Estimation of the Global Number of Nicotine Vapers in 2020

The combustion of tobacco is the main cause of tobacco-related morbidity and mortality. Nicotine vaping products (NVP) are potentially disruptive innovations with considerable potential for population health. A key question is whether NVP are replacing tobacco cigarettes, which requires mapping uptake of NVP. Collecting information on nicotine use is dicult for many countries due to cost. The objective of this study was to derive a global estimate of vaping. Since we information To estimate the prevalence of vaping in countries lacking we used the of the same region and Based on surveys, we calculated the average prevalence of vaping for each WHO region, World Bank income classication group and legal status of For each of these groups the average prevalence of vaping was calculated. These values were used as substitutes for the prevalence gures in the countries with absent data. The number of vapers was calculated by taking as the denominator the adult population.


Introduction
Nicotine is one of the world's most popular drugs, alongside alcohol and caffeine. Most users, around 68%, consume nicotine by smoking tobacco cigarettes. Safer nicotine products including Nicotine Vaping Products (e-cigarettes, NVP -in this paper) provide a non-combustible means for consuming nicotine. A key question for public health is the extent of uptake of NVP, which have been available on many markets since 2007.
The World Health Organization (WHO) estimates that there are 1.1 billion smokers. This gure has remained unchanged since 2000 and is equivalent to one in ve of the global population [1]. Declines in the prevalence of smoking in some countries and population groups have been offset by increases elsewhere and by population increases. Awareness of the risks of inhaling combusted tobacco products is increasing, but mainly con ned to middle and high economic groups.
80% of smokers live in low and middle income countries and 46% live in just three countries -China, India and Indonesia [1,2]. The health effects of smoking tobacco are well known with an estimate of 7 million premature deaths from smoking related disease each year [3], which is three times more than the number who die from malaria, HIV and tuberculosis combined.
The combustion of tobacco and consumption of Asian types of smokeless tobacco -the latter being a mixture of tobacco and other potentially toxic substances -is a major cause of tobacco-related morbidity and mortality. It has long been noted, to quote the UK psychiatrist Michael Russell, that people smoke for the nicotine but die from the tars and gasses that are inhaled [4]. Tobacco smoke contains an estimated 7000 chemicals, some of which are carcinogenic or have other negative health effects. In 2002, the UK Royal College of Physicians identi ed the bene ts of shifting the market in favour of less hazardous products [5]. But at the time, apart from medicinal nicotine replacement therapy (NRT) products, there were few viable safer alternatives to smoking, except for the oral tobacco "snus" used in Scandinavia.
Options for smokers changed with the arrival of the modern electronic cigarette which was patented in 2004. The rst device was invented by a Chinese pharmacist who wanted a safer alternative to smoking, and was facilitated by developments in battery miniaturisation and electronic circuitry. NVP came onto the market in Europe and North America from around 2007 onwards. Most NVP -the devices and the nicotine liquids -were and still are manufactured in China. Most are distributed and sold online and in 'bricks and mortar' shops by small and medium-sized companies. Tobacco companies were late arrivals to manufacturing nicotine vaping products and still account for only 10 to 20% of the market [6].
There has been rapid uptake of nicotine vaping products in some countries in Europe and in North America. In the UK the percent of the adult population using nicotine vaping products rose from 1.7% in 2012 to 7.1% in 2019 [7]. Data from the UK suggest a substitution effect whereby people who consume nicotine are moving to choose safer products, evidenced by the association between the increase in nicotine vaping products and the decline in smoking, and that nicotine vaping products are the main tool that smokers use when they want to quit smoking cigarettes. The prevalence of nicotine vaping in European Union countries varies between less than 0.2% and more than 7% [7][8][9].
The uptake of nicotine vaping products, and indeed of other non-combustible safer nicotine products such as heated tobaccos and Swedish snus, is an unusual public health phenomenon because it has been driven by consumers and products rather than by formal public health agencies. For example, in the UK, the main increase in uptake of nicotine vaping products predates endorsement by Public Health England. In many countries uptake has occurred despite objections and concerns from public health agencies. At a global public policy level, the WHO has held a negative position on nicotine vaping products, despite harm reduction being mandated in the global Framework Convention on Tobacco Control [10]. The legislative response to nicotine vaping products varies, with the predominant response being a regulatory model (75 countries), bans -most of which are the interpretation of legacy legislation (36 countries) or absence of regulatory control (85 countries) [11].
Mapping the use of tobacco and nicotine products is patchy. The WHO promotes the MPOWER tobacco control programme, the rst letter of which refers to Monitoring [12]. One of the key points of monitoring is the collection of up-to-date information on the use of tobacco and related products. This is an extremely di cult task for many countries due to the cost. Given this dearth of information on tobacco use, the WHO uses a variety of epidemiological techniques to develop estimates that are necessarily based on partial and sometimes unrepresentative data.
The situation regarding population measurement of NVP is even more challenging given -in epidemiological terms -the relative recent arrival and uptake of these products. Ideally information on vaping should come from nationally representative surveys, or extrapolation from geographically limited surveys. Some of the information on vaping comes from convenience samples including internet surveys.
The other main source of information comes from market research studies. However, the interpretation of these is di cult in population terms, because they focus on units sold or market value, rather than on the number of users. These studies also mainly derive from the store data, hence miss internet sales and small shops. There is an uncertain relationship between market values and number of users, which depends, amongst other things, on frequency and quantity of use. Most market research is commissioned by manufacturers and is not in the public domain.
NVP are potentially disruptive innovations and which in turn have considerable potential for the tobacco harm reduction. A key question is the rate at which safer nicotine products (Safer Nicotine Products -SNP -in this paper) are replacing tobacco cigarettes. It is therefore important to map the uptake of nicotine vaping products. The objective of this study was to derive a global estimate of the number of people using NVPs.

