Smokeless tobacco use: pattern of use, knowledge and perceptions among rural Bangladeshi adolescents

Sample size and sampling technique

To reach a 95% confidence level with 2.5% marginal error and 10.1% expected ever SLT prevalence, a sample size of 553 was required (Daniel, 2009; WHO, 2015). Of the 18 secondary schools in Ramgati Upazila, with a total population of 8,994 students, 790 (8.78%) were recruited from two schools using a stratified method according to the enrolment size in each school. Initially the number of students for all 18 schools was collected from the local administrative office. From there, five schools with highest enrolment size were selected and from those five schools two schools were chosen randomly from their registration number.

Survey tool and data collection

A self-administered questionnaire was adapted from the SLT module of the GYTS questionnaire version 1.00 (2012) (Global Youth Tobacco Survey Collaborative Group, 2012). Guided by the Social Cognitive Theory and Health Belief Model, the questionnaire was developed and comprised of closed-ended, open-ended and multiple response questions. Additional questions on types of SLT use, the reason for failing to quit SLT and eight knowledge questions were added from selected literature and validated surveys (Bhaskar et al., 2016; Abdullah et al., 2014; Kaduri et al., 2008).

The questionnaire was translated from English to Bengali and then back-translated to English by professional translators. It was also reviewed by the research team, local healthcare workers who had previously conducted a survey with students and teachers to ensure that the Bengali version was idiomatically appropriate for Bangladeshi adolescents. The questionnaire was piloted among five female and five male students from two secondary schools and was modified accordingly. Students took up to 15 min to complete the questionnaire.

Study measures

The dependent measure or outcome variable for the present study was SLT use, which was classified into three main categories: ‘ever users’, ‘current users’ and ‘never users’ (WHO, 2015). Ever users were individuals who had used SLT at least once in their life, even if this was a tiny portion. Current SLT users were those who used SLT at least once in the last 30 days (WHO, 2015). Never users had never used SLT before. Additionally, respondents were asked, “What type or brand of SLT have you tried?”

Social cognitive factors

“Personal characteristics” of the participants were measured by asking questions related to age, gender, father’s and mother’s profession, the age of initiation and SLT dependency. “Environmental factors” were assessed through querying about the source and access to SLT products. “Self-efficacy” was measured through respondent’s ability to deny SLT when offered by their friends. “Outcome expectancies” was measured by querying the reasons for SLT use.

Health belief model

“Perceived susceptibility” was measured through students’ knowledge about harmful effects of SLT use and its contents, such as the development of white patches in the mouth, oral cancer, gum diseases and heart disease from SLT use. “Perceived severity” was assessed through asking respondents’ perception about the harmfulness of SLT compared to smoked tobacco. “Perceived benefit” was assessed by querying about the potential benefits of SLT use. “Perceived barriers” was measured by asking about the difficulty of quitting SLT. “Cues to action” was assessed by questioning about their experience receiving help to quit SLT and noticing the health warning on SLT packages.

Data analysis

All the data were entered into MS Excel 2013 and exported into SPSS version 24 for analysis. Data analyses were conducted in multiple phases. In the first phase, a simple descriptive analysis (frequency, percentage, mean and median) was conducted to examine current SLT use behaviour as well as perception and knowledge about its ill effects. In the second phase, statistical analyses were performed to explore the association between different variables, such as socio-demographic variables, SLT use behaviours, perception and knowledge. Associations between categorical variables were assessed using a chi-square or Fisher exact test where appropriate. Knowledge was assessed using a scoring system where participants were given one point for each correctly answered question and zero otherwise. The sum of scores was calculated, and its relation to other variables was assessed using the student t-test or Mann–Whitney test when appropriate. The knowledge score was further categorised into poor knowledge (score of 0–2), average knowledge (score of 3–5), and good knowledge (score of 6–8). For the predictive model, the knowledge score was considered a nominal variable.

