A nationally representative study on socio-demographic and geographic correlates, and trends in tobacco use in Nepal

Tobacco control still poses an immense challenge for the government of Nepal. Updated knowledge on the current pattern of tobacco use and its associated factors will be helpful for policy makers to curb the tobacco epidemic. This study fills this gap by, (i) exploring demographic, socio-economic and geographic correlates of current tobacco use using a nationally representative sample of 15–49-year adults from Nepal Demographic Health survey 2016, and (ii) examining the prevalence and trends of both smoking and non-smoking forms of tobacco use in a nationally representative sample of 15–49-year adults drawn from three consecutive Demographic Health Surveys (DHS) between 2006 and 2016.Among males, the prevalence of smokeless tobacco use was higher than that of smoking (40.1% and 27.4% respectively), whereas among females smoking was more common than smokeless tobacco use (prevalence of 5.5% and 3.8% respectively). Both smoking and smokeless tobacco use were associated with older age and lower level of education. Among males, those living in urban areas were more likely to consume any form of tobacco. Residents of terai/plains were more likely to use smokeless tobacco. The concentration curves on cumulative proportion of tobacco use ranked by wealth quintiles showed tobacco use to be highest among the lowest socio-economic groups in both males and females in all three survey years. We found a decreasing trend of tobacco smoking and an increasing trend of smokeless tobacco use over the 10-year period. However, the consumption of both forms of tobacco increased in young males during the same period. Proper monitoring of adherence to directives of the anti-tobacco law should be ensured to curb the increasing burden of tobacco use among young males, and a similar effort is needed to sustain the decline in tobacco uses among other population groups in Nepal.


Results
Demographic and socioeconomic correlates of tobacco use in 2016. The overall prevalence rate for any form of tobacco use was higher in males (52.3%) than in females (8.4%). Similar differences were also reported for tobacco smoking (27.4% in males and 5.5% in females) and SLT use (40.1% in males and 3.8% in females). The unadjusted analysis showed that young males and females below 25 years had the lowest prevalence of tobacco use (Table 1). These findings correspond with adjusted analysis which found age to be a significant predictor of tobacco use. Compared to adolescent males and females aged 15-19 years, other age groups had higher adjusted odds of using all forms of tobacco (Table 2). Similarly, in unadjusted analysis males and females with no formal education had the highest prevalence of any form of tobacco use (77% for male and 18% for females) ( Table 1). The adjusted analysis also found lower educational status to be significantly associated with the use of any form of tobacco. Males with no formal education were 2.86 times (CI 1.92-4.17) more likely to smoke tobacco than males with higher education. Conversely, females with no formal education were 6.67 times (CI 2.  more likely to smoke tobacco than their counterparts with higher education ( Table 2).
In unadjusted analyses, the prevalence of tobacco smoking (34.2%) was highest for the men from the poorest households whereas the women from poorest households had highest prevalence for tobacco smoking (12.9%) as well as smokeless tobacco use (8.2%) and any forms of tobacco use (18.9%). After adjustment, odds of any tobacco use decreased with increasing wealth for both men and women, although the trend appears to be more marked for women. For instance, women from poorest households were four times (CI 2.38-7.14) more likely to use any form of tobacco than women from richest households (Table 2). Similarly men from poorest households were two times (CI 1.41-2.78) more likely to use any form of tobacco than men richest households. Media exposure particularly television, internet and newspaper/magazine showed no significant relationship with any kind of tobacco use. However, women who listened to the radio at least once a week had lower odds (AOR 0.69, CI 0.54-0.88) of using SLT compared to women who did not listen to the radio ( Table 2).
Geographic correlates of tobacco use in 2016. In unadjusted analysis, prevalence of any form of tobacco use among males was 51.5% in urban areas and 53.9% in rural areas. Similarly, prevalence of any form of tobacco use among females was 7.6% in urban areas and 9.7% in rural areas (Table 1). Multivariable analysis showed that males from urban areas had lower odds of using any form of tobacco, including both smoking (AOR 0.78, CI 0.63-0.96) and SLT (AOR 0.80, CI 0.66-0.98) than their counterparts from rural areas. On the contrary, females showed no such significant association of urban-rural residence with tobacco use (Table 2).
