Smoking and smokeless tobacco use in nine South and Southeast Asian countries: prevalence estimates and social determinants from Demographic and Health Surveys

1In 2010, globally, 54% of Disability adjusted Life Years (DALYS) were caused from non-communicable diseases (NCDs) and tobacco smoking including second-hand smoke was one of the leading risk factors for global disease burden accounting for 6.3% of global DALYS. If the current trend of tobacco use continues, it could cause approximately 8.3 million deaths annually by the year 2030, and more than 80% of them may occur in low- and middle-income countries (LMICs), where nearly two-thirds of the world’s smokers live. In 2012, there were an estimated 967 million smokers from 187 countries, with the highest burden of tobacco use in high-income countries (HICs), intermediate in middle-income countries, and lowest in low-income countries. The deaths attributable to tobacco use are 18%, 11%, and 4% respectively. Increasing rates of smoking in many LMICs and decreasing rates in HICs may lead to increased proportional tobacco-related mortality in LMICs,.

2The Southeast Asia region is home to nearly 400 million tobacco users, who experience about 1.2 million deaths annually. Although smokeless tobacco (SLT) use is common among women, smoking among women is increasing,. In Southeast Asia, tobacco is used in diverse forms, including cigarettes or bidis (dried tobacco rolled in paper or leaf), SLT such as chewing khaini (tobacco with slaked lime and aromatic spices), surti (dried tobacco leaves for chewing), or paan masala (tobacco with aromatic spices), sucking gutkha (mixture of tobacco and molasses available in small sachets), applying gul or gudaku as dentifrice, and inhaling nas and naswar (nasal inhalation of tobacco powder). Bidis are popular in Bangladesh, India, Maldives, Nepal, and Sri Lanka, whereas cheroots are popular in Myanmar, and roll-your-own cigarettes (in palm leaves or paper) are popular in Thailand and Timor Leste. In Bangladesh, India, and Nepal, use of gul, gudaku, mishri, masher, lal dantamanjan as dentrifice, and nas/naswar is common. SLT use in various forms is directly responsible for oropharyngeal cancers,.

3Socio-economic differentials in tobacco use have existed in both developed, and developing countries-. Studies have reported that tobacco consumption rates are higher in lower socioeconomic classes and less-educated groups,,. Moreover, smoking prevalence is lower among women worldwide, particularly in South and Southeast Asia. Therefore, assessing socio-economic differentials of tobacco use in Southeast Asia by population-based surveys will provide information about effectiveness of tobacco control measures and aid policymaking. The global tobacco surveillance system, World Health Organization (WHO) STEPS program, and WHO World Health Surveys (WHS) have provided such information. However, these surveys cover several countries from various regions, but not all of the countries in a region. Moreover, the literature from these surveys has emphasized smoking and reported determinants of tobacco use,,, but not about SLT use, which is prevalent in South and Southeast Asia. Demographic and Health Surveys (DHS) collect information about tobacco use in nationally representative samples of men and women and have provided national estimates of tobacco use for Nepal, India,, sub-Saharan Africa, and other countries. We aimed to provide national estimates on prevalence and social determinants of tobacco smoking and smokeless tobacco use in South and Southeast Asian countries.

4We used data from nationally representative samples of women and men from DHSs conducted between 2005–2006 and 2012–2013 in nine South and Southeast Asian countries: India, Pakistan, Nepal, Bangladesh, Maldives, Indonesia, Philippines, Cambodia, and Timor Leste (Table 1). DHS aims to provide reliable indicators about fertility, family planning, health, and nutrition of populations in developing nations,. DHSs were implemented by country-level statistical offices or other local institutions and are technically supported by ORC (Opinion Research Corporation) Macro International Inc. of Calverton, Maryland, USA and financially supported by the United States Agency for International Development (USAID).

5The final samples of households in DHSs were selected by two stages, stratified random sampling using population proportionate to size technique to include both rural and urban residents. Heads of each selected household answered general questions about the household and listed the household members. All women aged 15 to 49 years and men aged 15 to 49 years or more (up to 54 years in India, Indonesia, and Bangladesh and up to 64 years in Maldives) who were the usual residents were eligible to participate. Trained interviewers collected information about demographic and socio-economic factors and health status. Questions about tobacco use were asked of all eligible men and women,, except in Bangladesh and the Philippines (Table 1).

