Tobacco use in either smoking or smokeless forms is a major public health challenge with 80% of users living in low- and middle-income countries (LMICs). Smokeless tobacco (SLT), defined as a tobacco containing product that is consumed through chewing in the mouth, sniffing, or as a dissolvable has over 356 million users globally with 232 million in India and Bangladesh1,2. SLT is a nicotine containing addicting substance with other carcinogenic chemicals that increases the risk of cancers of the head and neck, especially mouth and throat, and is independently associated with incidence of various cardiovascular diseases3,4. As per the Global Burden of Disease 2019 estimates, SLT is attributed to cause over 8.6 million DALYs and about 350000 deaths annually5.

In LMICs, with most tobacco smokers being men and the prevalence of tobacco smoking being comparatively very low in women, there is a lack of policy attention and neglect of the impact of SLT use among women6. According to the Global Adult Tobacco Survey (GATS)-2 (2015–2016), India has 29.6% male and 12.1% female SLT users, signifying high prevalence in both genders compared to tobacco smoking in 19% men and only 2% women7. Women might have a higher risk of SLT addiction especially due to earlier initiation during adolescence and is also linked to a practice employed for suppression of hunger8. Furthermore, there is evidence to suggest that female SLT users have higher odds of developing oral cancer compared with males. SLT use during pregnancy is also linked to impairment of fetal lung and brain development, adverse maternal and infant nutritional outcomes, preterm delivery, low birth weight, and stillbirth9-12.

Evidence from the GATS is also indicative of a declining trend in smokeless tobacco use amongst women from 18.4% (2009–2010) to 12.8% (2016–2017). According to the National Family Health Survey-4 (2015–2016), a greater proportion of women in India have also reported an absence of intention to quit and also higher failure to quit SLT use compared to males13. Women in India are also less likely to notice health warnings related to tobacco including SLT use, suggestive of adverse social determinants such as reduced literacy contributing to higher risk of tobacco and SLT addiction compared to men. Consequently, three fifths of deaths attributable to SLT use occur in women14. In some cases, women may also lack confidence to access tobacco cessation services due to the perceived societal stigma linked to the phenomenon.

It is important to identify the prevalence and predictors of tobacco use through disaggregated data amongst women in India, particularly in the more vulnerable older population with prolonged addiction who, in the absence of interventions to promote cessation, have increased risk of developing cancer and heart disease due to the additive effect of multiple synergistic risk factors with tobacco14. However, there exists limited evidence on tobacco use characteristics by women in India. The evidence of tobacco use from demographic and health surveillance data is restricted to younger and middle-aged women who have comparatively improved educational parameters which limits generalizability of those findings to older women13. Consequently, the present study was conducted with the objective of ascertaining the prevalence and sociodemographic determinants of tobacco (smoking and SLT) use and the predictors of quit in SLT-using older women in India from analysis of nationally representative health survey data.


Data source

The present study compares findings from two rounds of the WHO Study on Global Ageing and Adult Health (WHO SAGE) for India, a national survey that collected survey data from adults aged ≥50 years. Our work is based on the cross-sectional survey data collected in the SAGE Wave 1 (2007–2008) and the SAGE Wave 2 (2015), India, which were implemented in the six selected states of Assam, Karnataka, Maharashtra, Rajasthan, Uttar Pradesh and West Bengal. Both SAGE Wave 1 and SAGE Wave 2 focused on data collection in persons aged ≥50 years and another smaller comparative sample of adults aged 18–49 years. The same primary sampling units (PSUs) and households covered in the SAGE Wave 1 in 2007 comprised the follow-up sample for SAGE Wave 2 in 2015, although the extent of loss to follow-up and proportion of new recruitment is not reported. Both SAGE Wave 1 and 2, India, employed a multi-stage stratified cluster sample design. Based on the selection probability at each stage of selection, household, and individual weights for analysis at the household level and personal level, respectively, were determined. A total of 9116 completed household interviews were included in SAGE Wave 2, of which 1998 interviews were of persons aged 18–49 years (1165 women and 833 men) and 7118 interviews of persons aged ≥50 years (3781 women and 3337 men). Data were collected using a standardized questionnaire having country-specific adaptations9,10.