In unadjusted analysis, prevalence of tobacco smoking was highest among males in Province 7 (40.2%) and highest among females in Province 6 (13.1%; Table 1, Fig. 1). But after adjustment, males in Province 3 (AOR 1.59, CI 1.08-2.36) and Province 7 (AOR 1.65, CI 1.18-2.32) were more likely to smoke tobacco than males in Province 1 (most advantaged in political and economic means). Similarly females in Province 3 (AOR 3.25, CI 2.09-5.06), Province 6 (AOR 3.02, CI 1.97-4.64) and Province 7 (AOR 2.67,  were more likely to smoke tobacco than females in Province 1 ( Table 2). In unadjusted analysis, SLT use was the highest among males in Province 2 (58.5%) and among females in Province 1 (7.6%). After adjustment, there was no significant difference in the likelihood of SLT consumption between males in Province 2 and Province 1 but females in province 1 had highest likelihood of SLT use compared to all other Provinces. The adjusted analysis also showed males in Province 3 (AOR 0.60, CI 0.40-0.90) and Province 4 (AOR 0.69, CI 0.49-0.98) were less likely to consume SLT than males in Province 1. Women in Province 2 (AOR 0.26, CI 0.14-0.50), Province 3 (AOR 0.29, CI 0.16-0.51) and Province 6 (AOR 0.31, CI (0.19-0.52) were less likely to consume SLT than women in Province 1 ( Table 2). The highest prevalence of any form of tobacco use was reported from Province 2 (62.2%) for males and from Province 6 (16.7%) for females (Table 1, Fig. 1). In unadjusted analysis, the disparity was also observed according to the ecological region where the use of SLT among males were highest in Terai region (50.7%) compared to other ecological regions (<30%) ( Table 1). In adjusted analysis, males from Terai region had nearly two times higher odds (AOR 1.93, CI 1.20-3.08) of SLT use compared to males from Mountainous region ( www.nature.com/scientificreports www.nature.com/scientificreports/ trends in tobacco use in Nepal. The overall prevalence of tobacco consumption has remained steady among males since 2006. However, the prevalence of tobacco smoking among males has declined whereas consumption of SLT has increased in the recent years (Fig. 2). There has been continuous decline in tobacco smoking and modest decline in SLT use among females over the last decade (Fig. 2). By age group, the prevalence of tobacco smoking declined most prominently among both males and females aged 35 and above. However, the prevalence of SLT increased in males aged 35 and above (Fig. 3). There was a negative gradient in tobacco uses across the wealth quintiles in both males and females (Fig. 4). In males, the bottom 20% constituted nearly 22% of tobacco consumers while the middle 60% constituted nearly 63%, and the top 20% constituted nearly 15% of tobacco users in 2016. (Fig. 4a). However the disparity was more evident in females, the bottom 20% constituted nearly 43% of tobacco users and the top 20% constituted nearly 7% of tobacco users in females (Fig. 4b). The adjusted analyses for 2016 also showed similar results with the odds ratios decreasing across quintiles for increasing wealth, both for men and women.

Discussion
This sub-analysis of the data collected by a nationally representative survey (NDHS 2016) provided the current status of tobacco use and its distribution among adults aged 15-49 years in Nepal. The overall prevalence of tobacco use was 52.3% in males and 8.4% in females. Non Communicable Diseases (NCD) risk factors survey 2013 reported lower prevalence in males (48.1%) but higher prevalence in females (14.1%). This could be attributable to difference in the sampling design between these surveys or perhaps is a true change in prevalence since 2013 13 . Our results showed marked gender difference in tobacco consumption which has been similarly reported from other South Asian countries [14][15][16] . We found that the distribution of tobacco consumption in Nepal varied not only across the population subgroups but geographic regions as well. The increasing prevalence of tobacco smoking by age, similar to that reported in previous demographic health surveys 2006 17 and 2011 18 , is a reminder of the historical burden of tobacco use in Nepal. Around 36.7% of adolescent males used to smoke tobacco products in Nepal in 2001 19 . These males could have continued smoking as evident from the similar prevalence of tobacco smoking (30.8%) in males aged 30-34 years in 2016. It shows the failure of aggressive tobacco control program targeting the younger generations. The higher prevalence of tobacco consumption among the lesser educated population subgroup regardless of age-groups accentuates the inequities in distribution of this important risk factor for chronic diseases. For example, males with no formal education were 2.86 times (CI 1.92-4.17) and females with no formal education were 6.67 times (CI 2.38-20) more likely to smoke tobacco than those with higher education. The higher burden of tobacco consumption among lesser educated population was also seen in previous demographic health surveys 9,10 .
