Global burden of disease due to smokeless tobacco consumption in adults: an updated analysis of data from 127 countries

Smokeless tobacco (ST) refers to various tobacco-containing products that are consumed by chewing, keeping in the mouth or sniffing, rather than smoking [1]. ST products of many different sorts are used by people in every inhabited continent of the world (Table 1) [1]. For example, in Africa, toombak and snuff are commonly used, while in South America, chimó is the product of choice. In Australia, indigenous people use pituri or mingkulpa [2], and in Central Asia, nasvay consumption is very common. In North America, plug or snuff are favoured, and even in Western Europe, where ST products are largely banned, there are exemptions allowing people in Nordic countries to use snus [3]. All the above products vary in their preparation methods, composition and associated health risks (Table 1), but it is in South and Southeast Asia where the greatest diversity of ST products exists, accompanied by the highest prevalence of use [4]. Here, the level of cultural acceptability is such that ST products are often served like confectionery at weddings and other social occasions.

ST products contain nicotine and are highly addictive. Often, they also contain carcinogens, such as tobacco-specific nitrosamines (TSNA), arsenic, beryllium, cadmium, nickel, chromium, nitrite and nitrate, in varying levels depending on the product [5, 6]. The pH of the products also varies widely, with some (e.g. khaini, zarda) listing slaked lime among their ingredients [7]. Raising the pH in this way increases the absorption of nicotine and enhances the experience of using the ST product, increasing the likelihood of dependence. The elevated pH also increases the absorption of carcinogens, leading to higher toxicity and greater risk of harm [7].

The harmful nature of many ST products, and the fact that 300 million people around the world use ST [8], make ST consumption a global public health issue. Many ST products lead to different types of head and neck cancers [9, 10]. An increased risk of cardiovascular deaths has been reported [11], and its use in pregnancy is associated with stillbirths and low birth weight [12, 13].

Because of the diversity described above, ST should not be considered as a single product, but rather as groups of products with differences in their toxicity and addictiveness, depending on their composition. As a consequence, it is difficult to estimate the global risks of ST to human health and to agree on international policies for ST prevention and control. Several country-specific studies [14, 15] have been carried out, and in 2015, we published an estimate of the global burden of disease associated with ST use [16]. We used a novel approach, whereby we classified ST products according to their availability in different geographical regions of the world. For example, ST products in South Asia pose a much greater risk to health than those available in Nordic countries, where the manufacturing process removes many of the toxins from the finished product [6, 17]. Using this approach, we estimated the worldwide burden of disease attributable to ST consumption, measured in terms of disability adjusted life years (DALYs) lost and the numbers of deaths in 2010 [16]. Here, we update this estimate to include data up to 2019, providing an indication of how the global ST arena has changed in the intervening years.

Our methods for updating the estimates of ST disease burden were broadly the same as those used in our earlier publication; these are well described elsewhere [16]. Here, we will summarise these methods and explain any modification made, particularly in relation to the revised timelines. We assessed disease burden for individual countries by varying their populations’ exposure to ST, using the comparative risk assessment method [15]. These individual estimates were then summarised for 14 World Health Organization (WHO) sub-regions (Additional file 1: Appendix 1) as well as for the world.

We first searched the literature to identify the latest point prevalence of ST use among adults ≥ 15 years in men and women for each country (see Additional file 1: Appendix 2 for detailed methods). We searched for the latest estimates for x countries included in our previous study as well as those additional y countries where estimates have been made available since 2014 for the first time. We derived single estimates for each country preferring nationally representative surveys using internationally comparable methods over non-standardised national or sub-national surveys.