Study population

The dataset consists of individuals aged 18–49 and ≥50 years. In households identified as ‘older’ for sampling purposes, all household members aged ≥50 years were invited to participate in the study. We included women aged ≥50 years in this analysis.

Outcome variables

The smoking status was assessed by the variable q3001: ‘Have you ever smoked any tobacco products’ and q3002, ‘Do you currently use any tobacco products?’. The SLT use was assessed by the variable q3002a: ‘Do you currently use smokeless tobacco?’, wherein both daily and non-daily users were considered as having SLT usage. SLT quit was assessed using the questions: ‘Used smokeless tobacco in the past?’, and ‘Do you currently use smokeless tobacco?’.

Explanatory variables

In our analysis, we considered a set of individual sociodemographic and lifestyle characteristics as controlling variables. These characteristics encompassed age, education level, marital status, body mass index (BMI), religion, ethnic group, place of residence, and wealth quintiles.

Education level had 4 categories ranging from ‘Not educated/less than primary’ to ‘College and higher’. Principal Components Analysis was used to generate scores that were transformed into ‘wealth quintiles’, where quintile 1 represents the lowest wealth and quintile 5 the highest. For marital status, categories of currently married and cohabiting were combined to form three categories. Similarly, the religious denomination variable was classified into three broad categories: Hinduism, Islam, and Other. Ethnic groups were classified as scheduled tribes, scheduled castes, other backward classes, and other. The WHO’s standard guidelines for BMI (kg/m2) were used to divide into three categories: normal (18.5), underweight (18.5–24.9), and pre-obese/obese (≥25.0).

Statistical analysis

Univariate analysis was done to assess the distribution of the sample. The prevalence and frequency of SLT usage were calculated after applying sampling weights. For bivariate associations, a chi-squared test was performed. Binary logistic regression was used for multivariable regression analysis. Variables having a p<0.20 in the bivariate association were included in the binary logistic regression models for multivariable analysis. Both crude odds ratios (ORs) and adjusted odds ratios (AOR) and their 95% CIs and p-values were calculated. A p<0.05 was considered significant in multivariable analysis. All assumptions were checked for the final logistic regression models. Data were analyzed using Stata version 15.1 (StataCorp, College Station, Texas).

Ethical considerations

The SAGE was approved by the World Health Organization’s Ethical Review Board (reference number RPC149). Written informed consent was obtained from all study respondents. The de-identified datasets were obtained after a formal permission from the IIPS. Since the SAGE datasets are anonymous and publicly available with no identifiable information about the participants, no separate ethical approval is required for this secondary data analysis.


The SAGE-2 (2015) dataset consisted of 4946 females, with a mean age of 54.62 years (SD=14.42). The mean age of initiation of smoking in women aged ≥50 years was 27.57 years (SD=21.74) and for SLT it was 28.76 years (SD=24.28). Table 1 gives the sociodemographic characteristics in the study sample consisting of females aged ≥50 years (n=3781).

Table 1

Sociodemographic characteristics of the participants (SAGE Wave-2)

CharacteristicsWomen (N=3781)
Age (years)
BMI (kg/m2) (n=3450)
Normal weight178449.65
Marital status
Never married260.45
Currently married/cohabiting235862.94
Education level
Not educated/lower than primary45137.37
Up to secondary school60344.95
High school11110.09
College and higher917.59
Ethnic group
Scheduled tribes2856.71
Scheduled castes63514.95
Other backward classes175149.77
Quintiles of wealth score
Mother tongue (n=3554)

In Table 2, it can be seen that the prevalence of tobacco ever use, in any form, among women decreased from 34.17% (n=3254; 95% CI: 31.78–36.64) (SAGE-1; 2007) to 18.17% (n=3772; 95% CI: 16.63–19.82) (SAGE-2; 2015). The prevalence of current tobacco product use in any form was 9.89% (n=352; 95% CI: 8.74–11.17). Overall prevalence of current tobacco smoking was 9.42% (n=331; 95% CI: 8.29–10.69) and the mean duration of usage among current tobacco users was 18.67 years (SD=15.96). Overall prevalence of SLT usage among women aged ≥50 years was estimated as 12.3% (n=454; 95% CI: 10.99–13.72) and the mean duration of usage among current SLT users was 22.93 years (SD=16.83).

