Study design and participants
Data on smokeless tobacco use were extracted from the GYTS, which is a cross-sectional and school-based program led by the WHO and the U.S. Centers for Disease Control and Prevention aiming to provide nationally representative estimates on tobacco use among adolescents and guide tobacco control programs worldwide. The GYTS used the same two-stage random cluster sampling framework in each participating country, and at the first stage, schools were randomly selected, and at the second stage, classes were also randomly selected from the target schools. All students in the selected classes were eligible to complete the standardized questionnaires, which were self-administered by students and included the same core set of questions to make them comparable across countries. More information on the GYTS can be found on the Centers for Disease Control and Prevention website (https://www.cdc.gov/tobacco/global/gtss/gtssdata/index.html) [28].
Data in the U.S. were from the NYTS, which is a national, annual, and school-based program to assess tobacco use among U.S. youth. The NYTS followed a similar methodology to the GYTS. The questions and possible closed-ended answers on smokeless tobacco use and related factors are consistent with the GYTS. More information on the NYTS can be found on the U.S. Centers for Disease Control and Prevention website (https://www.cdc.gov/tobacco/data_statistics/surveys/nyts/index.htm) [29]. All GYTSs and NYTSs were approved by the National Ethics Committee, and all students/guardians provided verbal consent.
Both the GYTS and NYTS are ongoing. We used the latest data from 138 countries in 2010-2019 to estimate the prevalence of smokeless tobacco use and its associated factors among adolescents aged 12-16 years. We also used data from 100 countries that had conducted two or more surveys between 1999 and 2019 to estimate the secular trend in the prevalence of smokeless tobacco use. Of note, the GYTS did not release the latest data in China, thus we extracted the related information from the Chinese Youth Tobacco Survey report in 2014 [30], which is a part of the GYTS. A flow chart of inclusion and exclusion of countries is shown in Additional file 1: Fig. S1. After excluding participants with missing data on sex, age and current smokeless tobacco use, and those aged <12 or >16 years, a total of 1,039,249 adolescents aged 12–16 years between 1999 and 2019 were included in the data analyses.
Dependent variable
Current smokeless tobacco use was defined as using any form of smokeless tobacco products during the past 30 days based on response to the question “During the past 30 days, did you use any form of smokeless tobacco products (such as snuff, chewing tobacco, dip, gutka)?”.
Independent variable
Demographic variables included sex and age. Other independent variables included cigarette smoking, other tobacco product use, parental smoking, smoking status of closest friends, tobacco advertisement exposure, being offered free tobacco products, being taught about dangers of smoking, and World Bank income level. Cigarette smoking was defined as smoking cigarettes on at least one day during the past 30 days based on response to the question “During the past 30 days, on how many days did you smoke cigarettes?”. Other tobacco product use was defined as using combustible tobacco products other than cigarettes during the past 30 days based on response to the question “During the past 30 days, did you use any form of smoked tobacco products other than cigarettes (such as cigars, pipe, waterpipe)?”. Parental smoking was assessed based on response to the question “Do your parents smoke tobacco?”, and it was divided into “Neither smoking,” “Only father smoking,” “Only mother smoking,” and “Both smoking” in our analyses. Smoking status of closest friends was assessed based on response to the question “Do any of your closest friend smoke tobacco?”, and it was divided into “None”, “Some”, “Most”, and “All”. Tobacco advertisement exposure was defined as exposure to more than one type of tobacco advertisement based on responses to the following three questions: “During the past 30 days, did you see any people using tobacco on TV, in videos, or in movies?”, “During the past 30 days, did you see any advertisements or promotions for tobacco products at points of sale (such as stores, shops, restaurant)?”, and “Do you have something with a tobacco product brand logo on it (such as t-shirt, pen, backpack)?”. Being offered free tobacco products was assessed based on response to the question: “Has a person working for a tobacco company ever offered you a free tobacco product?”. Being taught about dangers of smoking was assessed based on response to the question: “During the past 12 months, were you taught in any of your classes about the dangers of tobacco use?”. Income level categories for each country followed the standards of the World Bank's classification based on the survey year of the GYTS/NYTS in our study.
Statistical analysis
According to the complex sampling design of the GYTS, the weighted prevalence estimates and their 95% confidence intervals (CI) of smokeless tobacco use in each country were calculated using original sampling weights, strata, and primary sampling units provided in the datasets using the SAS PROC SURVEYFREQ procedure. The original weights were calculated by the following formula: W=W1*W2*f1*f2*f3*f4, where W1 is the inverse of the selection probability of each school; W2 is the inverse of selection probability of each class; f1 is the school-level non-response adjustment factor calculated by school enrolment size; f2 is the class-level non-response adjustment factor for each school; f3 is the student-level non-response adjustment factor for each class; f4 is the post-adjustment stratification factor calculated by grade and sex. We rescaled the original weights to calculate the overall and subgroups’ prevalence of smokeless tobacco use based on each country’s sample size. The rescaled weights were calculated as the maximum country sample size (i.e., U.S.: n=13,689) divided by the sample size of each country multiplied by the original weights. Chi-square analysis was used to test for differences in the prevalence between sexes, age groups, and other subgroups. Chi-square trend test was used to examine the secular trend in the prevalence with consideration of data from all surveys in each country between 1999 and 2019. The prevalence estimates of secular trends were calculated per 5 calendar years. Multivariable logistic regression models were used to assess the association of potential associated factors (sex, age, cigarette smoking, other tobacco product use, parental smoking, smoking status of closest friends, tobacco advertisement exposure, being offered free tobacco products, being taught about dangers of smoking, and World Bank income level) with current smokeless tobacco use, and the code for each variable is shown in Additional file 1: Table S1. In order to correct the overall probability of type I error in multiple statistical tests (α=0.05), the Bonferroni’s correction was used to adjust the critical significance level of each statistical test. The corrected critical significance level was equal to the original critical significance level (0.05) divided by the number of tests performed. A two-sided P-value less than the Bonferroni’s corrected critical significance level was regarded as statistical significance and SAS 9.4 (SAS Institute, Cary, NC, US) was used for all analyses.