Article Text

PDF

Global burden of all-cause and cause-specific mortality due to smokeless tobacco use: systematic review and meta-analysis
  1. Dhirendra N Sinha1,
  2. Rizwan A Suliankatchi2,
  3. Prakash C Gupta3,
  4. Thaksaphon Thamarangsi4,
  5. Naveen Agarwal4,
  6. Mark Parascandola5,
  7. Ravi Mehrotra6
  1. 1 Tobacco Free Initiative Unit, World Health Organization, New Delhi, India
  2. 2 Department of Community Medicine, Velammal Medical College Hospital and Research Institute, Madurai, Tamil Nadu, India
  3. 3 Healis—Sekhsaria Institute for Public Health, Mumbai, Maharashtra, India
  4. 4 Department of Noncommunicable Disease and Environmental Health, World Health Organization, New Delhi, India
  5. 5 Tobacco Control Research Branch, Behavioral Research Program, National Cancer Institute, Bethesda, Maryland, USA
  6. 6 Institute of Cytology and Preventive Oncology (ICMR), Noida, Uttar Pradesh, India
  1. Correspondence to Dr Dhirendra N Sinha, Regional Adviser, Surveillance (Tobacco Control), Tobacco Free Initiative Unit, World Health Organization, Regional Office for South-East Asia, I.P. Estate, New Delhi 110002, India; dhirendrasinha1{at}gmail.com

Abstract

Objectives To systematically review and meta-analyse the studies investigating the association between smokeless tobacco (SLT) use and all-cause mortality and cause-specific mortality outcomes among adult users of SLT and estimate the number of attributable deaths worldwide.

Methods Random-effects meta-analysis was used to estimate the pooled risk of death due to SLT use. Population attributable fractions were derived and used to calculate the number of attributable deaths. Observational studies published upto 2015 were identified through MEDLINE, IndMED, Google Scholar and other databases. Data on the prevalence of SLT use was obtained from latest reports or national surveys. Data on the disease burden were obtained from the Global Burden of Disease Study. Hospital-based or community-based case–control and cohort studies that adjusted for the smoking status were included.

Results 16 studies that provided estimates for mortality due to all cause, all cancer, upper aerodigestive tract (UADT) cancer, stomach cancer, cervical cancer, ischaemic heart disease (IHD) and stroke were included. A significant association was found for mortality due to all cause (1.22; 1.11–1.34), all cancer (1.31; 1.16–1.47), UADT cancer (2.17; 1.47–3.22), stomach cancer (1.33; 1.12–1.59), cervical cancer (2.07; 1.64–2.61), IHD (1.10; 1.04–1.17) and stroke (1.37; 1.24–1.51). Subgroup analysis showed major regional differences. Globally, the number of attributable deaths from all-cause mortality was 652 494 (234 008–1 081 437), of which 88% was borne by the South-East Asian region.

Conclusions SLT is responsible for a large number of deaths worldwide with the South-East Asian region bearing a substantial share of the burden.

  • Prevention
  • Non-cigarette tobacco products
  • Low/Middle income country
  • Global health
  • Advocacy

Statistics from Altmetric.com

Introduction

Conventionally, the computation of disease burden due to tobacco use has been largely based on smoking and secondhand smoke exposure and has often excluded the smokeless tobacco (SLT) component.1–3 This exclusion could be due to the fact that tobacco smoking is the most popular form of tobacco consumption in the world with widely acknowledged harms whereas the consumption of SLT is seen as being more common in the eastern hemisphere with ‘apparently’ unclear harms. However, any estimation of tobacco-related disease burden is incomplete without addressing the SLT component.

SLT use has become much more widespread than previous estimates and has been documented in more than 120 countries in every WHO region.4 The prevalence is also increasing in many countries such as Bangladesh and India, where it is one of the main forms of tobacco consumption.5 There has been a delay in the global response to address the SLT problem, in spite of the WHO Framework Convention on Tobacco Control frequently recommending equal attention to smoking and SLT control.6 Examples of such delay can be gleaned from the ad hoc amendments made to national tobacco laws in countries like Bangladesh, India and Nepal, which are some of the major consumers of SLT.