Table 2

Distribution of tobacco use in women in India

VariableSAGE Wave-2 (N=3781)SAGE Wave-1 (N=3534)
n% (95% CI)n% (95% CI)
Ever used any form of tobacco (either smoking or smokeless)37723254
Yes68718.17 (16.63–19.82)104734.17 (31.78–36.64)
No308581.83 (80.18–83.37)220765.83 (63.36–68.22)
Ever used smokeless tobacco in the past
Daily but not current user431.19 (0.84–1.69)
Non-daily but not current user70.14 (0.06–0.29)
Never initiated non-user373198.67 (98.16–99.04)
Current tobacco user
Yes3529.89 (8.74–11.17)98529.14 (26.98–31.40)
No342990.11 (88.83–91.26)254970.86 (68.60–73.02)
Types of current tobacco user
Smokeless only2676.59 (5.61–7.74)
Smoking only1443.72 (2.99–4.63)
Dual user1875.69 (4.85–6.69)
Not using318383.98 (82.39–85.46)

The prevalence of past SLT use in the women aged ≥50 years was 1.19% for daily but not current user (n=43; 95% CI: 0.84–1.69), while 98.67% women reported as never having initiated tobacco use (n=3731; 95% CI: 98.16–99.04) (Table 2). Among all women aged ≥50 years, 6.59% (n=267) were only SLT users, 3.72% (n=144) were only tobacco smokers, and 5.69% (n=187) were dual users. Within the tobacco users (n=598), 23.26% (n=144; 95% CI: 19.06–28.06) were only tobacco smokers, 41.17% (n=267; 95% CI: 36.11–46.42) were only SLT users, and 35.57% (n=187; 95% CI: 30.87–40.58) were dual users.

Table 3 gives the distribution of factors associated with smoking status among women aged ≥50 years. On adjusted analysis, increasing BMI and wealth quintiles were observed as having significantly lower odds of smoking in women. Furthermore, women belonging to ‘Other’ religions had significantly higher odds of smoking compared to Hindu women (AOR=1.95; 95% CI: 1.11– 3.44).

Table 3

Distribution of factors associated with smoking status in women (SAGE Wave-2, N=3772)

VariableNon-smoking (N=3085) n (weighted %)Smoking (N=687) n (weighted %)OR (95% CI)paAOR (95% CI)pb
Age (years)0.00670.0705
50–59 (Ref.)1450 (47.14)282 (38.76)11
60–691042 (33.19)249 (38.52)1.41 (1.10–1.81)1.34 (1.03–1.74)
≥70593 (19.67)156 (22.72)1.40 (1.08–1.82)1.23 (0.92–1.65)
BMI (kg/m2) (n=3450)0.0010.0003
Underweight (Ref.)696 (25.45)231 (40.62)11
Normal weight1460 (50.3)324 (46.71)0.58 (0.46–0.74)0.67 (0.52–0.87)
Pre-obese/obese663 (24.25)76 (12.67)0.33 (0.22–0.49)0.45 (0.29–0.68)
Ethnic group (n=3771)0.0010.1087
Scheduled tribes (Ref.)219 (6.383)65 (8.13)11
Scheduled castes481 (13.69)153 (20.69)1.19 (0.82–1.72)1.14 (0.77–1.71)
Other backward classes1458 (50.48)289 (46.6)0.72 (0.51–1.03)0.82 (0.56–1.20)
Other926 (29.45)180 (24.57)0.66 (0.46–0.94)0.84 (0.56–1.24)
Religion (n=3771)0.00050.0008
Hinduism (Ref.)2617 (86.71)559 (79.47)11
Islam354 (10.79)99 (16.2)1.64 (1.21–2.21)1.72 (1.22–2.41)
Other113 (2.506)29 (4.335)1.89 (1.16–3.08)1.95 (1.11–3.44)
Quintiles of wealth score0.0010.0001
First (Ref.)560 (19.15)191 (31.31)11
Second550 (17.59)147 (19.18)0.67 (0.50–0.90)0.68 (0.49–0.93)
Third546 (16.97)137 (21.58)0.78 (0.56–1.09)0.88 (0.61–1.26)
Fourth657 (22.27)135 (17.88)0.49 (0.36–0.67)0.55 (0.38–0.79)
Fifth772 (24.03)77 (10.05)0.26 (0.18–0.37)0.37 (0.23–0.58)
Urban (Ref.)714 (30.17)118 (22.86)11
Rural2371 (69.83)569 (77.14)1.46 (1.04–2.05)1.02 (0.69–1.51)
Mother tongue0.1788
Hindi1152 (41.38)219 (40.1)0.80 (0.62–1.03)
Bengali677 (19.67)151 (20.06)0.84 (0.65–1.10)
Marathi486 (20.59)110 (17.62)0.71 (0.49–1.02)
Other (Ref.)770 (18.36)207 (22.22)1