We also updated risk estimates for individual diseases caused by ST; however, we kept to the original list of conditions, i.e. cancers of the oral cavity, pharynx and oesophagus, ischemic heart disease and stroke. We only searched for papers published since our last literature search; our updated search strategies can be found in Additional file 1: Appendix 3. As before, all searches and data extraction were independently scrutinised by a second researcher and any discrepancies were arbitrated by a third researcher. All case definitions for diseases and exposure (ST use) used in the retrieved articles were checked for accuracy and consistency and all analyses undertaken in these studies were assessed to see if they controlled for key confounders (mainly smoking and alcohol). We assessed study quality using the Newcastle-Ottawa Scale for assessing non-randomised studies in meta-analysis [24]. For all new studies, we log transformed their risk estimates and 95% confidence intervals to effect sizes and standard errors and added these to the rerun of our random-effects meta-analyses to estimate pooled risk estimates for individual conditions. Where possible, we pooled effect sizes to obtain country-specific risk estimates. For all outcomes in the meta-analyses, we conducted a GRADE assessment to assess the quality of evidence. We also pooled these effect sizes to obtain non-specific global risk estimates. Given that the risk varies from country to country, depending upon which products are locally popular, we used country-specific risk estimates where possible. In countries with no estimates, we used estimates of those countries where similar ST products were consumed. For other countries without estimates that consumed ST products known to contain high levels of TSNAs, we applied non-specific global estimates. Where no information was available on the composition of ST, we did not apply any estimates. Details on how these statistically significant estimates were applied to each WHO sub-region can be found in web Additional file 1: Appendix 4.

Based on the extent to which the included studies adjusted for potential confounders, we categorised them as ‘best-adjusted’ and ‘others’. We carried out a sensitivity analysis for all risks and attributable disease burden estimates including only ‘best-adjusted’ studies. A sensitivity analysis was also carried out by estimating risk estimates separating out cohort from case-control studies.

For each country, we used their point prevalence of ST use and the allocated risk estimate for each condition to estimate its population attributable fraction (PAF) as below:

Using the 2017 Global Burden of Disease (GBD) Study, we also extracted the total disease burden (B) in terms of number of deaths and DALYs lost due to the conditions associated with ST use for both men and women. The attributable burden (AB) due to ST was then estimated in deaths and DALYs lost for these conditions for both men and women using the following equation.

ST consumption was reported in 127 countries (Fig. 1). These estimates were extracted from nationally representative cross-sectional surveys conducted either as part of international (97/127) or national (30/127) health and tobacco surveillance (Additional file 1: Appendix 5a). A variety of age ranges (as young as 15 or as old as 89, including no upper age limit) were used to define adults.

ST consumption was more common among males than females in 95 countries (Table 2). Among males, Myanmar (62.2%), Nepal (31.3%), India (29.6%), Bhutan (26.5%) and Sri Lanka (26.0%) had the highest consumption rates. Among females, Mauritania (28.3%), Timor Leste (26.8%), Bangladesh (24.8%), Myanmar (24.1%) and Madagascar (19.6%) had the highest consumption rates. Within Europe, Sweden (25.0% males, 7.0% females) and Norway (20.1% males, 6.0% females) had the highest ST (snus) consumption rates.

Our post-2014 systematic literature search identified an additional four studies demonstrating a causal association between ST and oral cancer; these included two Pakistan-based and one India-based case-control studies and one US-based cohort study (Table 3). No new studies were found for pharyngeal and oesophageal cancers. PRISMA flow diagrams describing the selection process of the studies identified in the literature searches are provided in Additional file 1: Appendix 5b,c. By adding the new studies to the list of studies selected in our first estimates and revising the meta-analyses, we found that the pooled estimates were statistically significant for cancers of the mouth (Fig. 2). The non-specific pooled estimate for oral cancers, based on 36 studies, were 3.94 (95% CI 2.70–5.76). The country-specific relative risk for oral cancers for India was higher (RR 5.32, 95% CI 3.53–8.02) than no-specific estimates and for the USA remained statistically insignificant (RR 0.95, 95% CI 0.70–1.28). Since no new studies were added for pharyngeal and oesophageal cancers, their non-specific risk estimates of 2.23 (95% CI 1.55–3.20) and 2.17 (95% CI 1.70–2.78) remained as per our original estimates, respectively. For cardiovascular diseases, we identified another three Swedish studies for ischaemic heart disease and another two (one in Asia and one in Sweden) for stroke (Table 3). In the absence of any new non-Swedish studies on ischaemic heart disease (Fig. 3), we considered the relative risk (adjusted odds ratio 1.57, 95% CI 1.24–1.99) of myocardial infarction due to ST identified in the 52-country INTERHEART study [35] (conducted across nine WHO regions) as a valid estimate. However, the country-specific (Sweden) relative risk for ischaemic heart disease (RR 0.94, 95% CI 0.87–1.03) and both country-specific (RR 1.02, 95% CI 0.93–1.13 [Sweden]) and non-specific relative risks for stroke (RR 1.03, 95% CI 0.94–1.14) remained statistically insignificant. The GRADE assessment was moderate for oral, pharyngeal and oesophageal cancers and low for IHD (see Additional file 1: Appendix 7).

