Smoking prevalence and attributable disease burden in 195 countries and territories, 1990–2015: a systematic analysis from the Global Burden of Disease Study 2015

1Smoking was the second leading risk factor for early death and disability worldwide in 2015. It has claimed more than 5 million lives every year since 1990, and its contribution to overall disease burden is growing, especially in lower income countries. The negative effects of smoking extend well beyond individual and population health as billions of dollars in lost productivity and health-care expenditure are related to smoking every year. Successfully combatting the tobacco industry's pursuit of new smokers has been further complicated by the substantive—and sometimes rapid—social, demographic, and economic shifts occurring worldwide.,, As the tobacco industry moves to target previously untapped markets,,, strong tobacco control policies and timely monitoring of smoking patterns are imperative.

2Evidence before this studySmoking is a widely recognised risk factor for premature morbidity and mortality, but adequate monitoring of smoking levels and trends throughout the world has been challenging. Increasing investments in multi-country survey series has improved the availability of data for smoking behaviours, especially in lower income countries, but such surveys are quite infrequent and differences in survey questions and definitions can hinder appropriate comparisons between countries and across time. Through the Global Burden of Diseases, Injuries, and Risk Factors 2013 Study (GBD 2013), researchers collated diverse data sources and synthesised them to produce comprehensive, comparable estimates of daily smoking prevalence, by sex and age group, for 188 countries from 1990 to 2013. Additional analyses, including those by Bilano and colleagues in 2015, have applied similar methods to project trends in tobacco use through 2025 in 173 countries for men and 178 countries for women.Added value of this studyWith the 2015 update to the GBD, the number of data sources included was substantially increased and the estimation process for both smoking prevalence and attributable disease burden, as measured by disability-adjusted life-years (DALYs), has been improved. Two novel analyses are also provided through the GBD 2015 study: a birth cohort analysis of smoking patterns over time and a decomposition analysis to parse out changes in total DALYs attributable to smoking to changes in population growth, population ageing, smoking prevalence, and risk-deleted DALY rates. The latter assessment can assist with identifying what factors are contributing to changes in disease burden due to smoking–demographic trends, efforts to address smoking, or some combination of these factors. Further, we used the Socio-demographic Index (SDI), a new summary measure of overall development from GBD 2015, to assess levels and trends in smoking prevalence and attributable burden across the development spectrum.Implications of all the available evidenceAmid gains in tobacco control worldwide, smoking remains a leading risk factor for early death and disability. Although there have been some success stories, for many countries and territories, faster annualised rates of decline in smoking prevalence occurred between 1990 and 2005 than between 2005 and 2015. Although smoking prevalence and risk-deleted DALY rates fell across SDI quintiles, population growth and ageing ultimately offset these gains and contributed to overall increases in smoking-attributable disease burden in low to middle SDI geographies. Intensified tobacco control and strengthened monitoring are required to further reduce smoking prevalence and attributable burden, especially in view of the fact that demographic factors like population ageing are not easily amenable to intervention.

3Evidence before this study

4Smoking is a widely recognised risk factor for premature morbidity and mortality, but adequate monitoring of smoking levels and trends throughout the world has been challenging. Increasing investments in multi-country survey series has improved the availability of data for smoking behaviours, especially in lower income countries, but such surveys are quite infrequent and differences in survey questions and definitions can hinder appropriate comparisons between countries and across time. Through the Global Burden of Diseases, Injuries, and Risk Factors 2013 Study (GBD 2013), researchers collated diverse data sources and synthesised them to produce comprehensive, comparable estimates of daily smoking prevalence, by sex and age group, for 188 countries from 1990 to 2013. Additional analyses, including those by Bilano and colleagues in 2015, have applied similar methods to project trends in tobacco use through 2025 in 173 countries for men and 178 countries for women.