Tobacco use was highest amongst the lowest socio-economic groups in both males and females. However, the disparity in burden of tobacco consumption between rich and poor is narrowing in males whereas widening in females since 2006. The increase in tobacco taxes might have compelled the poor to quit tobacco, contributing to reduction in disparity in tobacco use among males in these recent years. The aftermath of people's movement in 2006 brought a paradigm shift in socio-political sector giving women more autonomy and an equal status as men in the social and political sector 20 . This gender empowerment along with erosion of cultural and economic constraints might have favored tobacco use among women from low socioeconomic groups 21 . However the change in socio economic inequities over time might be different from that depicted by concentration curves because the concentration curve analysis doesn't adjust for other variables. The significant association observed between tobacco use and place of residence (urban/rural) for males might be due to the fact that social and peer pressure, is particularly high among young males from urban areas, and drives them to initiate tobacco use at early ages. After adjusting for other variables, the likelihood of tobacco smoking was highest amongstmales in Provinces 3 and 7 and amongst females in Provinces 3, 6 and 7.The high likelihood of tobacco smoking in males in province 7 corroborates with the highest likelihood of tobacco smoking reported in Far western region of Nepal in previous DHS 9 . However, the high likelihood of tobacco smoking among males in Province 3 does not corroborate with likelihood of tobacco smoking among males in central region reported in previous DHS 9 . The adjusted analysis is however not available for females from previous demographic health surveys to make such comparisons. Similarly, highest prevalence of SLT use among men was reported in Province 2 (Eastern Terai).
Noticeably, the tobacco smoking prevalence has been increasing among adolescent males while declining among females of the same age group (Fig. 2). In Nepalese context, smoking is considered as an acceptable behavior for men but not for women. So, under-reporting of tobacco smoking in adolescent girls in urban areas is very likely. The increasing prevalence in tobacco smoking among adolescent males is driven by thrill seeking or risk-taking behavior in this age group. To put this other way round, an increasing prevalence highlights the weaknesses of existing tobacco control program in Nepal to reach the young masses who are critically vulnerable to the aggressive branding and corporate sponsorships of sports and musical events. Such strategy employed by tobacco companies to evade marketing restrictions by targeting alternative advertising avenues is of particular concern. Additionally, WHO's Tobacco Assessment Report highlighted Nepal's weakness in governing advertisements from international TV and radio stations 22 23 . The adaptation by tobacco industries to changed regulations can be seen with the aforementioned details of expenditures into market research, a higher denomination in sponsorship and a lower total in actual advertising. This will make the youth more vulnerable as the industries will seek respite on corporate sponsorships and improve public imagery of their brands through heavy donations into public and social-enterprises which the country has failed to monitor and govern 24 Table 2. Correlates of tobacco smoking, smokeless tobacco use and any tobacco use.   www.nature.com/scientificreports www.nature.com/scientificreports/ the government has been unable to fully implement graphic health warnings that cover 90% of the package of smoked and SLT products in Nepal. This regulation of 90% graphic health warnings is the largest in the region 25 .
The consumption of SLT (40.1%) is even higher than tobacco smoking (27.4%). The use of SLT is increasing among adolescent males, males aged 35 and above over the last decade (Fig. 2). This increasing trend might be due to shift in preference from tobacco smoking to SLT usage due to high taxes on smoking tobacco products and low taxation on SLT products 26 . There has been substantial growth in the import of chewing tobacco products (khaini, gutka, zarda) from India over the years in Nepal. In addition, Nepal has also seen a rampant growth of illegal production of SLT within the country in recent times 27 . However, there is a lack of comprehensive information on production, sales and revenue collection of SLT products in Nepal 24 . SLT use was found to be more prevalent in Terai areas closer to the border with India 28,29 . In contrast, residents of mountains and hills were more likely to be tobacco smokers. The higher burden of SLT use in Terai, and tobacco smoking in mountainous and hilly regions were also evident in previous demographic health surveys 9,10 . Hence, SLT control efforts need to be considered separately in order curb its increasing consumption.