We found that most of the included studies adjusted for potential confounders (35/38 for oral, 10/10 for pharyngeal and 15/16 for oesophageal cancers; and 13/16 for IHD) and classified as providing ‘best adjusted’ estimates. According to a sensitivity analysis restricted to only ‘best-adjusted’ studies, the overall risk estimates (RR/OR) for oral cancer increased from 3.94 to 4.46 and for oesophageal cancer from 2.17 to 2.22 (see Additional file 1: sensitivity analysis #1). Separate risk estimates for cohort and case-control studies are included in the Additional file 1: sensitivity analysis #2).

The above risk estimates were included in the mathematical model to estimate the population attributable fraction (PAF), as follows (also see Additional file 1, Appendix 4 for detailed justification): For oral, pharyngeal and oesophageal cancers, Sweden- and US-based country-specific risk estimates were applied to Europe A and America A regions, respectively. Similarly, India-based country-specific risk estimates were applied to Southeast Asia B and D and Western Pacific B regions. No risk estimates were applied to Europe C due to the non-existence of any risk estimates or information about the toxicity of ST products. For all other regions, non-specific country estimates were applied. A few exceptions were made to the above assumptions: a Pakistan-based country-specific estimate was applied for oral cancers for Pakistan and an India-based estimate for the other two cancers; for the UK, India-based country specific estimates were applied due to the predominant use of South Asian products in the country. For ischaemic heart disease, the INTERHEART disease estimates were applied to all WHO regions except two, i.e. Europe A due to the availability of Sweden-based country specific estimates and Europe C due to the non-availability of relevant information. As previously stated, an exception was made for the UK and the INTERHEART estimates were applied.

According to our 2017 estimates, 2,556,810 DALYs lost and 90,791 deaths due to oral, pharyngeal and oesophageal cancers can be attributed to ST use across the globe (Table 4). By applying risk estimates obtained from the INTERHEART study, 6,135,017 DALYs lost and 258,006 deaths from ischaemic heart disease can be attributed to ST use. The overall global disease burden due to ST use amounts to 8,691,827 DALYs lost and 348,798 deaths. The attributable disease burden estimates when restricted to only ‘best adjusted’ studies, did not change significantly; the DALYs lost attributable to ST increased to 8,698,142 and deaths to 349,222.

Among these figures, three quarters of the total disease burden was among men. Geographically, > 85% of the disease burden was in South and Southeast Asia, India accounting for 70%, Pakistan for 7% and Bangladesh for 5% DALYs lost due to ST use (Additional file 1: Appendix 6).

ST consumption is now reported in at least two thirds of all countries; however, health risks and the overall disease burden attributable to ST use vary widely depending on the composition, preparation and consumption of these products. Southeast Asian countries share the highest disease burden not only due to the popularity of ST but also due to the carcinogenic properties of ST products. In countries (e.g. Sweden) where ST products are heavily regulated for their composition and the levels of TSNAs, the risk to the population is minimal.

