Association between density and proximity of tobacco retail outlets with smoking: A systematic review of youth studies

Abstract

Background:

Reducing the retail availability of tobacco has been proposed as a component of tobacco endgame, yet it is not known whether retail availability has a direct impact on smoking behaviours. A narrative review and a meta-analysis have been undertaken to examine the density and proximity of tobacco retail outlets, but were limited in scope, exposure and outcome variables. The aim of this current study was to undertake a systematic review of the international literature on the density and proximity of tobacco retail outlets to homes, schools and communities and their association with smoking behaviours among youth.

Methods:

We reviewed and critically appraised the evidence documenting the association between density or proximity of tobacco retail outlets and smoking behaviours among school-age youth (18 and under), between 1 January, 1990 and 21 October, 2019. We reviewed original quantitative research that examined the associations of tobacco retail outlet density and proximity with individual smoking status or population-level smoking prevalence; initiation of smoking; frequency of tobacco use; sales to minors; purchasing by minors; susceptibility to smoking among non-smokers; perceived prevalence of smoking, and quitting behaviours.

Findings:

Thirty-five peer-reviewed papers met the inclusion criteria. This review provided evidence of a relationship between density of tobacco retail outlets and smoking behaviours, particularly for the density near youths’ home. A study using activity spaces also found a significant positive association between exposure to tobacco retail outlets and daily tobacco use. The review did not provide evidence of an association between the proximity of tobacco retail outlets to homes or schools and smoking behaviours among youth.

Conclusions:

The existing evidence supports a positive association between tobacco retail outlet density and smoking behaviours among youth, particularly for the density near youths’ home. This review provides evidence for the development and implementation of policies to reduce the density of tobacco retail outlets to reduce smoking prevalence among youth.

Introduction

Most well-established tobacco control measures focus on reducing demand for tobacco,[1] yet supply-side interventions may also help decrease smoking prevalence. Reducing the retail availability of tobacco has been proposed as a component of tobacco endgame,[2] especially in countries where well-developed tobacco control measures, such as tobacco tax or smoking restrictions in public places, already exist.

Internationally, there is wide retail distribution of tobacco, which in some countries such as NZ and Australia, remain one of the few forms of tobacco promotion.[3] Smoking is significantly more prevalent in deprived communities[4] as is the density of tobacco retail outlets (TRO).[5] [6–8] [9] [10] [11] Adults who live in neighbourhoods with higher numbers of TROs are significantly more likely to smoke, or to have tried smoking, compared to those living in areas with fewer TROs.[12–14] The density of TROs has also been shown to be higher in areas of Chicago where a larger proportion of the population were younger than 18 years old.[15] This is concerning given this is a time when many young people experiment with smoking; 4.1% of 14 to 15 year olds in a NZ survey reported experimental smoking.[16] TRO density was highlighted in 2009 as an issue that needed addressing, and yet in 2018 TRO density is still considered the ‘new frontier’ in tobacco control.[17]

Place-based policies limit the quantity, type and location of retailers that can sell tobacco. Examples are caps on the number of retailers in a geographic region or relative to population size, prohibiting or only allowing particular types of retailers to sell tobacco, prohibiting retailers from being located near schools, or regulating distance between retailers.[18] Implementation of place-based policies to restrict availability is increasing at the local level in some US states.[19] For example, San Francisco, California restricts new TRO licenses being issued within 500 feet of a school, within 500 feet of an existing tobacco retailer, to a pharmacy, a tobacco shop, or business whose main purpose is offering food or beverage consumption on the premises. In addition, the City will not issue new licenses until the number of TRO in all districts is as low as the current minimum (an equity goal).[20] In 2013, Hungary introduced regulations that tobacco sales were only allowed at 7,000 government-licensed outlets which reduced the density by approximately 83%. Limited evaluation of these regulations has been undertaken, consequently it is yet to be seen whether these regulations have been effective in reducing smoking.

There is considerable public support among adults for policies to reduce tobacco availability, despite limited evaluation of their efficacy. Quantitative studies in New Zealand have found support for restricting tobacco sales within 500m of a school;[21] implementing mandatory licensing of tobacco retailers;[21] reducing the number of stores selling tobacco;[21, 22] and restricting access to tobacco to adult-only stores.[22] In addition, adult smokers perceive policy scenarios in which tobacco is only sold at half the existing liquor stores or only at pharmacies, as being at least as effective in reducing smoking initiation and supporting cessation as continued tax increases.[23] A study in the Netherlands found support for limiting tobacco sales to specialized stores increased as a function of support for the protection of children against tobacco.[24] In New York City three proposals were supported, including limiting the number of retailers allowed to sell tobacco, prohibiting tobacco sales in pharmacies and at stores located near schools, all of which the city enacted in 2017.[25, 26] A common themes in these findings is support for policies that have a focus on protecting youth, which is associated with greater support among adult smokers.[25]

Although there has been an increasing body of research examining the association between tobacco retail availability and smoking, the results are still inconclusive. For example, a narrative review found the density of and proximity to TROs were correlated with adolescent lifetime smoking, past 12-month smoking, past 30-day smoking, and susceptibility to smoking.[27] However, the authors did not differentiate between exposure to outlets around homes and schools, and only included studies from the USA. and Canada. A recent meta-analysis of 11 studies examined density and proximity of TROs and found a small, but meaningful, association between TRO density around adolescents’ home and their past-month smoking.[28] Although, no significant association was found for the density of TROs around students’ school and their past-month smoking. The meta-analysis was limited by only examining a single outcome variable of past-month smoking, and excluded other important behavioural outcomes.

To date, no systematic review of TRO density and proximity and adolescents’ smoking has been conducted. A systematic review is required in order to inform future policies to restrict TRO availability, by identifying the most salient dimensions of access (e.g., density vs. proximity) and areas of exposure (e.g. home vs. school vs. community). Therefore, the aim of this study was to undertake a systematic review of the international literature on the density and proximity of TROs to homes, schools and communities and their association with smoking behaviours among youth. This would provide important information on which to base decisions regarding the implementation of policies designed to reduce density.

Method

Literature search strategy

The literature searches were conducted in June 2017 by LM and TJ who independently reviewed the articles for eligibility for inclusion. We identified original quantitative studies published in a peer-reviewed journal between 1 January, 1990 and 1 June, 2017 and written in English. We included additional studies after our search date as identified through references and other online searches. Keyword and title searches were undertaken in PubMed, Scopus, and Web of science. The keywords search terms were: (Tobacco or smok* OR cigarette or quit* OR cessation OR relapse) AND (density OR proximity OR access OR availab* or environment or neighborhood* or neighbourhood*) AND (retail* OR store OR outlet OR shop OR grocer*). A simultaneous search process occurred to retrieve articles for three separate systematic reviews; one focused on neighborhood characteristics (e.g., economic disadvantage and race/ethnicity, one on adults only, and this current study on youth. This search was part of a wider study examining smoking outcome variables among adults and youth, however the study presented here focuses solely on school-age youth (18 years and under).

A total of 2,774 non-duplicate articles were retrieved for screening (Figure 1). The titles of these 2,774 retrieved articles were reviewed and 2,676 were discarded, as they were not related to tobacco control. The titles and abstracts of the remaining retrieved articles were imported into Covidence,[29] and these were subsequently screened to identify whether the articles were relevant and met the inclusion criteria below. The full text of articles was examined where further clarification on the measures and study objective was needed. Further searches were conducted using the reference lists and “cited by” lists of retrieved articles and through “related article” searches on Google Scholar. The online editions of Tobacco Control and Nicotine & Tobacco Research, the two highest impact specialist journals on tobacco research, were also scanned for relevant articles. These additional searches elicited an additional 6 studies, giving 104 studies eligible for the wider study. A further nine studies were excluded at this stage as they did not meet the eligibility criteria (Supplementary File 2). A total of 95 studies were selected to be included in the wider study, and 29 of these papers published between 2003 and 2017 were included in this youth systematic review.

Figure 1:

Literature review flow chart

A refreshed search of the literature to include articles through 21 October 2019 was undertaken using the same search strategy described above. This also included a direct search of Tobacco Control, Nicotine and Tobacco Research, and Tobacco Regulatory Science and Addictive Behaviors. An additional six journal articles were identified which met the eligibility criteria for the study, and 19 studies were excluded (Supplementary File 1).

Inclusion criteria

Any type of research design was eligible for this review, providing it met the inclusion and exclusion criteria outlined above. Research was eligible if it included a quantitative exposure measure of density or proximity of retailers of tobacco products (i.e., cigarettes, loose tobacco, cigars, and cigarillos). Exposure measures included proximity of retailers to home, school or an activity space, a count ratio or other method of calculating density (e.g., kernel density estimation (KDE)). Modelling studies that quantified hypothetical scenarios, and research which only assessed density indirectly through proxy variables such as urbanization or number of point of sale (POS) advertisements for tobacco were excluded. Research was also excluded if geographic information systems (GIS) were used to map TROs in relation to schools or other locations, but did not quantify TRO density or proximity.

Inclusion criteria for behavioural outcome measures included individual-level smoking status or population-level prevalence (current smoking, ever-smoker, ex-smoker, occasional smoker, non-daily smoker, experimental smoker, smokeless tobacco use); initiation of smoking; frequency of tobacco use (number of cigarettes per day, moderate/heavy smokers); sales to minors (including enforcement); purchase attempts by minors (including proxy purchases); susceptibility to smoking among non-smokers (susceptibility, intention to smoke in the future); perceived prevalence of smoking, and quitting behaviours (abstinence rates, sustained quitting, attitudes to quitting, readiness to quit, quit attempts). Studies were included if the behavioural outcome measures were for persons or populations characterized as minors or school-age youth, this included people aged 18 years and younger, and those attending primary or secondary schools. Although this age range may include participants who are able to legally purchase tobacco in some jurisdictions, they were included in many school-based studies. Studies were excluded if the outcome measure related only to Electronic Nicotine Delivery Systems (ENDS), perceived access of ENDS, or vape stores. These studies were excluded because all TROs sell tobacco, but many do not sell ENDS.

Critical Appraisal

The methodological strengths and weaknesses of each of the studies were systematically examined by the lead author (LM) and either LR, PV or IGR. Specifically, the validity and suitability of the exposure and outcome measures, the risk of bias and confounding at outcome and study levels, spatial components of the studies, external validity, effect sizes, and overall strength of evidence were assessed for each study. Evidence supporting a causal relationship was assessed according to strength, consistency, reversibility and plausibility of association, and evidence of a dose–response association and a temporal relationship.[30] Supplementary File 2 provides detailed information on each of these items of the critical appraisal of eligible peer-reviewed journal articles. No qualitative studies were included in the systematic review. Data were extracted to a summary table and final critical appraisals were agreed upon following discussion between LM, LR, PV and IGR. The review methods for this study were established prior to the conduct of the review. No significant deviations from the protocol were made.

Results

Thirty-five papers met the inclusion criteria for this review, which is divided into five areas: density of TROs around the home, school and community/activity spaces, and proximity of TROs to home and school. A summary of the critical appraisal of each study is provided in Supplementary File 3, and scoring of each study in Table 1.

Table 1:

Scoring of studies

Author	Overall study design	Exposure	Outcome	Random error	Measurement error	Selection bias	Confounding	GIS	Modelling	External validity	Effect size	Overall
Adachi-Mejia et al., (2012)1	S	M	S	W	M	M	M	M	M	M	W	M
Adams et al., (2013)2	S	S	S	W	M	W	M	M	M	W	W	M
Astuti et al.,(2019)3	S	S	M	M	W	M	W	S	W	M	W	M
Burgoon et al., (2019)4	S	M	M	W	M	W	W	S	S	M	M	M
Chan et al, (2011)5	S	M	M	M	S	M	M	M	M	M	M	M
Gwon et al., (2017)6	S	M	S	M	M	M	M	M	S	M	W	M
Henriksen et al., (2008)7	S	S	S	S	M	S	S	S	M	M	M	S
Kaai et al., (2013)8	S	M	S	S	M	S	M	M	M	S	M	M
Kaai et al., (2014)9	S	M	S	S	M	S	M	M	M	S	W	M
Kaai et al., (2014)10	S	M	S	S	M	S	S	M	M	S	W	M
Kirchner et al.,(2015)11	S	S	S	M	M	S	M	M	S	M	S	M
Kowitt and Lipperman-Kreda (2019)12	S	S	S	W	S	W	W	S	M	W	M	M
Larsen et al., (2017)13	S	M	S	S	M	M	W	M	M	S	M	M
Leatherdale and Strath (2007)14	M	S	M	M	S	M	M	M	M	M	M	M
Lipperman-Kreda et al., (2012)15	M	S	M	W	M	W	M	S	M	M	W	M
Lipperman-Kreda et al., (2012)16	S	M	M	M	M	M	M	W	M	M	S	M
Lipperman-Kreda et al., (2014)17	S	W	M	M	M	M	M	M	M	M	W	M
Lipperman-Kreda et al., (2015)18	M	S	W	W	M	W	W	M	M	W	W	W
Lipperman-Kreda et al., (2016)19	S	S	M	M	M	M	M	W	S	M	S	M
Loomis et al., (2012)20	S	S	S	M	M	M	M	W	M	M	M	M
Lovato et al., (2007)21	M	S	M	M	M	M	W	W	W	S	W	M
Marsh et al., (2016)22	S	S	S	M	M	M	M	S	M	M	M	M
Mason et al., (2016)23	S	M	S	W	M	W	W	M	M	W	M	M
McCarthy et al,, (2007)24	S	S	S	S	S	S	S	M	S	M	M	S
Mennis and Mason (2016)25	M	M	M	M	M	M	M	S	M	W	M	M
Mennis and Mason (2016)26	S	M	S	W	M	W	W	M	M	W	M	M
Mistry et al., (2015)27	S	S	S	S	S	S	M	S	M	M	M	S
Novak et al., (2006)28	S	M	M	M	M	M	M	M	M	M	M	M
Perez et al., (2017)29	M	M	M	M	M	M	M	M	M	W	W	M
Phetphum et al., (2019)30	S	S	W	N/A	W	M	W	M	W	M	W	W
Pokorny et al., (2003)31	S	S	S	M	M	M	M	W	S	M	M	M
Schleicher et al., (2016)32	S	S	S	M	M	M	S	M	S	S	M	S
Scully et al., (2013)33	M	M	M	M	M	W	M	W	M	M	M	M
Shortt et al., (2016)34	S	S	S	M	M	M	S	S	S	S	S	S
Tunstall et al., (2018)35	S	S	S	M	M	M	S	M	M	M	M	M

W=weak, M=moderate, S= strong

References

^1.Adachi-Mejia, A.M., H.A. Carlos, E.M. Berke, et al., A comparison of individual versus community influences on youth smoking behaviours: a cross-sectional observational study. BMJ open, 2012. 2(5): p. e000767.

^2.Lipperman-Kreda, S., C. Mair, J.W. Grube, et al., Density and proximity of tobacco outlets to homes and schools: Relations with youth cigarette smoking. Prev. Sci., 2014. 15(5): p. 738–744.

^3.Mason, M.J., J. Mennis, N.M. Zaharakis, et al., The dynamic role of urban neighborhood effects in a text-messaging adolescent smoking intervention. Nicotine Tob. Res., 2015. 18(5): p. 1039–1045.

^4.Mennis, J., M. Mason, T. Way, et al., The role of tobacco outlet density in a smoking cessation intervention for urban youth. Health & Place, 2016. 38: p. 39–47.

^5.Mennis, J. and M. Mason, Tobacco outlet density and attitudes towards smoking among urban adolescent smokers. Substance abuse, 2016. 37(4): p. 521–525.

^6.Schleicher, N.C., T.O. Johnson, S.P. Fortmann, et al., Tobacco outlet density near home and school: associations with smoking and norms among US teens. Prev. Med., 2016. 91: p. 287–293.

^7.Shortt, N., C. Tisch, J. Pearce, et al., The density of tobacco retailers in home and school environments and relationship with adolescent smoking behaviours in Scotland. Tob Control, 2014: p. tobaccocontrol-2013–051473.

^8.Tunstall, H., N.K. Shortt, C.L. Niedzwiedz, et al., Tobacco outlet density and tobacco knowledge, beliefs, purchasing behaviours and price among adolescents in Scotland. Soc. Sci. Med., 2018. 206: p. 1–13.