More and more harmful health effects caused by SLT are being recognised, from cancers, precancers, cardiovascular effects, adverse pregnancy outcomes, respiratory infections, addiction and poor oral health to early mortality.7–9 Unlike the smoking problem, the innumerable number of ways in which SLT products are manufactured, stored and consumed and the inadequacy of regulatory mechanisms make the control of SLT a particularly difficult public health problem.10–12 SLT is also widely exported through illegal routes between neighbouring countries, throughout the globe.13 These are some of the factors that make SLT a challenging but also an indispensable component of tobacco control efforts.

There are a number of reviews that have investigated the association between disease incidence or prevalence with SLT use.14–21 Although incident cases attributable to SLT provides a useful measure of burden, the attributable mortality is more relevant and robust because reliable cause of mortality data are more widely available across countries, especially with the availability of the Global Burden of Disease (GBD) databases. It becomes much easier to inform policy-level changes when discussions take into account the attributable mortality. There are a few reviews on SLT attributable mortality also22 ,23 but they are limited in the number of outcomes studied or the geographical region accounted for. This provided the justification for carrying out a much more comprehensive estimation of the SLT attributable mortality.

The aim of this systematic review was to examine all available evidence on the association between SLT use (I) and all-cause and selected cause-specific mortality outcomes (O) among adults using SLT products (P) as compared with non-users (C), across the world.

Materials and methods

A scoping literature review showed that SLT use has been investigated as a probable risk factor for several cancerous and non-cancerous mortality outcomes.7 ,24 From this list, we selected all-cause mortality, all-cancer mortality and mortality due to upper aerodigestive tract (UADT) cancer, stomach cancer, cervical cancer, ischaemic heart disease (IHD) and stroke. These outcomes were chosen based on the consensus achieved at the expert group consultation,13 which considered the availability of adequate number (a minimum of two estimates) of studies for each outcome, the need to address the bulk of the mortality burden and their relevance to the global population. A systematic review was then conducted for each of these outcomes separately using predefined study definitions and selection criteria.

Study definitions

The primary exposure of interest in this study was SLT use. SLT comprises a wide variety of products used globally (like snus, tobacco tooth powders, snuff, gutkha, khaini, tambaku, qiwam, dohra, kimam, tobacco powder, mawa and others) and all types were included for the current review. The main outcomes included for this study were all-cause mortality, mortality due to all cancers, mortality due to cancers of the UADT (including mouth, pharynx, larynx and oesophagus) and mortality due to IHD and stroke. The cause of death was recorded as reported by the study author or based on International Classification of Diseases (ICD) coding given in the study.

Literature search

Two authors (RAS and DNS) independently carried out the literature search. Disagreements on study inclusion, quality assessment and data extraction were resolved by deliberation or by the third author (PCG). We searched databases such as PubMed, IndMED, Google Scholar, reports of the WHO, government reports, Science Citation Index and Open Grey. Details of the study selection criteria, keywords and search strategy used are given in the web appendix boxes 1–3. Briefly, the inclusion criteria were framed in such a way that the effect size in each study represented the independent effect of SLT that is adjusted for confounders such as smoking. In addition to the PubMed strategy described, keywords were used in a variety of combinations for each outcome in Google Scholar and the first 50 pages were screened for relevant and non-duplicate articles. Similarly, various combinations of the keywords were used in each of the databases and the same process was repeated. The cross-references of all selected articles were scanned for additional studies. Attempts were made to retrieve grey literature such as unpublished data, dissertations and conference proceedings. To obtain publicly inaccessible data, a minimum of two email requests were sent to the corresponding author. If more than one article was published from a study, the article that provided the most updated data was selected. The last date of literature search was 31 December 2015.

Data extraction

By using appropriate critical appraisal checklists, each selected article was assessed for quality by two authors (RAS and DNS). Study characteristics such as first author, year of publication, date of data collection, country, study design, sample size, characteristics of cases and controls or cohort, methods of assessment of outcome and exposure, definitions of exposure, comparison groups and sex-wise risk estimates with 95% CI were extracted onto a precoded spreadsheet. Any conflicts in the data extracted were resolved by deliberation.