5Added value of this study

6With the 2015 update to the GBD, the number of data sources included was substantially increased and the estimation process for both smoking prevalence and attributable disease burden, as measured by disability-adjusted life-years (DALYs), has been improved. Two novel analyses are also provided through the GBD 2015 study: a birth cohort analysis of smoking patterns over time and a decomposition analysis to parse out changes in total DALYs attributable to smoking to changes in population growth, population ageing, smoking prevalence, and risk-deleted DALY rates. The latter assessment can assist with identifying what factors are contributing to changes in disease burden due to smoking–demographic trends, efforts to address smoking, or some combination of these factors. Further, we used the Socio-demographic Index (SDI), a new summary measure of overall development from GBD 2015, to assess levels and trends in smoking prevalence and attributable burden across the development spectrum.

7Implications of all the available evidence

8Amid gains in tobacco control worldwide, smoking remains a leading risk factor for early death and disability. Although there have been some success stories, for many countries and territories, faster annualised rates of decline in smoking prevalence occurred between 1990 and 2005 than between 2005 and 2015. Although smoking prevalence and risk-deleted DALY rates fell across SDI quintiles, population growth and ageing ultimately offset these gains and contributed to overall increases in smoking-attributable disease burden in low to middle SDI geographies. Intensified tobacco control and strengthened monitoring are required to further reduce smoking prevalence and attributable burden, especially in view of the fact that demographic factors like population ageing are not easily amenable to intervention.

9The past decade has brought a substantial expansion and strengthening of tobacco control initiatives, harnessing a wide range of effective interventions and policy instruments for addressing the tobacco epidemic.,,,,,,, Successful strategies include taxation of tobacco products, bans on smoking in public places and instituting smoke-free zones,, restrictions on the marketing and promotion of cigarettes, including plain packaging laws, community-wide and nation-wide smoking cessation interventions,, and enforcement of both text and pictorial warning labels on tobacco products., Efforts to implement comprehensive tobacco control policies culminated in the adoption of the WHO Framework Convention on Tobacco Control (FCTC) in 2003. The FCTC, the world's first public health treaty, is viewed as a key driver of recent progress in reducing tobacco consumption and smoking prevalence in many regions of the world. As of 2016, 180 parties have ratified the FCTC, and many use WHO's MPOWER measures, established in 2008, to guide national and local FCTC compliance. More recently, WHO introduced the 25×25 non-communicable disease (NCD) targets, which include decreasing tobacco use by 30% between 2010 and 2025. Several countries have committed to an even stronger anti-smoking goal, setting national targets to become tobacco-free. Additionally, strengthening FCTC implementation was explicitly included in the United Nations' Sustainable Development Goals (SDGs). With tobacco control's increasing prioritisation on the global stage, accurately monitoring patterns in smoking and associated health outcomes is critical for identifying optimal intervention strategies across geographies, demographic groups, and the development spectrum.

10Previous analyses of smoking prevalence and attributable disease burden often were hindered by poor data availability, methodological limitations, or both.,, Investments in survey series focused on tobacco, such as the Global Adult Tobacco Surveys (GATS) and the Global Youth Tobacco Surveys (GYTS), have supported more in-depth assessments of national tobacco use. Nonetheless, remaining data gaps across countries and time, as well as differences in smoking-related questions and definitions among available data sources, necessitated large analytical improvements to produce a systematic and consistent understanding of smoking patterns. As part of the Global Burden of Diseases, Injuries, and Risk Factors 2013 Study (GBD 2013), Ng and colleagues generated the first comprehensive, comparable estimates of smoking prevalence and tobacco consumption for 188 countries from 1980 to 2013. Since then, other studies have used similar data synthesis approaches to project smoking trends from 2010 to 2025 in 173 countries for men and 178 countries for women. Previous GBD studies, have assessed the contribution of smoking to overall disease burden through the comparative risk assessment framework developed by Murray and Lopez. Recent studies have quantified the global effects of tobacco on achieving NCD mortality targets and life expectancy, while several assessed smoking-attributable mortality and non-fatal health outcomes for specific locations.,