AOR: adjusted odds ratio.

a p<0.20 included in adjusted regression model (age, BMI, ethnic group, religious denomination, quintiles of wealth score, and residence).

b p<0.05 considered significant. Goodness of fit, p=0.1552.

On bivariate analysis, age, BMI, ethnicity, religion, and quintiles of wealth score, were significantly associated (p<0.05) with SLT usage (Table 4). In the multivariate logistic regression analysis, increasing age, BMI, ethnic group, religion, and quintiles of wealth score, were significantly associated (p<0.05) with SLT usage. Elderly women (aged ≥70 years) had higher odds of using SLT (AOR=1.46; 95% CI: 1.05–2.03), while increasing BMI was negatively associated with SLT usage (AOR=0.53; 95% CI: 0.31–0.90). With reference to scheduled tribes, scheduled caste women had higher odds of using SLT (AOR=1.20; 95% CI: 0.77–1.86). Muslim women had higher odds of using SLT compared to the Hindu women (AOR=1.86; 95% CI: 1.24–2.69). There were 7.62% (n=37; 95% CI: 5.23–10.99) women who reported a successful quit after initiation with SLT (Table 5). The adjusted odds of quitting SLT in women belonging to the scheduled caste ethnicity were 0.70 times less (95% CI: 0.09–5.81) compared to the scheduled tribe ethnicity group.

Table 4

Distribution of factors associated with current SLT use in women (SAGE Wave-2)

VariableNot current SLT user (N=3327) n (weighted %)Current SLT user (N=454) n (weighted %)*OR (95% CI)paAOR (95% CI)pb
Age (years)0.00570.0425
50–59 (Ref.)1548 (46.71)186 (10.09)11
60–691135 (33.61)158 (13.61)1.40 (1.04–1.89)1.33 (0.98–1.82)
≥70644 (19.68)110 (15.01)1.57 (1.17–2.21)1.46 (1.05–2.03)
BMI (kg/m2) (n=3450)0.00050.0453
Underweight (Ref.)784 (26.78)143 (16.89)11
Normal weight1559 (49.84)225 (12.09)0.68 (0.51–0.90)0.76 (0.56–1.03)
Pre-obese/obese686 (23.38)53 (7.51)0.40 (0.24–0.66)0.53 (0.31–0.90)
Ethnic group (n=3779)<0.0010.008
Scheduled tribes (Ref.)240 (6.45)45 (15.68)11
Scheduled castes527 (13.90)108 (18.49)1.22 (0.80–1.86)1.20 (0.77–1.86)
Other backward classes1561 (50.16)190 (11.61)0.71 (0.47–1.06)0.74 (0.48–1.14)
Other997 (29.50)111 (9.47)0.56 (0.37–0.85)0.66 (0.42–1.04)
Religion (n=3779)0.01150.0057
Hinduism (Ref.)2817 (86.11)365 (11.53)11
Islam387 (11.28)68 (16.23)1.49 (1.06–2.08)1.83 (1.24–2.69)
Other121 (2.61)21 (19.07)1.81 (1.05–3.12)1.48 (0.80–2.73)
Quintiles of wealth score0.0010.0029
First (Ref.)637 (20.35)115 (16.38)11
Second615 (18.01)84 (11.65)0.67 (0.47–0.97)0.67 (0.45–1.00)
Third579 (16.81)107 (17.26)1.06 (0.73–1.56)1.11 (0.72–1.72)
Fourth696 (21.7)96 (11.15)0.64 (0.45–0.92)0.69 (0.45–1.06)
Fifth800 (23.12)52 (5.78)0.31 (0.20–0.50)0.45 (0.26–0.77)
Urban (Ref.)745 (29.4)88 (10.42)1
Rural2582 (70.6)366 (13.05)1.29 (0.88–1.90)
Mother tongue0.07100.0619
Hindi1255 (42.02)117 (10.33)0.68 (0.50–0.91)0.66 (0.48–0.91)
Bengali715 (19.44)113 (13.44)0.91 (0.67–1.23)0.82 (0.59–1.15)
Marathi512 (19.95)88 (13.04)0.88 (0.58–1.34)1.01 (0.65–1.56)
Other (Ref.)845 (18.6)136 (14.54)11