The data on the prevalence of SLT use for each country was obtained from the most updated published reports or surveys such as Global Adult Tobacco Surveys, Demographic Health Surveys, STEPS Risk Factor Surveys, the Special Europe Barometric Surveys and other national surveys within each country. The number of SLT users was calculated by multiplying the prevalence with age-matched population data. The data on the total disease burden due to a particular mortality outcome in terms of death and disability-adjusted life years (DALYs) were obtained from the database of the Global Burden of Disease Study (GBDS, 2010).25 The countries that had no prevalence data and/or burden data could not be included in the calculation of attributable burden.

Statistical analysis

The meta-analysis was performed using Stata software V.12.0. All meta-analyses were performed in logit scale due to their desirable statistical properties and by using the random-effects model.26 The effect size of interest was OR/relative risk for the effect of SLT use on the particular mortality outcome. Forest plots were drawn to describe the individual studies and the pooled effect size. Heterogeneity between the studies was examined using I2 statistics. An I2 value of >50% was considered as evidence of heterogeneity.27 Publication bias was assessed by visual inspection of the funnel plots and by means of Begg-Mazumdar's test and Egger's test. Asymmetry in the funnel plots and significant p values in the two statistical tests were considered to be evidences of publication bias. Statistical significance was set at a p value of <0.05.

In order to estimate the attributable burden due to SLT use, we used the population attributable fraction (PAF) method.28 All country-specific data were kept as close to the year 2010 as possible. Based on the pooled OR, SLT prevalence and total burden derived as described above, we calculated the PAF and the annual number of SLT attributable deaths/DALYs for the year 2010 for each country. For each country, the country-specific prevalence, country-specific total burden and region-specific pooled OR were applied. The attributable burden calculation was restricted to the ≥30 years age group to allow for a meaningful delay between SLT exposure and outcome development, assuming the average age of initiation of SLT use as 15 years. If a particular outcome was not significantly associated with SLT use in a region, then the attributable burden for that particular outcome for that region was not calculated. Our methodology relies on statistical significance of association in order to calculate the burden for a country. This choice was made to prevent inappropriately calculating the burden for a country where the evidence was not strong enough to justify the inclusion. If a country did not have region-specific pooled OR, then the global OR was applied. More details of the methodology are given in the web appendix.

This study has been conducted and reported in accordance with the PRISMA guidelines.29

Patient involvement

No patients were involved in setting the research question or the outcome measures, nor were they involved in the design and implementation of the study. There are no plans to involve patients in the dissemination of results.

Results

Prevalence of SLT use

SLT prevalence data were available for 133 countries, which translated in to about 350 million users worldwide. The vast majority (95%) of users were from the developing countries, especially the South-East Asia region (SEAR). There were marked differences in the type of data sources, the age groups included and the year of survey between the countries. Also, some countries had data for both sexes whereas some had data only for one sex or only for both sexes combined (see web table 1 and web figure 1a,b). Of these 133 countries, we could use the data of only 128 countries for which GBD estimates of mortality and DALY figures were available.

Qualitative synthesis

All the selected outcomes had at least two estimates at the end of the systematic review process (see web figure 2a–g). In total, 16 studies were selected for the analysis. All-cause mortality and IHD had the maximum number of eligible articles. The detailed description and quality assessment of each selected article is given in web tables 2a–g and 3. Briefly, most of them were population-based cohort studies, which was the most appropriate study design for determining the hazard of dying due to SLT exposure. Most of the studies documented the cause of death either by linking with national registers or by conducting verbal autopsy during house-to-house visits. In most studies, the cause of death was codified according to the ICD. Almost all the studies were carried out in India, USA and Sweden but no eligible study could be retrieved from the African region (AFR), Eastern Mediterranean region (EMR) and Western Pacific region (WPR). Studies from the European region (EUR) were mostly focused on IHD and stroke mortality rather than cancer mortality whereas studies from the USA and India were equally focused on all outcomes. Almost all the cohort studies included young to middle-aged participants (mostly ≥15 years) at entry and most European and American studies analysed the risk for men only, probably because of the low prevalence of SLT use among women. Except for two case–control studies, the sample size in all the cohort studies ran into tens of thousands of participants indicating adequate internal validity to make valid estimates of the mortality ratios, especially for rare outcomes such as cancer. In most studies, the OR was calculated for a comparison between the exclusive SLT user and non-tobacco user, which represented the best possible scenario to delineate the independent effect of SLT. Each study was unique in its handling of confounders and a variety of factors such as age, race, residence, wealth, education, occupation, body mass index, smoking, drinking, exercise, diet and others were considered. However, our inclusion criteria constrained all included studies to be adjusted for tobacco smoking. As far as the quality of studies is concerned, it has to be borne in mind that the studies included here were observational epidemiological studies and they bear the inherent biases associated with such designs (see web table 3). Nonetheless, these were the only study designs capable of delineating the association between SLT use and mortality outcomes within the ethical framework, which does not allow for randomisation of individuals to a known carcinogen. References of the studies selected for the meta-analysis are given in the web appendix.

