, a number of risks have been identified, evidence on their efficacy for smoking cessation is limited and most smokers using e-cigarettes continue to smoke. 18 It is important that evidence regarding e-cigarettes is appropriately contextualized and issues are considered in perspective. This includes noting that in many countries, >80-90% of the population are non-smokers and the main driver of decreasing prevalence of smoking is declining smoking initiation in youth, rather than smoking cessation. 20 As noted above, smokers seeking medical assistance to quit and using medications are in the minority. 12,13 For example, in Australia in 2019, 11% of the population aged 14 and over were current daily smokers ( Figure 1). 8 Among current smokers, recent data indicate 36.3% make at least one quit attempt in a given year and, of these, 33.2% report use of (approved) nicotine replacement therapy. 21 In Australia, this amounts to around 1.3% of the population aged 14 and over being a current daily smoker who makes a quit attempt using nicotine replacement therapy; an example of a group that could use e-cigarettes. In many countries, potential benefits to smokers-an important minority-are used to justify widespread exposure to e-cigarettes, with risks to young people and the much larger population who are non-smokers.

Epidemiological and methodological challenges
Evaluation of health impacts focuses primarily on evidence regarding the causal relationship of use of e-cigarettes to health outcomes. In terms of establishing the safety of an exposure, it is also important to consider the extent to which adverse effects can be excluded, which relates to the statistical limits around risk estimates. The context of use is also important. E-cigarette exposure in the general population is increasingly common, which means that moderate elevations in relative risk-around 20% to 30% (i.e. relative risks of 1.2 to 1.3) -would have important implications for public health. Hence, evidence is needed that is capable of both detecting and excluding such elevations in risk. It is therefore necessary to focus on study designs able to provide evidence regarding causality, with sufficient statistical power and quality. Alongside standard issues relating to reliably establishing and excluding exposureoutcome effects, there are some e-cigarette-specific challenges, outlined below.

Tobacco and e-cigarette industry interests and influence
A major underlying issue in establishing the health impacts of e-cigarettes is the massive ongoing and extremely well resourced efforts of the closely-related and overlapping tobacco and e-cigarette industries 22,23 to promote their products and further their interests. These are well documented in the World Health Organization taxonomy of tobacco industry tactics and include efforts to distort and undermine science, influence outcomes and interpretation of findings, influence political and community priorities and divide the tobacco control community. 24,25 The e-cigarette industry markets aggressively to children and adolescents, at the same time as claiming that its sole intention is to provide a 'safer alternative to tobacco' and to support smokers to quit. Industry funding of scientists and studies provides an important potential source of bias and should be carefully documented and considered when reviewing the evidence. 26,27 Diversity and uncertainty of e-cigarette exposure and rapid change The health effects of e-cigarettes relate chiefly to: (i) the chemicals emitted by the e-cigarette, which originate from the e-liquid, chemical reactions in the heating element and the device itself 28 ; (ii) the device; and (iii) aspects impacting on smoking behaviour. There may also be summative and/or synergistic effects of the combination of these exposures.
E-liquids generally contain chemical solvents propylene glycol and vegetable glycerine, and the vast bulk of e-cigarette use is of those delivering nicotine 29 ; these commonalities underpin investigation of health effects. Bearing this in mind, there are many thousands of e-cigarette devices and e-liquids, with >17 000 flavours available currently and hundreds being introduced to the market on a monthly basis. Hence, e-cigarette use in the population results in exposure to many thousands of different chemical combinations, and varying doses of these chemicals. 28 Certain health effects may relate to specific chemical constituents or device properties, and risk may therefore vary between e-cigarettes. Currently, a major division is between use of freebase nicotine and nicotine salt products. Nicotine salt preparations make high concentrations of nicotine more palatable by reducing throat irritation, and are the main form of nicotine used in disposable and pod e-cigarettes. 28 These e-cigarette types are small and easily concealed and are heavily marketed to children and adolescents, including through their design. They have been identified as key drivers of widespread youth use of e-cigarettes, including in Canada and the USA. 3,30 Many commonly used disposable nicotine salt products deliver high concentrations of nicotine, which would increase the risk of nicotine toxicity via inhalation, including seizures. Evidence also suggests that nicotine salt products confer higher risks of dependency than other types. 31 EVALI has largely-but not exclusively-been attributed to e-liquids containing tetrahydrocannabinol/vitamin E acetate. 'Popcorn lung' is bronchiolitis obliterans attributed to inhalation of diacetyl used as a flavouring in microwave popcorn; concerns have been raised about use of diacetyl in e-liquids. 32,33 Users of 'open' e-cigarettes fill a reservoir in the device with the e-liquid. Such e-liquids are either premixed or involve the mixing of the different components, often including the dilution of more concentrated liquid nicotine. Concentrated liquid nicotine is highly toxic and at-home dilution of high-concentration nicotine increases the risk of accidental and intentional poisoning, including in children. 3 It also increases the risk of e-liquid adulteration. The risk of injuries related to exploding batteries varies according to the construction of the device and the quality of the batteries within it.
