Global burden of disease due to smokeless tobacco consumption in adults: analysis of data from 113 countries

Background Smokeless tobacco is consumed in most countries in the world. In view of its widespread use and increasing awareness of the associated risks, there is a need for a detailed assessment of its impact on health. We present the first global estimates of the burden of disease due to consumption of smokeless tobacco by adults. Methods The burden attributable to smokeless tobacco use in adults was estimated as a proportion of the disability-adjusted life-years (DALYs) lost and deaths reported in the 2010 Global Burden of Disease study. We used the comparative risk assessment method, which evaluates changes in population health that result from modifying a population’s exposure to a risk factor. Population exposure was extrapolated from country-specific prevalence of smokeless tobacco consumption, and changes in population health were estimated using disease-specific risk estimates (relative risks/odds ratios) associated with it. Country-specific prevalence estimates were obtained through systematically searching for all relevant studies. Disease-specific risks were estimated by conducting systematic reviews and meta-analyses based on epidemiological studies. Results We found adult smokeless tobacco consumption figures for 115 countries and estimated burden of disease figures for 113 of these countries. Our estimates indicate that in 2010, smokeless tobacco use led to 1.7 million DALYs lost and 62,283 deaths due to cancers of mouth, pharynx and oesophagus and, based on data from the benchmark 52 country INTERHEART study, 4.7 million DALYs lost and 204,309 deaths from ischaemic heart disease. Over 85 % of this burden was in South-East Asia. Conclusions Smokeless tobacco results in considerable, potentially preventable, global morbidity and mortality from cancer; estimates in relation to ischaemic heart disease need to be interpreted with more caution, but nonetheless suggest that the likely burden of disease is also substantial. The World Health Organization needs to consider incorporating regulation of smokeless tobacco into its Framework Convention for Tobacco Control. Electronic supplementary material The online version of this article (doi:10.1186/s12916-015-0424-2) contains supplementary material, which is available to authorized users.


Background
Smokeless tobacco (SLT) consists of a number of products containing tobacco, which are consumed-without burning-through the mouth or nose [1]. A diverse range of SLT products are available worldwide, varying in their composition, methods of preparation and consumption, and associated health risks (Table 1) [1]. Its use is most prevalent in South and South-East Asia where one-third of tobacco is consumed in smokeless form [2,3]. Wrapped in a betel leaf with areca nut, slaked lime, and catechu, SLT is often served at social occasions in this region. Other products (e.g. gutkha, khaini) contain slaked lime, areca nut, flavourings, and aromatic substances [4]. A number of products based on powdered tobacco (e.g. snus) are also consumed in Nordic countries and North America. In other parts of world, the most commonly used SLT products (Table 1) include Chimó (Venezuela), Nass (Uzbekistan, Kyrgyzstan), Tambook (Sudan, Chad), and Snuff (Nigeria, Ghana, South Africa).
In addition to nicotine, SLT products contain over 30 carcinogens [5] including tobacco-specific nitrosamines  [6], Lawler et al. [17], and NIH & CDC 2014 report on smokeless tobacco products [37] (TSNA), arsenic, beryllium, cadmium, nickel, chromium, nitrite, and nitrate. The level of nicotine and carcinogens vary between products (Table 1) [6]. For example, nicotine content among SLT products varies between 0.2 and 40.1 mg/g, compared to commercial filtered cigarettes which contain 16.3 mg/g of nicotine [7]. Their pH also varies, which, being a key determinant of the level of absorption of nicotine and carcinogens, determines its toxicity: the higher the pH, the higher the absorption and, consequently, the higher the toxicity [6]. Such considerations mean that there are substantial variations between different SLT products in the level of risk posed to human health [4,[8][9][10][11]. It is therefore important not to consider SLT as a single product, but rather as groups of products with differences in their toxicity and addictiveness depending upon their carcinogen, nicotine, and pH levels. The diversity in SLT toxicity has been an impediment not only in establishing its global risks to human health, but also in agreeing on international policies for its prevention and control. It is therefore perhaps unsurprising that despite several country-specific studies [12][13][14][15] no attempt has hitherto been made to estimate its global disease burden.
To overcome these challenges, we developed a novel approach to estimate the global burden associated with the use of SLT products. The determinants of their toxicity (carcinogens and pH) and addictiveness (nicotine) are dependent on preparation methods, ingredients that are added to SLT products, and consumption behaviours. Given that the SLT preparations and consumption patterns are determined by, and vary with, geography and culture [16], it is possible to group them according to their availability in different parts of the world (Table 1). These groups of SLT products, classified according to different geographical regions, will also be distinguishable from each other on the basis of their toxicity, addictiveness, and associated health risks. Hence, the risks were assumed to be highest in those regions and cultures where products are combined with other ingredients, and are prepared and consumed in a way that makes them very alkaline (i.e. a high pH), and rich in nicotine and TSNA [6,17]. Building on this assumption, we aimed to estimate the worldwide burden of disease attributable to SLT use, measured in terms of disability adjusted life years (DALYs) lost and number of deaths in 2010.

