Appendix B.

Summary of Tobacco Longitudinal Studies.

Study [ref. no.], location, survey dates, ages, completion %	Outcome measures & empirical model	Advertising-promotion measures & selective results	Covariates in final model
Alexander et al. [102], New South Wales AU, 1979 & 1980, 10–12 years, 87%.	Change in smoking status from baseline (onset, quit, continued, nonsmoker). Logistic regression, but log-odds ratios not reported.	Approval of cigarette ads at baseline. Onset (adoption) of smoking is positively related to approval of advertising. Quitting is negatively related to approval. Smoking education classes are marginally related to onset, but not to quitting.	Age, parental smoking, sibling smoking, peer smoking, weekly spending money, teacher’s smoking, teacher’s gender, urban location, private school, alcohol use, smoking education classes. Interactions.
Armstrong et al. [103], AU schools, 1981 & 1982/83, 11–13 years, 82% & 64%.	Change in smoking status in prior 12 months (onset, continued) by gender. Stepwise logistic regression, but log-odds ratios not reported.	Perceived attraction to cigarette ads at baseline. For boys and girls, advertising is unrelated to onset at one-year follow-up and positively related at two-year follow-up. Smoking education classes have a significant negative effect in one of four cases (girls’ teacher-led).	Father smokes, mother smokes, sister smokes, best friend ever smokes, best friend currently smokes, believes most adults smoke, parental approval, peer pressure, perceived effects of smoking, country of birth, smoking intentions, smoking education classes.
Audrain-McGovern et al. [126], northern Virginia US, 2000–2003 (five waves), 14 years, 41%.	Four trajectories for smoking (9–12th grades). Latent class growth model. Attrition analysis.	Binary index for high- & low-receptivity (2 items: favorite brand, CBI). Receptivity is significant in 2 of 6 comparisons at 9th grade & 3 of 6 comparisons at 12th grade.	Gender, race, academic performance, alcohol use, marijuana use, depressive symptoms, novelty-seeking, peer smoking, physical activity, team sports participation.
Biener & Siegel [104], Mass. US, 1993 & 1997/98, 12–15 years, 58%. Attrition analysis.	Progression to smoking (100+ smokes in past 4 years) by baseline non-smokers. Logistic regression, but controls for only selected covariates. Unclear how these are selected.	Baseline receptivity to tobacco marketing (2 items: ownership of CBI, can name favorite ad’s brand). High receptivity is a predictor of progression to smoking, but moderate receptivity is not. High susceptibility is not significant if controlled for smoking susceptibility (p409). Reports cross-tabulation for receptivity.	Age, gender, race, parent education, household income, adult smoker in house, peer smoking, rebelliousness, depression, baseline initiation continuum, susceptibility to smoking.
Biener & Siegel [105], Mass. US, 1993 & 1997/1998, 12–15 years, 58%.	Eleven-point smoking initiation-susceptibility index (never smoked to 100 smokes & regular smoking past month). Multilevel regression.	Knowledge of tobacco slogans (12-pt scale) at follow-up (p207). Knowledge of tobacco slogans is a predictor of position on the smoking continuum, but omits other advertising covariates, including receptivity.	Age, gender, race, parent education, household income, peer smoking, adult smoker in house, perceived social value of smoking at follow-up, baseline initiation continuum. Mediation considered for perceived value but could be moderator relationship.
Charlton & Blair [106], 3 towns in northern UK, 1/1986 and 5/1986, 12–13 years, 100%.	Onset of smoking by gender for baseline nonsmokers. Stepwise logistic regression, but log-odds ratios not reported.	Cigarette-brand awareness; favorite advertisement; imputed TV sports cigarette-brand advertising. None of the advertising covariates are predictors of boys’ smoking onset (p815). For girls, awareness of at least one cigarette brand is significant.	Gender, parental smoking, peer smoking, positive view on smoking, negative view on smoking, perceived health effects of smoking, smoking education classes.
Choi et al. [107], California US, CTS, 1993 & 1996, 12–17 years, 49%.	Progression to established smoking (100+ smokes in past 3 years) by experimenters at baseline. Stepwise logistic regression.	Receptivity to tobacco advertising (3 items: own or willing to use CBI; have a favorite ad; could name any cigarette brand). Receptivity is a predictor of smoking at the high level, but not at the moderate level.	Age, gender, race, family relationships, family smoking, peer smoking, perceived peer smoking, perceived ability to quit, religiosity, school performance. Significant interactions between receptivity & other risk factors.
Dalton et al. [108], NH & VT US, 1999 & 2000/2001, 10–14 years, 73%.	Onset of smoking by baseline nonsmokers. Generalized linear (log-link) regression for relative risk ratios.	Smoking exposure in movies (for random sample of 50 movies). Receptivity to tobacco promotions is unreported covariate in multivariate regression. Movie smoking exposure is a significant predictor of onset.	Grade, gender, parent education, parenting style, school performance, parental smoking, sibling smoking, peer smoking, sensation seeking, rebelliousness, self-esteem, parents’ disapproval. Interactions.