It is often difficult to know with accuracy what the components of an e-liquid are, as labelling is variable and often absent or inaccurate. Many studies do not report on the types of e-cigarettes used by participants and/or many participants are unable to report on the specific type of ecigarette or combinations of e-cigarettes that they have used. There can also be issues generalizing between e-liquid types, and rapid change creates a 'moving target' for monitoring and research. For example although much contemporary use is of nicotine salt products, 34 most of the studies to date, including randomized trials of smoking cessation and observational studies of smoking uptake, are of e-cigarettes delivering freebase nicotine. A major related factor influencing e-cigarette exposure and risks is user behaviour: use confined to smoking cessation would lead to very different population risks compared with use primarily among youth who are not established smokers. 35 Limited data and the relatively recent introduction of e-cigarettes as a population exposure Population exposure to e-cigarette use has only become substantial since around 2010. Certain immediate or short-term outcomes are already discernible, particularly those where cause and effect are able to be ascertained at an individual level and where appropriate surveillance systems are in place. For example, the addictive effects of e-cigarettes are well established, as is their ability to cause poisoning, trauma, burns and toxicity through inhalation. The EVALI outbreak was detected and addressed rapidly. Effects on the uptake of smoking in young people have also been able to be assessed fairly quickly, with >25 studies available in our recent meta-analysis. 5 For some e-cigarette components and e-liquid constituents, toxicological and other data on their impacts is informative-for example, many effects of nicotine are well established.
It will be some time before it is possible to reliably ascertain many important effects of e-cigarettes on health, due to the time needed to conduct research and because some conditions take many years to develop. This, along with the other issues outlined in this article, means that evidence on major clinical conditions such as cancer, cardiovascular disease, mental illness, respiratory conditions other than EVALI and adverse reproductive outcomes is very limited.
The need to focus on high-quality evidence, including clinical endpoints rather than intermediate markers Evidence should focus on clinically meaningful health outcomes including disease endpoints such as invasive cancer, cardiovascular disease events (e.g. myocardial infarction and stroke), respiratory diseases (e.g. asthma, infections and chronic obstructive pulmonary disease) and addiction and dependency-related outcomes. In the absence of this type of evidence, it is tempting to focus on biomarkers, socalled 'intermediate' outcomes (such as arterial wall thickness) and pathophysiological parameters (e.g. heart rate, blood pressure). Whereas these may be informative, they are not reliable substitutes for clinical outcomes and there are many examples of the risks relating to relying on this type of evidence for making decisions on safety.

Choice of comparator
In epidemiological studies, a comparator has two major related but distinct functions: (i) as a reference group supporting reliable quantification of the effect of an exposure; and (ii) as a counterfactual in understanding the likely impact of an exposure. For e-cigarettes, the comparator needs to factor in cigarette smoking status, for reasons detailed further in the section below.
The most appropriate reference group for analyses quantifying the distinct health effects and safety of e-cigarettes is people who have neither used e-cigarettes nor smoked. 3 Inclusion of people who smoke in the reference group causes major problems quantifying direct e-cigarette effects (see below). Neither smoking nor using e-cigarettes is the most appropriate counterfactual to using e-cigarettes for the considerable majority of the population who are non-smokers in virtually all countries, particularly for young people (Figure 1).