^9.Leatherdale, S.T. and J.M. Strath, Tobacco retailer density surrounding schools and cigarette access behaviors among underage smoking students. Ann. Behav. Med., 2007. 33(1): p. 105–11.DOI: 10.1207/s15324796abm3301_12.

^10.Lovato, C.Y., H.C. Hsu, C.M. Sabiston, et al., Tobacco Point-of-Purchase marketing in school neighbourhoods and school smoking prevalence: a descriptive study. Can. J. Public Health, 2007. 98(4): p. 265–70.

^11.Henriksen, L., E. Feighery, N. Schleicher, et al., Is adolescent smoking related to the density and proximity of tobacco outlets and retail cigarette advertising near schools? . Prev. Med., 2008. 47(2): p. 210–214.

^12.McCarthy, W., R. Mistry, Y. Lu, et al., Density of tobacco retailers near schools: effects on tobacco use among students. Am. J. Public Health, 2009. 99(11): p. 2006–13.DOI: 10.2105/AJPH.2008.145128.

^13.Chan, W.C. and S.T. Leatherdale, Tobacco retailer density surrounding schools and youth smoking behaviour: a multi-level analysis. Tob Induc Dis, 2011. 9(1): p. 9.

^14.Adams, M.L., L.A. Jason, S. Pokorny, et al., Exploration of the link between tobacco retailers in school neighborhoods and student smoking. J. Sch. Health, 2013. 83(2): p. 112–118.

^15.Kaai, S.C., S.T. Leatherdale, S.R. Manske, et al., Using student and school factors to differentiate adolescent current smokers from experimental smokers in Canada: a multilevel analysis. Prev. Med., 2013. 57(2): p. 113–119.

^16.Scully, M., M. McCarthy, M. Zacher, et al., Density of tobacco retail outlets near schools and smoking behaviour among secondary school students. Aust. N. Z. J. Public Health, 2013. 37(6): p. 574–578.

^17.Kaai, S., S. Manske, S. Leatherdale, et al., Are experimental smokers different from their never-smoking classmates? A multilevel analysis of Canadian youth in grades 9 to 12. Chronic diseases and injuries in Canada, 2014. 34(2–3).

^18.Mistry, R., M. Pednekar, S. Pimple, et al., Banning tobacco sales and advertisements near educational institutions may reduce students’ tobacco use risk: evidence from Mumbai, India. Tob Control, 2013: p. tobaccocontrol-2012–050819.

^19.Marsh, L., A. Ajmal, R. McGee, et al., Tobacco retail outlet density and risk of youth smoking in New Zealand. Tobacco Control, 2015. 25(e2): p. 71–75.DOI: 10.1136/tobaccocontrol-2015-052512.

^20.Kaai, S.C., K.S. Brown, S.T. Leatherdale, et al., We do not smoke but some of us are more susceptible than others: A multilevel analysis of a sample of Canadian youth in grades 9 to 12. Addict. Behav., 2014. 39(9): p. 1329–1336.

^21.Gwon, S., G. Yan, G. Huang, et al., The influence of tobacco retailers on adolescent smoking: prevention and policy implications. International nursing review, 2017.

^22.Larsen, K., T. To, H.M. Irving, et al., Smoking and binge-drinking among adolescents, Ontario, Canada: Does the school neighbourhood matter? Health & Place, 2017. 47: p. 108–114.

^23.Pérez, A., L.-C. Chien, M.B. Harrell, et al., Geospatial associations between tobacco retail outlets and current use of cigarettes and e-cigarettes among youths in Texas. Journal of biometrics & biostatistics, 2017. 8(5).

^24.Pokorny, J., V.C. Smith, and S.B. Pokorny, The relation of retail tobacco availability to initiation and continued smoking. Journal of Clinical Child & Adolescent Psychology, 2003. 32(2): p. 193–204.DOI: 10.1207/S15374424JCCP3202_4

^25.Novak, S., S. Reardon, S. Raudenbush, et al., Retail tobacco outlet density and youth cigarette smoking: a propensity-modeling approach. Am. J. Public Health, 2006. 96(4): p. 670–6.DOI: 10.2105/AJPH.2004.061622.

^26.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Local tobacco policy and tobacco outlet density: Associations with youth smoking Prev. Med., 2012.DOI: 10.1016/j.jadohealth.2011.08.015.

^27.Loomis, B.R., A.E. Kim, A.H. Busey, et al., The density of tobacco retailers and its association with attitudes toward smoking, exposure to point-of-sale tobacco advertising, cigarette purchasing, and smoking among New York youth. Prev. Med., 2012. 55(5): p. 468–474.

^28.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Contextual and community factors associated with youth access to cigarettes through commercial sources. Tobacco Control, 2012.DOI: 10.1136/tobaccocontrol-2012-050473.

^29.Lipperman-Kreda, S., C. Morrison, J.W. Grube, et al., Youth activity spaces and daily exposure to tobacco outlets. Health & Place, 2015. 34: p. 30–33.

^30.Lipperman‐Kreda, S., J.W. Grube, K.B. Friend, et al., Tobacco outlet density, retailer cigarette sales without ID checks and enforcement of underage tobacco laws: associations with youths’ cigarette smoking and beliefs. Addiction, 2016. 111(3): p. 525–532.

^31.Burgoon, M.L., T. Albani, B. Keller-Hamilton, et al., Exposures to the tobacco retail environment among adolescent boys in urban and rural environments. The American journal of drug and alcohol abuse, 2019. 45(2): p. 217–226.

^32.Kowitt, S.D. and S. Lipperman-Kreda, How is exposure to tobacco outlets within activity spaces associated with daily tobacco use among youth? A mediation analysis. Nicotine Tob. Res., 2019.

^33.Kirchner, T.R., A.C. Villanti, J. Cantrell, et al., Tobacco retail outlet advertising practices and proximity to schools, parks and public housing affect Synar underage sales violations in Washington, DC. Tobacco Control, 2014: p. tobaccocontrol-2013–051239.

^34.Astuti, P.A., K.H. Mulyawan, S.K. Sebayang, et al., Cigarette retailer density around schools and neighbourhoods in Bali, Indonesia: A GIS mapping. Tobacco induced diseases, 2019. 17.

^35.Phetphum, C. and N. Noosorn, Tobacco Retailers Near Schools and the Violations of Tobacco Retailing Laws in Thailand. Journal of Public Health Management and Practice, 2019. 25(6): p. 537–542.

Density

Eight studies examined the density of TROs around adolescents’ homes, including one longitudinal ecological momentary assessment (EMA) study.[31] An EMA study “involves repeated sampling of subjects’ current behaviors and experiences in real time, in subjects’ natural environments”.[32] Most studies were conducted in the US, [14, 31, 33–37] and two in Scotland[38] (Table 2). Overall, tobacco density around the home was found to be associated with increased smoking behaviour in six of the eight studies.[14, 31, 34–37] The effect size varied from odds ratio (OR) 1.01 (95% CI=1.00 to 1.2)[35] to OR 1.53 (95% CI=1.27 to 1.85) for ever smoking and OR 1.47 (95% CI=1.13 to 1.91) and OR 1.35 (95% CI=1.10 to 1.67) for current smoking in high and low TRO density areas, respectively.[34] All except one study[31] were cross sectional in design, and as such unable to reveal a causal relationship. There was a four-fold difference in distance buffers used to measure the density of TROs around homes from 400m[38] to one-mile (1600m).[14] Four studies derived data from surveys and some suffered from low response rates,[35] loss to follow up[14, 33] and acknowledged the potential for under-representation of certain populations at risk (e.g., African Americans and those from low socio-economic backgrounds). Mason et al.’s (2015) EMA study[31] is useful in exploring the smoking behaviour among understudied populations, however, the sample was small and therefore the results should be interpreted with caution. In addition the convenience sample utilised in three studies could have affected the representativeness of the results.[36] [37] [31] In these studies the location of adolescents’ homes may also have been clustered, potentially indicating they may not be independent observations.

Table 2:

Density of tobacco retailers to home

Reference Location Participants Funding	Study design Sample selection Sample size	TRO and Spatial	Exposure(s)	Outcome(s)	Analytical method	Adjustment	Results
Adachi-Mejia et al., (2012)1 USA national study. 13–18 years. This work was supported by the National Institutes of Health.	Cross-sectional HH telephone survey (2007). Data came from wave 5, n=3,055, RR 47%. Additional sample of African American adolescents included due to attrition=598. Geocoded sample = 3646.	Geocoded national dataset of retailers likely to sell tobacco (2007). Road network used. KDE used. ArcGIS used.	Density of TROs per 1000 people.	Ever tried smoking Smoking intensity scale.	Multiple and ordinal logistic regression.	Individual: Gender, age, race, exposure to movie smoking, team sports participation, sensation seeking. Interpersonal: sibling smoking and friend smoking. Community: SES, Proportion black and Hispanic, poverty.	1a. No significant association (OR=1.27, 95% CI=0.92–1.76) 1b. No significant association (OR=1.11, 95% CI=0.70–1.79)
Lipperman-Kreda et al., (2012)2 California, USA. 13–18 years. National Cancer Institute and Tobacco-Related Disease Research Program.	Cross-sectional HH telephone survey in 45 Californian cities (2010) 3,062 HH were sampled from a purchased list. N=1,543 participated in wave 1 (RR= 50%). N= 1,312 also completed wave 2 (RR= 85%). Geocoded sample= 832.	Field observations to document TRO addresses (2009). Euclidean distance used. GIS used.	TRO density 0.75-mile of home TRO density 1.0-mile of home.	Smoking frequency (days smoked in past 30).	Negative binomial model.	Individual: gender, age, and race. Community: Population density, percent Hispanic, percent African American, percent unemployed, percent under 18 years, median HH income, percent college graduates.	1a. Significant positive association (β=0.293, robust SE=0.069, p≤0.05) 2a. Significant positive association (β=0.340, robust SE=0.082, p≤0.05)
Mason et al., (2016)3 Richmond, Virginia, USA. 14–18 years. Virginia Foundation for Healthy Youth.	Ecological Momentary Assessment (2013–2014). 197 primarily urban African Americans (91%) recruited using a convenience sample. Adolescents with Fagerström screening score of 1 were included.	Geocoded North American Industry Classification System of businesses most likely to sell tobacco (year not stated). Road network used. ArcGIS used.	TRO density around home - number of TROs within 0.5-mile service area of each subject’s home.	Momentary smoking.	Time-Varying Effects Models.	Individual: Age, gender and baseline nicotine dependence.	1a. Over the 6 months the treatment group had reduced odds to smoke, the control group had increased odds over month 1, 2, 5, & 6. Separation between CI in month 2 & 6 showing TRO density and smoking was significantly different between groups. Trajectories diverging at month 6.
Mennis and Mason (2016)4 Richmond, Virginia, USA. 14–18 years. Virginia Foundation for Healthy Youth.	Randomised controlled trial of a text messaging based tobacco cessation intervention (2013–2014). 197 primarily urban African Americans (91%) recruited using a convenience sample. Adolescents with Fagerström screening score of 1 were included.	American Industry Classification System of businesses most likely to sell tobacco (year not stated). Road network used. ArcGIS used.	TRO density around home - number of TROs within 0.5-mile service area of each subject’s home.	Days smoked (past month). Mediating variable- intention to smoke in next 3 months (0–4). Moderating variable: smoking cessation intervention: treatment or control group.	Moderated mediation model.	Individual: days smoked at baseline, age, and gender. Interpersonal: peer smoking and family smoking	1a. The direct effect of tobacco outlet density on days smoked at one month was not significant. (β=−0.055, SE=0.106, p=0.605). 1b. Smoking intention mediates the effect of TRO density on days smoked at one month (β=0.046, SE=0.029, 95% CI=0.006–0.125). Experimental condition moderates the indirect effect of TRO density on days smoked at one month through the mediating variable smoking intention (β=0.261, SE=0.106, p=0.015).
Mennis and Mason (2016)5 Richmond, Virginia, USA. 14–18 years. Virginia Foundation for Healthy Youth.	Cross-sectional baseline data for an intervention study (2013–2014). 197 primarily urban African Americans (91%) recruited using a convenience sample. Adolescents with Fagerström screening score of 1 were included.	Geocoded North American Industry Classification System of businesses most likely to sell tobacco (year not stated). Road network used. ArcGIS used.	TRO density around home - number of TROs within 0.5-mile service area of each subject’s home.	Intend to smoke next 3 months Smoking in 5 years? How ready are you to stop smoking?	Ordinal regression.	Individual: age, gender NS nicotine dependence Interpersonal: smoker in residence, and friends who smoke.	1a. Significant positive association (β=0.130, 95% CI=0.034–0.225). 1b. No significant association (β=0.065, 95% CI=−0.033–0.163). 1c. Significant positive association (β=−0.152, 95% CI=−0.250-[−0.053]).
Schleicher et al., (2016)6 USA national. 13–16 years. NIH Public Health Service and the National Cancer Institute.	Cross-sectional internet panel of one eligible teen and parent from each HH (2011–2012). RR= 40%. N=2,771.	Purchased address data for likely TROs in all zip codes that contained or adjacent to each teen’s residence (year not stated). Road network used. ArcGIS used.	TROs per mile within 0.5-mile of home addresses.	Ever tried a cigarette.	Logistic regression analysis.	Individual: age, gender, and grades. Interpersonal: close friends smoke, household race/ ethnicity, smokers in HH, and HH income. Community: race, ethnicity, and percent population in poverty.	1a. Significant positive association (OR=1.01, 95% CI=1.00–1.02, p<0.01).
Shortt et al., (2016)7 Scotland. 13–15 years. Scottish Collaboration for Public Health Research and Policy.	Cross-sectional national school survey (2010–2011). Final nationally representative sample of 37,307 pupils (RR = 91%). N= 20,446 who had addresses and outcome variable included in study.	Addresses of all premises registered on the Scottish Tobacco Retailers Register (2012). Density estimated with KDE. Density of retailers per km2 then extracted to postcode area centroids.	TRO density around the home– 800m.	Ever smoking Current smoking – do you smoke nowadays.	Logistic regression.	Individual: age, sex, ethnicity, free school meals, and self-perceived family wealth. Interpersonal: family structure and parental smoking status. Community: urban/rural, SES.	1a. Significant positive dose- response: Pupils living in higher density areas had significantly increased odds of ever smoking (OR=1.53, 95% CI=1.27–1.85) compared with to those with no outlets in the area (OR 1.42, 95% CI 1.20 to 1.67). 1b. Significant negative dose- response: Pupils living in areas of high density had significant lower odds of current smoking (OR=1.47, 95% CI=1.13–1.91) compared with to those with no outlets in the area (OR 1.35, 95% CI 1.10 to 1.67).
Tunstall et al., (2018)8 Scotland, national study. 13–15 years. European Research Council. University of Glasgow Neighbourhoods and Communities Programme.	SALSUS national biennial survey of 37,307 from 1851 classes in 302 schools. RR= 62% (Sept 2010 – Feb 2011). 40.9% were missing postcode information and excluded from study. Final sample 22,049.	TROs data from Scottish Tobacco Retailers register (sept 2012). Density divided into zero (32%), low (34%) and High (34%). KDE was used. ArcGIS.	Density of TROS within 400m of the adolescents’ residential postcode.	Current smoker purchasing from a shop. Current smoker proxy purchase Mediators: Parent smoking Friend smoking.	Path analysis – structured equation model.	Individual: sex, age, ethnic group, and perception of family economic status. Community: Neighbourhood deprivation and urban/rural status.	1a. No significant standardised direct effect (standardized β=−0.01, p≥0.05). Outlet density had a significant direct effect upon parents’ smoking behaviour among current smokers buying tobacco from shops (standardized β=0.10, p<0.05). Outlet density had a significant direct effect upon the likelihood of adolescent friends’ smoking among current smokers buying tobacco from shops (standardized β=0.08, p<0.05). 1b. No significant standardised direct effect (standardized β=−0.03, p≥0.05). Outlet density had a significant direct effect upon parents’ smoking behaviour among current smokers buying tobacco through proxies (standardized β=0.10, p<0.05). Outlet density had a significant direct effect upon the likelihood of adolescent friends’ smoking among current smokers buying tobacco through proxies (standardized β= 0.08, p<0.05).