6The following four identical questions were asked to elucidate information about tobacco use in all countries, but response options varied between countries (see below).

71) Do you currently smoke cigarettes? (response as ‘yes’ or ‘no’)

82) In the last 24 hours, how many cigarettes did you smoke? (response as numerical)

93) Do you currently smoke or use any other type of tobacco? (response as ‘yes’ or ‘no’)

104) What (other) type of tobacco do you currently smoke or use? (options provided were pipe, chewing tobacco, snuff, country-specific options, and others)

11Additional country-specific options given were the use of hookah (sheesha), bidi, and cigars in Maldives; bidi in India, Pakistan, Nepal and Bangladesh; cigars in Philippines; hand-rolled tobacco in Timor Leste; Pan Masala and Gutkha in India; and nuswar in Pakistan. Each respondent was classified as ‘current smoker’ if the response to the first question was ‘yes,’ the response to the fourth question was ‘pipe,’ or if they were using hookah, bidi, cigars, or hand-rolled tobacco. The respondents were classified as ‘current SLT user’ if the response to the fourth question was any form of SLT, including ‘chewing’ tobacco, ‘Pan Masala’, ‘Gutkha,’ and ‘snuff.’

12Age, religion, and marital status were reclassified for logistic regression analyses. Religion was categorized as Hindu, Islam, Roman Catholic, Buddhist, and others. Each religion was classified into two categories, i.e., main religion of the country and others, except in India (Hindu, Muslim, and others). Age of the participant was recoded as 15–29, 30–39, and 40-49/≥40 years (for women and men, respectively). Marital status was classified as being married or single. Single constituted being never married, separated, or divorced. A cohabiting partner was included under single as its proportion was very small in most countries. In DHS, place of residence was classified as rural or urban; educational level was classified as ‘no education’, ‘primary’, ‘secondary,’ or ‘higher.’ Household wealth index, considered a reliable proxy for household economic status, was calculated based on a standard set of household assets, dwelling characteristics, and ownership of consumer items as observed by the interviewer. Participants were ranked on the basis of their household score by dividing them into quintiles where the first quintile was the poorest 20% of the households and the fifth quintile was the wealthiest 20%.

13The institutional review boards of ORC Macro International Inc. and participating institutions in each country provided ethical clearance for DHSs. In each survey, participants were informed about voluntary participation and confidentiality of information and could refrain from responding to any of the questions. Before each interview, details of the survey were explained and informed consent was obtained. Written consent was not obtained since no intervention was applied to the participants.

14All analyses were done for men and women separately in each country. Descriptive analyses were done for smoking and SLT use. Overall weighted prevalence estimates for tobacco smoking and SLT use were calculated by including sample weights to account for the complex sampling design of DHS. Weighted prevalence estimates of smoking and SLT use were calculated according to age groups, religion, place of residence, marital status, education, and wealth quintiles. Binary logistic regression analyses were done to assess demographic (age was entered as continuous variable) and socio-economic factors associated with smoking and SLT use by SVY command on STATA/IC version 10. Beta-coefficients, their 95% confidence intervals, and p-values were calculated.

15The sample size of women interviewed was greater than for men since DHS mainly aimed to estimate indicators related to mother and child health (Table 1). In the 2008 Philippines DHS, men were not sampled, whereas in 2007 Bangladesh DHS, women were not asked about tobacco use. In Pakistan, Indonesia, and Maldives, information about religion was not collected (Table 2). Both men and women were mainly from rural areas in Maldives, Cambodia, Nepal, and Timor Leste, whereas in India, Pakistan, and Indonesia urban–rural distribution was nearly equal. More than half of men and women were aged between 15 and 29 years in all countries except Bangladesh, Maldives, and Indonesia, where about 45% of men were aged ≥40 years. Most participants were educated through primary and secondary level, except in the Philippines, where a third of women had received higher education. Nearly a third of men were uneducated in Bangladesh and Maldives. A high proportion of women (56.2% to 17.1%) were not educated in all countries except Indonesia (3.6%) and the Philippines (1.6%). Two-thirds or more of both men and women were married, but in Pakistan, Bangladesh, Maldives, and Indonesia ≥95% of men were married. Participants were evenly distributed across all wealth quintiles except in Maldives, Indonesia, and Timor Leste where the richest quintile was the smallest.