Very few exposures are as harmful as smoking, and quitting tobacco smoking is highly beneficial to health, compared with the counterfactual of continuing to smoke. For people who currently smoke, multiple e-cigarette-related counterfactuals apply. For ongoing dual smokers/ecigarette users, there is the comparison with people who continue to smoke and who do not use e-cigarettes. Since it is unclear whether smokers who use e-cigarettes may prolong smoking, the counterfactual of quitting by means other than e-cigarettes may also apply. For ex-smokers who have used e-cigarettes to quit completely, there are counterfactuals of continuing to smoke and of quitting smoking by other means.
There are active efforts to shift attention to smoking as the comparator for any use of e-cigarettes-that is, use by smokers and non-smokers, and use for smoking cessation and for other purposes. This is a tactic of the tobacco and e-cigarette industry, typified by use of statements along the lines that e-cigarettes are 'safer than smoking' for the broad marketing of these products, including to nonsmokers. Comparisons with smoking are also implicit in the framing of e-cigarettes as a 'harm reduction' measure. The exceptional harm of tobacco smoking means that such comparisons are highly advantageous to industry, and cannot be used as evidence of safety in absolute terms. E-cigarette effects on disease risk in smokers remain important, but they must be recognized as distinct from total population impacts and impacts in non-smokers. In particular, comparisons of e-cigarettes with smoking should be avoided in relation to use in non-smokers (the comparison should be with breathing air), potential harm reduction in one population should not automatically be considered harm reduction at a population level and the World Health Organization is clear that use of alternative tobacco products-such as e-cigarettes-in non-smokers cannot be considered harm reduction. 36 Bias and confounding, particularly in relation to tobacco smoking Quantifying the health impacts of e-cigarettes requires differentiation of their likely effects from those of other factors that use is related to. Potential confounding factors include age, sex, socioeconomic status, ethnicity/race, adiposity, physical activity, reproductive factors, alcohol consumption, other drug use, environmental tobacco smoke and pre-existing physical and mental health conditions, as well as particular issues in relation to combustible tobacco smoking.
Smoking markedly increases the risk of a very wide range of health conditions. 37,38 Contemporary data from Australia demonstrate that lung cancer incidence in current smokers is 18 times that of never-smokers, with a 6-fold risk in former smokers. 39 Cardiovascular disease risk in current versus never-smokers is 2-3-fold 40 and the risk of dying of chronic obstructive pulmonary disease is more than 30-fold. 37 In addition, risk increases with increasing duration of smoking and there are large elevations in risk with increasing number of cigarettes smoked per day. For example, compared with never-smokers, the hazard ratio for lung cancer incidence is 9.22 [95% confidence interval (CI) 5.14-16.55] for current smokers of 1-5 cigarettes per day and 38.6 (95% CI 25.7-58.1) with >35 cigarettes per day. 39 Risks also vary according to age and background risk of disease and among ex-smokers, disease risk diminishes with time since quitting. 37,41 The tendency of current smokers to quit when they become ill (the 'sick quitter' effect) is well established and adds further complexity to quantifying smoking effects. 42 In the situation where an exposure is related to smoking behaviour and smoking has a strong relationship to the outcome, reliable quantification of exposure-outcome relationships independent of smoking is difficult if not impossible, if smokers are included in the analyses. Adjustment of analyses for smoking generally can only account for relatively broad categories of current, past and never smoking, with limited ability to account for fine categories of intensity and other smoking attributes, leading to major issues with residual confounding. 43 Hence the most-and often the only-reliable evidence comes from confining analyses to people who have never smoked. Examples of where this method is used include studies of the relationship of lung cancer to environmental tobacco smoke 44,45 and risk factors other than smoking. 46 E-cigarette use relates strongly to smoking. 5,47-49 E-cigarette use is more common in current or ex-smokers, compared with never-smokers, and concurrent (dual) use of both e-cigarettes and tobacco cigarettes is a frequent pattern of use in many places. 9,10,50 There are complex differences in tobacco smoking behaviour between dual users and exclusive tobacco smokers, including in tobacco smoking intensity and duration, as well as other factors. In addition, smokers who become ill may take up e-cigarette use, aiming to reduce or quit smoking (sometimes termed 'sick switching'). These issues are likely to impact on effect estimates when comparing ex-smokers who use e-cigarettes with those who do not; similar problems will occur when comparing dual users with exclusive cigarette smokers. If the reference group is neversmokers, then comparisons with e-cigarette users who are current or ex-smokers are particularly problematic.