Methods
Since 2018 we have been collecting information on the prevalence of nicotine vaping as part of the Global State of Tobacco Harm Reduction project (GSTHR). This project maps the global, regional and national availability and use of SNP, the regulatory responses to these products, and the public health potential of tobacco harm reduction. Information is collected from the literature and a network of national and regional correspondents on smoking prevalence and mortality, SNP and NRT use and regulations and controls across 201 countries (262 including territories and signi cant within-country regions). The database has been designed to compile country-based information over time and product categories Country level data are available on the GSTHR website (https://gsthr.org) and updated in real time as new information becomes available. Data on vaping are reported where available from nationally representative surveys of the general adult population.
Inclusion criteria for the studies used in our analysis were that they were recent (published between 2012 and 2020) general population surveys and included information on daily or current nicotine vaping (which was often de ned as use of e-cigarettes). We found surveys on nicotine vaping from 49 countries (see Results).
The question arises as to how to derive a global estimate of vaping. The problem of missing data is quite common in the social sciences and epidemiology. Data loss begins at the survey implementation stage. The selected sample is never fully complete. Certain groups of people are always unavailable.
Researchers try to avoid statistical error caused by missing data and the most commonly used of these is post-strati cation weighting [14][15][16][17].
For characteristics aggregated at the country level -as is the case for smoking or vaping prevalence -we have almost the same situation as for survey incompleteness. General characteristics can be used to estimate missing information. Information on the prevalence of NVP use comes from surveys that have not been widely conducted in all countries. Moreover, each survey may measure its interest in its own way.
In the case of vaping, it can be occasional use, current use, and daily use, each with different details. When making comparisons between countries, we must also consider cultural biases in de nitions.

Estimation methods
Given that there is information for only 49 countries, the question arises as to how to estimate the prevalence of vaping in countries for which information is unavailable. We have used an accepted epidemiological method of estimating country data by assumed similarity between countries in the same region and economic condition for which data points are available. This methodology is commonly used for estimating health status in the absence of national surveys. Aceijas [18] showed one of the variants of this method in her study about relationship between drug injecting and HIV infections. Dawood [19] used it to estimate the mortality associated with pandemic in uenza in 2009. And Verity [20] applied a statistical models based on the same assumptions to estimate severity of coronavirus disease. The WHO and other United Nations organisations publish global and national estimates from multiple agencies on a wide range of indicators, particularly mortality levels and trends [21]. A good example is the process used to impute incidence at the country level in the Global Burden of Disease study, which uses three different statistical models to produce reliable estimates [22].
The method is similar to other methods used for statistical inference and analysis with incomplete data. The main mathematical tool used for estimation is a method that can be called average similarity. It assumes that if certain characteristics in a given group of countries are similar, then these countries will also be similar in other respects. This allows us to impute unknown values of the characteristics with the average values of those characteristics in the countries where they are known [23]. This is a major assumption and has obvious limitations.
The main characteristics that can be used are general characteristics such as geographic regions and economic groups. However, health-related behaviours such as nicotine use are related to a wide range of social phenomena. For example, religious customs and rituals, historical events, cultural, economic, and political conditions. A well-conducted operationalization of several parameters that could cover the above spheres would be a great improvement of the estimation procedure. These parameters are rarely available.
The quality of such estimation depends on the ratio of known and unknown fractions (see Survey population coverage). The higher the level of input data and the more homogeneous they are, the better the result. Figure 1 illustrates the relationship between amount and internal differentiation of input data and quality of estimates.
To achieve better control over the quality and reliability of the process, a synthetic goodness-of-t characteristic can be calculated. It can be based on the standard deviation and the number of cases in the aggregated groups. This is analogous to a simpli ed con dence interval.
A further issue regards the availability and relevance of data on country characteristics from which similarity between countries may be inferred. We used the WHO region, the World Bank income classi cation and the legal status of NVP. Our database (https://gsthr.org) lists countries which ban the sale of NVP.
The second methodological problem is that surveys are conducted in different years and are often not repeated. Most of the available data we found allow for the calculation of an estimate for 2018 (32 studies are from 2017-2018, 14 are earlier and 3 are later). Therefore, the data need to be adjusted to arrive at estimates for 2020. This can be done by reference to data on market growth. Based on publicly available data on e-cigarette market value reported by Statista [24], we can track changes in global revenues in the e-cigarette market. The relationship between market value and vaping prevalence is