Quantitative synthesis

Meta-analysis was conducted for seven separate mortality outcomes after the combinability of studies was deemed plausible. All-cause mortality had the largest number of individual estimates, followed by IHD, all cancer, stroke and others. At the global level, all the mortality outcomes were significantly associated with SLT use. The strongest association was found for UADT cancer (2.17; 1.47–3.22), followed by cervical cancer (2.07; 1.64–2.61), stroke (1.37; 1.24–1.51), stomach cancer (1.33; 1.12–1.59), all cancer (1.31; 1.16–1.47), all cause (1.22; 1.11–1.34) and IHD (1.10; 1.04–1.17; table 1, figure 1). The highest heterogeneity was noted for all-cause mortality (I2=92%), followed by UADT cancer and cancer mortality (∼70%) and other outcomes (<50%).

Table 1

Pooled estimates of mortality outcomes associated with smokeless tobacco use by WHO region

Figure 1

Forest plot for the meta-analysis of all-cause mortality attributable to smokeless tobacco use by WHO region. AMR, American region; ES, effect size; EUR, European region; SEAR, South-East Asian region.

Subgroup analysis

Region-wise analysis showed that not all outcomes were significantly associated with SLT use in every WHO region. All-cause, all-cancer and stroke mortality were not statistically significant in the EUR, UADT cancer and IHD mortality were not statistically significant in the American region and IHD mortality was not statistically significant in the SEAR (figure 1, web figure 3a–f and web table 4). There was no obvious evidence of publication bias for any of the outcomes according to the tests for heterogeneity and visual inspection of funnel plots (see web figure 4a–w).

We also carried out sex-wise analysis within each region. Only in the SEAR, an adequate number of studies were available for each sex to enable meaningful comparisons. The magnitude of difference in estimates was slight to moderate between men and women for all cause (1.21 vs 1.38), all cancer (1.42 vs 1.62), UADT cancer (2.16 vs 2.95), stomach cancer (2.00 vs 1.61), IHD (0.97 vs 1.13) and stroke (1.50 vs 1.27). A similar scenario was also observed with estimates at the global level (see web table 4 and web figure 5a–j). Subgroup analysis in terms of other variables such as confounder adjustment, study design or study quality could not be meaningfully carried out due to limited number of studies in the subgroup categories.

Sensitivity analysis

For each mortality outcome, we carried out sensitivity analysis by removing studies one by one from the overall estimate to identify any one study, which affected the estimate markedly but did not find evidence of any such outlier studies.(data not shown). Further details on the sensitivity analysis are given in the web appendix.

Attributable burden

Deaths

Globally, the number of deaths that could be attributed to SLT, due to all causes, was 652 494 (234 008–1 081 437). This estimate does not include the EUR because studies showed no statistically significant association between all-cause mortality and SLT use in this region. The SEAR bore the major proportion (88%) of this burden (table 2, figure 2). With regard to cause-specific deaths, the highest burden was due to stroke mortality (103 090; 45 787–162 636), followed closely by all-cancer mortality (101 004; 57 937–141 353). Both these estimates excluded the EUR due to the statistically non-significant association. The burden due to other causes, in the descending order of magnitude were UADT cancer (46 917; 20 580–75 874), stomach cancer (14 873; 7072–23 145), cervical cancer (10 583; 6742–14 810) and finally IHD (8677; 3068–14 897). The American region was not included in the calculation UADT cancer and IHD, whereas the SEAR was not included in the calculation of IHD attributable mortality. For all these outcomes, except IHD, the highest proportion of burden was borne by the SEAR but for IHD the highest burden was borne by the EUR (table 2, web table 5). Sex distribution of the burden showed that the proportion of deaths was higher among women for all cause, all cancer and IHD, in majority of the regions (see web table 6).