11In this analysis, we assess smoking prevalence and smoking-attributable disease burden, based on deaths and disability-adjusted life-years (DALYs), by sex and age group for 195 countries and territories from 1990 to 2015. We also investigate differences in smoking prevalence and attributable burden according to the Socio-demographic Index (SDI), a summary measure of income per capita, educational attainment, and total fertility rate. Additionally, we assess age and sex patterns by birth cohort across levels of development. Finally, we perform a decomposition analysis of potential drivers of smoking-attributable disease burden over time.

12This study follows the overall GBD 2015 comparative risk assessment framework, details of which have been previously published. Here we summarise the main steps in the estimation process; the appendix provides more details about data inputs and modelling strategies (pp 5–9). This study fully adheres to the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER).

13Improving upon methods used by Ng and colleagues, we calculated two exposure measures: prevalence of daily smoking of tobacco and the smoking impact ratio. We defined a daily smoker as an individual using any type of smoked tobacco product on a daily basis., We used 2818 data sources, covering 2928 geography-years of data, identified through the Global Health Data Exchange (GHDx), WHO InfoBase Database, and International Smoking Statistics Database; the appendix provides additional details on data sources (pp 5, 6). For any data that did not match our exposure definition we adjusted for frequency of use or type of tobacco consumed to avoid potential biases. We adjusted for smoking frequency and type simultaneously, which allowed us to account for their mutual correlations with each other (appendix pp 7, 8). Second-hand smoke exposure is estimated separately in GBD and is not included in this analysis.

14We generated estimates of smoking prevalence by sex and 5-year age groups starting at age 10 years. Any data that spanned multiple age groups or were reported for both sexes combined were split based on the age-sex patterns recorded from data sources with multiple age-sex groupings. We then used spatiotemporal Gaussian process regression (ST-GPR), a data synthesis method widely used in GBD, which allowed us to draw strength across geography, time, and age, incorporate both data and model uncertainty, and produce a full time-series of estimates for all 195 geographies. The appendix provides full details on the modelling strategy (pp 5–9).

15The second exposure measure, the smoking impact ratio, was first described by Peto and Lopez as part of a method to estimate smoking-attributable burden in the absence of information about smoking patterns. The smoking impact ratio is defined as the population lung cancer mortality rate in excess of the background lung cancer mortality rate recorded in non-smokers in the population, relative to the excess lung cancer mortality rate recorded in a reference group of smokers. We computed the smoking impact ratio for each analytic unit using the geography-specific, year-specific, age-specific, and sex-specific population lung cancer mortality rates from GBD 2015, and reference group lung cancer mortality rates from prospective cohort studies (appendix p 9).

16We assessed all available evidence that supported causal associations between smoking and 38 health outcomes using a systematic approach adapted from Hill's criteria for causation and the World Cancer Research Fund evidence grading schema (appendix p 9). We added seven new outcomes to those used in GBD 2013: larynx cancer, peptic ulcer disease, rheumatoid arthritis, cataract, macular degeneration, hip fracture, and non-hip fracture.

17We used 5-year lagged smoking prevalence in estimating smoking attributable burden for cardiovascular diseases, tuberculosis, diabetes, lower respiratory infections, asthma, cataracts, macular degeneration, fractures, rheumatoid arthritis, and peptic ulcer disease. We chose a 5-year lag based on findings showing that most risk-reduction occurs within 5 years of quitting smoking. We used the smoking impact ratio in estimating smoking- attributable burden for cancers, chronic obstructive pulmonary disease (COPD), interstitial lung disease, other chronic respiratory diseases, and pneumoconiosis. The appendix provides a complete list of outcomes and their associated exposure metric (pp 31, 32).