Results
Vaping surveys -data availability Information on the prevalence of nicotine vaping from nationally representative surveys was available for 49 countries (see Fig. 2 and supplementary table). There are two publicly available international research programmes and eight national programmes on tobacco use and vaping. The international research from which we obtained data comprises: The Global Adult Tobacco Survey (GATS) [25], last conducted in 2017, providing data on six countries (there are 25 of them in the study, but only six were used due to outdated surveys or not covering the subject of e-cigarettes); and the Special Eurobarometer on Attitudes of Europeans towards tobacco and electronic cigarettes [8], last conducted in 2017, gathering data from 28 countries 1

Survey population coverage
To be more relevant, it is useful to look at the availability of data from a world population perspective.
According to the United Nations, the adult (15 years old or above) population of the countries for which vaping survey data were available was nearly 2.8 billion in 2018. The population of the remaining countries was over 2.9 billion. Thus, information on vaping was lacking for just over half of the world's population (see Table 1). It is even more interesting to look at what proportion of the population was surveyed by region and income group. The most complete vaping survey coverage -over 90% -was in the Western Paci c region, driven by Australia, China, Japan, Malaysia, New Zealand, the Philippines and Taiwan. In Europe and the Americas, more than three-quarters of citizens were surveyed. The rest of the world remains almost unknown. Only 8% of people in Asia were surveyed, 3.6% in Africa, and less than 2% in the Eastern

Mediterranean.
As might be expected, this is closely related to the economic status of countries. In high-income countries, more than 87% of citizens were surveyed. In upper middle-income countries, not much less -77%. There is a huge gap between them and lower income countries with only 9% of citizens surveyed. In the lowest income countries, it was less than 6%.
Estimation for countries without nicotine vaping survey data Based on the available information from the surveys, we calculated the average prevalence of vaping for each WHO region, World Bank income classi cation group and the legal status of the sale of e-cigarettes.
Unfortunately, as we expected, some groups are very poorly represented (see Table 2). Low-income countries are represented by survey data only from Uganda. Uganda is also the only data point for the African region. Similarly, we have only one data point from the South East Asia region with Bangladesh and the East Mediterranean region with the United Arab Emirates.  Fig. 3) -the most detailed, three-factorsubdivision we had information for only 13 subgroups, which allowed us to calculate estimates for 83 countries.
For the other countries, we used a two-factor breakdown covering all pairs of these three factors. A second (2) split was made on the basis of income groups and sales status, which gave us eight information cells covering 161 countries, a third (3) was made on the basis of income groups and regions with 10 information cells covering 142 countries and a fourth (4) was made on the basis of regions and sales status with nine information cells covering 102 countries. The last (5) subdivision was based only based on one income groups factor.
The results of the calculations have been placed successively in the blanks remaining after the previous step. This means that the countries remaining without an estimated value after the rst step have been assigned the values generated in the second step. In the third step we lled in the missing values remaining after the second step and in the fourth step remaining after the third step. All remaining gaps were lled with the fth step.
We started with 49 known countries. The rst step increased this number to 83, the next to 161, next to 190, the fourth one gave only three more countries, increasing the number of countries to 193, and the fth to 198. There were still three countries left. We attributed the average value obtained from all known countries to these countries. Figure 3 shows detailed information of this process.
The number of vapers was calculated by taking as the denominator the total adult population (over 15 years old) obtained from the UN database for 2018 [2].
Adjustment of the estimate for market value changes between 2018 and 2020 Given the time lapse between when most of the surveys were undertaken and 2020, it is necessary to estimate subsequent growth. The only indicator available is estimates of market growth i.e. the value of sales. The question is, what is the relationship between e-cigarette market growth and changes in nicotine vaping prevalence? We have information on both variables for the UK for the period 2011 to 2019. We have used market average e-cigarette market revenues per capita to avoid distortions related to population changes. As can be seen in Fig. 4, both trends were characterised by steady growth. The correlation between both time series was very strong with a Pearson correlation coe cient of 0.933. Based on this analysis, changes in e-cigarette market revenue are consistent with changes in vaping prevalence by 93%.
We assumed that this relationship is similar globally.

