Table 1

Included records, n=37, characteristics, effect sizes identified (n effects), effects included in meta-analysis (k) and risk of bias (RoB) index score

StudyDesignArea (date)StatisticsSampling (n)n effects (k)*RoB†
Barnes et al 19 Cross-sectionalWestern Australia (2003–2009 participants recruited; 1999–2011 hospital data; 2011 tobacco retailers)Logistic regressionWestern Australia adults: probability/representative stratified (n=12 270 smoker and non-smokers; n=1873 smokers)38 (38)0
Berg-Beckhoff et al 53 Cross-sectionalDenmark (2011 respondents; NR tobacco retailers)Multinomial logistic regressionDenmark respondents aged 15–79 years: probability (n=2122)2 (0)0
Brown et al 62 Cross-sectionalBaltimore, Maryland (2009 participants; 2009 tobacco retailers)Logistic regression (note study tests gender interactions)Young adult black men and women (aged 21–24 years) originally recruited in 1993 as first-grade students in nine public elementary schools: follow-up of cohort after attrition (n=283)15 (0)0
Cantrell et al 63 Cross-sectionalEight Designated Media Areas across the USA (2008–2010 respondents; retailers NR)Generalised estimating equationsUS smokers aged 18–49 years: randomly selected (n=2377)14 (14)1
Cantrell et al 64 Cross-sectionalUSA (2013 participants; 2012 tobacco retailers)Logistic regressionAdults aged 18–34 years participating in GfK KnowledgePanel: Participation Legacy Young Adult Cohort Study, which used a nationally representative sample (n=4288)6 (6)0
Chaiton et al 43 Cross-sectionalToronto, Canada (2008–2010 participants; 2011 tobacco retailers)Conditional fixed effects regressionHealthcare setting/cessation clinic: census (n=734)16 (12)1
Chaiton et al 44 LongitudinalOntario, Canada (2005–2011 participants; 2011 tobacco retailers)Mixed models/survival analysisOntario adult smokers in urban or suburban area: probability (n=2414)4 (4)0
Chuang et al 72 Cross-sectionalFour cities in Northern California (1979–1990 participants; 1979–1990 convenience stores)Mixed modelsStanford Heart Disease Prevention Programme, adults aged 25–74 years: (8121)15 (0)1
Clemens et al 59 Longitudinal within subjectScotland (2000–2015 maternity records; 2012 tobacco retailers)Conditional fixed effects logistic regressionMaternity records of women who changed smoking behaviour and had two or more pregnancies (n=55 234)4 (4)0
Dearfield (dissertation)+ Dearfield (paper)65 74 Cross-sectionalWashington, District of Columbia (2012–2014 participants; NR tobacco retailers)Multilevel modelHealthcare Setting (District of Columbia Primary Care Practice Based Research Network); convenience; current or former smokers (n=152)2 (0)2
Eng 201949 LongitudinalToronto (participants/tobacco retailers NR)Logistic regression, survival analysisHealthcare setting, patients with lung and head and neck cancer: convenience (721 patients with lung cancer, 445 patients with head and neck; 391 were smokers at baseline)6 (6)0
Eng 202020 LongitudinalToronto (2009–2012 participants; NR tobacco retailers)Survival analysisHealthcare setting, patients with lung cancer: convenience (n=1411)6 (6)0
Farley 2016 (dissertation)+ Farley 2019 (published paper)66 75 Cross-sectionalNew York, New York (2011–2013 respondents; 2012 tobacco retailers)Linear regression, geographically weighted regression; multilevel modelsNYC respondents aged 18+ years: probability, phone (NR)3 (3)0
Farley et al 21 Cross-sectionalLouisiana (1997–1998 maternity records; NR tobacco retailers)Multilevel modellingLive births from Louisiana vital records: census (n=105 111)See evidence table (k=0)0
Fleischer et al 45 LongitudinalCanada (2006–2011 respondents; 2007–2010 tobacco retailers)Generalised estimating equationsCanadian adult smokers: probability (cessation: 4388 observations, 2024 individuals; relapse n=866 observations, 448 individuals)8 (8)0
Ghenadenik et al 46 Cross-sectional with self-reported retrospective outcome variableCanada (2011–2012 participants; 2011 tobacco retailers)Mixed models with multinomial outcome variableAdults aged 18–25 years proficient in English or French with ≥1 year residence at current address in the Interdisciplinary Study of Inequalities in Smoking: Random (n=2070)9 (9)0
Ghenadenik et al 50 Longitudinal panelMontreal, Canada (2011–2012, 2013–2014 participants; 2011 tobacco retailers)Mixed modelsAdults aged 18–25 years with ≥1 year residence at current address and did not move across study period in the Interdisciplinary Study of Inequalities in Smoking: Random (n=1106)9 (9)0
Ghenadenik et al 51 Longitudinal panelMontreal, Canada (2011–2012, 2013–2014 participants; 2011 tobacco retailers)Mixed modelsAdults aged 18–25 years with ≥1 year residence at current address and did not move across study period in the Interdisciplinary Study of Inequalities in Smoking: Random (n=1025)30 (30)0
Halonen et al 55 LongitudinalFinnish towns (n=10) and hospital districts (n=6) (1997–2005 baseline, 2008–2009 follow-up participants; 2010 tobacco retailers)Generalised estimating equationsEmployees of Finnish towns and hospitals who had reported being current smokers in the Finnish Public Sector study (1997–2005): convenience (n=6663)32 (32)0