RR= response rate, TRO = tobacco retailer, HH= household SES= socioeconomic status

References

^1.Adachi-Mejia, A.M., H.A. Carlos, E.M. Berke, et al., A comparison of individual versus community influences on youth smoking behaviours: a cross-sectional observational study. BMJ open, 2012. 2(5): p. e000767.

^2.Lipperman-Kreda, S., C. Mair, J.W. Grube, et al., Density and proximity of tobacco outlets to homes and schools: Relations with youth cigarette smoking. Prev. Sci., 2014. 15(5): p. 738–744.

^3.Mason, M.J., J. Mennis, N.M. Zaharakis, et al., The dynamic role of urban neighborhood effects in a text-messaging adolescent smoking intervention. Nicotine Tob. Res., 2015. 18(5): p. 1039–1045.

^4.Mennis, J., M. Mason, T. Way, et al., The role of tobacco outlet density in a smoking cessation intervention for urban youth. Health & Place, 2016. 38: p. 39–47.

^5.Mennis, J. and M. Mason, Tobacco outlet density and attitudes towards smoking among urban adolescent smokers. Substance abuse, 2016. 37(4): p. 521–525.

^6.Schleicher, N.C., T.O. Johnson, S.P. Fortmann, et al., Tobacco outlet density near home and school: associations with smoking and norms among US teens. Prev. Med., 2016. 91: p. 287–293.

^7.Shortt, N., C. Tisch, J. Pearce, et al., The density of tobacco retailers in home and school environments and relationship with adolescent smoking behaviours in Scotland. Tob Control, 2014: p. tobaccocontrol-2013–051473.

^8.Tunstall, H., N.K. Shortt, C.L. Niedzwiedz, et al., Tobacco outlet density and tobacco knowledge, beliefs, purchasing behaviours and price among adolescents in Scotland. Soc. Sci. Med., 2018. 206: p. 1–13.

^9.Leatherdale, S.T. and J.M. Strath, Tobacco retailer density surrounding schools and cigarette access behaviors among underage smoking students. Ann. Behav. Med., 2007. 33(1): p. 105–11.DOI: 10.1207/s15324796abm3301_12.

^10.Lovato, C.Y., H.C. Hsu, C.M. Sabiston, et al., Tobacco Point-of-Purchase marketing in school neighbourhoods and school smoking prevalence: a descriptive study. Can. J. Public Health, 2007. 98(4): p. 265–70.

^11.Henriksen, L., E. Feighery, N. Schleicher, et al., Is adolescent smoking related to the density and proximity of tobacco outlets and retail cigarette advertising near schools? . Prev. Med., 2008. 47(2): p. 210–214.

^12.McCarthy, W., R. Mistry, Y. Lu, et al., Density of tobacco retailers near schools: effects on tobacco use among students. Am. J. Public Health, 2009. 99(11): p. 2006–13.DOI: 10.2105/AJPH.2008.145128.

^13.Chan, W.C. and S.T. Leatherdale, Tobacco retailer density surrounding schools and youth smoking behaviour: a multi-level analysis. Tob Induc Dis, 2011. 9(1): p. 9.

^14.Adams, M.L., L.A. Jason, S. Pokorny, et al., Exploration of the link between tobacco retailers in school neighborhoods and student smoking. J. Sch. Health, 2013. 83(2): p. 112–118.

^15.Kaai, S.C., S.T. Leatherdale, S.R. Manske, et al., Using student and school factors to differentiate adolescent current smokers from experimental smokers in Canada: a multilevel analysis. Prev. Med., 2013. 57(2): p. 113–119.

^16.Scully, M., M. McCarthy, M. Zacher, et al., Density of tobacco retail outlets near schools and smoking behaviour among secondary school students. Aust. N. Z. J. Public Health, 2013. 37(6): p. 574–578.

^17.Kaai, S., S. Manske, S. Leatherdale, et al., Are experimental smokers different from their never-smoking classmates? A multilevel analysis of Canadian youth in grades 9 to 12. Chronic diseases and injuries in Canada, 2014. 34(2–3).

^18.Mistry, R., M. Pednekar, S. Pimple, et al., Banning tobacco sales and advertisements near educational institutions may reduce students’ tobacco use risk: evidence from Mumbai, India. Tob Control, 2013: p. tobaccocontrol-2012–050819.

^19.Marsh, L., A. Ajmal, R. McGee, et al., Tobacco retail outlet density and risk of youth smoking in New Zealand. Tobacco Control, 2015. 25(e2): p. 71–75.DOI: 10.1136/tobaccocontrol-2015-052512.

^20.Kaai, S.C., K.S. Brown, S.T. Leatherdale, et al., We do not smoke but some of us are more susceptible than others: A multilevel analysis of a sample of Canadian youth in grades 9 to 12. Addict. Behav., 2014. 39(9): p. 1329–1336.

^21.Gwon, S., G. Yan, G. Huang, et al., The influence of tobacco retailers on adolescent smoking: prevention and policy implications. International nursing review, 2017.

^22.Larsen, K., T. To, H.M. Irving, et al., Smoking and binge-drinking among adolescents, Ontario, Canada: Does the school neighbourhood matter? Health & Place, 2017. 47: p. 108–114.

^23.Pérez, A., L.-C. Chien, M.B. Harrell, et al., Geospatial associations between tobacco retail outlets and current use of cigarettes and e-cigarettes among youths in Texas. Journal of biometrics & biostatistics, 2017. 8(5).

^24.Pokorny, J., V.C. Smith, and S.B. Pokorny, The relation of retail tobacco availability to initiation and continued smoking. Journal of Clinical Child & Adolescent Psychology, 2003. 32(2): p. 193–204.DOI: 10.1207/S15374424JCCP3202_4

^25.Novak, S., S. Reardon, S. Raudenbush, et al., Retail tobacco outlet density and youth cigarette smoking: a propensity-modeling approach. Am. J. Public Health, 2006. 96(4): p. 670–6.DOI: 10.2105/AJPH.2004.061622.

^26.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Local tobacco policy and tobacco outlet density: Associations with youth smoking Prev. Med., 2012.DOI: 10.1016/j.jadohealth.2011.08.015.

^27.Loomis, B.R., A.E. Kim, A.H. Busey, et al., The density of tobacco retailers and its association with attitudes toward smoking, exposure to point-of-sale tobacco advertising, cigarette purchasing, and smoking among New York youth. Prev. Med., 2012. 55(5): p. 468–474.

^28.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Contextual and community factors associated with youth access to cigarettes through commercial sources. Tobacco Control, 2012.DOI: 10.1136/tobaccocontrol-2012-050473.

^29.Lipperman-Kreda, S., C. Morrison, J.W. Grube, et al., Youth activity spaces and daily exposure to tobacco outlets. Health & Place, 2015. 34: p. 30–33.

^30.Lipperman‐Kreda, S., J.W. Grube, K.B. Friend, et al., Tobacco outlet density, retailer cigarette sales without ID checks and enforcement of underage tobacco laws: associations with youths’ cigarette smoking and beliefs. Addiction, 2016. 111(3): p. 525–532.

^31.Burgoon, M.L., T. Albani, B. Keller-Hamilton, et al., Exposures to the tobacco retail environment among adolescent boys in urban and rural environments. The American journal of drug and alcohol abuse, 2019. 45(2): p. 217–226.

^32.Kowitt, S.D. and S. Lipperman-Kreda, How is exposure to tobacco outlets within activity spaces associated with daily tobacco use among youth? A mediation analysis. Nicotine Tob. Res., 2019.

^33.Kirchner, T.R., A.C. Villanti, J. Cantrell, et al., Tobacco retail outlet advertising practices and proximity to schools, parks and public housing affect Synar underage sales violations in Washington, DC. Tobacco Control, 2014: p. tobaccocontrol-2013–051239.

^34.Astuti, P.A., K.H. Mulyawan, S.K. Sebayang, et al., Cigarette retailer density around schools and neighbourhoods in Bali, Indonesia: A GIS mapping. Tobacco induced diseases, 2019. 17.

^35.Phetphum, C. and N. Noosorn, Tobacco Retailers Near Schools and the Violations of Tobacco Retailing Laws in Thailand. Journal of Public Health Management and Practice, 2019. 25(6): p. 537–542.

Eighteen studies examined the association between the density of TROs around schools and smoking behaviours. All of these involved cross-sectional, school-based surveys, except two, which involved a household telephone[14] and an internet survey.[35] Seven of these studies were undertaken in Canada,[39–45] six in the US,[14, 35, 46–49] and one in each NZ,[16] India,[50] Scotland,[34] South Korea[51] and Australia[52] (Table 3). The available studies do not provide strong evidence of an association between density of TROs around schools and ever or current smoking among adolescents at those schools. The most consistent evidence of an association was found for susceptibility to future smoking among non-smokers.[16, 41] The effect size ranged from adjusted odds ratio (AOR) 1.03 (95%CI=1.01 to 1.05)[41] to AOR 1.09 (95% CI=1.03 to 1.14).[16] The only other study to examine this outcome did not find an association.[44] For all other outcome variables there were no consistent findings across the studies, this was particularly evident for ever and current smoking.

Table 3:

Density of tobacco retailers to school

Reference Location Participants Funding	Study design Sample selection Sample size	TRO and Spatial	Exposure(s)	Outcome(s)	Analytical method	Adjustment	Results
Leatherdale and Strath (2007)9 2001 Ontario, Canada. 14–18 years. Canadian Cancer Society/ National Cancer Institute of Canada’s Centre for Behavioral Research and Program Evaluation. Population Health Research Group - University of Waterloo. Social Sciences and Humanities Research Council of Canada. Ontario Tobacco Research Unit.	Cross-sectional school survey (2000–2001). N= 19,464 eligible students (96% RR) at 29 Ontario schools completed the survey.	TROs counted by observers within a 6 block radius of schools (year not stated). Road network used. No GIS used.	TRO in a 6 block of schools.	Smoker – occasional or regular Buys own cigarettes .	Multilevel logistic regression modelling.	Individual: gender, age, , smoking status, no of cigarettes smoked per day, smokes with family, frequency of smoking during school day, frequency of smoking on weekends, and frequency of being asked age when purchasing. Interpersonal: number of close friends who smoke, parent who smokes, and siblings smoke, School: school smoking rate.	1a. No significant association (OR=1.01, 95% CI=0.99–1.03). 1b. Positive association (OR=1.04, 95% CI=1.01–1.08).
Lovato et al., (2007)10 Five provinces of Canada. 15–19 years. The Canadian Institute for Health Research.	Cross-sectional school survey (year not stated). Cluster sampling used to select municipalities, then 81 secondary schools randomly selected. N= 22,318.	Store observations of TROs (year not stated). Euclidean used. No GIS used.	TROs within 1km of school.	School smoking prevalence (number of students at school who smoked at least a few puffs on ≥2 days past month divided by total number of student participants at school) “High” versus “low” school smoking prevalence (median split of school smoking prevalence as defined above).	Bivariate association assessed using Pearson correlation coefficients. ANOVA.	No adjustment made.	1a. No significant association (Pearson’s r=0.16, p>0.05) 1b. No results reported.
Henriksen et al., (2008)11 California, USA. High school students (no age given). California Tobacco-Related Disease Research Program.	Cross-sectional California Student Tobacco Survey (2005– 2006). N= 24,875 students from 135 randomly selected high schools. School RR = 87.4%, student RR= 79.4%.	TRO data from state licensing data (year not stated). Euclidean distance used. GIS likely, no software mentioned.	Density of TROs within 0.5-mile of school.	School smoking prevalence past 30 days. Number of cigarettes smoked (past 30 days – school level weighted average).	Ordinary least squares regression.	School: race, ethnicity and proportion qualified for free or reduced price meal. Community: median HH income, population density, and neighbourhood type.	1a. Smoking prevalence significantly higher at schools in neighbourhoods with the highest TRO density (>5) than neighbourhoods with no TROs (β=3.2, 95% CI=0.8–5.6). Smoking prevalence not significantly higher at schools in neighbourhoods with moderate (1–4) TRO density than neighbourhoods with no TROs (β=1.7, 95% CI=−0.3–3.7) 1b. The number of cigarettes smoked was not associated with density of TROs (no data given).
McCarthy et al,, (2007)12 California, USA. Average age 14.9 years. Middle and High school students. California Tobacco-Related Disease Research Program.	Cross-sectional California Student Tobacco Survey (2003–2004). N= 19,306 students from 245 randomly selected high schools. School RR = 85%, student RR= 66%.	California Board of Equalization on tobacco retail licensees (2006). Average straight-line distance. Used batch geocoder and Arc GIS.	Density – number of TROs within 1.0-mile radius of the school.	Experimental smoking Established smoking High school students experimental smoking Middle school students experimental smoking No. of cigs smoked on days smoked Purchase tobacco from a store.	Multilevel logistic and random-intercept models in a generalized linear mixed-model framework.	Individual: age, gender, ethnicity, English use at home, previous years grades, hopelessness in past year, perceived ease of access, and depressive symptoms. Interpersonal: peer smoking and best friend smoking. School: school level: average parental education, rural/urban, type of school.	1a. A positive association (OR=1.11, 95% CI=1.02–1.21). 1b. No significant association (OR=1.06, 95% CI=0.94–1.20). 1c. A positive association (OR=1.18, 95% CI=1.06–1.30). 1d. No significant association (OR=0.91, 95% CI=0.76–1.08). 1e. No significant association (no data given). 1f. No significant association (no data given).
Chan et al, (2011)13 Ontario, Canada. Grade 9–12 students (ages 14–17 years). Ontario Ministry of Health and Long-Term Care/Ministry of Health Promotion and by Cancer Care Ontario. Canadian Cancer Society	Cross-sectional School survey (2005–2006). Of the 36,175 students eligible to complete the survey in 76 Ontario schools, 74% completed the survey, and 72% provided complete data. N = 25,893	TRO data obtained from a national database of Canadian business (2005–2006). Euclidean buffers used. Arcview used. Possibly spatial auto-correlation.	TRO within 1km of school.	Smoking susceptibility Occasional smokers Daily smokers.	Multi-level logistic regression.	Individual: sex and grade. Interpersonal: older sibling who smokes, parent who smokes, and number of close friends who smoke. Community: neighbourhood disadvantage.	1a. Significant positive association (OR=1.03, 95% CI=1.01–1.05). Sex interaction found. 1b. No significant association (OR=0.99, 95% CI=0.97–1.01). 1c. No significant association (OR=1.00, 95% CI=0.98–1.02).
Adams et al., (2013)14 2002 Illinois, USA. Grade 7–10 students (ages 15–16 years). National Cancer Institute.	Cross -sectional school survey (2000). Of 19,837 eligible students, 9,704 had parental consent, completed the survey and went to a school in the area where tobacco purchase attempts were made. Final sample included 21 middle and 13 high schools.	TRO data obtained from Illinois liquor Control Commission’s merchant compliance program (2002). Euclidean was used. ArcGIS used	TRO density within 0.5-mile of school.	Ever smoking Current smoking.	A random-effects regression analysis was performed.	Individual: grade, sex, and race/ethnicity. Community: Illegal sale rates, median neighbourhood income, mean neighbourhood density.	1a. No significant association (OR=1.10, 95% CI=0.99–1.20). 1b. No significant association (OR=1.04, 95% CI=0.95–1.14).
Kaai et al., (2013)15 Ten Canadian Provinces. Grade 9–12 students (ages 14–17 years). Health Canada and the Ontario Ministry of Health and Long-term Care.	Cross-sectional school survey (2008–2009). Nationally representative sample of 29,296 students attending 133 schools from all 10 Canadian provinces. This study used 5,440 sub sample of current and experimental smokers.	TRO data obtained from a national database of Canadian business (2008–2009). Euclidean was used. ArcGIS used.	TRO density within 1km of the school.	Current smoking (occasional or daily) vs experimental smoking.	Multilevel logistic regression.	Individual: gender, grade, alcohol use, marijuana use, school connectedness, knowledge, and beliefs. Interpersonal: parents smoke, siblings smoke, friends smoke, School: school location. Community: SES (median HH income).	1a. Significant positive association (OR=1.03, 95% CI=1.01–1.05).
Scully et al., (2013)16 Victoria, Australia. 12–17 years. Victorian Department of Health.	Cross-sectional school survey (2008). A total of 4,398 students from 67 randomly selected Victorian schools. This study includes data for 2,044 students from 35 schools for which TRO density and cigarette price information was available. Student RR = 78%.	TRO data collected by field workers (2008). Euclidean radius used. No GIS used.	TRO density within 500m of school.	Past month smoking Number of cigarettes smoked in past week.	Multilevel logistic regression.	Individual: age, sex, SES, spending money, and perceived ease of buying cigarettes. Interpersonal: parental smoking status. School: school type.	1a. No significant association (OR=1.06, 95% CI=0.90–1.24). 1b. Significant positive association (OR=1.03, 95% CI=1.02–1.26).
Kaai et al., (2014)17 Ten provinces of Canada. Grade 9–12 students (ages 14–17 years). Health Canada. Propel Centre for Population Health Impact. Prince Edward Island Department of Education and Early Childhood Development. Ontario Ministry of Health and Long-term Care.	Cross-sectional school survey (2008–2009). Survey administered to 29,296 students attending 133 schools from all 10 Canadian provinces. The student RR= 73%. This study used 18,072 students who were experimental or never smokers.	TRO data obtained from a national database of Canadian business (2008–2009). Euclidean was used. ArcGIS used.	TRO density within 1km of school.	Experimental smoking.	Multilevel logistic regression.	Individual: gender, grade, alcohol use, marijuana use, pocket money, school connectedness, knowledge and attitudes towards tobacco, perception of school smoking status. Interpersonal: parents smoke, siblings smoke, friends smoke, School: school location. Community: SES.	1a. No significant association (OR=0.99, 95% CI=0.97–1.02).
Lipperman-Kreda et al., (2014)2 45 California cities. 13–18 years. National Cancer Institute and Tobacco-Related Disease Research Program.	Cross-sectional HH telephone survey (2010). 3062 HH were sampled from a purchased list. N=1,543 participated in the wave 1 (RR= 50%). N= 1,312 also completed wave 2 (RR= 85%). Geocoded sample= 832.	Field observations to document TRO addresses (year not stated). Euclidean distance used. GIS used.	TRO density 0.75-mile of school TRO density 1.0-mile of school.	Smoking frequency (no of days smoked in past 30).	Negative binomial model.	Individual: gender, age, and race. Community: population density, percent Hispanic, percent African American, percent unemployed, percent under 18 years, median HH income, and percent college graduates.	1a. No significant association (β=0.124, robust SE=0.073, p>0.05) 2a. No significant association (β=0.064, robust SE=0.058, p>0.05)
Mistry et al., (2015)18 Mumbai, India. Grade 8–10 students (ages unknown). Fulbright-Nehru Scholar Program and the Jonsson Cancer Center Foundation.	Cross-sectional school survey (2010). 26 schools sampled from all public and private high schools in Mumbai. One 8th, one 9th and one 10th class was randomly sampled at each school. The school-level and class-level response rates were 100% and the student-level response rate was 99%. N =1,533.	Field GIS data were collected to obtain TRO addresses (2010). Euclidean radius. GIS used.	TRO within 100m of school TRO within 200m of school TRO within 300m of school TRO within 400m of school TRO within 500m of school.	Ever tobacco use (ever tried tobacco (list given)) Current tobacco use (smoked or chewed tobacco past 30 days) Current smokeless tobacco use (chewed tobacco past 30 days).	Random-effects logistic regression .	Individual: gender, age, religion, pocket money, parent use, peer use, ease of access, hopelessness, and school fee structure.	1a. No significant association for medium density (OR=1.01, 95% CI=0.53–1.91) or high density (OR=0.72, 95% CI=0.36–1.45), compared with low density. 2b. Significant positive association for medium density (OR=2.25, 95% CI=1.19 −4.25) and high density (OR=2.05, 95% CI=1.02–4.13), compared with low density. 2c. Significant positive association for high density (OR=1.49, 95% CI=1.01–2.82), compared with low density. 3b. Significant positive association for high density (OR=2.16, 95% CI=1.06–4.40), compared with low density. 3c. Significant positive association for high density (OR=2.12, 95% CI=1.04–4.32), compared with low density. 4b. A significant positive association for medium density (OR=2.40, 95% CI=1.18–4.87) and high density (OR=2.74, 95% CI=1.50–5.01), compared with low density. 4c. Significant positive association for high density (OR=2.60, 95% CI=1.36–4.97), compared with low density. The remaining relationships were non-significant.
Marsh et al., (2016)19 National, New Zealand. 14–15 years. Ministry of Health. Cancer Society of New Zealand. New Zealand Lottery Health Scholarship.	All 621 secondary schools with a Year 10 class were invited and 298 schools took part (RR=48%) (2012). 59,627 Year 10 students were enrolled nationally and 28,443 participated (RR= 48%). N= 27,238 completed the survey and were included in the analysis.	TRO obtained from District Health Boards (2012). Road network used. ArcGIS used.	TRO density within 1km of school TRO density within 500m of school.	Current smoking (past 30 days +100 cigarettes) Experimental smoking (smoked past 30 days, but not 100 cigarettes) Susceptible smoking Purchasing tobacco past 30 days Tried to purchase tobacco past 30 days (current smokers only).	Logistic regression.	Individual: sex, age, ethnicity, Interpersonal: parental and friend smoking. School: school decile. Community: urban/rural.	1a. Significant negative association for medium density TRO 1km around the school (OR=0.79, 95% CI=0.65–0.96) and high density TRO (OR=0.8, 95% CI=0.67–0.96) for current smoking, compared with zero density. 1c. Positive association for high density TRO within 1km of school (OR=1.07, 95% CI=1.01–1.16) and susceptibility to smoking, compared with zero density. 1e. Positive association for high density TRO 1km of school (OR=1.52, 95% CI=1.26–1.84) and tried to purchase tobacco, compared with zero density. 2a. Negative association for medium density TRO 500m of school (OR=0.89, 95% CI=0.8–0.99) and high density TRO (OR=0.75, 95% CI=0.65–0.87) for current smoking, compared with zero density. 2c. Positive association for high density TRO 500m of school (OR=1.09, 95% CI=1.03–1.14) and susceptibility to smoking, compared with zero density. The remaining relationships were non-significant.
Schleicher et al., (2016)6 USA national. 13–16 years. NIH Public Health Service and the National Cancer Institute.	Cross-sectional internet panel of one eligible teen and parent from each HH (2011–2012). RR= 40%. N=2,771.	Purchased address data for likely TROs in all zip codes that contained or adjacent to each teen’s residence and school (year not stated). Road network used. ArcGIS used	TROs per mile within 0.5-mile of school addresses.	Ever tried a cigarette.	Logistic regression analysis.	Individual: age, gender, grades, Interpersonal: close friends smoke, peer-parent household race/ ethnicity, smokers in household, and HH income. Community: race, ethnicity, and percent population in poverty.	1a. No significant association (OR=1.0, 95% CI=1.00–1.01).
Shortt et al., (2016)7 Scotland. 13–15 years. Scottish Collaboration for Public Health Research and Policy.	Cross-sectional national school survey (2010–2011). Final nationally representative sample of 37,307 pupils (RR = 91%). N= 20,446 who had addresses and outcome variable included in study.	Addresses of all premises registered on the Scottish Tobacco Retailers Register (20102). Density estimated with KDE. Density of retailers per km2 then extracted to postcode area centroids.	TRO density around the school – 800m	Ever smoking Current smoking – do you smoke nowadays.	Logistic regression.	Individual: age, sex, ethnicity, free school meals, and self-perceived family wealth. Interpersonal: family structure and parental smoking status. Community: urban/rural and SES.	1a. Significant negative dose- response: Pupils attending school in highest density areas had significant lower odds of ever smoking (OR=0.66, 95% CI=0.50–0.86) compared with those in lowest density areas (OR=0.70, 95% CI=0.53–0.93). 1b. Significant negative dose- response: Pupils attending school in higher density areas had significantly increased odds of current smoking (OR=0.75, 95% CI=0.59–0.95) compared with those in lower density areas (OR=0.77, 95% CI=0.61–0.98).
Kaai et al., (2014)20 Ten provinces in Canada. Grades 9–12 students (ages 14–17 years). Health Canada. Propel Centre for Population Health Impact. Prince Edward Island Department of Education and Early Childhood Development. Ontario Ministry of Health and Long-term Care.	Cross-sectional school survey (2008–2009). The survey was administered to 29,296 students (RR=73%) attending 133 secondary schools (RR=59%) from all 10 provinces in Canada. This study used the subset of students who were never smokers (n = 15,982).	TRO data obtained from a national database of Canadian business (2008–2009). Euclidean radius used. Arcview used.	TRO density within 1km of school.	Smoking susceptibility.	Multilevel logistic regression.	Individual: sex, grade, alcohol use, marijuana use, tobacco knowledge and attitudes. Interpersonal: friends smoking, rules about smoking in the home. Community: urban/rural, median HH income.	1a. No significant association (OR=1.00, 95% CI=0.99–1.01).
Gwon et al., (2017)21 4 boroughs in South Korea. 13–15 years. Center for Global Inquiry and Innovation Grants, University of Virginia.	Cross-sectional school survey (2015) at 14 middle schools in 4 boroughs. A class in each grade was selected; all students asked to complete the survey. RR = 94% N=698 students	Address of TROs obtained from each borough (2015). Unknown if straight line or road network used. ArcGIS used.	TRO density within 0.5-mile around school.	Lifetime smoking Current smoking –smoked in past month.	Multivariable modelling.	Individual: age, gender, perceived economic class, and weekly allowance. Interpersonal: parental smoking, sibling smoking, and peer smoking.	1a. No significant association (OR=1.00, 95% CI=0.92–1.10). 1b. No significant association (OR=1.05, 95% CI=0.92–1.20).
Larsen et al., (2017)22 Ontario, Canada. Grades 9–12 (ages 14–17 years). Hospital for Sick Children Research Training Centre. Canadian Respiratory Research Network. Canadian Institutes of Health Research. Canadian Lung Association. Canadian Thoracic Society. British Columbia Lung Association and Industry Partners: Boehringer-Ingelheim Canada Ltd AstraZeneca Canada Inc., Novartis Canada Ltd. Canadian Institutes of Health Research-Public Health Agency of Canada.	Cross-sectional school survey (2013). N= 6,142 students from 109 high schools representative of Ontario schools. RR = 63%.	TRO obtained from DMTI Spatial Route Logistics dataset (2013). Road network used. GIS used	Density of TROs per square km within 1.6km of school.	Smokers – smoked one cigarette in past year.	Sequential multilevel logistic regression.	Individual: age and sex. Community: SES and population density.	1a. Significant positive association (OR=1.11, 95% CI=1.00–1.23).
Perez et al., (2017)23 Four counties in Texas, USA. Students in grades 6, 8, and 10 (ages 11–16 years). National Cancer Institute of the National Institutes of Health. Center for Tobacco Products of the Food and Drug Administration of the United States Department of Health and Human Services.	TATAMS uses a random sample of students in grades 6, 8, and 10 from five counties in Texas, 2104 −2015. One county excluded due to small size (5 schools and 142 students), and 24 students excluded due to lack of data. 42 schools and 3,741 students.	TRO data obtained from Texas Comptroller of public accounts (Nov 2104). Density divided into schools with no TRO and those with 1 or more TRO. R2BayesX, Spacestat, and ArcGIS.	Presence of at least one TRO with 0.5-mile of the school in: Dallas/Tarrant counties Harris counties Travis counties.	Past 30 days use of cigarettes.	Bayesian structured additive regression models and univariate ordinary Kriging method.	Individual: sex, race/ethnicity, grade, family standard of living. School: school zip code level characteristic, including percentage of high school graduates, median HH income, percent below poverty level.	1a. There was no evidence of a geospatial association between the presence of a TRO and smoking in Tarrant county. There was evidence of a geospatial association between the presence of a TRO and smoking in the eastern area of Dallas county where 5 schools are located. 2a. There was evidence of a geospatial association between the presence of a TRO and smoking in the southeastern area of Harris county where 2 schools are located. 3a. There was no evidence of a geospatial association between the presence of a TRO and smoking in Travis county. Note: We were unable to verify these results as no confidence intervals or p-values accompany the relative risk estimates.

References

^1.Adachi-Mejia, A.M., H.A. Carlos, E.M. Berke, et al., A comparison of individual versus community influences on youth smoking behaviours: a cross-sectional observational study. BMJ open, 2012. 2(5): p. e000767.

^2.Lipperman-Kreda, S., C. Mair, J.W. Grube, et al., Density and proximity of tobacco outlets to homes and schools: Relations with youth cigarette smoking. Prev. Sci., 2014. 15(5): p. 738–744.

^3.Mason, M.J., J. Mennis, N.M. Zaharakis, et al., The dynamic role of urban neighborhood effects in a text-messaging adolescent smoking intervention. Nicotine Tob. Res., 2015. 18(5): p. 1039–1045.

^4.Mennis, J., M. Mason, T. Way, et al., The role of tobacco outlet density in a smoking cessation intervention for urban youth. Health & Place, 2016. 38: p. 39–47.

^5.Mennis, J. and M. Mason, Tobacco outlet density and attitudes towards smoking among urban adolescent smokers. Substance abuse, 2016. 37(4): p. 521–525.

^6.Schleicher, N.C., T.O. Johnson, S.P. Fortmann, et al., Tobacco outlet density near home and school: associations with smoking and norms among US teens. Prev. Med., 2016. 91: p. 287–293.

^7.Shortt, N., C. Tisch, J. Pearce, et al., The density of tobacco retailers in home and school environments and relationship with adolescent smoking behaviours in Scotland. Tob Control, 2014: p. tobaccocontrol-2013–051473.

^8.Tunstall, H., N.K. Shortt, C.L. Niedzwiedz, et al., Tobacco outlet density and tobacco knowledge, beliefs, purchasing behaviours and price among adolescents in Scotland. Soc. Sci. Med., 2018. 206: p. 1–13.

^9.Leatherdale, S.T. and J.M. Strath, Tobacco retailer density surrounding schools and cigarette access behaviors among underage smoking students. Ann. Behav. Med., 2007. 33(1): p. 105–11.DOI: 10.1207/s15324796abm3301_12.

^10.Lovato, C.Y., H.C. Hsu, C.M. Sabiston, et al., Tobacco Point-of-Purchase marketing in school neighbourhoods and school smoking prevalence: a descriptive study. Can. J. Public Health, 2007. 98(4): p. 265–70.

^11.Henriksen, L., E. Feighery, N. Schleicher, et al., Is adolescent smoking related to the density and proximity of tobacco outlets and retail cigarette advertising near schools? . Prev. Med., 2008. 47(2): p. 210–214.

^12.McCarthy, W., R. Mistry, Y. Lu, et al., Density of tobacco retailers near schools: effects on tobacco use among students. Am. J. Public Health, 2009. 99(11): p. 2006–13.DOI: 10.2105/AJPH.2008.145128.

^13.Chan, W.C. and S.T. Leatherdale, Tobacco retailer density surrounding schools and youth smoking behaviour: a multi-level analysis. Tob Induc Dis, 2011. 9(1): p. 9.

^14.Adams, M.L., L.A. Jason, S. Pokorny, et al., Exploration of the link between tobacco retailers in school neighborhoods and student smoking. J. Sch. Health, 2013. 83(2): p. 112–118.

^15.Kaai, S.C., S.T. Leatherdale, S.R. Manske, et al., Using student and school factors to differentiate adolescent current smokers from experimental smokers in Canada: a multilevel analysis. Prev. Med., 2013. 57(2): p. 113–119.

^16.Scully, M., M. McCarthy, M. Zacher, et al., Density of tobacco retail outlets near schools and smoking behaviour among secondary school students. Aust. N. Z. J. Public Health, 2013. 37(6): p. 574–578.

^17.Kaai, S., S. Manske, S. Leatherdale, et al., Are experimental smokers different from their never-smoking classmates? A multilevel analysis of Canadian youth in grades 9 to 12. Chronic diseases and injuries in Canada, 2014. 34(2–3).

^18.Mistry, R., M. Pednekar, S. Pimple, et al., Banning tobacco sales and advertisements near educational institutions may reduce students’ tobacco use risk: evidence from Mumbai, India. Tob Control, 2013: p. tobaccocontrol-2012–050819.