Distefan et al. [109], California US, CTS, 1996 & 1999, 12–15 years, 67%.	Any smoking by baseline never-smokers. Popular stars’ movies in 3 years before baseline are reviewed. Logistic regression.	At baseline, respondents named their 2 favorite male & female movie stars. Favorite stars’ smoking predicts smoking for girls (but not boys). High receptivity also predicts smoking. Reports cross-tabulation.	Age, gender, race, school performance, family smoking, peer smoking, parents’ disapproval, susceptibility to smoking. Interactions with age, gender, etc.
Gidwani et al. [110], nationwide NLSY US, 1990 & 1992, 10–15 years, na.	Onset of smoking by baseline nonsmokers. Logistic regression.	TV viewing hours per day (0 to 5+ hours) at baseline. Statistically significant effects for 4-5 hours and more than 5 hours per day. Confidence intervals are unclear and some variable are excluded (p507).	Age, gender, race, math score, reading score, vocabulary score. Additional factors are household income, maternal education, mother’s age, maternal IQ, number of children in household.
Gilpin et al. [111], California US, CTS, 1993–1999, 1996–2002, 12–17 years & 18–23 years, 47% & 48%	Established smoking at follow-up by baseline experimenters and nonsmokers. Logistic regression. Attrition analysis	Receptivity to tobacco advertising at baseline (3 items: own or willing to use CBI; named highly advertised brand; name of brand in favorite ad). Receptivity is significant at moderate and high levels for both cohorts	Age, gender, race, school performance, parental smoking, sibling smoking, peer smoking, baseline smoking status. Interactions between receptivity and smoking status, peer smoking
Hanewinkel et al. [81], Schleswig-Holstein DE, 2005 & 2006, 10–16 years, 80%.	Onset of smoking by baseline never-smokers. Generalized logistic (log-link) regression and path analysis model.	Frequency of exposure to movies or videos that are rated as appropriate for ages 16 and older (FSK-16 rating). Significant results for two higher levels of viewing FSK-16 movies.	Age, gender, school performance, school type, parental smoking, sibling smoking, peer smoking, parenting style, sensation seeking.
Hanewinke & Sargent [122], Schleswig-Holstein DE, 2005 & 2006, 10–16 years, 82%.	Any smoking at follow-up by nonsmokers at baseline. Generalized linear model (log link), with school type as cluster variable.	Frequency of exposure to smoking in 50 popular US movies (extrapolated from 398 films); favorite tobacco ad. Significant results for movie exposure quartiles & favorite tobacco ad. Reports cross-tabulation.	Age, gender, school performance, school type, parental smoking, sibling smoking, peer smoking, parenting style, sensation seeking. Interactions between exposure & age, gender, etc.
Jackson et al. [112], North Carolina US, 2002 & 2004, 12–14 years, 85%. Attrition analysis.	Onset of smoking by baseline nonsmokers. Stepwise logistic regression, with separate results for blacks and whites.	Exposure to movies by rating; TV set in bedroom; hours of TV use; frequency of TV use; parental program rule for TV. In final model, R-rated movies & private TV are significant for whites. No variables are significant for blacks.	Grade, gender, race, school grades, parents’ education, family smoking, peer smoking, parental engagement, parental relationship, college aspirations, sensation seeking.
Lopez et al. [113], Asturias ES, base & 3 follow-ups, 13–14 years, 64%. Attrition analysis.	Progression to regular smoking (one per week) by baseline nonsmokers. Stepwise logistic regression.	Number of brands identified in 3 commonly displayed billboard ads at baseline. Significant effect of number of brands on regular smoking at 6, 12, & 18 month follow-up.	Age, gender, SES, family smoking, peer smoking, school. Other variables are missing full description (attitude, social influence, intentions to smoke). Interactions.
Pierce et al. [114], California US, CTS, 1993 & 1996, 12–17 years, 61%.	Susceptible to smoking (combines nonsmokers & experimenters). Logistic regression. See [131] for attrition analysis.	Receptive to tobacco advertising (3 items: own or willing to own CBI; have a favorite ad; named brand in favorite ad). Receptivity is significant at moderate and high levels.	Age, gender, race, school performance, family smoking, peer smoking. Interactions between exposure to smokers & susceptibility are not significant.
Pierce et al. [115], California US, CTS, 1996 & 1999, 12–14 years, 65%.	Onset of smoking by never-smokers at baseline. Logistic regression.	Receptive to tobacco advertising at baseline (3 items: own or willing to use CBI; have a favorite ad; named brand in favorite ad). Receptivity is positive if more-authoritative parents.	Age, gender, race, school performance, parental education, family smoking, peer smoking susceptibility to smoking, authoritative parenting style. Interactions with age & gender.
