Table 1 Characteristics of included studies

European Smoking Prevention Framework Analysis (ESFA)

Mercken et al. (2007) [25]

1998

12–13

1886

9

Netherlands

Longitudinal

Mplus 4.1

Structural Equation Modelling (SEM)

To examine the effect of influence and selection for reciprocal and non-reciprocal friendship on smoking

Medium

Social selection and influence

Mercken et al. (2009a) [26]

1998

12–13

1886

9

Netherlands

Longitudinal

Mplus 4.1

SEM

To examine the specific contribution of influence and selection for reciprocal and non-reciprocal friendship and deselection on smoking changes.

Medium

Social selection and influence

Mercken et al. (2009b) [27]

1998

Mean 13

7704

17 Danish, 11 Finnish, 9 Dutch, 8 Portugese, 4 UK & 21 Spanish

Denmark, Finland, Netherlands, Portugal, UK, Spain

Longitudinal

SIENA

Stochastic Actor Oriented Model (SAOM)

To examine smoking-related friendship selection and friends’ influence within the same school grade, while controlling for alternative selection mechanisms.

Medium

Social selection and influence

Mercken et al. (2010a) [28]

1998

13–16

1326

11

Finland

Longitudinal

SIENA

SAOM

To examine the strength of influence and selection processes on smoking for reciprocal and non-reciprocal friendship

High

Social selection and influence

Mercken et al. (2010b) [29]

1998

13–16

1163

9

Finland

Longitudinal

SIENA

SAOM

To examine gender differences in the strength of influence and selection processes on smoking for reciprocal and non-reciprocal friendship

High

Social selection and influence;

Network position

Teenage Health in Schools (THiS) study

Turner et al. (2006) [30]

2001

13–15

489 baseline, 407 follow-up

2

Scotland

Cross-sectional

NEGOPY 4.50, SPSS

×2 test and F ratio (not multivariate)

To investigate whether peer structures and influences affect smoking rates

Low

Socioeconomic status;

Network position

Pearson et al. (2006) [31] 

2001

13–15

3379

9

Scotland

Cross-sectional

NEGOPY

Logistic regression

Do associations between network measures and substance use differ according to context

Low

Socioeconomic status;

Network position

ASSIST- A Stop Smoking In Schools STudy

Steglich et al. (2012) [32]

2001

12–16

596 baseline, 585 follow-up

3

UK

Longitudinal

SIENA

SAOM

To compare results of different approaches to SABM in measuring link between network structure and smoking

Medium

Social selection and influence

Mercken et al. (2012) [33]

2001

12–14

1677 baseline, 1614 follow-up

11

UK

Longitudinal

SIENA

SAOM

To examine how smoking based selection and influence processes change over time

High

Social selection and influence

Promoting School-Community-University Partnerships to Enhance Resilience (PROSPER)

Copeland et al. (2017) [34]

2002

13–18/19

11,802

28 school districts

USA (Iowa)

Longitudinal

Not specified

Autoregressive Latent Trajectory Models (ALT)

To examine whole and ego network effects on smoking, particularly isolation

Medium

Network position

Ragan (2016) [35]

2002

13–18/19

Mean 6200 at each wave

27 school districts

USA (Iowa)

Longitudinal

SIENA

SAOM

To examine the effect of peer beliefs on smoking-

Medium

Social selection and influence

McMillan et al. (2018) [36]

2002

13–18/19

9135

51

USA (Iowa)

Longitudinal

SIENA

SAOM

To investigate the effect of gender on peer influence and selection

High

Social selection and influence

Osgood et al. (2014) [37]

2002

11–14

9500 at each wave

27 (rural, low SES)

USA (Iowa)

Longitudinal

HLM 6.08

Multi-level regression

To examine network positive in cohesive peer groups and its association with substance use

Medium

Network position

Context of Adolescent Substance Abuse study

Ennet et al. (2008 [38])

2002

11–17

6579

13 middle schools W1, 18 high schools W2/3

USA (North Carolina)

Longitudinal

SAS V9

Hierarchical Growth Models (HLM)

