Table 2

Risk of bias of included studies

Study nameDeveloping and applying appropriate eligibility criteriaMeasurement of exposureMeasurement of outcomeControlling for confoundingCompleteness of data
High, low, unclearHigh, low, unclearHigh, low, unclearHigh, low, unclearHigh, low, unclear
Watkins 201821 Low
‘The PATH youth sample consists of individuals whose parents were sampled for the PATH adult survey’
Selection of exposed and unexposed from same population.
Low
‘The PATH Study adapted many items from well-established existing national surveys’.
‘Nicotine dependence items in the PATH Study’s questionnaires were based in part on the National Epidemiological Survey on Alcohol and Related Conditions Survey’.
‘Other health-related items in the PATH Study questionnaires came from validated screeners including the Global Appraisal of Individual Needs (GAIN) and Patient Reported Outcomes Measurement Information System (PROMIS)’.
‘The PATH Study questionnaires also included items from international, state and privately funded tobacco surveillance surveys, such as the ITC Project, the Visual Media Influences on Adolescent and Young Adult Smoking Behaviour surveys and the Minnesota Adult Tobacco Survey’.
30
Low
‘The PATH Study adapted many items from well-established existing national surveys’.
‘Nicotine dependence items in the PATH Study’s questionnaires were based in part on the National Epidemiological Survey on Alcohol and Related Conditions Survey’.
‘Other health-related items in the PATH Study questionnaires came from validated screeners including the Global Appraisal of Individual Needs (GAIN) and Patient Reported Outcomes Measurement Information System (PROMIS)’.
‘The PATH Study questionnaires also included items from international, state and privately funded tobacco surveillance surveys, such as the ITC Project, the Visual Media Influences on Adolescent and Young Adult Smoking Behaviour surveys and the Minnesota Adult Tobacco Survey’.
30
Low
‘To account for variation in smoking across sociodemographic groups: adjusted for sex, age, race/ethnicity (black or African American, Latino or other), parental educational level (bachelor’s degree or higher) and urban residence’
Low
‘Of 13 651 wave 1 youth participants, 11 996 completed wave 2 (unweighted retention, 87.9%), including 1915 individuals who reached 18 years of age before follow-up and were administered the wave 2 adult survey’.
Table 2 within 21:
For cigarette use ever, … the largest fraction of missing information was 0.011 in model 1 and 0.0186 in model 2
For past 30 day cigarette use, … and the largest fraction of missing information was 0.028 in model 1 and 0.032 in model 2
Table 3 within 21:
For cigarette use ever, … the largest fraction of missing information was 0.009 in model 1 and 0.021 in model 2.
For past 30 day cigarette use, … the largest fraction of missing information was 0.018 in model 1 and 0.033 in model 2’.
Jaber 20159 18 29 Low
‘60 Irbid schools were stratified by sex (boys, girls, and mixed sex schools) and type (public and private). A random sample of 19 schools was selected with probability proportionate to size’.
Low
‘Smoking behaviour was assessed using a pilot-tested questionnaire developed in accordance with WHO international guidelines, and several instruments validated in Arabic, such as the GYTS’.
‘To ensure the validity of the responses, parents or school personnel were not allowed to attend the data collection session’.
Low
‘Smoking behaviour was assessed using a pilot-tested questionnaire developed in accordance with WHO international guidelines, and several instruments validated in Arabic, such as the GYTS’.
‘To ensure the validity of the responses, parents or school personnel were not allowed to attend the data collection session’.
Low
‘Modelling included: sex; age; pocket money; educational level of parents; whether parents, siblings, friends, teacher smoked cigarettes; relation with parents, siblings, teachers and classmates; intention to smoke; refusal of self-efficacy; beliefs (cigarette smoker has more friends, cigarette smoking is more attractive, cigarette smoking reduces weight, cigarette smoking harms health, easy to quit cigarette after smoking a year); intention to smoke next year; and whether the student has seen advertisements promoting or warning against cigarette smoking or actor smoking in the media, and warning label on cigarette packs’.
Low
Censored in interval:
  • Baseline to year 1 (grade 7): 6.9%.