Establishing the safety of e-cigarettes requires studies able to detect and exclude relative risk increases from e-cigarette exposure of 20-30%. However, this magnitude of elevation in disease risk is small in comparison with that resulting from relatively minor differences in the number of cigarettes smoked per day among smokers-particularly for diseases such as lung cancer and chronic obstructive pulmonary disease. Hence for many conditions, residual confounding with tobacco smoking is likely to overwhelm the ability to reliably quantify-and exclude-direct effects of e-cigarettes, even following adjustment. This means that never-smokers who use e-cigarettes and who do not go on to smoke, are the most reliable exposed population in which to quantify the direct health effects of e-cigarette use. 4,51 The fact that some never-smokers who initiate e-cigarette use ultimately start smoking further limits evidence about health outcomes attributable to e-cigarette use. 4,51 Effect modification/statistical interaction Exposure-outcome relative risks tend to differ in magnitude between subgroups which vary in absolute rates of disease. For example, cardiovascular disease mortality rates vary by age. Reductions in cardiovascular mortality are observed with blood pressure-lowering treatments in all age groups and the magnitude of this effect varies with age; greater relative risk reductions are observed in younger age groups and greater absolute risk reductions are observed in older age groups. 52 Rates of disease differ markedly between current, past and never-smokers. 37 Even if it was possible to reliably ascertain relative risks in populations that included smokers, the relative effect of e-cigarettes would be likely to differ between them. Moreover, the combined effects of smoking and e-cigarettes may differ from their separate effects. This highlights the importance of conducting stratified analyses, including quantifying the effects of e-cigarettes separately in people who currently smoke, ex-smokers and neversmokers, bearing in mind the issues with confounding and bias outlined above.

Statistical power
Broadly speaking, reliable quantification of an exposureoutcome relationship requires sufficient numbers of outcome events among the exposed and unexposed to estimate the parameters of interest with required precision, taking account of matters such as measurement error, confounding and bias. All of the issues noted earlier in the paper have a bearing on statistical power.
In addition, whereas certain important health outcomes, such as addiction, mental health issues and asthma occur across the lifespan, most of the major clinical disease outcomes of interest in examining the effects of e-cigarettes-such as cancer, cardiovascular disease and chronic obstructive pulmonary disease-largely occur at older ages. Use of e-cigarettes at older ages (e.g. those aged >40 years) is almost always among people who currently smoke or have smoked in the past; there is very little use among older never-smokers. 8 E-cigarette use among neversmokers and people who are not established smokers is a common pattern in youth and, since smoking habits are generally not established until people are in their mid-20s, use below this age is chiefly not aimed at smoking cessation.
Hence, at the age where the vast majority of disease events occur, use of e-cigarettes is almost exclusively in smokers, making it very difficult to separate the effects of e-cigarettes from those of variations in smoking behaviour (see above). At ages where e-cigarette use among neversmokers is more common, chronic disease events-apart from those mentioned above-are generally rare. For example, in a major large cohort study of e-cigarettes and respiratory outcomes, 99.4% of e-cigarette users were either current or former smokers. 43 Even in large studies, the stratification necessary to be assured of effects independent of smoking may leave limited power to detect effects in never-smokers, particularly as disease events will tend to occur disproportionately in people who smoke. 53

Implications
Despite the methodological and other issues identified, evidence regarding the health impacts of e-cigarettes is accumulating rapidly. The risks that have been identified are sufficient to conclude that use in non-smokers, particularly youth, is harmful to health. For most major health outcomes, the effects of e-cigarettes are uncertain.