16Among men, weighted prevalence of smoking varied between the countries; the highest prevalence was found in Indonesia (72.3%), followed by Timor Leste (69.5%), Bangladesh (60.0%), and Maldives (47.3%), but prevalence was lower in India (34.1%), Nepal (33.6%), Cambodia (34.7%), and Pakistan (31.6%). Prevalence of SLT use among men also varied between countries, with the highest prevalence in India (36.7%), followed by Nepal (34.8%) and Bangladesh (21.4%), and the lowest in Indonesia (0.46%) and Timor Leste (2.5%) (Figure 1). Among women, weighted prevalence of smoking was much lower than men in all countries; the highest prevalence was in Nepal (9.8%), followed by Philippines (5.2%), Maldives (4.6%), and Pakistan (4.02%). Prevalence of SLT use among women was highest in India (9.0%), followed by Cambodia (5.1%), Nepal (4.8%), and Maldives (4.2%) (Figure 1). The most common form of tobacco consumed was cigarettes in all countries except India and Timor Leste. Cigarettes/bidis were smoked in India, and hand-rolled cigarettes were smoked in Timor Leste. However, Indian men and women used diverse forms of SLT including ‘gutkha,’ ‘pan masala,’ and other chewing tobacco (unspecified). Chewing tobacco was also common in Nepal, Bangladesh, and Cambodia. Indonesian men mostly smoked cigarettes, while women there were also using chewing tobacco. Smoking a pipe/cigar was only seen among Filipino and Nepalese women, while women from Maldives and Pakistan mostly smoked hookah (Figure 2).

17Differentials in smoking and SLT use according to socio-economic and demographic variables are shown separately for men and women in Tables 3 and 4, respectively. Prevalence of both smoking and SLT use among men was higher in rural areas than urban areas in all countries (for example, among Indian men 36.8 vs. 29.5 and 39.9 vs. 31.3, respectively). Smoking and SLT use among women was higher in rural areas in India, Pakistan, Nepal, and Cambodia, but these differentials varied for prevalence of smoking and SLT use among women in other countries. Prevalence of smoking and SLT use among both men and women varied according to wealth quintiles and educational level. In all countries among men and women, prevalence of smoking and SLT use was generally highest among the least educated and the lowest among those with higher education. Prevalence of tobacco use was highest among the poorest and lowest among the richest. For example, among Nepalese men, prevalence rates for smoking and SLT use in those with no education vs. the highest education was 57.9 vs. 20.5 and 57.8 vs. 18.4, respectively. Among Bangladeshi men prevalence rates for smoking and SLT use in the poorest group vs. the richest was 70.1 vs. 46.8 and 26.6 vs. 15.7, respectively. The only exception was seen in Maldivian men, where differentials were very small for prevalence of smoking (poorest vs. richest was 47.6 vs. 52.9) (Table 3). Prevalence of both smoking and SLT use was higher among married men and women in all countries except the Philippines, Pakistan, Indonesia, and Maldives. Prevalence of both smoking and SLT use was higher among men and women aged ≥40 years in all countries, except for smoking among men and women in Maldives and SLT use among Pakistani and Indonesian men. Among Indian, Pakistani, Nepalese, and Maldivian women differentials by age were very high. For example, among Maldivian women, prevalence of smoking and SLT use in the 15–29 age group vs. those 40–49 years old was 1.1 vs. 10.4 and 0.2 vs.10.6, respectively (Table 4).