^19.Marsh, L., A. Ajmal, R. McGee, et al., Tobacco retail outlet density and risk of youth smoking in New Zealand. Tobacco Control, 2015. 25(e2): p. 71–75.DOI: 10.1136/tobaccocontrol-2015-052512.

^20.Kaai, S.C., K.S. Brown, S.T. Leatherdale, et al., We do not smoke but some of us are more susceptible than others: A multilevel analysis of a sample of Canadian youth in grades 9 to 12. Addict. Behav., 2014. 39(9): p. 1329–1336.

^21.Gwon, S., G. Yan, G. Huang, et al., The influence of tobacco retailers on adolescent smoking: prevention and policy implications. International nursing review, 2017.

^22.Larsen, K., T. To, H.M. Irving, et al., Smoking and binge-drinking among adolescents, Ontario, Canada: Does the school neighbourhood matter? Health & Place, 2017. 47: p. 108–114.

^23.Pérez, A., L.-C. Chien, M.B. Harrell, et al., Geospatial associations between tobacco retail outlets and current use of cigarettes and e-cigarettes among youths in Texas. Journal of biometrics & biostatistics, 2017. 8(5).

^24.Pokorny, J., V.C. Smith, and S.B. Pokorny, The relation of retail tobacco availability to initiation and continued smoking. Journal of Clinical Child & Adolescent Psychology, 2003. 32(2): p. 193–204.DOI: 10.1207/S15374424JCCP3202_4

^25.Novak, S., S. Reardon, S. Raudenbush, et al., Retail tobacco outlet density and youth cigarette smoking: a propensity-modeling approach. Am. J. Public Health, 2006. 96(4): p. 670–6.DOI: 10.2105/AJPH.2004.061622.

^26.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Local tobacco policy and tobacco outlet density: Associations with youth smoking Prev. Med., 2012.DOI: 10.1016/j.jadohealth.2011.08.015.

^27.Loomis, B.R., A.E. Kim, A.H. Busey, et al., The density of tobacco retailers and its association with attitudes toward smoking, exposure to point-of-sale tobacco advertising, cigarette purchasing, and smoking among New York youth. Prev. Med., 2012. 55(5): p. 468–474.

^28.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Contextual and community factors associated with youth access to cigarettes through commercial sources. Tobacco Control, 2012.DOI: 10.1136/tobaccocontrol-2012-050473.

^29.Lipperman-Kreda, S., C. Morrison, J.W. Grube, et al., Youth activity spaces and daily exposure to tobacco outlets. Health & Place, 2015. 34: p. 30–33.

^30.Lipperman‐Kreda, S., J.W. Grube, K.B. Friend, et al., Tobacco outlet density, retailer cigarette sales without ID checks and enforcement of underage tobacco laws: associations with youths’ cigarette smoking and beliefs. Addiction, 2016. 111(3): p. 525–532.

^31.Burgoon, M.L., T. Albani, B. Keller-Hamilton, et al., Exposures to the tobacco retail environment among adolescent boys in urban and rural environments. The American journal of drug and alcohol abuse, 2019. 45(2): p. 217–226.

^32.Kowitt, S.D. and S. Lipperman-Kreda, How is exposure to tobacco outlets within activity spaces associated with daily tobacco use among youth? A mediation analysis. Nicotine Tob. Res., 2019.

^33.Kirchner, T.R., A.C. Villanti, J. Cantrell, et al., Tobacco retail outlet advertising practices and proximity to schools, parks and public housing affect Synar underage sales violations in Washington, DC. Tobacco Control, 2014: p. tobaccocontrol-2013–051239.

^34.Astuti, P.A., K.H. Mulyawan, S.K. Sebayang, et al., Cigarette retailer density around schools and neighbourhoods in Bali, Indonesia: A GIS mapping. Tobacco induced diseases, 2019. 17.

^35.Phetphum, C. and N. Noosorn, Tobacco Retailers Near Schools and the Violations of Tobacco Retailing Laws in Thailand. Journal of Public Health Management and Practice, 2019. 25(6): p. 537–542.

Overall, the studies of TRO density around schools were of high quality with generally large sample sizes (10 studies had a sample of over 15,000 students) and response rates greater than 50%, although some suffered from selection bias such as differences in how students were selected.[14, 34, 47] Although the most common size buffer was 1,000m, the size of buffers varied from 100m[50] to 1,600m,[45] making comparisons between studies difficult. However, Marsh et al[16] showed that the level of density is more important than the distance of polygon used. Inconsistent findings of the density of TROs around the school and youth smoking may have been due to limitations inherent in the outcome variables used. Of the 18 studies in this section, two reported population-level outcome variables. In two school-based studies of past-month smoking, Lovato et al., (2007) dichotomized prevalence of smoking 2 or more days in the past 30 to compare two categories of schools[40] and Henriksen et al., (2008) modeled prevalence of any smoking in the past 30 days.[48] Ten studies in this section used past 30 day smoking as a measure of current smoking, which includes daily and occasional smoking in the same category. This may have resulted in non-differential misclassification and underestimate of the association with TRO density. A further study combined two types of tobacco products (i.e. smoked and smokeless tobacco).[50]

Seven cross-sectional studies[15, 53–58] one longitudinal study,[59] and one geographic EMA study[60] measured TRO density in adolescents’ communities in the USA (Table 4). The size of the community measured in these studies varied from neighbourhoods of combined census tracts with a mean of 3, 479 people per tract[15] to mid-sized cities[54, 57, 59] and counties,[55] two studies focused on activity spaces,[56, 60] and one study assessed TRO exposure along the path between home and school.[58] TRO density in these communities was found to be significantly associated with higher odds of recent, past-year and lifetime smoking.[15, 54, 59] The first study examining activity spaces found a significant positive association between TRO density and daily tobacco use.[56] However, a second study about activity space, found no direct association between daily exposure to TROs and daily tobacco use, only indirect associations were found through daily exposure to peer smoking.[60] Of the nine community studies, six calculated the density of TROs at a city or county level which may have masked associations within adolescents’ neighbourhoods and contributed to the lack of significant findings, as no variation in density within the city/county would have been represented.[15, 53–55, 57, 59] These factors may have resulted in non-differential error in the exposure measurements, which were potentially misclassified for all areas, and which could have biased the results towards the null.

Table 4:

Density of tobacco retailers to community, activity spaces, and travel path

Reference Location Participants Funding	Study design Sample selection Sample size	TRO and Spatial	Exposure(s)	Outcome(s)	Analytical method	Adjustment	Results
Pokorny et al., (2003)24 11 towns in Illinois, USA. Grades 6–8 (ages 10–16 years). Robert Wood Johnson Foundation Substance Abuse Policy Research Program.	Cross-sectional school survey (1999). Students were sampled within towns using a cluster sampling procedure. If there was more than one school district within a town, one district was selected to include schools with the greatest proportion of grade 6–8 students who lived within the target community. N= 5,234 (75% RR)	TRO data from licensing list, or directory of business that usually sell tobacco, audited with phone call (year not stated). No GIS used.	Number of TROs as a function of youth population in each community.	Smoking initiation (tried smoking) Continued smoker (past 30 days).	A random-effects regression analysis.	Individual: age, grade, sex, race, perceived access to tobacco, and ability to purchase tobacco. Interpersonal: adult user in home, peer users.	1a. No statistically significant association (no data given). 1b. No statistically significant association (no data given).
Novak et al., (2006)25 Chicago, USA. 11–23 years. John D. and Catherine T. MacArthur Foundation. The National Institute of Justice. The National Institute of Mental Health. The Tobacco Etiology Research Network. National Institute on Alcohol Abuse and Alcoholism	Cross-sectional HH survey (1995–1999). Census blocks within 80 randomly selected neighborhood were sampled. HH with youth meeting criteria were interviewed n=2,116.	TROs identified from field observation (year not stated). Unclear if spatial analyses was used.	Number of blocks with at least one TRO per observed block faces per census tract.	Recent smoking (past 30 days).	Generalized estimating equations.	Individual: demographics. Community: commercial land use, racial composition, neighbourhood poverty. Propensity score stratification.	1a. Positive association (OR=1.20, 95% CI=1.00–1.44).
Lipperman-Kreda et al., (2012)26 California, USA. 13–16 years. National Cancer Institute. The Tobacco-Related Disease Research Program.	Cross-sectional computer assisted HH phone interview (2009–2010). 50 mid-size non-contiguous California cities selected using geographic sampling. HH sampled from a purchased list of phone numbers. Sample size 1,491 (RR – 50%).	TRO in each city obtained from State of California Board of Equalization data-files (2009). No spatial analysis.	Licensed TROs per 10,000 people.	Ever smoked Smoked past 12 months Smoked past 30 days.	Multilevel logistic and linear regression.	Individual: gender, ethnicity, and age. Community: population density, ethnicity, female heads of HH with minors, unemployment, education, and median HH income.	1a. Significant positive association (OR=1.312, 95% CI=1.041=1.655). 1b. Significant positive association (β=.010, SE=0.003, p≤0.005). 1c. No significant association (β = .002, SE= 0.002, p>0.05).
Loomis et al., (2012)27 New York, USA. 9–17 years. New York State Department of Health Tobacco Control Program.	Pooled cross-sectional data from New York Youth Tobacco Survey (NY-YTS) (2000–2009). The NY-YTS uses a multistage probability sampling method to select middle and high schools. Within selected schools, classes were randomly selected to participate. Students in the selected classes take the survey. N=70,427	List of licensed TROs from New York State (NYS) Dept (2000–2008). Arc GIS used.	TROs per 1,000 youth aged <17 years in each NYS county.	Susceptible never smokers Current smoker Frequent smoking Cigarettes per day Refused sale due to age when buying in a store in past 30 days Asked to show proof of age when buying cigarettes in a store in past 30 days.	Logistic and linear regression.	Individual: gender, ethnicity, age, and weekly income of the student Interpersonal: whether the student lives with a smoker, School: prevalence of smoking in student’s school.	1a. No significant association in NY City (NYC) (OR=1.05, 95% CI=0.95–1.15) or NYS (OR=1.14, 95% CI=0.84–1.53). 1b. No significant association in NYC (OR= 0.99, 95% CI=0.88–1.11) or NYS (OR=1.07, 95% CI=0.77–1.46). 1c. Negative association in NYC (OR=0.50, 95% CI=0.29–0.84). No significant association in NYS (OR=0.74, 95% CI=0.20–2.78). 1d. No significant association in NYC (OR=1.27, 95% CI=0.80–2.01) or NYS (OR=1.89, 95% CI=0.57–6.19). 1e. No significant association in NYC (OR=0.89, 95% CI=0.67–1.16) or NYS (OR=0.93, 95% CI=0.35–2.49). 1f. No significant association in NYC (OR=1.18, 95% CI=0.89–1.58) or NYS (OR=0.60, 95% CI=0.22–1.62).
Lipperman-Kreda et al., (2014)28 California, USA. 18 year old confederate buyers. National Cancer Institute. The Tobacco-Related Disease Research Program.	Cross-sectional confederate buyers study at 20 randomly selected TROs in each city (2011). A geographic sampling method was used to select 50 mid-size non-non-contiguous California cities.	The number of licensed TRO in each city obtained from State of California Board of Equalization data-files (2011). TRO addresses were identified using field observations. No spatial analysis. GIS used	TROs per 10,000 people.	Retailer non-compliance Asked for ID/age.	Multilevel logistic regression analyses.	Community: Pop density, % minors, median HH income, education, African-American, Hispanic, prevalence of adult smokers, TRO density, TRO licensing, and cigarette tax. Retail: Young buyer gender, buyer actual age, type of TRO, female clerk, clerk age, min age signs, and no. of customers in line behind buyer.	1a. No significant association (OR=1.05, 95% CI=0.96–1.14) 1b. No significant association (OR=0.95, 95% CI=0.89–1.01).
Lipperman-Kreda et al., (2015)29 San Francisco, USA. 14–18 years. Pacific Institute for Research and Evaluation.	Cross-sectional study using GPS to monitor movement of youth (2014). Convenience sample of 11 youth recruited through youth organisations.	TRO data obtained from field observations (2010–2011). Euclidean buffers used. Arc GIS used.	50m from youth activity space 100m from youth activity space.	Daily tobacco user.	Generalized Estimating Equations.	None	1a. Positive association (Wald χ2=3.9, p<0.05) 2a. Positive association (Wald χ2=4.4, p<0.05).
Lipperman-Kreda et al., (2016)30 2009 – 2011 California, USA. 13–16 years. National Cancer Institute. National Institute on Alcohol Abuse and Alcoholism. National Institutes of Health. Tobacco-Related Disease Research Program.	Longitudinal study using cross-sectional lifetime smoking data across three waves of surveys (2009–2011). A geographic sampling method was used to select 50 mid-size non-non-contiguous California cities. Data from 1,478 youths who participated in at least one wave of data collection, lived in same city across all study years, and provided complete data. 72% follow up from wave 1.	Licensed TROs in each city obtained from State of California Board of Equalization data-files (2011). No GIS used.	Licensed TROs per 10,000 people.	Lifetime smoking.	Multi-level random-effects logistic and linear regression analyses.	Individual: gender, age, race and ethnicity. Community: SES, population density, percent of population under <18 years, percent white, percent African American, and percent Hispanic.	1a. Positive association (OR=1.12, 95% CI=1.04–1.22).
Burgoon et al., (2019)31 9 counties in Ohio, USA. 11–16 years. National Cancer Institute and FDA Center for Tobacco Products.	Boys recruited through either address based probability sampling or community sampling, to participate in interviewer administered survey (n=1220). Year of data collection unknown. 140 students excluded from analysis. RR not given.	Tobacco license data – source not disclosed. A 1.0-mile buffer was created around a Euclidian distance between each participant’s school and home addresses. All licensed tobacco retailers within the created buffer were summed. ArcGIS used.	Potential exposure to stores that are likely to sell tobacco products between each participant’s school and home addresses.	Ever vs Never tobacco use.	Weights were calculated for the two samples. Negative binomial regression.	Individual: age, race, ethnicity. Community: neighbourhood poverty.	1a. No unadjusted association between potential exposure to TROs and ever smoking (rate ratio=0.96, 95% CI: 0.72–1.26). Tobacco use not included in final adjusted models.
Kowitt and Lipperman-Kreda (2019)32 8 cities in California, USA 16–20 years California Tobacco-Related Disease Research Program. National Institute on Alcohol Abuse and Alcoholism of the National Institutes of Health.	Cross-sectional study using GPS to monitor movement of youth, combined with EMA. Brief daily text surveys (Feb 17 – May 18). 11 Californian cities within 8-mile radius of Oakland. 8 of these cities selected based on stratification of SES, TRO density, and then random selection. Multi-tiered approach to recruit 85 eligible participants with tracked GPS data for at least 360mins per day.	TRO within 10-mile buffers of each city were identified from NAICS codes of commercial lists. Verified by phone call (year of data not given). GPS locations geocoded. Activity spaces constructed by joining GPS locations into a polyline, which was buffered and overlaid with TROs.	Daily exposure to the number of tobacco outlets −50m. Daily exposure to the number of tobacco outlets −100m. Daily time spent within 50m of tobacco outlets. Daily time spent within 100m of tobacco outlets.	Daily tobacco use Mediators: Daily exposure to peers who use tobacco Daily exposure to adults who use tobacco.	Multilevel structural equation modeling.	Individual: sex, race, ethnicity, and perceived SES.	1a. No association between daily tobacco use and number of tobacco outlets within 50m (probit regression coefficient: 0.01, p=.82) 2a. No association between daily tobacco use and number of tobacco outlets within 100m (probit regression coefficient: 0.01, p=.62) 3a. No association between daily tobacco use and daily number of minutes participants were within 50m of tobacco outlets (probit regression coefficient: 0.01, p=.34) 4a. No association between daily tobacco use and daily number of minutes participants were within 100m of tobacco outlets (probit regression coefficient: 0.002, p=.65) Indirect association in models of daily exposure: Greater daily exposure to tobacco outlets within 50 m and 100 m of activity polylines was positively associated with seeing peers use tobacco on a given day (probit regression coefficient: 0.10, p<.001; probit regression coefficient: 0.08, p<.001, respectively). Greater daily exposure to tobacco outlets within 50 m and 100 m of activity polylines was positively associated with seeing adults use tobacco on a given day (probit regression coefficient: 0.02, p<.001; probit regression coefficient: 0.07, p < .001, respectively). Seeing peers use tobacco on a given day was positively associated with tobacco use on that day in both the 50 m and 100 m models (probit regression coefficient: 2.23, p<.001; probit regression coefficient: 2.24, p<.001). Indirect association in models of time spent near TROs: Seeing peers use tobacco on a given day was positively associated with tobacco use on that day in both the 50 m and 100 m models (probit regression coefficient: 2.24, p<.001; probit regression coefficient: 2.24, p<.001). No other mediation results were significant.

