This systematic review of seven cohort studies on over 13,000 participants shows some evidence for an association between prior alcohol advertising and marketing exposure and subsequent alcohol drinking behaviour in young people. All seven studies demonstrated significant effects across a range of different exposure variables and outcome measures. These included exposure to direct advertising using broadcast and print media and indirect methods such as in-store promotions and portrayal of alcohol drinking in films, music videos and TV programmes. The consistency of effect across a heterogeneous group of studies may be considered a strength.
Notably, three studies showed that onset of drinking in adolescent non-drinkers at baseline were significantly associated with exposure. Robinson [19
] showed that for each additional hour of TV viewing per day the risk of starting to drink increased by 9% during the following 18 months. Sargent [21
] found that for additional hour of exposure to alcohol use depicted in popular movies there was a 15% increase in likelihood in having tried alcohol 13 to 26 months later. Ellickson [16
] showed that exposure to in-store beer displays significantly predicted drinking onset two years later. Effects were less clear in baseline drinkers, whilst greater exposure predicted greater drinking frequency, analyses adjusting for possible confounding factors failed to detect significant relationships.
In studies on mixed groups of drinkers and non-drinkers, increased frequency of TV viewing and music video viewing was highly significantly related to the amount of alcohol consumed while going out [20
]. In the study by Snyder [17
] of US individuals aged 15 to 26 years, for each additional advertisement seen the number of drinks consumed increased by 1%.
Of interest, to our knowledge, at least two more prospective cohort studies meeting our inclusion criteria have been published since our review was completed [39
]. Since updating our searches for all new studies is beyond the original scope of the project, we have not incorporated these two studies into the main body of the review. Nevertheless, it is important to note that both of these studies also showed significant relationships between receptivity to alcohol marketing or alcohol advertising in young people. Eleven year olds in the highest centile of exposure to TV beer advertisements, alcohol ads in magazines, in-store beer displays and beer concessions, radio listening time and ownership of beer promotional items were 50% more likely to be drinkers than youth in the lowest centile of exposure one year later controlling for demographic and psychosocial factors and prior drinking [39
]. In a sample of non-drinkers aged 11 to 15 years, those reporting high receptivity to alcohol marketing defined as owning or wanting to own alcohol branded promotional items were 77% more likely to initiate alcohol use one year later compared with youth reporting minimal receptivity adjusted for demographic and psychosocial factors and social influences to drink [40
There are several limitations that should be considered when interpreting the results of this review. Whilst we made an a priori
decision to only include and review cohort studies which potentially are less likely to suffer from systematic bias than less robust study designs such as cross-sectional surveys or interrupted time series studies, it is nonetheless important to note that cohort studies are also susceptible to bias if not designed and executed using rigorous standards. One of the biggest threats to the validity of observational studies such as cohort studies is the issue of confounding, whereby the outcome of interest is influenced by some other factor or factors in addition to the exposure of interest. Whereas all of the studies controlled for a variety of confounding factors possibly related to alcohol drinking behaviour, unmeasured or unknown confounders cannot be adjusted for and it is not possible to know if residual confounding influenced the analysis. For example, alcohol expectancies, family history, peer influence and personality characteristics may act as confounders in the relationship between exposure to advertising and marketing and subsequent alcohol use. Given the magnitude of the effect sizes shown in these studies, we cannot rule out the possibility that they were due to the effects of residual and unmeasured confounding [41
]. However, previous work evaluating smoking exposure in movies and smoking behaviour in adolescents using a simulation model showed that effects of unknown or unmeasured confounders would need to be large in order to overturn the results [42
]. Given that no observational study can control for all unmeasured or unknown confounders, researchers may wish to consider using similar approaches to determine the potential impact of such confounders.