18For each outcome included in this analysis we used relative risk estimates derived from prospective cohort studies comparing smokers to never smokers (appendix p 9). Population attributable fractions were calculated based on estimates of exposure, relative risks, and the theoretical minimum risk exposure level for smoking (zero smoking). Following population attributable fraction calculation, we multiplied estimates of deaths and DALYs by outcome-specific population attributable fractions, and then summed them across all 38 outcomes to compute overall disease burden attributable to smoking (appendix p 9).

19We captured and propagated uncertainty through all steps of the analysis, including sampling uncertainty from data extraction, uncertainty from models used to adjust data reported in non-standard frequency-type combinations, uncertainty in the ST-GPR model, and uncertainty in deaths and DALYs for the 38 included outcomes. Ultimately, we produced 1000 draws of exposure and attributable burden estimates, for each geography, year, age, and sex, from which 95% uncertainty intervals (UIs) were taken using the 2·5 percentile and 97·5 percentile of the distribution.

20To parse out the drivers of changes in smoking-attributable DALYs from 2005 to 2015, we assessed the relative contribution of four factors: population growth, population age structure, risk-deleted DALY rates, and smoking exposure. Risk-deleted rates are defined as the DALY rates that would have been recorded had we removed smoking as a risk factor. We estimated risk-deleted DALY rates by multiplying the observed cause-specific DALY rates by one minus the cause-specific population attributable fractions. For the decomposition analysis, we used the methods developed by Das Gupta (appendix p 10).

21We present results aggregated by level of SDI, a composite indicator of development estimated for each geography based on lag-distributed income per capita, average educational attainment among individuals over age 15 years, and total fertility rate. SDI values were scaled to a range from 0 to 1. The appendix provides SDI values for each geography (pp 21–25).

22The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

23Worldwide in 2015, the age-standardised prevalence of daily smoking was 25·0% (95% uncertainty interval [UI] 24·2–25·7) in men and 5·4% (5·1–5·7) in women (table 1). 51 countries and territories had significantly higher prevalence of smoking than the global average for men, and these countries were located mainly in central and eastern Europe and southeast Asia (figure 1). For women, 70 countries, mainly in western and central Europe, significantly exceeded the global average. Among men, prevalence of daily smoking was highest in middle SDI countries, whereas for women high SDI countries had the highest prevalence of daily smokers (figure 2). Compared with other SDI levels, low SDI geographies generally had the lowest prevalence of daily smoking for both sexes.

24Between 1990 and 2015, the global age-standardised prevalence of daily smoking fell significantly for both sexes, decreasing by 28·4% (95% UI 25·8–31·1) for men and 34·4% (29·4–38·6) for women (table 2). 13 countries (Australia, Brazil, China, Denmark, Dominican Republic, Iceland, Kenya, the Netherlands, New Zealand, Norway, Sweden, Switzerland, and the USA) recorded significant annualised rates of decline both between 1990 and 2005 and 2005 and 2015, suggesting sustained progress in tobacco control (table 1). 18 countries showed a faster annualised rate of reduction in daily smoking in the most recent decade compared with between 1990 and 2005. Focusing on the most recent decade, since 2005, 53 (27%) of 195 countries and territories recorded significant decreases in age-standardised prevalence of male daily smoking, whereas only 32 (16%) recorded significant reductions for women.

25In 2015, there were 933·1 million (95% UI 831·3–1054·3) daily smokers in the world, 82·3% of whom were men (768·1 million [690·1–852·2]). The ten countries with the largest number of smokers together accounted for 63·6% of the world's daily smokers. China, India, and Indonesia, the three leading countries in total number of male smokers, accounted for 51·4% of the world's male smokers in 2015. On the other hand, the USA, China, and India, which were the leading three countries in total number of female smokers, accounted for only 27·3% of the world's female smokers. Together, these results suggest that the tobacco epidemic is less geographically concentrated for women than for men.