RR= response rate, TRO = tobacco retailer, HH= household SES= socioeconomic status

References

^1.Adachi-Mejia, A.M., H.A. Carlos, E.M. Berke, et al., A comparison of individual versus community influences on youth smoking behaviours: a cross-sectional observational study. BMJ open, 2012. 2(5): p. e000767.

^2.Lipperman-Kreda, S., C. Mair, J.W. Grube, et al., Density and proximity of tobacco outlets to homes and schools: Relations with youth cigarette smoking. Prev. Sci., 2014. 15(5): p. 738–744.

^3.Mason, M.J., J. Mennis, N.M. Zaharakis, et al., The dynamic role of urban neighborhood effects in a text-messaging adolescent smoking intervention. Nicotine Tob. Res., 2015. 18(5): p. 1039–1045.

^4.Mennis, J., M. Mason, T. Way, et al., The role of tobacco outlet density in a smoking cessation intervention for urban youth. Health & Place, 2016. 38: p. 39–47.

^5.Mennis, J. and M. Mason, Tobacco outlet density and attitudes towards smoking among urban adolescent smokers. Substance abuse, 2016. 37(4): p. 521–525.

^6.Schleicher, N.C., T.O. Johnson, S.P. Fortmann, et al., Tobacco outlet density near home and school: associations with smoking and norms among US teens. Prev. Med., 2016. 91: p. 287–293.

^7.Shortt, N., C. Tisch, J. Pearce, et al., The density of tobacco retailers in home and school environments and relationship with adolescent smoking behaviours in Scotland. Tob Control, 2014: p. tobaccocontrol-2013–051473.

^8.Tunstall, H., N.K. Shortt, C.L. Niedzwiedz, et al., Tobacco outlet density and tobacco knowledge, beliefs, purchasing behaviours and price among adolescents in Scotland. Soc. Sci. Med., 2018. 206: p. 1–13.

^9.Leatherdale, S.T. and J.M. Strath, Tobacco retailer density surrounding schools and cigarette access behaviors among underage smoking students. Ann. Behav. Med., 2007. 33(1): p. 105–11.DOI: 10.1207/s15324796abm3301_12.

^10.Lovato, C.Y., H.C. Hsu, C.M. Sabiston, et al., Tobacco Point-of-Purchase marketing in school neighbourhoods and school smoking prevalence: a descriptive study. Can. J. Public Health, 2007. 98(4): p. 265–70.

^11.Henriksen, L., E. Feighery, N. Schleicher, et al., Is adolescent smoking related to the density and proximity of tobacco outlets and retail cigarette advertising near schools? . Prev. Med., 2008. 47(2): p. 210–214.

^12.McCarthy, W., R. Mistry, Y. Lu, et al., Density of tobacco retailers near schools: effects on tobacco use among students. Am. J. Public Health, 2009. 99(11): p. 2006–13.DOI: 10.2105/AJPH.2008.145128.

^13.Chan, W.C. and S.T. Leatherdale, Tobacco retailer density surrounding schools and youth smoking behaviour: a multi-level analysis. Tob Induc Dis, 2011. 9(1): p. 9.

^14.Adams, M.L., L.A. Jason, S. Pokorny, et al., Exploration of the link between tobacco retailers in school neighborhoods and student smoking. J. Sch. Health, 2013. 83(2): p. 112–118.

^15.Kaai, S.C., S.T. Leatherdale, S.R. Manske, et al., Using student and school factors to differentiate adolescent current smokers from experimental smokers in Canada: a multilevel analysis. Prev. Med., 2013. 57(2): p. 113–119.

^16.Scully, M., M. McCarthy, M. Zacher, et al., Density of tobacco retail outlets near schools and smoking behaviour among secondary school students. Aust. N. Z. J. Public Health, 2013. 37(6): p. 574–578.

^17.Kaai, S., S. Manske, S. Leatherdale, et al., Are experimental smokers different from their never-smoking classmates? A multilevel analysis of Canadian youth in grades 9 to 12. Chronic diseases and injuries in Canada, 2014. 34(2–3).

^18.Mistry, R., M. Pednekar, S. Pimple, et al., Banning tobacco sales and advertisements near educational institutions may reduce students’ tobacco use risk: evidence from Mumbai, India. Tob Control, 2013: p. tobaccocontrol-2012–050819.

^19.Marsh, L., A. Ajmal, R. McGee, et al., Tobacco retail outlet density and risk of youth smoking in New Zealand. Tobacco Control, 2015. 25(e2): p. 71–75.DOI: 10.1136/tobaccocontrol-2015-052512.

^20.Kaai, S.C., K.S. Brown, S.T. Leatherdale, et al., We do not smoke but some of us are more susceptible than others: A multilevel analysis of a sample of Canadian youth in grades 9 to 12. Addict. Behav., 2014. 39(9): p. 1329–1336.

^21.Gwon, S., G. Yan, G. Huang, et al., The influence of tobacco retailers on adolescent smoking: prevention and policy implications. International nursing review, 2017.

^22.Larsen, K., T. To, H.M. Irving, et al., Smoking and binge-drinking among adolescents, Ontario, Canada: Does the school neighbourhood matter? Health & Place, 2017. 47: p. 108–114.

^23.Pérez, A., L.-C. Chien, M.B. Harrell, et al., Geospatial associations between tobacco retail outlets and current use of cigarettes and e-cigarettes among youths in Texas. Journal of biometrics & biostatistics, 2017. 8(5).

^24.Pokorny, J., V.C. Smith, and S.B. Pokorny, The relation of retail tobacco availability to initiation and continued smoking. Journal of Clinical Child & Adolescent Psychology, 2003. 32(2): p. 193–204.DOI: 10.1207/S15374424JCCP3202_4

^25.Novak, S., S. Reardon, S. Raudenbush, et al., Retail tobacco outlet density and youth cigarette smoking: a propensity-modeling approach. Am. J. Public Health, 2006. 96(4): p. 670–6.DOI: 10.2105/AJPH.2004.061622.

^26.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Local tobacco policy and tobacco outlet density: Associations with youth smoking Prev. Med., 2012.DOI: 10.1016/j.jadohealth.2011.08.015.

^27.Loomis, B.R., A.E. Kim, A.H. Busey, et al., The density of tobacco retailers and its association with attitudes toward smoking, exposure to point-of-sale tobacco advertising, cigarette purchasing, and smoking among New York youth. Prev. Med., 2012. 55(5): p. 468–474.

^28.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Contextual and community factors associated with youth access to cigarettes through commercial sources. Tobacco Control, 2012.DOI: 10.1136/tobaccocontrol-2012-050473.

^29.Lipperman-Kreda, S., C. Morrison, J.W. Grube, et al., Youth activity spaces and daily exposure to tobacco outlets. Health & Place, 2015. 34: p. 30–33.

^30.Lipperman‐Kreda, S., J.W. Grube, K.B. Friend, et al., Tobacco outlet density, retailer cigarette sales without ID checks and enforcement of underage tobacco laws: associations with youths’ cigarette smoking and beliefs. Addiction, 2016. 111(3): p. 525–532.

^31.Burgoon, M.L., T. Albani, B. Keller-Hamilton, et al., Exposures to the tobacco retail environment among adolescent boys in urban and rural environments. The American journal of drug and alcohol abuse, 2019. 45(2): p. 217–226.

^32.Kowitt, S.D. and S. Lipperman-Kreda, How is exposure to tobacco outlets within activity spaces associated with daily tobacco use among youth? A mediation analysis. Nicotine Tob. Res., 2019.

^33.Kirchner, T.R., A.C. Villanti, J. Cantrell, et al., Tobacco retail outlet advertising practices and proximity to schools, parks and public housing affect Synar underage sales violations in Washington, DC. Tobacco Control, 2014: p. tobaccocontrol-2013–051239.

^34.Astuti, P.A., K.H. Mulyawan, S.K. Sebayang, et al., Cigarette retailer density around schools and neighbourhoods in Bali, Indonesia: A GIS mapping. Tobacco induced diseases, 2019. 17.

^35.Phetphum, C. and N. Noosorn, Tobacco Retailers Near Schools and the Violations of Tobacco Retailing Laws in Thailand. Journal of Public Health Management and Practice, 2019. 25(6): p. 537–542.

All the studies mentioned above were adjusted for a range of individual, social and community variables, although socio-economic status (SES) was often omitted,[14, 39, 50, 51] or income was used as a proxy for SES.[42–44]

The spatial units for density in these three types of studies range from 100m buffers around a school[50] to whole counties, the largest of which is 6,954 square kilometres.[61] The type of density measure also varied, with studies using a circular buffer,[14, 40–44, 47, 48, 50–52] a walking distance along the road network,[16, 31, 35–37, 45] or more complex technique of KDE.[33, 34] Variance in size, inclusion criteria of communities studied, and the variance in controls on advertising and product displays which adolescents would be exposed to, may also preclude comparison between studies and generalisability to communities outside the study area.[53, 54, 57, 59]

Different methodological decisions may have impacted the results of these studies, including geocoding the location of homes, school and TROs, dealing with missing data, and exclusion of TROs within a walking distance, but outside of study areas. For example, Adachi-Mejia et al., (2012) reported that 13% of participants who did not provide a home address had their residence geocoded to the centroid of their ZIP code.[33] This may have resulted in non-differential error in the exposure measurement and could have biased the results towards the null. Similarly, Lipperman-Kreda et al., (2014) only geocoded TROs within city or borough boundaries, which could have resulted in measurement error for some participants.[14] In two studies, a large proportion of participants (40 to 45%) were excluded from the analysis because they did not provide a postcode, which could have biased the results.[34, 38]

TRO data were obtained from a range of sources e.g., licensed data, local health boards, and purchased commercial lists. The accuracy of the data was generally not stated; therefore the data may be a source of unknown size of error, which could have limited the research findings. Only one-third of the studies[14, 39, 40, 50, 52, 53, 57] [56] [15] used field observations to collect or audit their data to ensure that tobacco was sold at the store at the time of the study. In addition, the types of retailers included in the study was not always reported, or the proportion these represented of the total TROs.[45, 52] This may have led to an under or over estimate of the density of TROs, and therefore the associations examined. In some studies it was unclear what year the location of TRO data was collected,[31, 36, 37] whether the location of TROs was concurrent in time with the survey data collection,[46, 56] or if multiple data collection of TRO locations was undertaken for studies with more than one wave of outcome data.[15]

Proximity

It was hypothesized that students would be more likely to engage in smoking behaviours if there was a shorter distance between their home and TROs. Data on this association were sourced from cross-sectional data from two cohort studies,[14, 33] and one intervention study.[36] Two studies used surveys of large randomly selected households[14, 33] and one study used baseline data of a community-based intervention[36] (Table 5). Eight cross-sectional studies examined the association between the proximity of TROs to schools and smoking behaviours.[14, 45, 46, 48, 51, 62–64] Four studies were undertaken in the USA,[14, 46, 48] one in South Korea, [51] one in Canada,[45] one in Indonesia[63] and one in Thailand[64] (Table 6).

Table 5:

Proximity of tobacco retailers to home

Reference Location Participants Funding	Study design Sample selection Sample size	TRO and Spatial	Exposure(s)	Outcome(s)	Analytical method	Adjustment	Results
Adachi-Mejia et al., (2012)1 US national study. 13–18 years. This work was supported by the National Institutes of Health.	Cross-sectional HH telephone survey (2007). Data came from wave 5, n=3,055, RR 47%. Additional sample of African American adolescents included due to attrition=598. Geocoded sample = 3,646.	Geocoded national dataset of retailers likely to sell tobacco (2007). Road network used. ArcGIS used.	Proximity from home address to closest TRO.	Ever tried smoking Smoking intensity scale.	Multiple and ordinal logistic regression.	Individual: gender, race, age, SES, exposure to movie smoking, team sports participation, and sensation seeking. Interpersonal: sibling smoking, friend smoking Community: Proportion black and Hispanic, and poverty.	1a. No significant association (OR=0.96, 95% CI=0.67–1.36). 1b. No significant association (OR=0.74, 95% CI=0.45–1.20).
Lipperman-Kreda et al., (2014)2 45 California cities, USA. 13–18 years. National Cancer Institute and Tobacco-Related Disease Research Program.	Cross-sectional HH telephone survey (2010). 3062 HH were sampled from a purchased list. N=1,543 participated in wave 1 (RR= 50%). N= 1,312 also completed wave 2 (RR= 85%). Geocoded sample= 832.	Field observations to document TRO addresses (year not stated). Euclidean distance used. GIS used.	Proximity to the closest TRO from each participant’s home.	Smoking frequency (no of days smoked in past 30).	Negative binomial model.	Individual: gender, age, and race. Community: Population density, percent Hispanic, percent African American, percent unemployed, percent under 18 years, median HH income, and percent college graduates.	1a. No significant association (β=−1.025, robust SE=0.729, p>0.05)
Mennis and Mason (2016)5 Richmond, Virginia, USA. 14–18 years. Virginia Foundation for Healthy Youth.	Cross-sectional baseline data for an intervention study (2013–2014). 197 primarily urban African Americans (91%) recruited using a convenience sample. Adolescents with Fagerström screening score of 1 were included.	Geocoded North American Industry Classification System of businesses most likely to sell tobacco (year not stated). Road network used. ArcGIS used.	Proximity to nearest TRO from home.	Intend to smoke next 3 months Smoking in 5 years? How ready are you to stop smoking?	Ordinal regression.	Individual: age, gender, nicotine dependence. Interpersonal: smoker in residence and friends who smoke.	1a. No significant association (CO β=−0.377, 95% CI=−1.628–0.874). 1b. Significant positive association (β=−2.095, 95% CI=−3.480--[−0.711]). 1c. No significant association (β=0.647, 95% CI=−0.628–1.922).

RR= response rate, TRO = tobacco retailer, HH= household SES= socioeconomic status

References

^1.Adachi-Mejia, A.M., H.A. Carlos, E.M. Berke, et al., A comparison of individual versus community influences on youth smoking behaviours: a cross-sectional observational study. BMJ open, 2012. 2(5): p. e000767.

^2.Lipperman-Kreda, S., C. Mair, J.W. Grube, et al., Density and proximity of tobacco outlets to homes and schools: Relations with youth cigarette smoking. Prev. Sci., 2014. 15(5): p. 738–744.

^3.Mason, M.J., J. Mennis, N.M. Zaharakis, et al., The dynamic role of urban neighborhood effects in a text-messaging adolescent smoking intervention. Nicotine Tob. Res., 2015. 18(5): p. 1039–1045.

^4.Mennis, J., M. Mason, T. Way, et al., The role of tobacco outlet density in a smoking cessation intervention for urban youth. Health & Place, 2016. 38: p. 39–47.

^5.Mennis, J. and M. Mason, Tobacco outlet density and attitudes towards smoking among urban adolescent smokers. Substance abuse, 2016. 37(4): p. 521–525.

^6.Schleicher, N.C., T.O. Johnson, S.P. Fortmann, et al., Tobacco outlet density near home and school: associations with smoking and norms among US teens. Prev. Med., 2016. 91: p. 287–293.

^7.Shortt, N., C. Tisch, J. Pearce, et al., The density of tobacco retailers in home and school environments and relationship with adolescent smoking behaviours in Scotland. Tob Control, 2014: p. tobaccocontrol-2013–051473.

^8.Tunstall, H., N.K. Shortt, C.L. Niedzwiedz, et al., Tobacco outlet density and tobacco knowledge, beliefs, purchasing behaviours and price among adolescents in Scotland. Soc. Sci. Med., 2018. 206: p. 1–13.