In today's society, the media plays a crucial role in shaping people's understanding and attitude towards a certain issue. Exposure to mass media has been reported to cause increase in tobacco smoking 30 and decline in SLT use 31 in various studies. In this study, there was no significant association between television and internet use and likelihood of any forms of tobacco use. However, for radio use, frequent listeners were significantly less likely to consume any forms of tobacco, and particularly smokeless tobacco.
In Nepal, tobacco products can still be purchased in the shops and consumed even in public places 6,32 . This poses an immense challenge for the effective implementation of anti-tobacco law. We need further researches exploring the adherence to directives of the anti-tobacco law by the vendors especially outside the schools and colleges in Nepal. This would be the starting step in addressing the growing burden of tobacco use among the adolescents. Investing in tobacco control measures (tobacco tax increases, smoke-free policies, package warnings, and advertising bans, awareness campaigns) have proven to be cost effective in terms of saving health care costs and human productivity 33,34 . In addition, integrating tobacco cessation in routine primary health care settings also needs to be considered for those who desire to quit smoking 24 . Implementing these various tobacco control measures would require interdisciplinary and inter-sectoral collaborative endeavors among professionals and institutions that have traditionally worked separately.
Graphic health warning (2011-present) and way forwards for plain packaging. Introduction of graphic health warning in tobacco products was started in 2011 with 75% coverage, and it was subsequently increased to 90% in 2014 35 . On the positive side again, the government of Nepal is planning to introduce plain www.nature.com/scientificreports www.nature.com/scientificreports/ packaging by the end of 2018. New tobacco products will not have any color, imagery, corporate logo and trademark 36 . However, there is a mismatch between the tobacco control efforts and monitoring of compliance of packaging rules. One such example is the use of sticker over the tobacco graphic warning (Fig. 5). Further, a proper monitoring of tobacco control in the changed scenario relied excessively on public system with no independent assessment and documentation. Detail examination on the effect of graphic warning and proposed plain packaging in tobacco use is necessary in future, especially among the young population in Nepal who are particularly vulnerable to the tobacco uses (any product). Consequently, the increasing tobacco smoking among adolescent males is a tale-tale sign of the relative inefficiency in the tobacco control program to target the young masses, and could also be due to better reporting of tobacco use in recent surveys.
Based on a large national sample that consists of both urban and rural populations in Nepal, findings of this study may be utilized for monitoring tobacco epidemic and assessing the prevalence among population subgroups at country-level. However, as in any cross-sectional study, we could not establish causal association between tobacco consumptions and its determinants. In a conservative society of Nepal, it is very likely that tobacco use based on self-report may be under-reported and needs to be verified by measuring urinary cotinine levels in future studies. We don't have information on prevalence of tobacco use among those aged 50 years and over as NDHS has limited the age of males and females respondents from 15-49 years only. Though prevalence rates of tobacco consumption is declining among Nepalese women, very little is known about the prevalence and determinants of exposure to secondhand smoke from male smokers. This dimension of exposure to secondhand tobacco smoke needs to be considered in future demographic health surveys. Finally, our study informs the implementation of Tobacco Product Control and Regulatory Act 2010: reduction in SLT use and focus on adolescent males should become the priority for tobacco control efforts in Nepal.

Conclusion
More than half of the Nepalese males and around 8% of Nepalese females aged 15-49 years were reported to have consumed tobacco. People from lower socioeconomic status and less education were more likely to consume tobacco. This study highlights a need for a greater vigilance on implementation of Tobacco Product Control and Regulatory Act 2010, targeting young males. www.nature.com/scientificreports www.nature.com/scientificreports/ Methods Data sources. We used data from a nationally representative Nepal Demographic Health Surveys carried out in 2006, 2011 and 2016 for this study. These surveys used a stratified two-stage cluster sampling in rural areas and three stages sampling in urban areas. Primary sampling units (wards) were selected with probability proportional to their size in the first step. Then, households for interviews were selected based on equal probability systematic selection criterion. A detailed study methodologies and description of survey variables for each survey year has been presented in respective DHS reports 17,18,37 . These surveys collected information about demographic factors, socio-economic factors and health status from all members aged 15 and above in the sample of households selected. Subsequently, all females aged 15-49 who stayed in the households the night before the survey were interviewed. In every alternate household selected, all males aged 15-49 years, who stayed in the households the night before the survey, were interviewed for tobacco using practices.