26Among the ten countries with the largest number of total smokers in 2015, seven recorded significant decreases in male smoking prevalence and five had significant decreases in female smoking prevalence since 1990 (table 2). Of these countries, Brazil recorded the largest overall reduction in prevalence for both male and female daily smoking, which dropped by 56·5% (51·9–61·1) and 55·8% (48·7–61·9), respectively, between 1990 and 2015. Indonesia, Bangladesh, and the Philippines did not have significant reductions in male prevalence of daily smoking since 1990, and the Philippines, Germany, and India had no significant decreases in smoking among women. All of the three countries with female age-standardised smoking prevalence less than 3·0% (China, India, and Bangladesh) succeeded in keeping smoking prevalence low in women. Notably, female prevalence of daily smoking significantly increased in Russia and Indonesia since 1990 (table 2).

27Delving into the smoking patterns of adolescents can shed light on trends in smoking initiation. Between 1990 and 2015, the global prevalence of daily smoking for this age group significantly decreased for each sex, falling from 16·1% (95% UI 14·4–18·0) to 10·6% (9·3–12·1) for men and from 4·8% (4·3–5·6) to 3·0% (2·6–3·7) for women (table 2). Despite global decreases, several countries still had a high prevalence of smoking among individuals aged between 15 and 19 years. In 2015, there were 22 countries with female smoking prevalence in this age group higher than 15·0%, 18 of which were located in western or central Europe. Countries with high male smoking prevalence were much more dispersed. Of the 24 countries with male smoking prevalence higher than 20·0%, six were in eastern Europe, and the remainder were spread across ten other regions (appendix pp 13, 14). The rank of countries with the largest smoking populations for the 15–19 years age group was mostly consistent with the rank for all-age smoking populations (table 2).

28Although no country had a significant increase for men or women in this age group since 2005, only three countries saw smoking prevalence in 15 to 19 year-olds significantly drop for both men and women since 2005 (New Zealand, Iceland, and the USA). Iceland had the largest significant decrease among men, decreasing from 14·8% (95% UI 11·7–18·5) in 2005 to 9·0% (5·6–13·3) in 2015. New Zealand had the largest significant decline among women, decreasing from 20·8% (18·1–23·8) in 2005 to 12·5% (10·1–15·5) in 2015 (available to view through GHDx).

29Parsing out daily smoking prevalence by age group and birth cohort allows for a more fine-grained examination of smoking prevalence, age patterns, and temporal trends by level of development (figure 2). Male age patterns of smoking were fairly consistent across levels of SDI, with prevalence generally peaking between the ages of 25 and 35 years. For women, however, age patterns varied more by SDI; female smoking prevalence typically peaked around age 25 years for high and high-middle SDI countries, while prevalence generally increased until age 60 years in low to middle SDI countries. Across birth cohorts, smoking prevalence decreased by age group, sex, and SDI level. The most notable decreases were recorded in high and high-middle SDI countries for men, where sizeable reductions in smoking prevalence in 15 to 24 year-olds occurred across birth cohorts. Middle SDI countries, which have the highest levels of daily smoking among men, had minimal changes in prevalence across birth cohorts, suggesting far less progress in curbing smoking initiation or promoting cessation. For women, prevalence is consistently lower than in men; nevertheless, reductions in smoking prevalence across birth cohorts generally were smaller than those recorded for men.

30In 2015, 6·4 million deaths (95% UI 5·7–7·0) were attributable to smoking worldwide, representing a 4·7% (1·2–8·5) increase in smoking-attributable deaths since 2005. More than 75% of these deaths were in men, and 52·2% took place in four countries (China, India, the USA, and Russia). Smoking was the second-leading risk factor for attributable mortality among both sexes in both 2005 and 2015, following high-systolic blood pressure. The relative ranking of smoking-attributable disease burden, as measured in DALYs, increased from third to second between 2005 and 2015. In 2015, there were 148·6 million (95% UI 134·2–163·1) smoking-attributable DALYs worldwide, and smoking was the leading risk factor for attributable disease burden in 24 countries, an increase from 16 countries in 1990 (figure 3). Further, smoking was ranked among the leading five risk factors for 109 countries in 2015. Between 2005 and 2015, only Egypt recorded a significant increase in the age-standardised smoking-attributable mortality rate among both sexes, increasing by 11·4% (95% UI 0·3–24·7) over that time period. On the other hand, 82 countries had significant decreases in their age-standardised smoking-attributable mortality rates since 2005.