^9.Leatherdale, S.T. and J.M. Strath, Tobacco retailer density surrounding schools and cigarette access behaviors among underage smoking students. Ann. Behav. Med., 2007. 33(1): p. 105–11.DOI: 10.1207/s15324796abm3301_12.

^10.Lovato, C.Y., H.C. Hsu, C.M. Sabiston, et al., Tobacco Point-of-Purchase marketing in school neighbourhoods and school smoking prevalence: a descriptive study. Can. J. Public Health, 2007. 98(4): p. 265–70.

^11.Henriksen, L., E. Feighery, N. Schleicher, et al., Is adolescent smoking related to the density and proximity of tobacco outlets and retail cigarette advertising near schools? . Prev. Med., 2008. 47(2): p. 210–214.

^12.McCarthy, W., R. Mistry, Y. Lu, et al., Density of tobacco retailers near schools: effects on tobacco use among students. Am. J. Public Health, 2009. 99(11): p. 2006–13.DOI: 10.2105/AJPH.2008.145128.

^13.Chan, W.C. and S.T. Leatherdale, Tobacco retailer density surrounding schools and youth smoking behaviour: a multi-level analysis. Tob Induc Dis, 2011. 9(1): p. 9.

^14.Adams, M.L., L.A. Jason, S. Pokorny, et al., Exploration of the link between tobacco retailers in school neighborhoods and student smoking. J. Sch. Health, 2013. 83(2): p. 112–118.

^15.Kaai, S.C., S.T. Leatherdale, S.R. Manske, et al., Using student and school factors to differentiate adolescent current smokers from experimental smokers in Canada: a multilevel analysis. Prev. Med., 2013. 57(2): p. 113–119.

^16.Scully, M., M. McCarthy, M. Zacher, et al., Density of tobacco retail outlets near schools and smoking behaviour among secondary school students. Aust. N. Z. J. Public Health, 2013. 37(6): p. 574–578.

^17.Kaai, S., S. Manske, S. Leatherdale, et al., Are experimental smokers different from their never-smoking classmates? A multilevel analysis of Canadian youth in grades 9 to 12. Chronic diseases and injuries in Canada, 2014. 34(2–3).

^18.Mistry, R., M. Pednekar, S. Pimple, et al., Banning tobacco sales and advertisements near educational institutions may reduce students’ tobacco use risk: evidence from Mumbai, India. Tob Control, 2013: p. tobaccocontrol-2012–050819.

^19.Marsh, L., A. Ajmal, R. McGee, et al., Tobacco retail outlet density and risk of youth smoking in New Zealand. Tobacco Control, 2015. 25(e2): p. 71–75.DOI: 10.1136/tobaccocontrol-2015-052512.

^20.Kaai, S.C., K.S. Brown, S.T. Leatherdale, et al., We do not smoke but some of us are more susceptible than others: A multilevel analysis of a sample of Canadian youth in grades 9 to 12. Addict. Behav., 2014. 39(9): p. 1329–1336.

^21.Gwon, S., G. Yan, G. Huang, et al., The influence of tobacco retailers on adolescent smoking: prevention and policy implications. International nursing review, 2017.

^22.Larsen, K., T. To, H.M. Irving, et al., Smoking and binge-drinking among adolescents, Ontario, Canada: Does the school neighbourhood matter? Health & Place, 2017. 47: p. 108–114.

^23.Pérez, A., L.-C. Chien, M.B. Harrell, et al., Geospatial associations between tobacco retail outlets and current use of cigarettes and e-cigarettes among youths in Texas. Journal of biometrics & biostatistics, 2017. 8(5).

^24.Pokorny, J., V.C. Smith, and S.B. Pokorny, The relation of retail tobacco availability to initiation and continued smoking. Journal of Clinical Child & Adolescent Psychology, 2003. 32(2): p. 193–204.DOI: 10.1207/S15374424JCCP3202_4

^25.Novak, S., S. Reardon, S. Raudenbush, et al., Retail tobacco outlet density and youth cigarette smoking: a propensity-modeling approach. Am. J. Public Health, 2006. 96(4): p. 670–6.DOI: 10.2105/AJPH.2004.061622.

^26.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Local tobacco policy and tobacco outlet density: Associations with youth smoking Prev. Med., 2012.DOI: 10.1016/j.jadohealth.2011.08.015.

^27.Loomis, B.R., A.E. Kim, A.H. Busey, et al., The density of tobacco retailers and its association with attitudes toward smoking, exposure to point-of-sale tobacco advertising, cigarette purchasing, and smoking among New York youth. Prev. Med., 2012. 55(5): p. 468–474.

^28.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Contextual and community factors associated with youth access to cigarettes through commercial sources. Tobacco Control, 2012.DOI: 10.1136/tobaccocontrol-2012-050473.

^29.Lipperman-Kreda, S., C. Morrison, J.W. Grube, et al., Youth activity spaces and daily exposure to tobacco outlets. Health & Place, 2015. 34: p. 30–33.

^30.Lipperman‐Kreda, S., J.W. Grube, K.B. Friend, et al., Tobacco outlet density, retailer cigarette sales without ID checks and enforcement of underage tobacco laws: associations with youths’ cigarette smoking and beliefs. Addiction, 2016. 111(3): p. 525–532.

^31.Burgoon, M.L., T. Albani, B. Keller-Hamilton, et al., Exposures to the tobacco retail environment among adolescent boys in urban and rural environments. The American journal of drug and alcohol abuse, 2019. 45(2): p. 217–226.

^32.Kowitt, S.D. and S. Lipperman-Kreda, How is exposure to tobacco outlets within activity spaces associated with daily tobacco use among youth? A mediation analysis. Nicotine Tob. Res., 2019.

^33.Kirchner, T.R., A.C. Villanti, J. Cantrell, et al., Tobacco retail outlet advertising practices and proximity to schools, parks and public housing affect Synar underage sales violations in Washington, DC. Tobacco Control, 2014: p. tobaccocontrol-2013–051239.

^34.Astuti, P.A., K.H. Mulyawan, S.K. Sebayang, et al., Cigarette retailer density around schools and neighbourhoods in Bali, Indonesia: A GIS mapping. Tobacco induced diseases, 2019. 17.

^35.Phetphum, C. and N. Noosorn, Tobacco Retailers Near Schools and the Violations of Tobacco Retailing Laws in Thailand. Journal of Public Health Management and Practice, 2019. 25(6): p. 537–542.

Table 6:

Proximity of tobacco retailers to school

Reference Location Participants Funding	Study design Sample selection Sample size	TRO and Spatial	Exposure(s)	Outcome(s)	Analytical method	Adjustment	Results
Henriksen et al., (2008)10 California, USA. High school students (ages unknown). California Tobacco-Related Disease Research Program.	Cross-sectional California Student Tobacco Survey (2005–2006). N= 24,875 students from 135 randomly selected high schools. School RR = 87.4%, student RR= 79.4%.	TRO data from state licensing data (year not stated). Euclidean distance used. GIS likely, no software mentioned.	Proximity of at least one TRO within 1000 ft. of school.	School smoking prevalence past 30 days. Number of cigarettes smoked (past 30 days – school level weighted average).	Ordinary least squares regression.	School: race, ethnicity and proportion qualified for free or reduced price meal. Community: median HH income, population density, and neighbourhood type.	1a. No significant association Positive association (β=1.1, 95% CI=−0.9–3.0). 1b. No significant association (no data given).
McCarthy et al., (2007)11 California, USA. Average age 14.9 years. Middle and High school students. California Tobacco-Related Disease Research Program.	Cross-sectional California Student Tobacco Survey (2003–2004). N= 19,306 students from 245 randomly selected high schools. School RR = 85%, student RR= 66%.	TRO data from California Board of Equalization on tobacco retail licensees (2006). Average straight-line distance. Used batch geocoder and Arc GIS.	Proximity – distance from school office to each retailer in 1.0-mile radius.	Experimental smoking Established smoking High school students experimental smoking Middle school students experimental smoking No. of cigs smoked on days smoked Purchase tobacco from a store.	Multilevel logistic and random-intercept models in a generalized linear mixed-model framework.	Individual: age, gender, ethnicity, English use at home, previous years grades, hopelessness in past year, perceived ease of access, and depressive symptoms. Interpersonal: peer smoking and best friend smoking, School: average parental education, rural/urban, type of school.	1a-f. No significant associations were found between proximity of TRO to schools and any of the outcome variables. No data given.
Lipperman-Kreda et al., (2014)2 45 California cities, USA. 13–18 years. National Cancer Institute and Tobacco-Related Disease Research Program.	Cross-sectional HH telephone survey (2010). 3062 HH were sampled from a purchased list. N=1,543 participated in the wave 1 (RR= 50%). N= 1,312 also completed wave 2 (RR= 85%). Geocoded sample= 832.	Field observations to document TRO addresses (year not stated). Euclidean distance used. GIS used.	Proximity - distance to the closest TRO from each school using straight-line method.	Smoking frequency (days smoked in past 30).	Negative binomial model.	Individual: gender, age, and race. Community: Population density, percent Hispanic, percent African American, percent unemployed, percent under 18 years, median HH income, percent college graduates.	1a. No significant association (β=1.281, robust SE=1.078, p>0.05)
Kirchner et al.,(2015)33 Washington DC, USA. High schools. National Institute on Drug Abuse. National Institutes of Health. Food and Drug Administration. Department of Health and Human Services. American Legacy Foundation. Centers for Disease Control and Prevention.	Sales violation data under the Synar inspection programme. Spatial data on the location of all public, charter, and independent high schools (n=45) in Washington DC.	Sample of 347 TROs of 750 in DC area, which were randomly selected for inspection by the Synar Inspection Program (2009–2010). Used the road network. ArcGIS used.	Proximity to closest DC high school within a 1.0-mile radius.	Sales violation.	Multivariate logistic regression.	Retail: store type, block group demographics, advertising practices.	1a. Among TROs in block groups with a greater than 56% African–American population, each 100m decrease in proximity to a high school was found to increase the likelihood of a sales violation (OR=1.29, 95% CI=1.07–1.58).
Gwon et al., (2017)20 4 boroughs in South Korea. 13–15 years. Center for Global Inquiry and Innovation Grants, University of Virginia.	Cross-sectional school survey (2015). At 14 middle schools in 4 boroughs, a class in each grade was selected; all students asked to complete the survey. RR = 94% N=698 students	Address of TROs obtained from each borough (2015). Unknown if straight line or road network used. ArcGIS used.	TRO proximity – mean proximity of TROs in 0.5-mile radius from school.	Lifetime smoking Current smoking (30days).	Multilevel logistic model.	Individual: age, gender, perceived economic class, and weekly allowance. Interpersonal: parental smoking, sibling smoking, and peer smoking.	1a. No significant association (OR=1.00, 95% CI=0.99–1.02). 1b. No significant association (OR=1.00, 95% CI=0.98–1.02).
Larsen et al., (2017)21 Ontario, Canada. Grades 9–12 (ages 14–17 years). Hospital for Sick Children Research Training Centre. Canadian Respiratory Research Network. Canadian Institutes of Health Research. Canadian Lung Association. Canadian Thoracic Society. British Columbia Lung Association and Industry Partners: Boehringer-Ingelheim Canada Ltd AstraZeneca Canada Inc., Novartis Canada Ltd. Canadian Institutes of Health Research-Public Health Agency of Canada.	Cross-sectional school survey (2013). N= 6,142 students from 109 high schools representative of Ontario schools. RR = 63%.	TRO obtained from DMTI Spatial Route Logistics dataset (2013). Road network used. GIS used	Distance from school to closest 3 TROs (mean distance) within 1.6km.	Smokers – smoked one cigarette in past year.	Sequential multilevel logistic regression.	Individual: age and sex. Community: population density and SES.	1a. No significant association (OR=1.145, 95% CI=0.945–1.387).
Astuti et al.,(2019)34 Denpasar, Indonesia. 6–18 years. Australia Indonesia Centre Health. Cluster. Australian Commonwealth Government. Indonesia Endowment Fund for Education scholarship.	Retailer audit survey – face to face survey with n=1,000 of 4, 114 randomly selected retailers. RR = 99%. Dec 2017 – Jan 2018.	Field observations to document TRO addresses with GPS coordinates. Dec 2017 – Jan 2018. Radius around schools. No information on whether the road network was used. ArcGIS used.	Proximity of TROs to 379 schools with radius of: <100m 100.1–250m 250.1 – 500m >500m.	Selling to young people.	Logistic regression.	School: school level (primary, Junior and senior), school type (public and private), and retailer type.	1a. Reference group 2a. No significant association (OR=0.84, 95% CI=0.56–1.26). 3a. No significant association (OR=0.70, 95% CI=0.47–1.04). 4a. Negative association (OR=0.53, 95% CI=0.33–0.87).
Phetphum and Noosorn (2019)35 One town in Thailand. Primary school through to university. Tobacco Control Research and Knowledge Management Center. Thai Health Promotion Foundation.	Surveys with 121 tobacco retailers regarding experiences of violation of tobacco retailing laws. RR 100%.	Source of TRO data unknown. Field observations or auditing unknown. Proximity between TRO and school measured using the road network within 500m of 14 schools (primary (2), middle (6), secondary (3), and vocational schools (2), and one university). QGIS used	Proximity from TRO to nearest school, categorized as >500 meters or ≤500 meters.	Did you sell cigarettes to teenagers or person younger than 18 years old? Checking buyer’s age before selling.	Chi-squared test.	Analyses not adjusted for potential confounders.	1a. No association between proximity of TROs and selling cigarettes to minors (χ2=1.748, p=0.20). 1b. No association between proximity of TROs and checking buyer’s age before selling (χ2=0.060, p=0.84).

RR= response rate, TRO = tobacco retailer, HH= household SES= socioeconomic status

References

^1.Adachi-Mejia, A.M., H.A. Carlos, E.M. Berke, et al., A comparison of individual versus community influences on youth smoking behaviours: a cross-sectional observational study. BMJ open, 2012. 2(5): p. e000767.

^2.Lipperman-Kreda, S., C. Mair, J.W. Grube, et al., Density and proximity of tobacco outlets to homes and schools: Relations with youth cigarette smoking. Prev. Sci., 2014. 15(5): p. 738–744.

^3.Mason, M.J., J. Mennis, N.M. Zaharakis, et al., The dynamic role of urban neighborhood effects in a text-messaging adolescent smoking intervention. Nicotine Tob. Res., 2015. 18(5): p. 1039–1045.

^4.Mennis, J., M. Mason, T. Way, et al., The role of tobacco outlet density in a smoking cessation intervention for urban youth. Health & Place, 2016. 38: p. 39–47.

^5.Mennis, J. and M. Mason, Tobacco outlet density and attitudes towards smoking among urban adolescent smokers. Substance abuse, 2016. 37(4): p. 521–525.

^6.Schleicher, N.C., T.O. Johnson, S.P. Fortmann, et al., Tobacco outlet density near home and school: associations with smoking and norms among US teens. Prev. Med., 2016. 91: p. 287–293.

^7.Shortt, N., C. Tisch, J. Pearce, et al., The density of tobacco retailers in home and school environments and relationship with adolescent smoking behaviours in Scotland. Tob Control, 2014: p. tobaccocontrol-2013–051473.

^8.Tunstall, H., N.K. Shortt, C.L. Niedzwiedz, et al., Tobacco outlet density and tobacco knowledge, beliefs, purchasing behaviours and price among adolescents in Scotland. Soc. Sci. Med., 2018. 206: p. 1–13.

^9.Leatherdale, S.T. and J.M. Strath, Tobacco retailer density surrounding schools and cigarette access behaviors among underage smoking students. Ann. Behav. Med., 2007. 33(1): p. 105–11.DOI: 10.1207/s15324796abm3301_12.

^10.Lovato, C.Y., H.C. Hsu, C.M. Sabiston, et al., Tobacco Point-of-Purchase marketing in school neighbourhoods and school smoking prevalence: a descriptive study. Can. J. Public Health, 2007. 98(4): p. 265–70.

^11.Henriksen, L., E. Feighery, N. Schleicher, et al., Is adolescent smoking related to the density and proximity of tobacco outlets and retail cigarette advertising near schools? . Prev. Med., 2008. 47(2): p. 210–214.

^12.McCarthy, W., R. Mistry, Y. Lu, et al., Density of tobacco retailers near schools: effects on tobacco use among students. Am. J. Public Health, 2009. 99(11): p. 2006–13.DOI: 10.2105/AJPH.2008.145128.