18The association of tobacco use among men and women with socio-economic and demographic factors was assessed using binary logistic regression analysis. Men from rural areas had greater likelihood of smoking and SLT use in India only. Women from rural areas had greater likelihood of smoking in all countries except Nepal and Cambodia and SLT use in India, Cambodia, and Timor Leste (Table 5). Older men were more likely to smoke and use SLT in most countries (except for SLT use in Indonesia and Timor Leste). Older men were less likely to smoke in Maldives (β = −0.03) and Indonesia (β = −0.01).Older women were more likely to smoke and use SLT in all countries (β coefficients ranged from 0.04 to 0.12, Table 5). Married men were more likely to smoke and use SLT in all countries except Bangladesh (for both smoking and SLT use), Maldives, Indonesia, Cambodia, and Timor Leste (for SLT use only). Married women were more likely to smoke in the Philippines but were less likely to smoke in Indonesia and Maldives. Married women were more likely to use SLT in Nepal only (Table 5). Smoking and SLT use among both men and women were strongly associated with education (protective effect) in all countries, except for SLT use among men from Bangladesh and Timor Leste. Individuals who were educated were less likely to smoke or use SLT. Smoking and SLT use among men was associated with wealth in all countries except Nepal, Pakistan, Timor Leste, and Maldives, while the association of wealth with smoking and SLT use among women was seen in all countries except Timor Leste (for smoking). Wealthier individuals were less likely to smoke or use SLT. Smoking among men was associated with religion in India, Nepal, Cambodia, and Timor Leste, while SLT use among men was associated with religion in India and Nepal. Smoking among women was associated with religion in India, Philippines, and Cambodia, while SLT use among women was associated with religion in India and Nepal (data about religion was not collected in Indonesia and Maldives).

19Our report using DHS datasets provided national-level estimates and information about the pattern of tobacco use in nine countries in the South and Southeast Asia region. Our disaggregated analyses by gender and type of tobacco use demonstrated that pattern of tobacco consumption has cross-country and intracountry variations. In each country, tobacco consumption among men and women was unequally distributed in all demographic and socio-economic groups. Tobacco use among women was very low in all countries, but smoking was higher in Nepal and SLT use higher in India than other countries. Prevalence of smoking and SLT use among men was almost equal in India and Nepal, but among Bangladeshi men, smoking was higher than SLT use. Prevalence of smoking among men was very high in Indonesia, Timor Leste, and Maldives while SLT use was very low. In all countries, significant associations between age, education, and wealth for both smoking and SLT use highlights the existence of social disparities in tobacco use.

20Prevalence estimates were comparable to DHS-based estimates for India and Nepal and were much higher than estimates for India and Nepal in GATS,, but prevalence in Cambodia was lower compared to another national survey. Only three (India, Bangladesh, and the Philippines) of the nine countries that had also participated in the first wave of GATS did not allow comparison of prevalence in all GATS countries,. Moreover, our estimates cannot be compared with those of GATS and WHS, which defined current smoking as smoking of any form of tobacco either daily or occasionally,, while the Global Burden of Disease (GBD) study defined daily smoking as smoking any type of tobacco product at least once per day. Our estimates for the current smoking rate among men in Indonesia and Timor Leste were approximately 70%, whereas GBD reported rates as <61%. Similarly, among Nepalese women, our estimate for current smoking was 10%, but GBD results put daily smoking rate at 17%. Differences as large as these between our estimates and those from GBD may have occurred due to differences in data sources, definition of smoking, and statistical methods used for estimating prevalence rates. GBD used multiple sources of microdata and aggregate data from major multicountry survey programs such as DHS, GATS, WHS, WHO STEPwise Approach to Surveillance program, and others, national-level, multiyear surveys on health, addiction, and risk factors, and three large health databases. To estimate prevalence of daily smoking, GBD analyzed data from the sources listed above for 2,102 country-years from 181 countries for a 33-year period, while we used cross-sectional samples of men and women from one DHS in each of the nine South and Southeast Asian countries for 2005–2013. GBD defined a daily smoker as someone who smokes any type of tobacco product at least once per day, but we defined current smoker as someone who responded as “yes” to the question, “Do you currently smoke cigarette?”. Moreover, GBD adopted robust statistical techniques such as regression analyses to adjust for data providing varying or non-standard definitions for smoking, but we relied on participants’ response to three questions to define current smokers. GBD estimates are more realistic, as the study used a spatial-temporal regression model and Gaussian process regression to create a complete time-series for all data from multiple sources, followed by computation of age-standardized prevalence rates, while our study computed the weighted point prevalence of current smoking among men and women, which was not age-standardized. Such methodological heterogeneity of tobacco surveys and the importance of more systematic design of surveys, questionnaires, and definitions have been previously underscored,. Comparison with DHS-based reports for India and Nepal shows an increase in prevalence of tobacco use,. Low prevalence of smoking and SLT use among women as reported earlier, is not surprising since it is socially unacceptable for women to smoke in South Asian communities. However, SLT use was common in India, Nepal, and Cambodia, confirming the results of previous studies,.