31Overall, in 2015, cardiovascular diseases (41·2%), cancers (27·6%), and chronic respiratory diseases (20·5%) were the three leading causes of smoking-attributable age-standardised DALYs for both sexes. Of all risk factors, smoking was the leading risk factor for cancers and chronic respiratory diseases, but only the ninth leading risk factor for cardiovascular diseases. The appendix shows the 30 leading causes of DALYs attributable to smoking, including changes over time (pp 19, 20). For women, the leading cause of smoking-attributable DALYs was COPD, whereas the leading cause for men was ischaemic heart disease.

32Relative to changes in smoking exposure, the main drivers of overall changes in attributable burden due to smoking varied by both sex and SDI level (figure 4). Since 2005, all-cause DALYs attributable to smoking for men decreased by 11·8% (95% UI 10·0–13·9) in high-SDI countries, the only SDI level with a significant decrease in attributable burden for men. For women, only middle-SDI countries had a significant reduction in all-cause DALYs attributable to smoking (a 22·6% decrease [9·0–32·8]) between 2005 and 2015. In both instances, a combination of reduced smoking exposure and reduced risk-deleted DALY rates contributed to overall reductions. Conversely, all-cause burden due to smoking significantly increased in low SDI and low-middle SDI countries since 2005 for men. This rise in attributable DALYs was driven mainly by a combination of population growth and population ageing for both sexes. In women, while rising exposure to smoking has resulted in increased DALYs due to smoking for low-middle SDI countries, this increase was not significant. Generally, population growth was the leading factor for increasing attributable burden due to smoking among the low SDI countries between 2005 and 2015. For countries of middle to high SDI, more pronounced sex differences emerged. For instance, decreases in male smoking prevalence propelled an overall reduction in attributable burden for high SDI countries, whereas changes in smoking exposure had minimal effects on overall burden for women at similarly high levels of SDI.

33A complete dataset of all results by geography, year, sex, and age group can be downloaded through GHDx, and an interactive data visualisation of smoking prevalence results can be found online.

34Despite more than half a century of unequivocal evidence of the harmful effects of tobacco on health,, in 2015, one in every four men in the world was a daily smoker. Prevalence has been, and remains, significantly lower in women—roughly one in every 20 women smoked daily in 2015. Nonetheless, much progress has been accomplished in the past 25 years. Specifically, the age-standardised global prevalence of daily smoking fell to 15·3% (95% UI 14·8–15·9), a 29·4% (27·1–31·8) reduction from 1990, with smoking rates decreasing from 34·9% (34·1–35·7) to 25·0% (24·2–25·7) in men and from 8·2% (7·9–8·6) to 5·4% (5·1–5·7) in women. These reductions were especially pronounced in high SDI countries and Latin America, probably reflecting concerted efforts to implement strong tobacco control policies and programmes in Brazil and Panama, among others.

35Yet amid these gains, many countries with persistently high levels of daily smoking recorded marginal progress since 2005, and smoking remained among the leading risk factors for early death and disability in more than 100 countries in 2015, accounting for 11·5% of global deaths (95% UI 10·3–12·6) and 6·0% (5·3–6·8) of global DALYs. Smoking patterns diverged by geography, level of development, sex, and birth cohort, emphasising the need for tailored approaches to change smoking behaviours. Although male smoking prevalence still far exceeded that of female smokers in 2015, the most pronounced reductions in smoking prevalence since 1990 were generally found for men—and more places saw minimal changes or increases in smoking among women. These trends highlight how the tobacco epidemic, and corresponding industry forces, has expanded beyond a male-centred health challenge.