Pierce et al. [116], California US, CTS, 1996 & 1999, 12–15 years, 67%.	Experimented with smoking by never-smokers at baseline; susceptible to smoking Logistic regression.	Receptive to tobacco advertising (3 items: own or willing to use CBI; have a favorite cigarette ad; named brand in favorite ad). Neither moderate nor high receptivity predicts experimentation or susceptibility.	Age, gender, race, school performance, family smoking, peer smoking, susceptibility to smoking, curious about smoking at baseline. Interactions with age and gender (not significant).
Pucci & Siegel [117], Mass. US, 1993 & 1997/1998, 12–15 years, 59%. Attrition analysis.	Brand of initiation for experimenters; brand of regular smokers. Simple correlation analysis.	Individual exposure to brand-specific advertising in sample of 14 magazines (307 of 627 youth read one or more magazines in sample). Brand exposure is correlated with smoking.	Gender, race (only two covariates reported).
Sargent et al. [118], rural VT US, 1996 & 1997, 1998, 8–17 years, 66%. Attrition analysis.	Smoking status index on 0-5 scale (0 = never-smoker, 5 = 100+ cigarettes in lifetime). Logistic regression.	Own or willing to own CBI. Receptivity to CBI predicts progression on smoking index scale. Change in receptivity predicts progression in subsample. Reports determinates of receptivity.	Grade level, gender, school performance, parental education, family smoking, peer smoking, baseline smoking status, tobacco prevention program intervention.
Thrasher et al. [123], Cuernavaca & Zacatecas MX, 2006 & 2007, 11–14 years, 83%.	Smoking onset (past yr) & current smoker (past 30 days) by never-smokers at baseline. Logistic regression. Attrition analysis.	Exposure to movie smoking in standard list of 42 movies (minutes); own CBI. Mixed results for movie exposure & smoking onset. Movie exposure predicts current smoking. CBI insignificant for both outcomes.	Age, gender, school type, parental smoking, sibling smoking, peer smoking, parental approval, parenting style, sensation-seeking, self-esteem, TV in bedroom.
Titus-Ernstoff et al. [124], NH & VT US, 2002 & 2003 (3 waves), 9–12 years, 90%. Attrition analysis.	Onset of smoking by baseline nonsmokers. Poisson regression (relative risk ratios) for each wave of exposure.	Exposure to smoking in 50 movies (assessed at each wave). Baseline & later exposures predict smoking initiation.	Age, gender, race, school performance, parental smoking, peer smoking, sensation seeking, rebelliousness, self-regulation, self-esteem, parent education, maternal responsiveness, maternal monitoring.
Weiss et al. [119], California US, 2000, 2002 & 2003, 10–13 years, 80%.	Smoking susceptibility (combines nonsmokers & smokers). Multilevel model. Attrition analysis.	Exposure to pro-tobacco media (TV portrayals & displays at tobacco outlets). Pro-tobacco media predicts smoking susceptibility. Reports cross-tabulation for exposures.	Gender, race, immigration status, acculturation status, anti-tobacco media exposure. Interactions with pro- & anti-tobacco exposure; interactions with race & acculturation.
Wilkinson et al. [125], Houston, TX US, 2001 & 2003 (4 waves), 11–13 years, 90%.	Experimentation with smoking (ever, new). Stepwise logistic regression.	Movie-smoking exposure in a sample of 50 movies. For experimentation (new), movie-exposure is significant for Mexican-born, but not US born. Reports exposure means.	Age, gender, country of birth, family smoking, peer smoking, acculturation, parental education, risk taking, anxiety, detention. Interactions with country of birth & acculturation.
Wills et al. [120], NH & VT US, 1999 & 2000/2001, 9–13 years, 69%. Attrition analysis.	Onset of smoking by baseline never-smokers. Structural model with movie exposure at baseline as exogenous variable.	Movie-smoking exposure (number of occurrences in sample of 50 movies). Movie exposure has an indirect effect on onset through increased affiliation with peer smoking as well as a direct effect. Reports table of correlations.	Age, gender, race, school performance, parental education, parental smoking, sibling smoking, peer smoking, maternal responsiveness, mother’s rules, rebelliousness, sensation seeking, self-esteem, baseline smoking status.
Wills et al. [121], nationwide US, 2003 & 2004, 10–14 years, 85%. Attrition analysis.	Onset of smoking (ever smoked) by baseline never-smokers. Structural model with movie exposure at baseline as exogenous variable.	Movie-smoking exposure (number of occurrences in sample of 50 movies). Movie exposure has indirect effects on onset through smoking expectancies and peer smoking as well as a direct effect. Reports table of correlations.	Age, gender, race, school performance, family structure, parental education, parenting style, household income, parental smoking, sibling smoking, peer smoking, rebelliousness, sensation seeking, self esteem, self control, baseline smoking status.