21Our findings that current smoking is prevalent from the age of 15 years onwards but was higher in older age groups which is similar to results of previous surveys-,. This may be due to a cohort effect (i.e., smoking was less likely to be initiated in more recent decades). This means that more attention should be paid to young men in Indonesia, Maldives, and Bangladesh, where prevalence rates of smoking were alarmingly high. Some think that tobacco companies have been aggressively marketing to young people in these countries, particularly in Indonesia. A protective effect of education on smoking, after controlling for other factors, was consistent with results of previous studies-. However, association of smoking with wealth index was consistent for women in most countries, but not for men, highlighting that smoking behavior may be context-specific, needing country-level analysis like that seen in the GATS report but not in other reports from WHS and DHS. In developed countries, the smoking epidemic began among the rich and educated and later spread to lower socio-economic groups, but in developing countries the less educated may have taken up smoking, due to lack of awareness about health risks. Lack of association of wealth index with smoking in some countries could be explained by parental influences, peer pressure, and cultural acceptance of smoking. Significantly higher rates of smoking among urban residents have been reported,; in our study a higher prevalence of tobacco use among rural residents was not significant for men (with multivariate analysis). We found that socio-demographic factors associated with SLT use were increasing age, lower education, and poverty among both men and women, which is similar to determinants of current smoking in our study and studies from India, Nepal, and Bangladesh, which analyzed SLT use separately. These findings are also similar to multicountry surveys that reported social determinants of tobacco use or smoking only,.

22An advantage of using DHS data was that DHS included large representative samples of men and women and allowed cross-country comparison of tobacco prevalence. Our analysis has revealed, albeit with some limitations, that comparable estimates of current tobacco use can be obtained from DHS. However, future DHS in more than 85 LMICs should include more questions about previous tobacco use and quit attempts to provide data for monitoring of the global tobacco epidemic,. Although GATS intends to include more countries in future surveys, DHS cover most LMICs where the majority of tobacco-attributable deaths occur. Our analysis fulfills the need for assessing the social disparities in tobacco use, and studying the pattern of tobacco use in South and Southeast Asia, which already has a high burden of communicable and nutrition-related diseases. Urgent actions are necessary to reduce the burden of NCDs. If nothing is done, health inequalities health among the socially disadvantaged may widen further.

23There are some limitations of estimates from DHS data due to its design and the questions asked about tobacco use. First, the sample of men and women aged 15–49 years leads to under-estimation, if men and women aged >49 years had higher rates of tobacco use. Second, from the limited questions asked we could only estimate current use, unlike WHS and GATS, which have provided insights into other stages of smoking behavior such as never user and former user. Third, participants were asked to quantify tobacco use by cigarettes smoked during the last 24 hours, but other forms of smoking (bidi, pipe, hookah) and SLT use were not quantified. Fourth, the associations with social factors lack temporal relationships, as DHS were cross-sectional studies. Lastly, in conservative Asian societies it is very likely that tobacco use based on self-report may be under-reported, and DHS did not verify this by measuring urinary cotinine levels.

24Our study provides information about prevalence and patterns of tobacco use among men and women in South and Southeast Asian countries not covered in other multicounty surveys and confirms that tobacco use was higher among men, the less educated, and the poor, particularly those living in rural areas. Policymakers should consider the diverse forms of tobacco used and social distribution in each country to provide context-specific tobacco prevention and control strategies and target vulnerable groups. Policymakers need to consider SLT use separately in tobacco control efforts, since the economic and health effects of SLT use are different from that of smoking.

25The authors declare that they have no competing interests.

26CTS was principal author who conceptualized the manuscript, performed data analysis, interpreted the results, and wrote the results and discussion sections of the manuscript. PMS assisted in conceptualization of the manuscript, wrote the background and methods sections, and revised the final drafts of the manuscript. IAM assisted in conceptualization of the manuscript, assisted in data analyses, prepared the tables and graphs, and revised the final draft of the manuscript. SS assisted in conceptualization of the manuscript, assisted in revised data analyses, prepared the revised tables and results section, and revised the initial version of the manuscript. All authors read and approved the final manuscript.