In the third phase, a multivariate logistic regression model was used to predict current SLT use. However, initially a univariate logistic regression was performed to determine variables that could be used to predict current SLT use. Akaike information criterion (AIC) was also extracted for each variable. Some variables were presented with different cut-off values. Each of them was assessed, and the variable with the lowest AIC was included in the final model. Only variables that were significant in the initial screening (univariate logistic regression) were included in the multivariate logistic regression. A correlation matrix was also constructed to ensure the absence of multicollinearity between variables included in the model. Exponentiated coefficients (odds ratio) were extracted from the model as well as the overall AIC of the model. Confidence intervals (95%) and Wald statistics were used to assess whether regression coefficients were significantly different from zero (null hypothesis). A receiver operator characteristics curve was plotted to assess whether the model could accurately classify the data. The Area Under the Curve (AUC) was 0.84 which is a good indicator of the model’s predictive power.

The odds ratio with a 95% CI was used to quantify the association. For the purpose of the analysis, a P-value of less than 0.05 was considered significant. Continuous variables were expressed as means or medians, whereas categorical variables were presented in frequencies or percentages. Data were missing for some of the questions; therefore, percentages were expressed according to the number of valid responses.

Sample characteristics

Of the 820 enrolled students, 790 attended school on the day of the survey and all the students gave consent and participated in the study. More than half of the participants were male (63.3%). The mean age was 13.8 ± 0.07 years, and the median age was 14 years. There was no significant difference in the mean age of male (14.0 ± 0.07) and female students (13.7 ± 0.07). The predominant occupation of the participants’ fathers was farming (62.4%; 493), and almost all the participants’ mothers were housekeepers (98.9%; 781) (Table 1).

SLT use behaviour

A total of 9.5% (75) reported ever using SLT. Approximately 3.7% (29) of the participants in this study or 38.7% of the ever users reported that they were currently using SLT. Males had a higher incidence of both ever and current SLT use with 68% (51) and 65.5% (19), respectively. Among the ever users, participants reported they started SLT use as early as the age of 7 years or younger (5.3%; four). However, most SLT users reported using SLT at the age of 12 or 13 years (34.7%; 25). Male SLT users (35.3%; 18) started using SLT at an earlier age (10–11 years) compared to female users (12–13 years) (41.7%; 10) (Table 2). Only two types of SLT were popular among the SLT users. Zarda was the most common type of SLT used (80%) followed by Pan Masala (20%) (Table 2). The majority used SLT less than once per day (65.5%; 19). Overall SLT dependency was low, where 89.3% (67) and 84% (63) of SLT users did not feel the need to use SLT first thing in the morning and did not feel a strong desire to use it again after using it once, respectively. Many respondents could not articulate their reason for using SLT (36.7%; 29). However, only male users (9.3%; five) and respondents aged 14 years and above (7.9%; five) used SLT because their friends also used. Regarding SLT cessation, 89.7% (26) of current SLT users wanted to quit, but none of them received help or advice from a program or professionals to do so. Regarding the attempt to quit SLT in the past 12 months, 34.5% (10) of current SLT users tried to stop SLT but failed. Between both genders, male students (42.1%; eight) had a higher rate of failure in quitting SLT compared with female students (20.0%; two) (Table 2).

Regarding access and availability, many students obtained SLT from social sources (41.4%; 12) and from stores and shops (34.4%; 10). Among current users who bought SLT products, 86% (12) were not refused because of their age. In terms of their exposure to the anti-tobacco messages, 45.3% (34) of respondents reported not seeing any health warning on SLT packages. In contrast, 29.3% (22) students had seen the health warnings and thought of quitting SLT.

The overall susceptibility to SLT use was 2.8% (22) when offered by friends and was significantly associated with ever SLT use. Specifically, 8% (six) of ever SLT users compared to 2.2% (16) of never users were likely to use SLT if offered by their friend (p = 0.01). The majority of students (59%; 466) thought it would not be difficult to quit once someone started using SLT and this was significantly associated with age. Specifically, 52.9% (128) of respondents 15 years or older believed it would be difficult to quit SLT compared to 35.8% (196) of respondents less than 15 years old (p < 0.001).