A key ongoing challenge for epidemiology and public health is not only the obtaining of reliable evidence on ecigarettes, but also its interpretation and use as well as its misuse. Effective public health action requires meticulous consideration of evidence, the assessment of risks and benefits overall and to multiple population groups and effective measures to manage risk. Like aviation and surgery, it is not about 'taking a punt' on something that seems intuitively appealing or hoping that certain measures will work.
In particular, it is not about assuming that scaling up something that appears to work for an individual or smaller group would automatically be beneficial for the population as a whole. Mistakes in public health cost millions of lives and span generations. Public health also requires active opposition to forces-such as the tobacco industry-which seek to profit at the expense of the health and wellbeing of the community.
Public health action on e-cigarettes, including regulatory decision making, must manage their identified risks as well as uncertainty. The precautionary principle highlights the importance of avoiding exposure to an agent when there is uncertainty combined with scientifically plausible evidence of serious risks to human health, especially those affecting future generations. 54 It clarifies that the onus is on proving safety-including by the proponents of an intervention-before allowing widespread exposure, and emphasizes the need to take action to prevent harm, rather than solely reacting once harm is done. 55 Governments are important stakeholders and regulation of e-cigarettes varies widely internationally, with at least 32 countries banning them totally, 79 having regulations that do not include a total ban and 84 having no regulations in place, at the time of writing. 56 Overall, the evidence is supportive of efforts in many jurisdictions to avoid e-cigarette use in the general population, particularly in non-smokers and youth. It is also supportive of the World Health Organization position that non-pregnant smokers who switch completely and promptly to an appropriately regulated [preferably medically approved] e-cigarette may benefit, bearing in mind the lack of evidence on long-term impacts (brackets ours). 56 It is important that evidence regarding the health impacts of e-cigarettes is interpreted and applied independent of industry influence. Industry profits from maximizing sales and hence promotes widespread e-cigarette use, and research to understand and efforts to control e-cigarette use are hampered by its activities. Industry actively seeks to conflate evidence of potential efficacy for smoking cessation in some individuals, groups or settings, with the idea that such use is beneficial overall in smokers and that widespread e-cigarette use would improve public health. If the tobacco and e-cigarette industry would 'stay in its box' and truly only target smokers who want to use their products to quit, things would be very different. It is worth conducting a thought experiment to consider what this would look like: no marketing to children and adolescents; no flavours designed to appeal to children (such as bubble gum and candy floss); use primarily for smoking cessation; and submission of e-cigarette products for regulatory approval for smoking cessation.
The issues outlined in this article highlight the importance of investment in high-quality independent research to address outstanding and ongoing issues regarding e-cigarettes, including but not limited to: data to effectively monitor e-cigarette and tobacco product use, including tobacco smoking, particularly among youth; studies of health impacts, able to separate e-cigarette and smoking effects; evidence informing the balance of quality, safety and efficacy of e-cigarettes as aids for smoking cessation; research on countering tobacco and e-cigarette industry influence; and research on effective measures to manage identified risks related to e-cigarettes, including avoiding e-cigarette use in non-smokers, especially youth. Evidence on how to effectively control inappropriate marketing to children and youth, particularly via social media, is a priority. 22 It is also important to acknowledge uncertainty and to resist the temptation to draw inappropriately on substandard evidence, in circumstances where high-quality evidence is lacking. Research should make the most of policy heterogeneity in the approach to e-cigarettes, to inform effective regulatory responses to maximize the health of current and future generations, including through accelerated tobacco control.

Ethics approval
No ethical approval was required as this paper is based on publicly available data.

Data availability
This paper is based on publicly available data.
Author contributions E.B. conceptualized the paper. E.B. and A.Y. wrote the first draft. G.J. provided statistical expertise. All authors contributed important intellectual expertise and approved the final version.

Funding
This paper draws on an independent report commissioned by the Australian Department of Health. E.B. is supported by an Investigator Grant (reference 2017741) from the National Health and Medical Research Council of Australia.