36Low to middle SDI countries saw increased disease burden attributable to smoking since 2005, a trend that that occurred despite variable decreases in smoking prevalence and risk-deleted DALY rates. Population growth or ageing, or a combination of both, ultimately contributed to increased disease burden attributable to smoking in these countries. In higher SDI countries, population growth and ageing offset the potential for larger gains in places where notable declines in smoking prevalence and risk-deleted DALY rates occurred. This finding points to a crucial challenge ahead for tobacco control: unless progress in reducing current smoking and preventing initiation can be substantially accelerated, demographic forces, which are far less amenable to immediate intervention, are poised to heighten the disease burden associated with smoking's global toll.

37Since 2005, the year when the FCTC entered into force, it has redefined global, regional, and national approaches to tobacco control and policy., Case studies point to the successful uptake and enforcement of FCTC components in many countries with especially prominent reductions in smoking prevalence. Pakistan, Panama, and India stand out as three countries that have implemented a large number of tobacco control policies over the past decade and have had marked declines in the prevalence of daily smoking since 2005, compared with decreases recorded between 1990 and 2005.,, At the same time, many countries, including Australia, Brazil, Canada, South Korea, and the USA, among others, achieved sizeable declines in smoking prevalence well before FCTC adoption.,,, Altogether, 18 countries recorded a faster annualised rate of decline from 1990 to 2005 than from 2005 to 2015.

38Brazil, which has achieved the third largest significant decline in age-standardised smoking prevalence since 1990, is a noteworthy success story. Brazil accomplished this reduction through a combination of tobacco control policies that began with advertising restrictions and smoking bans in some public places starting in 1996 and culminated with Brazil achieving the highest level of achievement in all MPOWER measures except for monitoring by 2011. Policies were comprehensive and were supplemented with fiscal interventions that included raising taxes and establishing minimum prices for tobacco products. Finally, Brazil has achieved high levels of compliance through enforcement.,,,,

39Critics of the FCTC argue that the treaty's effectiveness may be limited in various settings, especially since compliance has lagged in many countries.,, The FCTC, while necessary and vital for creating the policy environment for more effective tobacco control worldwide, is not sufficient to fully address each country's tobacco control needs. Rather, countries will need to both implement FCTC-stipulated measures and supplement such policies and programmes with strong enforcement and high rates of compliance. For example, India, where 11·2% of the world's smokers live, supplemented the Cigarettes and Other Tobacco Products Act (COTPA) with the creation of a National Tobacco Control Programme (NTCP) in 2007. NTCP was created to strengthen implementation and enforcement of the various provisions of COTPA at the state and district level. It has been rolled out in phases and currently covers about 40% of all districts in India.

40Despite concerted efforts to control tobacco around the world, there remain a number of countries where current levels and recent trends raise concern. For example, Indonesia, a country with very high levels of smoking, particularly among men, has not yet ratified the FCTC and scores very poorly on the MPOWER indicators. Also, in Russia, prevalence among women has been increasing, and, until recently, there were very few laws related to tobacco control. Russia passed a comprehensive tobacco control policy in 2014 and has the potential to achieve progress on tobacco control. As a region, eastern Europe has seen a statistically significant increase in smoking prevalence among women since 1990. Increases among women, along with a sustained high prevalence of male smokers, can be linked to tobacco industry targeting during the 1990s. The tobacco industry is now turning its focus toward emerging markets in sub-Saharan Africa, seeking to exploit the continent's patchwork tobacco control regulations and limited resources to combat industry marketing advances., Given the large effects of population growth and ageing on smoking-attributable disease burden—and Africa's rapidly changing demographic profile—a renewed dedication to strong, proactive tobacco policies and monitoring will be vital for the continent.