Methods
We used the comparative risk assessment method, which evaluates changes in population health (burden of disease) that result from modifying a population's exposure to a risk factor [18,19]. For this, we used 2010 datasets, which provided the most recent global estimates of burden of disease [20]. The estimates were calculated for individual countries and then grouped into 14 World Health Organization (WHO) sub-regions (Additional file 1: Appendix 1) [21]. These were generated through estimating the following: 1. The prevalence of SLT consumption 2. Diseases caused by SLT use 3. The relative risks of acquiring these diseases 4. The population attributable fraction (PAF) for each of these diseases 5. The overall burden of these diseases in terms of DALYs lost and deaths 6. Proportion of this burden attributable to SLT use

Prevalence of smokeless tobacco use
We carried out a systematic literature search (see Additional file 1: Appendix 2 for a detailed description of the methods employed) for the point prevalence (current use) of SLT consumption among all adult (≥15 years) populations, and also for men and women separately. Only one prevalence report was included for one country. Latest national prevalence data collected as part of an international or regional survey were preferred over an older isolated national or a sub-national survey. We used data from the Global Adult Tobacco Survey (GATS), where available [22]. In its absence, other international (WHO STEPwise approach to Surveillance, The Demographic and Health Surveys), regional (Special Europe Barometer), national, and/or sub-national surveys were used to extract prevalence data.

Diseases caused by smokeless tobacco use
A scoping review was carried out to identify associated diseases. A series of focused literature reviews were subsequently carried out to find and assess the evidence of causation between each of these diseases and SLT use. Our search strategies and selection criteria are provided in Additional file 1: Appendix 3. One researcher ran the searches, which were then independently scrutinised by another independent researcher who considered the search results against the pre-specified inclusion and exclusion criteria. Similarly, one researcher extracted data, which were independently crosschecked by another researcher. In particular, we appraised the studies for case definitions for diseases and for assessment methods for measuring exposure to SLT and for investigating the effects of potential confounders. We excluded those diseases (and respective studies) where evidence was not supportive of a causal relationship. Only studies that adequately controlled for smoking and/or alcohol as potential confounders either at the design or the analysis stage were carried forward into the next stage of the analysis (discussed below). Quality was assessed using the Newcastle-Ottawa Scale for assessing the quality of nonrandomised studies in meta-analyses [23].