Han et al 54 LongitudinalBirmingham and Bristol, UK (2007–2009 participants; 2009 assumed tobacco retailers)Continuation ratio logit modelsPrimary care research setting (29 general practices), current smokers: convenience (n=611)6 (6)2
King et al 76 Cross-sectionalNorth Carolina and Virginia (2017, participants; 2018 tobacco retailers)Mixed generalised linear modelsFirst-year college students in 2010 (Assessment of the Post-College Experience [ACE II] cohort) present in wave 10 of data collection and who live in North Carolina or Virginia (n=1099)12 (12)0
Kirchner et al 67 Cross-sectionalUSA (2012 participants; 2012 tobacco retailers)Generalised categorical regressionOnline panel (English-speaking, aged 25 years or older), current smoking status: quota (n=2376)See evidence table (k=0)2
Kirst et al 52 Cross-sectionalToronto, Canada (2009–2011 participants; 2011 tobacco retailers)Mixed modelsAdult residents who could communicate in English and had been a resident in neighbourhood for at least 6 months (Neighbourhood Effects on Health and Well-being Study): probability based (n=2412)2 (2)0
Marashi-Pour et al 42 Cross-sectionalNew South Wales, Australia (2008–2010 participants; 2009–2011 tobacco retailers)Mixed modelsAdults: probability (n=29 375)2 (2)0
Pearce et al 60 LongitudinalScotland (2000–2015 participants; NR tobacco retailers)Mixed modelsMaternity records: census (~750 000)2 (0)0
Pearce et al 58 Cross-sectionalNew Zealand (2002–2003; 2004 tobacco retailers)Mixed modelsAdults: probability but weights not applied to these analyses (n=12 529)12 (12)0
Pearce et al 61 Cross-sectionalScotland (2008–2011, survey; 2012 tobacco retailers)Logistic regression, weightedScottish adults aged 16 years or older: probability (n=24 387)12 (12)0
Pulakka et al 56 LongitudinalFinland (2003–2012)Conditional logistic regression meta-analysed across studyTwo Finnish studies of participants who had changed their smoking status: one convenience and one representative (n=3641)7 (0)0
Pulakka et al 57 Longitudinal between subjects and case-crossoverFinland (2003, 2008-2013 participants; tobacco retailers 2010 and 2013)Logistic regression (between subjects) and conditional logistic regression (case-crossover)Two Finnish studies: Finish Public Sector study (convenience) and Health and Social Support study (representative) (n=20 729)22 (22)0
Reitzel et al 68 LongitudinalHouston, Texas (2005–2007 participants; 2006 tobacco retailers)Continuation ratio logit modelsAge 21 years or older, smoked five cigarettes per day for last year, English proficiency at sixth-grade reading level or higher, motivation to quit in next 30 days: convenience (n=414)5 (5)1
Rodriguez et al 73 Cross-sectionalUSA (2015 participants; NR tobacco retailers)Logistic regressionPanel Study of Income Dynamics Transition to Adulthood Supplement: Restricted to African-American young adults aged 18–22 years (n=692)4 (0)2
Scheuermann et al 69 Cross-sectionalKansas City, Missouri (participants/tobacco retailers NR)Mixed modelsBlack and white current smokers enrolled in smoking cessation intervention (counselling, varenicline) cohort who had income <400% of the federal poverty level: cohort (n=440)5 (2)1
Shareck et al 48 Cross-sectionalMontreal, Canada (2011–2012 participants; 2011 retailers)Generalised estimating equationsYoung adult (18–25 years) in Interdisciplinary Study of Inequalities in Smoking who were fluent in French or English and had lived at current address for ≥1 year: probability (n=1994)8 (8)0
Shareck et al 47 Cross-sectionalMontreal, Canada (2011–2012 participants; 2011 tobacco retailers)Log-binomial regression modelsYoung adults (aged 18–25 years) in Interdisciplinary Study of Inequalities in Smoking: random, restricted to those having smoked one full cigarette or more in lifetime (n=921)16 (16)0
Vyas et al 77 Cross-sectionalSan Francisco Bay, California (2015–2017 participants; tobacco retailersGeneralised estimating equationsChinese or Vietnamese male, adult (aged 18 years or older), daily smokers (at least one cigarette daily in previous 7 days) who could speak and read Chinese or Vietnamese and who were enrolled in lifestyle intervention study (designed to promote smoking cessation, healthy eating and physical activity): convenience (n=340)1 (1)2
Watkins et al 70 Longitudinal/Ecological momentary assessmentDallas, Texas (2011–2012 participants; 2012 tobacco retailers)Mixed modelsSmoking cessation clinic in safety net healthcare setting; aged 18 years or older, English-speaking at seventh-grade level, willing to quit smoking a week following initial visit, smoked at least five cigarettes per day: convenience (n=47)5 (0)1
Young-Wolff et al 71 Cross-sectionalSan Francisco Bay Area (2006–2013 participants; 2010 tobacco retailers)Mixed modelsInpatient psychiatric healthcare setting, adult smokers living with serious mental illness recruited for three tobacco treatment clinical trials: convenience (n=1061)11 (4)2

  • *n indicates effect sizes reported in record and k indicates effect sizes used in meta-analysis.

  • †RoB assessed for each study (higher numbers indicate higher risk). See the online repository evidence table for full details of predictor and outcome variables.

  • ACE II, Assessment of the Post-College Experience; NR, not reported.