Global estimation of nicotine vaping -2018
The number of vapers in the 49 countries from which the survey data are derived is 40.3 million. Using our estimation procedures we estimate that 17.8 million vapers live in the remaining 152 countries. This gives a total of 58.1 million vapers worldwide in 2018.

Page 11/20
More than half of them live in high income countries. We estimate that 2.1 million vapers live in low income countries, 7.8 million in lower middle income countries, 19 million in upper middle income countries and 29.3 million in high income countries.  Projection of the estimated global number of vapers -2020

million vapers live in the African region, 4.2 million in the Eastern
The projection for year 2020 was made on the basis of information from the E-Cigarettes Worldwide Statista Market Forecast [24]. The global number of vapers estimate was tted to revenue data at the 2018 time point (meaning revenue data from 2018 was directly paired with estimated number of vapers).
The market trend was zeroed (calibrated) at 2018. The proportions series was adjusted with the abovementioned coe cient of the correlation between average e-cigarette market revenue and number of vapers. For the year 2020, the projection is 68 million vapers globally (see Fig. 5). 1 A new special Eurobarometer 506 was published in February 2021 [9]. Comparing our projection to 2020 (GNV) and the Eurobarometer data from 2020 (EB), we have 16 countries where the prevalence of vaping are greater in EB than GNV; 12 countries where prevalence are less in EB than GNV. The differences are small. On average, GNV is underestimated relative to EB by about 0.29 percentage points.

Discussion
We estimated that there were approximately 68 million users of nicotine vaping products worldwide in 2020. This estimate should be treated with caution, as it is based on extrapolation of data from 49 countries; data were missing for 152 countries and territories (over half of the world population). We use a method commonly used in global epidemiological studies that assumes similarity between countries with shared characteristics. The robustness of this assumption needs to be tested and whether the method is suitable for estimating consumer behaviour where there are many factors at play including social, economic, cultural, commercial and legal considerations. The small number of data points for some world regions and World Bank classi cations is a further weakness. We consider that the methodological limitations are more likely to overestimate of the global number of vapers.
There is also wide variation in survey de nitions of vaping. It is obvious from the data collected and assessed in our database that the available information is often of poor quality -survey samples are sometimes insu ciently large to provide su cient precision to measure such small sections of the population. The sampling method of many studies is not probabilistic, which makes it impossible to generalise the measured values correctly per population. It is well known that monitoring risk behaviours in relatively small or hidden populations is not straightforward.
These limitations indicate a surprising de ciency in the national and hence global monitoring of nicotine vaping. Despite WHO's call for countries to implement the MPOWER strategy -recall that the rst letter in the strategy's name refers to monitoring -global monitoring of tobacco use is weak. Only 38% of the global population is covered by surveys of tobacco use [27] and only 1 in 3 countries monitors tobacco use by repeating nationally representative youth and adult surveys at least once every 5 years [28].
Moreover, despite the fact there is considerable global and international debate about vaping nicotine products, most countries are taking no steps to monitor use at a national level. Evidence based policymaking is very di cult in the absence of evidence. Large sums of money have been invested by international philanthropic organisations including Bloomberg Philanthropies to oppose tobacco harm reduction using NVP and other safer nicotine products. It is perhaps ironic that these initiatives do not include any impact measure. Efforts to improve the coverage and accuracy of information about nicotine vaping should be encouraged.

Conclusions
Our best estimate of 68 million nicotine vapers indicates considerable uptake given that the products have only been available on most markets for 10 years or less, that there is either no support or is opposition to vaping by many countries, and that in countries which regulate e-cigarettes there may be strict controls over advertising and promotion.
But, given the global scale of tobacco smoking at 1.1 billion progress in adoption of alternative products is slow. The number of people using nicotine vaping products is still a small fraction of the number of people who smoke.
Historically, changes in nicotine consumption can take many decades. For example the last major disruptive innovation for tobacco was the invention of the cigarette rolling machine in the 1880s but it took a further 60 years or so for the machine rolled cigarette to replace most other forms of tobacco use in richer countries. Given the public and individual health imperative to either quit or switch from combustible to non-combustible nicotine products, there is an urgent need to scale up tobacco harm reduction.
GSTHR -Global State of Tobacco Harm Reduction. Figure 1 Schematic representation of input data and quality of estimates.

Figure 2
Surveys on vaping. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors. Correlation between average e-cigarette market revenue per capita and prevalence of vaping in UK Figure 5 Estimated trends in the worldwide number of nicotine vapers

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. supplementarytable.docx