Table 2

Global and regional smokeless tobacco attributable all-cause and cause-specific mortality and DALY (in thousands) among adults aged ≥30 years in 2010

Figure 2

Global burden of absolute number of all-cause deaths attributable to smokeless tobacco use in 2010. This map is for representational purposes only and does not reflect the position of the authors or their affiliated organisations on the legal statuses of the boundaries between countries or territories.

Disability-adjusted life years

Globally, the attributable DALYs were estimated to be 45.9 million due to all causes and a major share (80%) of it was borne by the SEAR. The burden due to cause-specific outcomes in the descending order was 3.6 million for all cancer, 2.4 million for stroke, 1.5 million for UADT cancer, 420 000 for stomach cancer, 416 000 for cervical cancer and 209 000 for IHD. Similar to deaths, the major proportion of cause-specific DALYs was borne by the SEAR except for IHD, which was borne by the EUR (table 2). Also, proportion of DALYs was higher among women for all cause, all cancer and IHD (see web table 6).

Discussion

Lim et al 30 estimated that the attributable burden due to tobacco smoking (including secondhand smoking) was 6.3 million deaths and 156 million DALYs) globally in 2010 and that it was the one of the top three causes for global disease burden. However, their estimation did not include the SLT component but in order to understand the total burden due to tobacco products, it is absolutely essential to consider the burden due to SLT, which is consumed by about 350 million users worldwide.7 In this paper, we have attempted to provide an estimate of the all-cause and selected cause-specific mortality burden that can be attributed to SLT, globally.

By combining estimates made by Lim et al with ours, we found that SLT use contributed to 9% of all deaths and 23% of all DALYs that can be attributed to all forms of tobacco use, globally. However, in the countries of the SEAR where SLT is the major form of tobacco used as compared with smoking, these proportions could be higher. The estimation of disease burden attributable to a risk factor such as SLT at the global level provides a framework to enable policy-level discussions and set in motion the implementation of national control programmes aimed at this specific risk factor, to reduce tobacco attributable disease burden.

A recent study by Siddiqi et al 22 estimated the SLT attributable burden due to select outcomes such as oral, pharyngeal and oesophageal cancers and IHD. We differ from their methodology in some significant ways. Their meta-analysis included studies that measured the risk of having or developing the condition (ie, prevalence or incidence). On the contrary, we have used studies that have directly estimated the risk of dying from SLT use instead of the disease incidence/prevalence studies and therefore likely to produce more refined estimates of mortality burden. The number of studies that assess the SLT-associated mortality risk may be fewer than those that assess the risk of disease prevalence/incidence but using the former will give a more precise but conservative estimate of burden because mortality ORs tend to be smaller than incidence/prevalence ORs. For instance, Siddiqi et al estimated the attributable deaths due to UADT cancer as 62 283 (1.7 million DALYs) whereas our estimate was 46 917 (1.5 million DALYs, which unlike theirs also included nasopharyngeal and laryngeal cancer); whereas the former excluded countries of American subregion D, European subregion C and Western Pacific subregion A, we excluded the American region due to a lack of statistically significant association. Evidently, this wide variation was mainly due to the difference in the total number of countries included in the calculation and the pooled OR applied.

Many systematic reviews have been conducted on the association between SLT use and the risk of developing or having cancers, IHD and stroke but reviews, systematic or otherwise, on the risk of mortality due to SLT use were scarce.14–17 ,20 ,31–36 In that sense, our study has been a new addition to the body of evidence for the mortality-associated effects of SLT.

Regional differences

Examination of the region-wise results indicate that only certain outcomes were statistically significant in certain regions. It can be clearly seen that in the EUR, there was no significant association with mortality due to all cause, all cancer and stroke but a significant association was seen with IHD. Interestingly, we found no significant association with IHD mortality in the SEAR and the Americas, where we found significant associations with mortality due to all cause, all cancer, stomach cancer, cervical cancer and stroke. For UADT cancer there was a significant association only in the SEAR. SLT is not a single entity but encompasses a plethora of products, the composition of which varies from country to country.11 ,12 ,37 In regions where the SLT products would be expected to have a high content of carcinogens, their carcinogenic effects might be pronounced, and where they are high in the nicotine content with low amounts of carcinogens, their cardiac and vascular effects may be prominently expressed. Apart from this straightforward explanation, other factors like variations in baseline population risk, sex composition, manner and route of SLT consumption, SLT storage conditions, diagnostic methods and the level of healthcare availability might play a role in explaining the differences observed.