AOR: adjusted odds ratio.

a p<0.20 included in adjusted regression model (age, BMI, ethnic group, religious denomination, quintiles of wealth score, and mother tongue)

b p<0.05 considered significant.

* Row-wise percentages given. Goodness of fit, p=0.1861.

Table 5

Determinants of quitting behavior among women ever initiated on SLT (SAGE Wave-2, N=491)

VariableQuit SLT (N=37) % (95% CI)OR (95% CI)paAOR (95% CI)pb
Age (years)0.664
50–59 (Ref.)42.7 (25.12–62.34)1
60–6938.94 (22.49–58.35)0.90 (0.36–2.29)
≥7018.36 (9.08–33.62)0.65 (0.24–1.72)
BMI (kg/m2) (n=452)0.3037
Underweight (Ref.)32.76 (17.36–53.05)1
Normal weight40.13 (22.15–61.23)0.97 (0.37–2.55)
Pre-obese/obese27.11 (11.81–50.82)2.37 (0.70–7.98)
Ethnic group0.0410.05
Scheduled tribes (Ref.)4.35 (1.06–16.21)11
Scheduled castes5.70 (1.75–17.02)0.5 (0.08–3.16)0.70 (0.09–5.81)
Other backward classes54.35 (35.64–71.92)2.27 (0.49–10.63)3.40 (0.51–22.54)
Other35.59 (19.89–55.16)3.18 (0.66–15.41)4.37 (0.64–30)
Hinduism (Ref.)62.07 (40.57–79.68)11
Islam23.69 (9.785–47.05)1.96 (0.66–5.85)1.35 (0.40–4.49)
Other14.24 (4.099–39.22)4.18 (0.98–17.80)4.77 (1.28–17.76)
Quintiles of wealth score0.737
First (Ref.)28.85 (14.39–49.44)1
Second19.26 (9.20–35.96)1.12 (0.37–3.39)
Third20.75 (9.29–40.11)0.82 (0.25–2.67)
Fourth13.26 (5.85–27.34)0.67 (0.21–2.12)
Fifth17.87 (6.26–41.50)1.74 (0.42–7.17)
Urban (Ref.)33.02 (16.19–55.70)1
Rural66.98 (44.30–83.81)0.65 (0.24–1.79)
Mother tongue0.295
Hindi26.74 (12.03–49.35)1.10 (0.35–3.45)
Bengali37.12 (20.75–57.09)2.45 (0.93–6.48)
Marathi20.27 (9.942–36.93)1.35 (0.48–3.80)
Other (Ref.)15.87 (7.89–29.35)1
Type of past-use (n=50)0.368
Daily (Ref.)87.77 (73.62–94.86)1
Non-daily12.23 (5.14–26.38)2.87 (0.28–29.40)
Type of addiction (n=36)0.859
SLT only (Ref.)81.60 (61.99–92.34)1
Dual user18.40 (7.66–38.01)1.18 (0.18–7.80)

AOR: adjusted odds ratio.

a p<0.20 included in the adjusted regression model (ethnic group and religious denomination).

b p<0.05 considered significant. Goodness of fit, p=0.89.