  • Year 1 to year 2 (grade 8): 0.7%.

  • Year 2 to year 3 (grade 9): 8.4%.

Soneji 201519 Low
‘Our data come from the first 2 waves of the Dartmouth Media, Advertising, and Health Study’.
‘At baseline 3342 individuals 15–23 years old were recruited from 6466 eligible households via a random digit- dialling telephone survey using landline (67.0%) and cell phone frames (33.0%). Of 3342 respondents, 2541 also completed a web-based visual survey at baseline. In wave 2 of the study we followed up 2541 respondents who had completed both portions of the base- line survey; 1596 completed the web-based visual survey at wave 2’.
Unclear
No relevant information reported by authors.
Unclear
No relevant information reported by authors.
Low
‘Covariates: we assessed Demographic characteristics of respondents: their age, sex, race/ethnicity, region, urbanicity’.
Socioeconomic status: maternal educational level, annual parental household income
Behavioural risk factors: whether any of the respondents’ friends smoked cigarettes, their parental cigarette smoking status
‘to address the possibility that the results were driven by a subset of deviant-prone adolescents and young adults, we controlled for another risky behaviour, namely, problem drinking (≥6 drinks per occasion): considered binge drinking’.
Low
‘We performed multiple imputation to address missing data for maternal educational level (4.2% missing) and annual parental household income (29.5% missing)’.
‘Our multiple imputation method assumed that these data were missing at random’.
Fielder 201322 Unclear
‘Interested women signed up on a website, were screened to ensure eligibility’.
This might indicate a risk of selection bias.
‘Eligible participants were first-year female college students (N=483) engaged in a 13 month longitudinal study from August 2009 to August 2010 designed to investigate health behaviours and relationships during the transition to college’.
Unclear
No relevant information reported by authors.
Unclear
No relevant information reported by authors.
Low
‘We tested demographics as potential covariates of hookah, cigarette, and marijuana initiation using logistic regression’.
Low
‘There were no differences between included participants and women excluded due to missing data (n=59) on age, race, ethnicity, impulsivity, sensation-seeking, or frequency of pre-college hookah, cigarette, mari- juana use, or binge drinking (all ps>0.05)’.
Treur 201720 Low
‘Cohort I consists of 6819 adolescents aged 11–17 years (mean age=13.8 (SD=1.1), 48.2% female) who were enrolled in a study that investigated the impact of school smoking policy on changes in adolescents’ smoking behaviour. Data were collected in 2014–2015 from 19 secondary schools randomly selected across the Netherlands’.
Unclear
No relevant information reported by authors.
Unclear
No relevant information reported by authors.
Low
‘Besides age, sex and educational attainment, a composite score of smoking propensity at T0 was added as covariate as well as an interaction term between propensity to smoke and ever use of e-cigarettes with nicotine/e-cigarettes without nicotine/waterpipe’.
Unclear
Loss to follow-up not reported.
Creamer 201813 Low
‘Students attending one of 24 2-year and 4-year colleges in Texas and involved in the first two waves of the Marketing and Promotions Across Colleges in Texas Project (Project M-PACT) were recruited to participate in the online survey via email invitation’.
Low
‘Measures for the current study were modelled after existing surveys and were initially reviewed by nine tobacco control experts who provided guidance on revisions. Final item modifications were conducted through an iterative process of cognitive interviewing with young adults who were not part of the present study’.
Low
‘Measures for the current study were modelled after existing surveys and were initially reviewed by nine tobacco control experts who provided guidance on revisions. Final item modifications were conducted through an iterative process of cognitive interviewing with young adults who were not part of the present study’.
Low
‘Sex, age, race/ethnicity, type of college and parental education were included as covariates’.
Low
‘Attrition analyses were conducted to determine if students who were eligible to participate at wave two but did not participate were significantly different from those who did participate. Because missing data were related to observed variables, it was assumed that the data were missing at random’.