Credibility of the estimates

In the estimation of tobacco-related disease burden, the accuracy of estimates depends on a number of factors. Among the most important is the pooled OR obtained from the meta-analysis, which is affected by many confounding and competing risk factors like tobacco smoking, alcohol drinking and others. In order to reduce this effect, we designed our inclusion criteria to make sure that the OR from each study was adjusted for at least tobacco smoking. Another major concern with the meta-analysis was the high level of heterogeneity (I2>50%) seen for certain outcomes such as all-cause mortality, cancer mortality and UADT cancer mortality. Although we used the random-effects model to counteract the heterogeneity, the unexplained heterogeneity around the estimates does exist still. We did not estimate the burden in regions with non-significant associations as per the existing literature but future studies may change our understanding of that association and affect the estimates. We estimated the burden in regions such as WPR and EMR, where there was no literature, by applying the overall global pooled OR which might have introduced a degree of imprecision into the overall estimation. Among the outcomes considered here, the estimates for stomach and cervical cancers were derived from a single study carried out in India and therefore caution must be exercised during their interpretation. It should be understood that no casual inference is implied through these estimates, which can better be achieved through mechanistic studies.

SLT products vary widely in composition, from simple tobacco flakes packaged nicely into bite size packets commonly seen in Europe to crude point-of-sale reconstitution of several components called Paan in the South-East Asian countries. Every region has its distinct range of SLT products which reflect their long-standing traditions—for example, the several types of traditional snuff products (like Tawa, Neffa, Snuif) and Toombak in the AFR, Khaini, Gutka, Betel quid and Zarda in the SEAR and WPR, Rapé and Chimó in the South Americas, dry snuff, dissolvable tobacco, Iqmik and moist snuff in the USA, Tombol, Nas, Shammah and Zarda in the EMR and lastly the popular snus in the EUR. These products have different toxicological profiles and consequently are associated with varying risks of disease incidence and mortality but none of the products is safe to use. These variations in composition are complicated by the different ways of consumption, for example, apart from the most popular method of chewing, users may resort to snuff dipping, snorting, drinking tobacco water and using tobacco laced dentifrices. It is a challenge to estimate the risk of disease incidence or mortality attributable to such a heterogeneous risk factor. Any review involving SLT will be limited by these issues, unless a single product is studied such as the European snus or the Asian Naswar. An alternative solution would be to group studies by the type of product and apply this estimate regionally or globally but such risk and prevalence data by type of product are very rarely reported in literature, precluding this approach. In addition to these, SLT is also heterogeneous with reference to manufacture, processing, trade and regulations. The 2014 report on ‘Smokeless tobacco and public health: a global perspective’ extensively documents these concerns.7 The results of this review have to be interpreted with these concerns kept in context. In spite of these limitations, describing the harmful health effects of such a heterogeneous group of products under the umbrella term of ‘SLT’ provides a simple but effective way to inform tobacco control policies at the regional and national levels, appraise the lawmakers of the attributable burden and generate awareness among the population.