Knowledge about the harmful effects of SLT

To assess respondents’ knowledge about the adverse effects of SLT, students were asked eight questions. The overall knowledge about the harmful effects of smokeless tobacco use was good among respondents. The majority of adolescents (75.6%; 597) thought SLT use is bad for health. However, 21.2% (169) of study participants did not know whether SLT use is bad or good for the health. Over 32% (254) of students did not know whether SLT is less harmful compared to smoked tobacco (ST) and 25% (202) thought it is less harmful than ST. Additionally, 29.2% (231), 25.3% (200), 26% (206), and 25.4% (201) of students did not know SLT causes white patches in the mouth, oral cancer, gum diseases, and heart diseases respectively. Also, 32.5% (257) of respondents did not know SLT contains Nicotine.

The mean knowledge score was 5.19 ± 0.15. Male students (5.36 ± 0.09) were more likely to be knowledgeable compared to female students (4.9 ± 0.12). Students whose fathers’ profession was teaching (6.21 ± 0.37) had the highest mean score, and the lowest score was seen from students whose fathers’ occupation was farming (5.08 ± 0.09). Moreover, the average knowledge score was lowest among current SLT users (4.31 ± 0.40) followed by ever users (4.48 ± 0.20), and the highest average score was from the group who never tried SLT (5.27 ± 0.08).

Table 3 represents the knowledge category across various factors. Respondents were grouped into three main categories based on the total score distribution (poor knowledge, average knowledge, good knowledge). Overall, 54.2% (428) of respondents had good knowledge about the harmful effects of SLT. The majority of never SLT users (55.4%; 396) scored in the good knowledge category compared to ever users (42.7%; 32). Male respondents were more familiar with the SLT use hazards (56%) compared to females (51%) (Table 3). Knowledge was significantly associated with witnessing anti-tobacco messages on the SLT packages. Specifically, 61.8% (21) of ever SLT users who did not witness anti-tobacco messages on SLT packages had good knowledge compared to 26.8% who did see the anti-tobacco messages (χ2 = 9.87, p = 0.007, φ = 0.363). Knowledge was also significantly associated with respondent’s self-efficacy, where 55.2% (424) of respondents who would refuse to use SLT if offered by a friend had good knowledge compared to 18.2% (four) of those who would use SLT (χ2 = 11.98, p = 0.003, φ = 0.123). In addition, perceived barriers were significantly associated with knowledge, as 21.9% (102) of respondents who thought it would not be difficult to quit SLT once started had poor knowledge compared to 9.6% (31) of those who thought the opposite (χ2 = 20.95, p =  < 0.001, φ = 0.163) (Table 3).

Predictors of current SLT use

Initial univariate logistic regression analysis showed that age (p = 0.005), father’s profession (p = 0.046), self-efficacy (p = 0.001), perceived barriers (p = 0.029), perceived benefit (p = 0.014) and perceived severity (p = 0.008) were significant variables (Table 4). In the next stage, these variables were included in the multivariate logistic regression.

The multivariate regression analysis indicated that students who were 14 years old and older were 6.5 times more likely to be current SLT users compared to those younger than 14 years (p = 0.002) (Table 4). Perceived benefits and perceived severity of SLT use was associated with lower odds of current SLT use (OR = 0.21, 95% CI [0.059–1.03] and OR = 0.36, 95% CI [0.12–0.97], respectively) (Table 5). Self-efficacy was also a significant predictor of current SLT use where students who would use SLT if offered by a friend were more likely to be current SLT users compared to those who would not (OR = 5.78, 95% CI [1.46–19.65]). Lastly, perceived barriers regarding the difficulty of quitting SLT use was also a significant predictor of current SLT use where those who perceived quitting as easy were more likely to be current users compared to those who did not (OR = 0.30, 95% CI [0.10–0.74]) (Table 5). The model correctly classified 82% of the dependent variables.