We investigated socio-demographic and geographic correlates only for 2016 as such analysis for 2006 and 2011 were already conducted in previously published papers 9,10 . All the three surveys were conducted using same methodology and collected information on the same variables.
Variables. Information about tobacco use was obtained from the following questions that were asked in both male's and females's questionnaires of DHS: 1. Do you currently smoke tobacco every day, somedays, or not at all? 2. In the past, have you ever smoked tobacco every day, some days, or not at all? 3. Do you currently use smokeless tobacco every day, some days, or not at all? 4. On average, how many of the following products do you currently smoke? (Multiple response: Manufactured cigarettes/hand rolled cigarettes/pipes full of tobacco/cigars, cheroots or cigarillos/water pipe/any other).
For our analysis, dichotomous outcome variables were created from the information obtained from above questions. Any respondent answering yes to question 1, was categorised as "Tobacco smoker". Similarly any respondent answering yes to question 3, was categorised as then "Smokeless tobacco user". "Any form of tobacco user" in this study is defined as any person consuming cigarette/pipe/bidi/other smoking forms which were derived from the responses to question 1 and 3 above. If a respondent answers yes to either of question 1 or 3, then, he/she was categorised as any form of tobacco user.
Explanatory variable. We selected three categories of explanatory variables: socio-demographic, geographic and access to information. The socio-demographic variables were age of the participants, sex, educational status and wealth index. Educational status was grouped into four categories: no formal schooling, attended primary schooling, attended secondary schooling and higher (includes high school, college/university, and post graduate and above education). Wealth index was derived using principal component analysis which included information on number and kinds of consumer goods such as bicycle or car, housing characteristics such as source of drinking water, and availability of toilet facilities. Geographic variables included place of residence (rural/urban), province and ecological zone (mountain/hill/terai). Province was categorized into seven administrative divisions, according to the current administrative structure of Nepal. Access to information included the frequency of reading newspapers, watching television and listening to the radio. A detailed study methodology and description of survey variables has been presented elsewhere 37 . statistical analysis. The STATA version 14 (STATA corp) was used for analyses. We used sample weights to calculate national estimates from reported values. The association of explanatory and outcome variables was measured using Chi-square test. The multivariate logistic regression adjusting for all covariates was used to calculate adjusted odds ratio (AOR). The reference categories for calculation of adjusted odds ratios was selected on the basis of a research article from Nepal demographic health survey 2016 38 . The covariates (age, education, ecological zone, place of residence, wealth quintile and access to information) was entered simultaneously in the model. The cluster sampling design was adjusted in all the statistical analysis considering place of residence as a stratum and enumeration area as a cluster for this study. The geographic analysis of provinces of Nepal was performed in SAS v9.4 using gmap procedure 39 . We plotted graphs using all three DHS surveys data to visualize trend of tobacco use among Nepalese males and females. The existence of socio economic inequality in tobacco consumption was assessed by concentration curve. This curve has been used to measure and compare the degree of socio-economic status (SES) related inequality in various health indicators [40][41][42] . We plotted cumulative percentage of the sample, ranked by wealth quintile from lowest to highest on X-axis and the corresponding cumulative percentage of any form of tobacco use was plotted on Y-axis. For each of the three surveys, the percentage of the sample who lie within each quintile differs slightly from the nominal value, but the deviations are only of a few percent. A curve above the line of equality indicates a greater concentration of any form of tobacco use among the poor and a curve below the line indicates a greater concentration of tobacco use among the rich. ethics statement. The survey protocol was reviewed and approved by the Nepal Health Research Council (NHRC) and the ICF Institutional Review Board. Written informed consent was taken from the household head to carry out the interviews. Participants were informed about voluntary participation and confidentiality of information and could refrain from responding to any of the questions. It was ensured that the information provided by the participants or collected about the participants was protected and non-identifiable. All methods were performed in accordance with the relevant guidelines and regulations.