Background Social media (SM) may influence adolescents’ perceived social norms and subsequent health risk behaviours, although the evidence base around this is still developing. We conducted a systematic review of the relationship between SM and adolescent health risk behaviours: alcohol/tobacco/drug use, e-cigarettes, diet, physical activity, antisocial behaviours, gambling, sexual risk behaviours and multiple health risk behaviours, in adolescents aged 10–19 years.

Methods We searched CINAHL, EMBASE, MEDLINE, APA PsycINFO, SocINDEX, preprint repositories and Google Scholar for studies published post-1996 reporting at least one relevant outcome with an SM measure (PROSPERO: CRD42020179766). Exposures of interest were time on SM, frequency of use, and exposure to health risk behaviour content (HRBC). Screening and risk of bias (RoB) were completed independently by two reviewers using a modified Newcastle Ottawa Scale. Following Cochrane guidance, we conducted synthesis based on direction of effects (benefit vs harm), sign testing and estimation of the proportion of datapoints reporting adverse effects (presented). Meta-analyses will produce average effect sizes (underway).

Results Of 13,150 hits, 84 studies were included. Twenty studies were low RoB, 27 moderate, and 38 high. Between studies all outcomes were addressed, the most common being alcohol use (n=25) and sexual risk behaviours (n=20). Twelve studies investigated >1 outcome. For alcohol use, most datapoints reported harmful effects of time spent (88.9%; 95% CI 56.5–98.0%, p=0.04), frequency (79.3%; 61.6–90.2%, p=0.002), exposure to HRBC (100.0%; 75.8–100.0%, p<0.001) and other SM activity measures (81.8%; 52.3–94.9%, p=0.07). Datapoints examining sexual risk behaviours mostly reported harmful effects of time (75.0%; 30.1–95.4%, p=0.63), frequency (91.7%; 64.6–98.5%, p=0.006), HRBC (100.0%; 43.9–100.0%, p=0.25), and other SM activity (76.2%; 61.5–86.5%, p<0.001). For e-cigarettes (n=8) and antisocial behaviour (n=17), all datapoints reported harmful effects of SM (e-cigarettes 95% CI 67.6–100.0%, p=0.008; antisocial behaviour 95% CI 81.6–100.0%, p<0.001). Across all outcomes, exposure to HRBC on SM was most likely to report a harmful effect (100.0 vs 83.0% for other exposures, p=0.0062). Harmful effects were similar for datapoints at high (85.6%) and low/moderate (86.7%) RoB.

Conclusion SM use is adversely associated with adolescent health risk behaviours, particularly exposure to content pertaining to these behaviours. The current evidence base is limited by methodological weaknesses, including a lack of longitudinal data (risking reverse causation) and future robust research to assess causality is needed. Given the increasing targeting of SM by unhealthy commodity industries, available evidence suggests action to reduce the risk adolescents face from exposure to health risk behaviours is needed.