Strengths and limitations

Among the strengths of this study were the wide range of outcomes we included, the thorough literature review process and the worldwide coverage of the estimates. This was also the first study that estimated the all-cause mortality burden due to SLT use on global, regional and national levels. However, some important limitations should be borne in mind while interpreting the results. Most of the limitations of this paper were similar to the limitations detailed previously.21 First, although we investigated seven mortality outcomes here, we did not examine outcomes such as tuberculosis and respiratory deaths, which have been found to be associated with SLT use. Also, we did not investigate the burden due to other SLT-associated disease conditions such as non-fatal myocardial infarction, non-fatal stroke, oral precancerous conditions, oral health issues and others; a complete picture of the disease burden due to SLT has to include these conditions but were beyond the scope of this paper. Second, not all outcomes had a uniformly adequate number of studies or quality of evidence. Third, the lack of literature in every WHO region and the approximations carried out thereof affect the accuracy of the estimates. Some studies which were available in such regions could not be included due to ineligibility; for example, studies from Taiwan by Lin et al 38 and Lan et al 39 could have provided precise estimates for the WPR but could not be included because the former study estimated the risk from betel quid use and not SLT whereas the latter study was carried out in an exclusively elderly cohort. Even in regions where studies were available, not all the countries of the region had conducted association studies. But it would be reasonable to assume that what applied for one country would also apply for other countries in that region. This limitation highlights the need for more local-level association studies. We have calculated the attributable DALYs based on the PAF obtained from the mortality studies because studies directly estimating the DALYs due to SLT use are not available and therefore the related findings should be cautiously interpreted as an approximation. Finally, the definitions of outcomes were not uniform across the studies included in the systematic review and the data sources used such as the GBDS data, which could have led to misclassification errors. This may be especially so in the case of cancer mortality where the deaths in individual studies may not uniformly represent the entire range of cancer diagnoses. It was, however, not possible to predict their effect and has to be taken into account while interpretation.

Conclusion

The process of recognising SLT as an important component of the global tobacco control efforts has been slow and laborious. This was mostly because the literature implicating SLT was scattered, contradictory and rarely used to provide a quantitative assessment of the attributable burden. That is why in this study we have come to the evidence-based conclusion that a large number of deaths could be attributed to SLT use; in other words, SLT may be the single largest risk factor for intervention in the prevention of about 650 000 deaths, which is about 10% of all deaths that can be attributed to all forms of tobacco use, worldwide.

Recommendations

Although we have been able to provide burden estimates at the global level, several knowledge gaps still remain and more studies are required to highlight the effects of SLT, especially in the WPR, EMR and AFR. Studies of SLT products from the AFR have shown very high levels of carcinogens.7 Future burden estimations should also consider the non-fatal effects of SLT in their calculations.20 It would come as no surprise to tobacco researchers worldwide that SLT is responsible for such a large disease burden. Now that we have a robust estimate of the burden in concrete terms, it would be easier to convince policymakers to allocate additional resources for curbing this problem. For far too long, the tobacco control activities have ignored the SLT component. It has to be realised that without a planned strategy against SLT, the fight against tobacco will remain incomplete, especially in countries where SLT is woven in to the cultural fabric of the people. In contrast to cigarette smoking, the users of SLT are more likely to be socioeconomically disadvantaged40 ,41 and in order to protect them from further spiralling down into ill health, the governments should urgently focus their attention to tackle this harmful behavioural choice.

What this paper adds

  • Globally, the prevalence of smokeless tobacco (SLT) use has been increasing, while the smoking epidemic has showed signs of decline. There is still no widespread acceptance of the harmful effects of SLT use, in spite of the several studies generating evidence on this topic over the past several years.

  • There is a strong need to rigorously analyse the literature implicating the role of SLT in causing excess mortality and objectively quantify the number of attributable deaths, globally and regionally. Previous reviews on this subject have limited themselves to a small number of outcomes or to a specific country or region.

  • This study has showed that nearly 650 000 deaths due to all causes could be attributed to the use of SLT among adults, worldwide. This constituted about 10% of all the deaths that could be attributed to all forms of tobacco use, globally. About 88% of this burden was borne by the South-East Asian region alone. SLT was also found to be associated with specific causes of deaths such as all cancers and cancers of the upper aerodigestive tract predominantly in the eastern hemisphere, ischaemic heart disease in the European region and stroke in all the regions except Europe.

Acknowledgments

The authors thank Dr Deneshkumar V, Bio-statistician in VMCHRI, for helping with the data extraction procedures.

References

View Abstract

Footnotes

  • Twitter Follow Rizwan Suliankatchi Abdulkader at @sarizwan1986

  • Contributors RAS, DNS and PCG developed the concept and study protocol. RAS and DNS carried out the literature review and PCG served as the arbitrator. NA carried out prevalence data extraction. Analysis plan was developed by all authors and carried out by RAS. RAS wrote the first draft and DNS, PCG, TT, MP and RM provided critical inputs to the analysis plan, interpretation and manuscript writing. All authors read and approved the final manuscript.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Data of this study are available to anyone who wishes to perform additional analysis with any objective, from RAS, by sending an email to sarizwan1986@gmail.com.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.