41The 2030 agenda features tobacco control as a key component to sustainable development, with SDG Target 3.a calling for stronger FCTC implementation. Nonetheless, the utility and potential impact of the SDGs on tobacco control may be hindered by the vagueness of Target 3.a (“Strengthen the implementation of the WHO FCTC in all countries, as appropriate”) and absence of defined targets for reducing smoking prevalence by 2030. Ultimately, to move all countries toward stronger tobacco control by 2030, improvements in policy formulation, enforcement and compliance, and the routine monitoring of smoking behaviour are urgently needed. Without valid and reliable data, these efforts risk being more aspirational than grounded in evidence-informed action. Multi-country survey series have substantially improved data availability on smoking prevalence, yet the disadvantages associated with such surveys—high cost, time lags, inconsistent questions across survey series, sample restrictions for young populations, and a reliance on self-reported smoking behaviour—necessitate the development of robust, locally focused, timely, objective, and low-cost methods of tracking smoking trends. Supplementing surveys with biomarker collection is essential because self-reported smoking prevalence is believed to be severely underestimating true smoking prevalence,,,, especially in population subgroups or places where tobacco use may not be culturally acceptable.

42Our findings should be interpreted taking into consideration the study's limitations. First, our exposure estimation focused on smoked tobacco and did not include smokeless tobacco products and e-cigarettes. Second, our definition of smoking exposure pertained to current daily smokers, and did not include occasional or former smokers, which might underestimate the attributable disease burden to smoking, especially in populations who tend to be less likely to smoke every day, such as women, children and young adults, and individuals with less disposable income. Third, we did not account for the intensity or duration of smoking. Fourth, the study relied on self-reported data, and it is possible that reporting biases varied across countries and over time. Fifth, for long-term effects of smoking on cancers and chronic respiratory diseases, we used the smoking impact ratio method, which estimates the lifetime cumulative effect of cigarette smoking using the proxy of recorded lung cancer mortality rates. This method provides robust estimates of the burden of cancers and chronic respiratory diseases related to tobacco but is not fully consistent with the GBD approach of estimating exposure independently of the outcomes affected by exposure. Also, the smoking impact ratio method is based on the cumulative effect of cigarette smoking rather than all types of tobacco smoking, and might be less robust for geographies in which non-smoker lung cancer might be significantly affected by air pollution or other factors. Sixth, our estimates of DALYs are probably underestimates because relative risk values used for estimating population attributable fractions might not fully represent all possible risk-outcome pairs experienced by sex, age group, and over time. Also, burden estimates did not account for the effect of both indoor and outdoor air pollution potentiating risks. Finally, minimal risk-outcome data were available for populations younger than 30 years, and therefore burden attribution was limited to age groups 30 years and older.

43Despite more than 50 years of anti-tobacco efforts, smoking remains a leading global risk factor. Its toll will remain substantial without more concerted policy initiatives, policy compliance and enforcement, and sustained political will to offset commercial interests. Despite progress in some settings, the war against tobacco is far from won, especially in countries with the highest numbers of smokers. The staggering toll of smoking on health echoes well beyond the individual, especially as tobacco threatens to exact long-term financial and operational burdens on already resource-constrained health systems. To significantly and permanently bend the global tobacco epidemic's trajectory, a renewed and sustained focus is needed on comprehensive tobacco control policies around the world. Success is possible, but requires effective and aggressively enforced policies and laws. Intensified efforts are also greatly needed to keep smoking prevalence rates low in populations which have not experienced a devastating epidemic yet, and to prevent children, adolescents, and young adults from starting to smoke.

44Correspondence to: Dr Emmanuela Gakidou, Institute for Health Metrics and Evaluation, University of Washington, Seattle, WA 98121 gakidou@uw.edu

45Correspondence to: Dr Emmanuela Gakidou, Institute for Health Metrics and Evaluation, University of Washington, Seattle, WA 98121 gakidou@uw.edu