We found ST prevalence figures in 12 countries that did not previously report ST use; new figures were also obtained for 55 countries included in the previous estimates [16]. Among these 55 countries: 19 reported a reduction in ST use among both men and women (e.g. Bangladesh, India, Nepal), 14 only among men (e.g. Laos, Pakistan) and eight only among women (e.g. Bhutan, Sri Lanka) (Fig. 4a, b). On the other hand, 13 countries showed an incline in ST use among both men and women (e.g. Indonesia, Myanmar, Malaysia, Timor Leste) and one country (Sweden) among men only. Overall, our updated ST-related disease burden in 2017 was substantially higher than that for 2010—by approximately 50% for cancers and 25% for ischaemic heart disease. This occurred despite a substantial reduction in ST prevalence in India (constituting 70% of the disease burden) and little change in the disease risk estimates. We are now reporting ST use in 12 more countries; however, the main reason for the increased burden of disease was a global rise in the total mortality and DALYs lost—oral, pharyngeal and oesophageal cancers, in particular. The disease burden due to these cancers lags several decades behind the risk exposure. Therefore, a significant reduction in ST-related disease burden as a result of a reduced prevalence will not become apparent for some time to come. Among other studies estimating ST-related global disease burden, our mortality estimates were far more conservative than those reported by Sinha et al. (652,494 deaths); however, their methods were different from ours [9]. Moreover, Sinha et al.’s estimates included a number of additional diseases such as cervical cancer, stomach cancer and stroke. None of these risks were substantiated in our systematic reviews and meta-analyses. On the other hand, our estimates of 2,556,810 DALYs lost and 90,791 deaths due to cancers are close to those estimated by the GBD Study for 2017, i.e.1,890,882 DALYs lost and 75,962 deaths due to cancers [91]. A reason for the slight difference between these two estimates might be that ours included pharyngeal cancers in the estimates while GBD Study only included oral and oesophageal cancers.

Our methods have several limitations. These have been described in detail elsewhere [16] but are summarised here. Our estimates were limited by the availability of reliable data and caveated by several assumptions. The ST use prevalence data were not available for a third of countries despite reports of ST use there. Where prevalence data were available, there were very few studies providing country-specific disease risks—a particular limitation in Africa and South America. In the absence of country-specific risk estimates, the model relied on assuming that countries that share similar ST products also share similar disease risks. For example, oral cancers risk estimates were only available from five countries (India, Norway, Pakistan, Sweden and the USA). For other countries, the extrapolated risks were based on similarities between ST products sold there and in the above five countries. The estimates for ischemic heart disease must be interpreted with caution, in particular, as the risk estimates for most countries were extrapolated from a single (albeit multi-country) study (INTERHEART). However, we excluded those regions from the above extrapolation where the INTERHEART study was not conducted. As previously noted, the total disease burden observed in 2017 is a consequence of risk exposure over several decades. Therefore, the attributable risk based on the prevalence figures gathered in the last few years may not be accurate. If ST prevalence has been declining in a country over the last few decades, the disease burden obtained by applying more recent prevalence figures may underestimate attributable disease burden. This may well be the case in India where ST use has declined by 17% between the 2009 and 2017 GATS surveys [92]. On the other hand, if ST use is on the rise (e.g. in Timor Leste), the attributable disease burden for 2017 could be an overestimate.

While we found a few more recent ST prevalence surveys and observational studies on the risks associated with ST use, big evidence gaps still remain. The ST surveillance data for many countries are either absent or outdated. The biggest gap is in the lack of observational studies on the risks associated with various types of ST used both within and between countries. While longitudinal studies take time, global surveillance of ST products, their chemical composition and risk profile can help improve the precision of future estimates. As cancer registries become more established around the globe, their secondary data analysis can also provide opportunities to estimate ST-related risks.

ST is the main form of tobacco consumption by almost a quarter of all tobacco users in the world. Yet, its regulation and control lags behind that of cigarettes. The diversity in the composition and toxicity of ST products and the role of both formal and informal sectors in its production, distribution and sale make ST regulation a particular challenge. In a recent policy review of 180 countries that are signatories to WHO FCTC, we found that only a handful of countries have addressed ST control at par with cigarettes [93]. The regulatory bar is often much lower for ST than cigarettes [94]. Where ST control policies are present, there are gaps in their enforcement [95]. On the other hand, Sweden has demonstrated what can be achieved through strong regulations; ST-related harm has not only been reduced significantly, but snus is now used to reduce harm from smoking. Countries where ST use is popular and poses risks to health need to prioritise ST control and apply WHO FCTC articles comprehensively and evenly across all forms of tobacco.

ST is consumed across the globe and poses a major public health threat predominantly in South and Southeast Asia. While our disease risk estimates are based on a limited number of studies with modest quality, the likely disease burden attributable to ST is substantial. In high-burden countries, ST use needs to be regulated through comprehensive implementation and enforcement of the WHO FCTC.