Assessing risk and meta-analyses
Risk estimates (relative risks/odds ratios) and their confidence intervals (CI) were log transformed to produce effect sizes and standard errors, respectively [24]. We carried out random effects meta-analysis using RevMan version 5 to estimate pooled risk estimates. We first obtained country-specific risk estimates (relative risks/odds ratios) for individual diseases by pooling data from the included studies carried out in respective countries. We then extrapolated non-specific global risk estimates by pooling respective country-specific risk estimates. We were mindful that the risk of acquiring diseases varies between countries owing to differences in SLT products used. Therefore, for each disease where good countryspecific risk estimates (pooled estimate from a metaanalysis of three or more studies in respective country) were available, we applied these to respective countries and also to those countries and regions where similar SLT products are used. In the absence of good countryspecific risk estimates, we used either one of the following two approaches: (a) In countries and regions that use SLT products with moderate to high pH and TSNAs levels, we applied non-specific global estimates (pooled estimate from a meta-analysis of all studies); and (b) in countries and regions where there was either no information available on the SLT products or the information available indicates low levels of pH and TSNA, we did not apply any estimates. Further details on the application of these assumptions across all 14 WHO regions are provided in web Additional file 1: Appendix 4. We only used those pooled relative risks (country or nonspecific) that were found to be statistically significant.
Where associations were presented for more than one SLT product in the same paper, we considered these as separate studies for the purpose of meta-analysis. Similarly, where risks were given separately for former and current SLT users, these were also treated as separate studies. We did not attempt to group risks according to gender because very few studies had such sub-group analysis.

Population attributable fraction
PAF is the proportional reduction in disease or mortality that would occur if exposure were reduced to zero [25,26]. PAF was estimated for each disease for each country for both males and females, using the following formula:

Overall burden
The overall number of DALYs and deaths for each associated disease for both males and females for each country were extracted from the 2010 Global Burden of Disease study [27,28].

Attributable burden
The attributable burden (AB), in deaths and DALYs, was estimated for each associated disease for each country for both males and females by multiplying PAF by the overall burden of the disease (B):

Prevalence of smokeless tobacco use
We found adult prevalence figures for SLT consumption in 115 countries (Fig. 1). The definition for 'adult' ranged from 15, 16, 25, or 35 years at one end to 49, 64, 65, 70, 74, 84, 85, 89, or no age limit at the other. The PRISMA diagram describing the selection of the prevalence reports is provided in Additional file 1: Appendix 5a.
In general, SLT consumption was higher among males than females (Table 2)

Diseases caused by smokeless tobacco use
The initial scoping review identified a number of associated diseases, including a range of cancers, cardiovascular diseases (ischaemic heart disease and stroke), periodontal conditions, and adverse pregnancy outcomes. The subsequent more focused systematic reviews identified 53 studies (Table 3) reporting association between SLT consumption and cancers of mouth, pharynx, larynx, oesophagus, lung, and pancreas (39 studies); and cardiovascular diseases, such as ischaemic heart disease and stroke (14 studies). PRISMA flow diagrams describing the selection process of the studies identified in the literature searches are provided in Additional file 1: Appendix 5b,c. The pooled non-specific relative risks were statistically significant for cancers of the mouth, pharynx, and oesophagus (Figs. 2, 3, 4, and 5). Only statistically significant relative risks (country-specific or non-specific) were included in the model to estimate attributable risks. For example, the pooled nonspecific relative risk for laryngeal cancer was 1.42 (95 % CI 0.77-2.59), and hence excluded (Additional file 1: Appendix 6). Likewise, none of the country-specific estimates for the USA were statistically significant   (Additional file 1: Appendix 4). Based on the above reviews, we assumed that a causal association exists between some SLT products and cancers of the mouth, pharynx, and oesophagus, and ischaemic heart disease.

Relative risks
Based on 32 studies, the estimated pooled non-specific relative risk for mouth (oral cavity, tongue, and lip) cancers was 3.43 ( (Fig. 4). For ischaemic heart disease, no good country-specific risk estimates were available (Fig. 5). However, we found one large casecontrol study (INTERHEART study) [29] conducted in 52 countries from all regions showing a statistically significant risk of ischaemic heart disease (adjusted odds ratio 1.57, 95 % CI 1.24-1.99) among SLT users.