^13.Chan, W.C. and S.T. Leatherdale, Tobacco retailer density surrounding schools and youth smoking behaviour: a multi-level analysis. Tob Induc Dis, 2011. 9(1): p. 9.

^14.Adams, M.L., L.A. Jason, S. Pokorny, et al., Exploration of the link between tobacco retailers in school neighborhoods and student smoking. J. Sch. Health, 2013. 83(2): p. 112–118.

^15.Kaai, S.C., S.T. Leatherdale, S.R. Manske, et al., Using student and school factors to differentiate adolescent current smokers from experimental smokers in Canada: a multilevel analysis. Prev. Med., 2013. 57(2): p. 113–119.

^16.Scully, M., M. McCarthy, M. Zacher, et al., Density of tobacco retail outlets near schools and smoking behaviour among secondary school students. Aust. N. Z. J. Public Health, 2013. 37(6): p. 574–578.

^17.Kaai, S., S. Manske, S. Leatherdale, et al., Are experimental smokers different from their never-smoking classmates? A multilevel analysis of Canadian youth in grades 9 to 12. Chronic diseases and injuries in Canada, 2014. 34(2–3).

^18.Mistry, R., M. Pednekar, S. Pimple, et al., Banning tobacco sales and advertisements near educational institutions may reduce students’ tobacco use risk: evidence from Mumbai, India. Tob Control, 2013: p. tobaccocontrol-2012–050819.

^19.Marsh, L., A. Ajmal, R. McGee, et al., Tobacco retail outlet density and risk of youth smoking in New Zealand. Tobacco Control, 2015. 25(e2): p. 71–75.DOI: 10.1136/tobaccocontrol-2015-052512.

^20.Kaai, S.C., K.S. Brown, S.T. Leatherdale, et al., We do not smoke but some of us are more susceptible than others: A multilevel analysis of a sample of Canadian youth in grades 9 to 12. Addict. Behav., 2014. 39(9): p. 1329–1336.

^21.Gwon, S., G. Yan, G. Huang, et al., The influence of tobacco retailers on adolescent smoking: prevention and policy implications. International nursing review, 2017.

^22.Larsen, K., T. To, H.M. Irving, et al., Smoking and binge-drinking among adolescents, Ontario, Canada: Does the school neighbourhood matter? Health & Place, 2017. 47: p. 108–114.

^23.Pérez, A., L.-C. Chien, M.B. Harrell, et al., Geospatial associations between tobacco retail outlets and current use of cigarettes and e-cigarettes among youths in Texas. Journal of biometrics & biostatistics, 2017. 8(5).

^24.Pokorny, J., V.C. Smith, and S.B. Pokorny, The relation of retail tobacco availability to initiation and continued smoking. Journal of Clinical Child & Adolescent Psychology, 2003. 32(2): p. 193–204.DOI: 10.1207/S15374424JCCP3202_4

^25.Novak, S., S. Reardon, S. Raudenbush, et al., Retail tobacco outlet density and youth cigarette smoking: a propensity-modeling approach. Am. J. Public Health, 2006. 96(4): p. 670–6.DOI: 10.2105/AJPH.2004.061622.

^26.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Local tobacco policy and tobacco outlet density: Associations with youth smoking Prev. Med., 2012.DOI: 10.1016/j.jadohealth.2011.08.015.

^27.Loomis, B.R., A.E. Kim, A.H. Busey, et al., The density of tobacco retailers and its association with attitudes toward smoking, exposure to point-of-sale tobacco advertising, cigarette purchasing, and smoking among New York youth. Prev. Med., 2012. 55(5): p. 468–474.

^28.Lipperman-Kreda, S., J.W. Grube, and K.B. Friend, Contextual and community factors associated with youth access to cigarettes through commercial sources. Tobacco Control, 2012.DOI: 10.1136/tobaccocontrol-2012-050473.

^29.Lipperman-Kreda, S., C. Morrison, J.W. Grube, et al., Youth activity spaces and daily exposure to tobacco outlets. Health & Place, 2015. 34: p. 30–33.

^30.Lipperman‐Kreda, S., J.W. Grube, K.B. Friend, et al., Tobacco outlet density, retailer cigarette sales without ID checks and enforcement of underage tobacco laws: associations with youths’ cigarette smoking and beliefs. Addiction, 2016. 111(3): p. 525–532.

^31.Burgoon, M.L., T. Albani, B. Keller-Hamilton, et al., Exposures to the tobacco retail environment among adolescent boys in urban and rural environments. The American journal of drug and alcohol abuse, 2019. 45(2): p. 217–226.

^32.Kowitt, S.D. and S. Lipperman-Kreda, How is exposure to tobacco outlets within activity spaces associated with daily tobacco use among youth? A mediation analysis. Nicotine Tob. Res., 2019.

^33.Kirchner, T.R., A.C. Villanti, J. Cantrell, et al., Tobacco retail outlet advertising practices and proximity to schools, parks and public housing affect Synar underage sales violations in Washington, DC. Tobacco Control, 2014: p. tobaccocontrol-2013–051239.

^34.Astuti, P.A., K.H. Mulyawan, S.K. Sebayang, et al., Cigarette retailer density around schools and neighbourhoods in Bali, Indonesia: A GIS mapping. Tobacco induced diseases, 2019. 17.

^35.Phetphum, C. and N. Noosorn, Tobacco Retailers Near Schools and the Violations of Tobacco Retailing Laws in Thailand. Journal of Public Health Management and Practice, 2019. 25(6): p. 537–542.

However, the available studies do not provide strong evidence of an association between proximity of TROs to the home or the school, and smoking behaviours among adolescents. The only significant finding for proximity to the home was from Mennis and Mason’s (2016) study; that finding suggests nicotine-dependent adolescent smokers who live in close proximity to a TRO are more likely to be pessimistic about their chances of long-term abstinence than those who live further from a TRO.[36] The only significant findings for proximity to schools was that sales violations increased near schools in urban areas with a high African–American population, with each 100m decrease in proximity between the TRO and the school,[62] and retailers located more than 500m from a school in Indonesia, compared with those located within 100m, were significantly less likely to sell cigarettes to young people.[63]

Limitations of the exposure measures could have biased results towards the null. In Adachi-Mejia et al., many stores that could potentially have sold tobacco were not geocoded,[33] and it is unclear in Larsen et al (2017) if convenience stores and gas stations were the only retailers included in the study, and what proportion of the total TROs these categories represent.[45] However, research suggests these two types of retailers comprise approximately one-half of tobacco retailers in the USA.[65] Four studies used field validation of TROs near schools to validate that the stores actively sold tobacco;[14, 62–64] however, there was no direct inspection of TROs in the remaining studies.[33, 36, 45, 46, 48, 51] This may have resulted in an under- or over-estimate of the distance to the nearest TRO from the participants home or school.

Similar to studies on the density of TROs, issues with geocoding only within city boundaries and coding residences to the centroid of the zip code when addresses were not available, limited the findings of the studies.[14, 33, 51] In measuring the exposure variable, four studies used the straight-line method to measure the distance to a TRO,[14, 46, 48, 51] rather than walkable distances or the road network, which was utilised in five studies. [33, 36, 45, 62, 64] Varying buffer sizes were used ranging from 0.2-mile to 1-mile, but generally, only one buffer size was used per study. Two studies used only a single point for a school campus to define proximity to a tobacco store, which could have under estimated the number of TROs.[46, 48]

A number of sampling limitations may have affected the findings. Adachi-Mejia et al., (2012) had a small national sample with a high loss to follow up,[33] the small sample size and response bias in Gown et al., (2017)[51] and the 63% response rate with no follow up of individual non-responders in Larsen et al., (2017)[45] may have underestimated smoking rates among the sample. While a large sample was selected from the initial process for Lipperman-Kreda et al., (2014), the final sample represented only 27% of households sampled in the first wave, which may have led to selection bias.[14] Mennis and Mason’s (2016) study had a small sample size of volunteer participants, a high proportion of African American participants, and was based in an urban setting.[36] Two studies used self-reported sales to minors to measure sales violations.[63, 64] The generalizability of the findings from these studies may be limited to the jurisdictions in which the data were collected, or to similar jurisdictions with a comparable density level and TRO policy environment.

Discussion

This systematic review examined the association between the density and proximity of TROs to homes, schools and communities and smoking behaviours among youth. The 35 studies reviewed provided evidence of a relationship between density of TROs and smoking behaviours, particularly for the density of TROs near youths’ home which found significant associations in six of eight studies. This appears to be more important than the density near their school, a finding consistent with Finan et al., (2018).[28] The largest effect (OR 1.53, 95% CI=1.27 to 1.85) found in this review was reported by a study with one of the largest number of participants (approx. 20,000 participants) and a reliable tobacco licensing scheme.[34] An association was also found between TRO density around schools and susceptibility to future smoking among non-smokers, but there were no consistent findings for other smoking behaviours and the density of TROs around schools. The review did not provide evidence of an association between the proximity of TROs to homes or schools and smoking behaviours among youth. These findings suggest that the number of TROs in vicinity to the home or school may be more influential than proximity to a retailer.[46] The findings in this review may be limited by the fact that studies which report significant outcomes are much more likely to be published; potentially biasing the reporting of our study.[66]

Many of the studies in the ‘community’ section examined the density of TROs at the city or county level, rather than much smaller geographic areas such as neighbourhoods. Data at this larger level may mask local level associations, which could reduce the ability to detect significant differences in the association between the density of TROs and youth smoking.[55] However, Lipperman-Kreda et al., (2012) point out that the optimal geographic unit for investigating density effects is unclear. The authors suggest that given the mobility of youth, traditional definitions of neighbourhoods may be too small a unit of analysis.[54] Increased mobility and autonomy mean that youth may have increased involvement in extended social networks which cross neighbourhood boundaries, and often extend past the home or school buffers that are used in most studies in this review.[56, 59] Studies should also consider the type of transportation that young people use, which may affect the number of tobacco stores that students are exposed to. Those who walk or bike to school may be exposed to a greater number of TROs than those who take the bus or drive to school.[14, 46] Only two studies included in our review used an activity space approach to measure TRO exposure. Additional research is needed to understand the strengths and limitations of this approach.

There was also a general lack of studies that focus on, or incorporate, rural areas, which is important because of geographic inequalities in tobacco use among rural youth.[67–69] Although buffer sizes varied, none was over a 1-mile radius, and many were smaller. Larger buffers for rural areas are common in spatial research, but not in the literature we reviewed. Therefore, the current literature may underestimate the relevant density and proximity of TRO to rural schools.[46] This raises the issue of a need for more studies to investigate the relationship between tobacco density and smoking behaviours in rural areas as these may differ to urban and suburban areas.

The lack of significant findings in many of these studies may also suggest that for youth populations commercial TROs are not their main source of tobacco, and that other sources such as social supply are more important. New Zealand research has shown that social sources account for a large proportion of tobacco supply for 14–15 year olds, particularly from friends or peers. Although only 10% purchased their tobacco from a shop or vending machine, regular smokers were significantly more likely than intermittent smokers to report having purchased tobacco.[70] Research on youth access has suggested that youth only need to know where one store is that will sell to minors, regardless of how many stores there are in an area.[71] However, all stores remain a source of exposure to availability, and advertising and product display in jurisdictions without regulation of these marketing activities.

Most papers included spatial analysis, allowing the geographic components of density and proximity to be considered. While there were two primary modes of calculating density (number of retailers within an area around a home or school address or number of retailers per capita) it remains an open question as to which is more suitable. Adachi-Mejia et al. (2012) compared the results of different methods of measuring proximity of TROs including buffers, network, and driving time and found no differences between the methods.[33] Ten papers considered retailer proximity, seven of which considered the distance to the first retailer only, though Phetphum et al., (2019)[64] measured the distance from the TRO to the nearest school, Larsen et al., (2017)[45] considered the mean distance to the first three retailers, and Gwon (2017)[51] examined the mean proximity to TROs. Many of the studies relied on a simple straight-line method to measure the proximity or density of TROs to homes and schools.[46] Using the distance along the road network, or walkable distances, to define the Euclidean (or circular) areas around home or school addresses may produce more accurate results as they represent actual access pathways. It is important to note that network routes were calculated based on shortest distances rather than reflecting actual routes that young people may take. It is worthwhile considering GPS monitoring in future studies to more accurately reflect true routes, especially where exposure to outlets or advertising is an important consideration.

All except one of the 35 studies in this review were cross-sectional in design, precluding any causal relationships from being identified. As Loomis and colleagues (2012) suggest, it is more likely that the location and density of TROs has a causal effect on youth smoking than youth smoking having an effect on the number of TROs.[55] Greater TRO density in an area could promote youth smoking through a number of pathways. The increased availability may increase the opportunity for young people to purchase tobacco, and contribute to the perception of the normality of smoking in a community. Some studies hypothesized that the relationship between density of TROs and smoking could also be due to increased marketing visual cues, rather than purchase accessibility. This may relate especially to countries where advertising is not restricted at the POS, such as the USA. Discrepancies in study findings in the relationship between density of TROs and smoking could be in part due to policy variation between countries.[52]

It has also been proposed that increased tobacco density may encourage greater competition between retailers.[14] Increased competition could lead to more sales to minors, drive cigarette prices down[39],[72] and raise the chances of single cigarette sales. A greater availability of retailers drives the cost of tobacco down due to reduced travel cost, and reduce the full cost of tobacco to the customer.[73] A greater number of TROs increases enforcement requirements, and may reduce the likelihood that any one of them is inspected.

Although this review did not find an association between proximity of a TRO and quitting, or relapse during a quit attempt among youth, this may be an important consideration for adult populations. Research has shown that TROs can create an environmental cue to smoke or purchase tobacco on impulse.⁵⁴ Halonen et al (2013) showed that quitters living within close proximity (e.g. 500m) of a TRO are more likely to relapse,[74, 75] and the presence of an outlet is a sufficient cue to prompt tobacco cravings and impulse purchases.[76]

This review found the density of TROs around the home, and to an extent schools, are the most strongly associated with youth smoking outcomes. Collectively, these studies suggest that we need to think more broadly about young people’s exposure to the tobacco retail environment. Moving beyond policies that limit the sale of tobacco near schools, regulations should consider multiple mechanisms in which easier access to TROs could promote smoking. Rather than policy options which target a specific area or type of store, this review suggests that policy options which reduce the overall density of TROs in a neighbourhood, would contribute to reducing smoking rates among youth. These options may include mazimizing distance between TROs which has been implemented in multiple jurisdictions,[77–79] limiting the number of TROs per population and with the goal of creating a more equitable distribution, such as that introduced in San Francisco,[80] restricting tobacco sales to a limited number of government-licensed stores similar to Hungary,[81] or to age-restricted stores where youth do not have access.[82, 83] Notably, this review suggests that the focus of existing literature is more problem-oriented (i.e., establishing whether TRO density/proximity contributes to smoking outcomes) than solution-oriented (i.e., whether regulating density/proximity prevent smoking uptake and improve cessation). Research is needed to evaluate the impact of density reduction policies on youth outcomes (smoking behaviors and sales to minors) as well as theorized mediators, such as exposure to marketing, price, anti-smoking norms and risk perceptions.

Limitations

This systematic review focused on peer reviewed journal articles, and we did not attempt to capture grey literature. It is possible that by narrowing the eligibility for inclusion of research, we may have potentially omitted non-peer reviewed research from this study. In synthesizing the studies for this review, we report odds ratios and other effect sizes to provide readers with additional information relevant to the interpretation of study findings. However, we acknowledge that odds ratios are not directly comparable across studies with different samples, populations, or explanatory variables.[84] Comparisons between studies with a large variance in sources of data, measurement methods, and statistical analysis can be difficult. To allow for better comparisons between results and quality of studies examining the density and proximity of TROs, future studies would benefit from performing a sensitivity analysis of different sizes of density and proximity buffers.

Conclusion

The existing evidence supports a positive association between TRO density and smoking behaviours among youth, particularly for the density near youths’ home. There is a substantial lack of longitudinal studies that would help to clarify the causal relationship and strengthen the evidence base for policy interventions. Pathways and policies to address these mechanisms, for the relationship between high density of TROs and youth smoking have been proposed.