Whilst these studies suggest that exposure to advertising and alcohol portrayal in the media increase likelihood of later alcohol consumption, they are unable to inform us how exposure brings about these changes, or what aspects of advertising and marketing are the active components. The extent to which psychological factors determine subsequent behaviours is a worthwhile topic for further study. One study [43
] has examined how persuasive alcohol media messages were associated with concurring beliefs and behaviours among youth, concluding that existing exposure based studies do not adequately account for the complex psychological causal mechanisms that may moderate or mediate the relationship between exposure and outcome. However, this analysis is based on cross-sectional data; further studies with longitudinal analyses are desirable. If a better understanding of the relationship of the intermediate steps between exposure and subsequent behaviours can be obtained, then our understanding of the mechanisms of action of alcohol advertising and marketing would be improved. This question, together with lessons learned from the collective experiences of conducting cohort studies [44
], should inform the design of future cohort studies.
One other serious threat to the validity of these studies was the degree of attrition in some of the studies. Losses to follow-up between assembly of the cohort and follow-up are inevitable but the aim is to keep this to a minimum as attrition bias may be introduced if reasons for missing data or loss to follow-up are related to exposure or outcome. If adolescents who were lost to follow up were more likely to be drinkers, or at high risk of drinking as found in three of the studies [17
], then this may then lead to underestimating the relationship between advertising and drinking. Generalisability of the results is also affected if losses are in one specific subgroup of participants, and the subsequent loss of power is also a problem with attrition. Of note, none of the studies reported how they estimated sample sizes required. In general, assessment of the design and conduct of the cohort studies reviewed was hampered by the lack of important methodological detail, and fell short of the current recommendations as set out in the STROBE statement [45
We cannot rule out the possibility of publication bias, whereby studies failing to detect significant relationships were not published, or studies for which selective reporting of only positive associations were published. Of course it is also possible that studies showing positive associations, if sponsored by the alcohol industry or other commercial organisations with a vested interest in advertising or marketing of alcohol, have not been published. Therefore, it is not possible to predict the likely impact of unpublished data on the results of this review. It is also possible that published studies were not found by our search as a fully comprehensive search of databases other than Medline and Embase and other sources only covering the social science literature was not possible within the scope of the limited funding for this review. Attempts, however, were made to locate all available studies by supplementing searches of databases with hand searching reference lists of key reviews and primary studies, which identified many articles published in journals not covered by Medline and Embase.
The results of these cohort studies are supported by findings in cross-sectional surveys which consistently report associations between increased exposure to alcohol advertising or marketing and drinking behaviour [2
], intentions to drink [46
] or advertising awareness and liking [2
]. Although, in one interrupted time-series study countries with advertising bans had lower levels of alcohol consumption and road traffic fatalities [50
], others failed to demonstrate significant effects [51
]. The rationale for the exclusion of these studies is outlined in the methods, and their exclusion would only be a concern if they generally showed a strong effect in the opposite direction.
One question that remains is whether early drinking behaviour shown in these cohort studies is predictive of risky or harmful drinking or alcohol-related problems in the future. Drinking onset at an earlier age has been shown to be associated with a greater likelihood of alcohol dependence in several cross-sectional studies [53
]. More recently, prospective cohort studies have also shown clear and significant associations between age of onset of drinking and subsequent heavy drinking and alcohol-related problems [56
Given the large budgets allocated to advertising and promotional activity by the alcohol industry, a paucity of research exists evaluating the effects of this advertising. Further research exploring the potential causal impact is warranted; the role of mass media as a potential source of influence on alcohol related knowledge and behaviour of young people has been neglected in many countries [60
The data from these studies suggest that exposure to alcohol advertising in young people influences their subsequent drinking behaviour. The effect was consistent across studies, a temporal relationship between exposure and drinking initiation was shown, and a dose response between amount of exposure and frequency of drinking was clearly demonstrated in three studies [17
]. It is certainly plausible that advertising would have an effect on youth consumer behaviour, as has been shown for tobacco [61
] and food marketing [62
Does this systematic review provide evidence that limiting alcohol advertising will have an impact on alcohol consumption amongst young people? Not directly: as we noted earlier we can not rule out that the effects demonstrated in these studies is due to residual confounding. Counter-advertising [30
], social marketing techniques [63
] or other prevention options such as parenting programmes [64
], price increases and limiting availability may offer more potential to limit alcohol problems in young people. Nonetheless, we now have stronger empirical evidence to inform the policy debate on the impact of alcohol advertising on young people, and policy groups may wish to revise or strengthen their policy recommendations in the light of this stronger evidence [1