We found that there was an increase in reported exposure to drinking guidelines in January 2016, which did not reoccur in January 2017. Following this, reported exposure fell, although remaining significantly higher than in December 2015. Following this rise in exposure, we found increases in measures of capability (proportion who reported tracking units of alcohol consumption and considered it easier to drink safely) and opportunity (proportion who perceived their lifestyle as conducive to drinking within guidelines). However, the change did not persist over subsequent months. Additionally, we observed some marginal changes in other measures.
A key strength of this research is its use of a theory-based framework for studying behaviour change (COM-B) in order to explore the short term effects of promoting lower risk DG [9]. Further strengths include those of the Alcohol Toolkit Study, which collects a nationally-representative sample of drinkers living in private households in England using consistent methods on a monthly basis which enabled analysis of pre- and post- intervention trends across multiple data points [13]. The relative robustness of the methods compared to previous studies and the theoretically informed approach are likely to strengthen the generalisability of the results to culturally-similar high income countries.
Among the key limitations are the potential confounding effects of December and January being traditionally heavy and light drinking months. However, data were collected over 15 months which facilitated comparison of results at guideline publication to those from the following year. Data collection was by self-report which is subject to known biases in studies on alcohol consumption; even though this study is not measuring consumption directly, issues such as social desirability may still have impacted responses [16, 17]. Since the results presented aggregate the responses of all participants, conclusions regarding the effect of publication of the new lower risk DGs on population sub-groups cannot be made. The question wording did not explicitly refer to the UK official government guidelines but we are unaware of other advice being widely interpreted as guidelines. Lastly, the proportion of non-drinkers in the ATS sample appears somewhat higher than other surveys in England: for example, the Adult Psychiatric Morbidity Survey (APMS) – which is the only other survey in England to also include the AUDIT – reported that 23% of respondents did not drink at all, while the equivalent figure for the ATS in 2014 was 29% [18]. Unlike the APMS, however, the ATS does not include a clarifying question and likely misclassifies a proportion of people who drink very rarely as ‘never’ [19].
Although previous evaluation of the efficacy of drinking guidelines to change behaviour is limited, this result is consistent with findings of low impact on those measures which have been explored [7, 8]. A contributing factor in this study may be the lack of large-scale organised promotion of the new low-risk DGs. For example, there have been no national mass media campaigns following the initial announcement of the guidelines and, despite over 80% of UK alcoholic product labels including the DGs [20], these labels have not been updated to give the new recommendations with guidance on how to do so only being published in March 2017 [21]. Additionally, reviews of the effectiveness of behaviour change efforts have consistently shown that achieving exposure is not sufficient to achieve effectiveness where interventions and campaigns are poorly designed. Promotion of drinking guidelines should therefore be designed with reference to prior theory and evidence on effective communication of messages and techniques for changing behaviours [22, 23]. This could be supported by the finding of this research that although reported exposure to drinking guidelines increased in January 2016 and remained above the level observed in December 2015, sustained change in the theoretical mediators of behaviour change was not demonstrated.
There was no difference in respondents’ knowledge of the number of units per day which it is safe to drink regularly over time (see Additional file 1). This could be seen as inconsistent with existing literature, which suggests that DGs can improve public knowledge of alcohol harms [10,11,12, 24]. Furthermore, the percentage of people who gave one or two units as the most units they could regularly drink on a single day before doing significant harm to their health was low in January 2016 (35.4%), meaning that most people thought that the level for low risk drinking was above that given in the new low-risk DGs. However, the measure of knowledge used here does not ask what the guideline figure is – rather it asks for the number of units that the respondent can regularly drink without significant health risk. It may be that this is interpreted as being different to the low-risk DG. Additionally, the number of units that it is safe to drink ‘regularly’ according to the low risk DGs is open to interpretation. It was not within the scope of this study to explore lay interpretations of the new guidelines or the COM-B measures used.
Given the results presented here, and the findings of low impact of DGs on alcohol consumption in the previous literature [7, 8], policy makers should consider the process of guideline implementation as well as additional or alternative methods to DGs when working to produce change in alcohol consumption. It is important to provide accurate information on the risks of alcohol consumption [4]. However, guidelines do not implement themselves; they require active, evidence-based strategies to support implementation [25, 26]. Furthermore, in order to build on the current findings, it is important to consider the impact of drinking guidelines on higher risk drinkers, who may view 14 units per week as unobtainable, and on health inequalities given the stark differences in risk faced by those in different socioeconomic groups [27,28,29].