Applying risk estimates
For cancers in general, pooled country-specific risk estimates obtained from Sweden and the USA were applied       NOS for assessing the quality of non-randomised studies in meta-analyses based on selection, comparability, and exposure/outcome. Number of stars (*) indicates the number of criteria met for each of these three categories [23] b Effect sizes are for oral and pharyngeal cancers combined and were included in the meta-analysis for oral cancer only c Based on parameter estimate and standard error reported in paper to Europe A and Americas A, respectively. For South-East Asia B and D and Western Pacific B regions, country-specific estimates from India were applied.
There were a few exceptions to this rule, because some countries (UK, Mexico, Pakistan, China, Mongolia) differed in their SLT consumption patterns from their respective regions (see Additional file 1: Appendix 4 for details). In short, country-specific risk estimates for cancers could only be fully applied to five regions. For the remaining nine regions, our findings were imputed either by applying statistically significant non-specific risk estimates or none at all (Additional file 1: Appendix 4). In case of ischaemic heart disease, Sweden was the only country with a pooled country-specific relative risk (0.98, 95 % CI 0.90-1.07) obtained from a good number (more than three) of studies. For 11 out of 14 regions, we used a large multi-country study (INTERHEART)-conducted in 52 countries-to apply and deduce risk estimates. The three regions (Europe A and C and Americas D) were excluded, as these were not among those regions included in the INTERHEART study (Additional file 1: Appendix 4). There was one exception (UK) where INTERHEART study estimates were applied because SLT products consumed in the UK commonly originate from South Asia.