To investigate peer networks and context for substance abuse

Medium

Social selection and influence;

Network position

Ennet et al. (2006) [39]

2002

11–17

5104

13 middle schools W1, 18 high schools W2/3

USA (North Carolina)

Longitudinal

SAS IML, UCINET, HLM

Hierarchical Generalized Linear Models (HGLM)

To investigate peer networks and context for substance abuse

Medium

Network position

FINEdu (Finnish Educational Transitions)

DeLay et al. (2013) [40]

2004

15–17

1419

9 (4 vocational, 5 academic)

Finland

Longitudinal

SIENA

SAOM

To investigate the effect of selection, deselection and socialisation on smoking

High

Social selection and influence

Kiuru et al. (2010) [41]

2005

15–18

1419

9

Finland

Longitudinal

RSIENA

SAOM

To examine changes in smoking in relation to changing or stable peer groups

High

Social selection and influence

Unnamed study

Huisman & Bruggeman (2012) [42]

2008

13–14

961

5

Netherlands

Longitudinal

RSIENA

SAOM

To examine how networks mediate the relationship between smoking and SES

Medium

Socioeconomic status;

Social selection and influence

Huisman (2014) [43]

2008

13–14

857

4

Netherlands

Longitudinal

RSIENA

SAOM

To examine the link between network and smoking while accounting for selection effects

Medium

Social selection and influence

SILNE (Smoking Inequalities – Learning from Natural Experiments)

Lorant et al. (2017) [44]

2013

14–16

10,604

50

Europe (6 countries)

Cross-sectional

SAS 9.3

Logistic regression

To investigate the role of social ties in socioeconomic differences in smoking

Medium

Socioeconomic status

Robert et al. (2019) [45]

2013

14–17

11,015

50

Europe (6 countries)

Cross-sectional

SAS 9.3

Multi-level logistic regression

To investigate the association between academic performance, smoking and SES

Medium

Socioeconomic status

Mulassi et al. (2012) [46] (cross-sectional)

2010

14–18

285

1

Argentina

Cross-sectional

Pajek, Epi info, SPSS

Kamada-Kawai algorithm

To study the association between network structure and smoking

Low

Network position

Valente et al. (2013) [47]

2010

15–16

1707

5

USA (LA)

Cross-sectional

Not specified

Exponential Random Graph Models (ERGMS)

To compare the association between adolescent smoking and friend smoking across different types of network

Medium

Social selection and influence

Forster et al. (2015) [48]

2012

12–14

184

1

USA (LA)

Cross-sectional

UCINET, Stata

Logistic regression

To investigate the interplay of individual characteristics and peer influences on substance use

Low

Network position

Hall & Valente (2007) [49]

2001

11–13

1960 baseline, 880 follow-up

6

USA (LA)

Longitudinal

Stata and LISREL

SEM

To evaluate the relative strength of selection and influence on adolescent smoking over two timepoints

Medium

Social selection and influence

Ramirez-Ortiz et al. (2012) [50]

2003

15–19

486 baseline, 399 follow-up

1

Mexico

Longitudinal

NetMiner II 2.4.0, SPSS, Stata

Chi squared and logistic regression

To investigate the effect of centrality on smoking

Low

Network position

Lakon & Valente (2012) [51]

2004

12–21 (97% 12–18 years old)

851

14

USA (LA)

Cross-sectional

SAS

HLM

To investigate social integration and smoking

Medium

Social selection and influence

Van Ryzin et al. (2016) [52]

2000

11–14

1289

8

USA (Pacific Northwest)

Longitudinal

RSIENA

SAOM

To investigate whether being well-liked can serve as a risk factor for substance use

Medium

Network position

Valente et al. (2005) [18]

2001

10–12

1486

16

USA (LA)

Longitudinal

Not specified

Multi-level logistic regression

To investigate popularity, network position and smoking

Medium

Network position

Kobus & Henry (2010) [53]

1997

11–14

163

1

USA (Illinois)

Cross-sectional

FNET

Generalised Linear Models

To investigate the effect of network position, peer substance use and their interaction on adolescents’ own use

Medium

Network position