The present study evaluated the prevalence and sociodemographic determinants of tobacco use patterns and predictors of quit among older women aged ≥50 years. Age, BMI, lower wealth quintiles, and ethnicity were found to be the determinants associated with tobacco smoking in the study sample, in line with the results of prior studies12,13. However, the prevalence of tobacco smoking in women aged ≥50 years in this analysis from the SAGE survey (9.42%) was nearly two-fold higher compared to observations from the GATS survey data (5.14%), suggesting the need for repeated surveys to validate the extent of this public health problem in India, especially in women12.

We found that the majority of tobacco users among females aged ≥60 years were SLT users, a finding similar to a study based on GATS I and II15. On bivariate analysis, age, BMI, ethnicity, quintiles of wealth index, and religious denomination were significantly associated with SLT usage. These findings agree with studies conducted in other LMICs such as Pakistan16, Bangladesh17 and Nepal18. SLT usage is often initiated at an early age when children are exposed and normalized to it and are frequently involved in its purchase for family members19. In this study, SLT use was found to have significantly declined among women, a finding which also corroborates the trend from GATS-1 and GATS-212. Similar to previous studies, age was found to be directly associated with SLT use, whereas BMI was found to have an inverse relation in our analysis15,20.

Furthermore, we observed that women belonging to the poorest section of society (first quintile) had higher odds of using SLT, a finding similar to that observed in the GATS 2 survey21. In this study, spoken language did not have a statistically significant association with SLT use unlike a previous study suggesting that the association was likely to have been coincidental22.

Nevertheless, our study findings indicate that ethnicity is possibly linked with SLT use in India as the scheduled caste ethnicity women had significantly higher odds of consuming SLT compared to women from other caste groups. A study based on the second round of the Indian National Family Health Survey (NFHS-2; 1998–1999) had also reported that tobacco consumption was significantly more prevalent among scheduled caste populations, signifying their persistent vulnerability13. Prior evidence also suggests that women belonging to socioeconomically disadvantaged populations employed in hard labor activities often use tobacco to suppress hunger8,23 and reduce perceived stress. Further, our finding suggests that Muslim women were more likely to use SLT, a finding consistent with evidence from NFHS-213 and Bangladesh24.

The present study findings suggest that a very small proportion of older women who are SLT users in India successfully quit tobacco. Improving the awareness of the harmful effects of tobacco and especially SLT use among women is necessary to reduce its initiation and persistence25. Evidence from GATS 2 had shown that nearly half of the women in India (45.3%) fail to take notice of health warnings on SLT product packages26. Furthermore, the use of quitline/helpline/direct counselling is very low overall, due to difficult accessibility, lack of social support, and associated stigma27. Our study findings suggest the need for enhanced tobacco use surveillance and targeted interventions for promoting tobacco cessation, especially SLT, amongst older women users who experience the double impact of adverse social determinants such as lower SES, and belonging to marginalized communities that contribute to reduced access to tobacco quit services. Future research should also explore the evolving dynamics of determinants and cultural factors shaping tobacco use, since a nuanced understanding of the motivations for tobacco use in this vulnerable population can inform the design of tailored interventions which are effective in reducing tobacco consumption among women in India.

Strengths and limitations

The key strength of this study is that it used data from the SAGE survey which used standardized questionnaires, a robust sampling strategy, and validated data collection methods, with high representativeness for older populations. However, this study has a few limitations. First, the data in this study are mostly cross-sectional, and therefore causation cannot be assumed in any direction. Second, recall and social desirability bias may have led to the likely underestimation of the tobacco burden, especially due to social stigma related to tobacco use among women in India28. Third, data points on willingness to quit and frequency and type of quit attempts were not available in this survey.


Nearly one in five older women were found to use tobacco in some form. The higher prevalence of tobacco smoking and smokeless tobacco use compared to GATS data calls for continuous surveillance and focused public health efforts. Furthermore, quit rates in female SLT users continue to be very low (nearly one in ten) suggesting the need for strengthening access, availability, and affordability of tobacco cessation services to promote successful quitting behavior.