Attributable burden
The attributable burden of SLT use is outlined in

Discussion
We have found that SLT is consumed worldwide and that its use results in substantial, potentially avoidable, morbidity and mortality. However, owing to marked differences in the types of products available, patterns of consumption, and associated risks, there are substantial differences in the attributable burden between regions and countries. In particular, SLT consumption in South-East Asia leads to a much greater burden of disease than in Sweden, despite its use being equally prevalent. This is due to the much lower levels of TSNA and pH in SLT products in Sweden compared to those found in SLT in South-East Asia [6]. Similarly, SLT products used in the USA have lower risk estimates than for those used in South-East Asia. We found that more than six million DALYs were lost and over a quarter of a million deaths occurred in 2010 owing to SLT consumption. However, our estimates require cautious interpretation because of a number of potential limitations.
First, our analysis was limited to those countries and diseases for which reliable prevalence and risk data were available, respectively. Most global tobacco surveys that reported on SLT consumption did not include all countries in the world. While global figures on smoking prevalence were available, we did not find any SLT prevalence figures for almost half of all countries. Where SLT prevalence figures were available, two countries (Micronesia and Saint Kitts & Nevis) were excluded from the final estimates owing to an absence of data for cancers in the 2010 Global Burden of Disease study. Moreover, for certain disease outcomes, e.g. adverse reproductive and oral health effects, poor quality as well as limited quantity of evidence precluded their inclusion.
Second, lack of country-specific risk estimates leads to considerable uncertainty. Despite several countries reporting SLT consumption, most did not have any reliable information on the types of SLT products used and on their associated health risks. For example, studies from several African countries reported high SLT consumption (Table 2), but provided little information on their hazard profile. There is some evidence, mainly from Sudan [30], that products used in Africa tend to have a higher pH than those used in Europe or in the USA. However, we did not find any data on the risks associated with widespread SLT use in southern parts of Africa. Likewise, various forms of SLT have been used in parts of South America (Brazilian rapê or Venezuelan chimó) for many years, yet there are no studies on the health effects of such products. In the absence of country-specific risk estimates, we assumed that in general those populations that consume similar SLT products are likely to share similar health risks and susceptibilities. We extrapolated and applied risk estimates to most countries included in our analysis on that basis (Additional file 1: Appendix 4). For cancer, our extrapolation was based on estimates obtained from several studies; for ischaemic heart diseases, extrapolations were mostly based on a single although large multicountry study (INTERHEART). As a result, almost three-quarters of the estimated SLT disease burden, which is attributed to ischaemic heart disease, is uncertain. Therefore, a cautious interpretation would be to exclude ischaemic heart disease burden figures from our estimates. However, in estimating these figures we had already excluded those regions and their respective countries that were not included in INTERHEART study. As a pointer on future research, our study highlights the need to study risk of SLT consumption on ischaemic heart diseases across the spectrum of SLT products and consumption behaviours. In time, this will produce more country-specific risk estimates, which would undoubtedly improve the reliability of our estimates presented here.
Third, the disease burden observed in 2010 is unlikely to be a consequence of SLT consumption in recent years. Therefore, our prevalence figures, obtained in surveys carried out in the last decade and used in the estimates, could be problematic. However, we assumed that the SLT consumption rates have remained stable over the last 30-40 years in these countries. We consider this as a safe assumption given that SLT use is not a new trend and historically embedded in culture and tradition in many countries, most remarkably in South Asia [31]. Consumption trends based on repeated youth surveys in India and Bangladesh suggest that SLT use has remained stable over the last decade [32]. Evidence from Sweden suggests that while more people are using snus now than 25 years ago, the consumption trends, compared to cigarette use, have essentially remained stable in this period [33,34].
Finally, the age range of the adult sampling frames used in different SLT prevalence surveys varied, which could also increase uncertainty. The main difference between two of the key categories used was in the adult range starting from either ≥15 years or ≥25 years. Given that the risk of cancers and ischemic heart disease accumulates after many years of use well beyond young adult age, it may not have made much of a difference to our burden of disease estimates.
For the seven countries in South-East Asia region D, we estimated that 55,060 deaths caused by cancers of  mouth, pharynx, and oesophagus, could be attributed to SLT in 2010. This is a little higher than the estimates from a recent study in which 50,000 deaths were attributed to SLT in eight South Asian countries [4]. This discrepancy may be explained by the fact that we used the most recent, updated prevalence and burden of disease figures. Our estimate does not include economic impact. However, given the nature of the associated diseases, it is likely that the SLT use imposes a huge economic burden on weak health systems and poor economies. Moreover, owing to higher consumption of SLT among people of lower socio-economic status and inequitable access to health care in low-income and middle-income countries, its use is likely to contribute to driving disadvantaged sections of these societies into further poverty. A disproportionate impact on the male population (more than 70 % of disease burden due to SLT is in males) is also likely to have a disproportionate economic impact on societies in terms of reduced workforce contributions by men. On the other hand, effective legislation, policy, and preventive programmes could avert this burden due to SLT.
The signatories of the WHO's Framework Convention on Tobacco Control should, in addition to the focus on reducing smoking consumption and related harm, now also consider the need to regulate production, marketing, and labelling of SLT products. This is particularly necessary in those countries where prevalence is high and SLT products are manufactured at a large scale without any checks on the carcinogenic level of their ingredients [35]. In countries where its use is largely limited to immigrant populations (such as in the UK) [36], strict regulation and taxation policies should be enforced which prevent import of SLT products and sale by local shops.
SLT is an important health issue, applying to a large part of the world. The data presented here are the most comprehensive gathered and brought together thus far. However, considerable uncertainties remain pertaining to risk estimation of different diseases associated with SLT use. Therefore more research is needed to investigate the newly established and previously known adverse health outcomes pertaining to SLT, particularly within countries where prevalence is high but no research evidence of risk estimation is available. Moreover, more descriptive questions about the type of SLT products and the pattern of use should be introduced into national surveys and publications of such findings encompassing all the regions.

Conclusions
Our study, a first attempt to assess global burden of disease due to SLT, estimates that more than six million DALYs are lost and over a quarter of a million deaths occur each year owing to its consumption. There is a need to build on the insights obtained from efforts to reduce cigarette smoking-related harm and to investigate strategies to reduce use of SLT and decrease the substantial associated burden of harm.