Smoking in movies and smoking initiation in adolescents: systematic review and meta-analysis.

Background and aims Preventing young people from initiating smoking is a vital public health objective. There is strong evidence that exposure to smoking imagery in movies is associated with an increased risk of smoking uptake. However, the estimate of the magnitude of effect is not clear, as previous reviews have synthesized estimates of cross-sectional and longitudinal associations. Therefore, we have performed a systematic review to quantify cross-sectional and longitudinal associations between exposure to smoking in movies and initiating smoking in adolescents. Methods Four electronic databases (MEDLINE, EMBASE, PsycINFO and International Bibliography of the Social Sciences, IBSS) and grey literature were searched from inception to May 2015 for comparative epidemiological studies (cross-sectional and cohort studies) that reported the relation between exposure to smoking in movies and smoking initiation in adolescence (10-19 years). Reference lists of studies and previous reviews were also screened. Two authors screened papers and extracted data independently. Results Seventeen studies met our inclusion criteria. Random-effects meta-analysis of nine cross-sectional studies demonstrated higher exposure (typically highest versus lowest quantile) to smoking in movies was associated significantly with a doubling in risk of ever trying smoking [relative risk (RR) = 1.93, 95% confidence interval (CI) = 1.66-2.25]. In eight longitudinal studies (all deemed high quality), higher exposure to smoking in movies was associated significantly with a 46% increased risk of initiating smoking (RR = 1.46; 95% CI = 1.23-1.73). These pooled estimates were significantly different from each other ( P = 0.02). Moderate levels of heterogeneity were seen in the meta-analyses. Conclusions The cross-sectional association between young people reporting having seen smoking imagery in films and smoking status is greater than the prospective association. Both associations are substantial, but it is not clear whether or not they are causal. [ABSTRACT FROM AUTHOR]