This report includes direct quotations from media and in places, where copyright permission could not be obtained, comments have been paraphrased by the authors (see Chapter 3). These statements include a range of opinions that relate to the quoted source and do not reflect the views of authors or funders.
Alcohol is a major contributor to global mortality and ill health and is a causal factor in over 200 health conditions including liver disease, heart disease, injuries and seven types of cancer.1 The Global Burden of Disease study attributed an estimated 2.3 million deaths and 85 million disability-adjusted life-years (DALYs) lost to alcohol in 2015.2 In England, alcohol caused an estimated 24,208 deaths in 2017 and 1.2 million hospital admissions in 2017–18 and was reported by victims as a factor in 39% of violent crimes.3,4 Estimates of the economic cost to society of alcohol-related health and social harms vary, but the most commonly used estimate is £21B (UK annual estimate).5
In response to this, successive UK Governments have published a series of alcohol strategies aiming to reduce alcohol consumption and the harm and disruption it causes to drinkers and those around them.5–7 The most recent strategy was published in 20125 and included the aim to:
Support people to make informed choices about healthier and responsible drinking, so it is no longer considered acceptable to drink excessively.
It further sought to:
Ensure everyone understands the risks around excessive alcohol consumption to help them make the right choices for themselves and their families.
A key policy for achieving these goals was to ask the Chief Medical Officer (CMO) to oversee a review of the country’s low-risk drinking guidelines.
Guidelines for low-risk drinking
Drinking guidelines have proliferated internationally since the 1980s and health authorities in most high-income countries now issue guidelines for alcohol consumption.8 These usually recommend a quantified level of alcohol consumption that drinkers should not exceed over a specific time period (e.g. per day or per week). The consumption level is usually expressed in standard drinks or units of alcohol. In the UK, 1 unit is 8 g or 10 ml of pure ethanol and is approximately half a pint of beer, 75 ml of wine or one 25-ml shot of spirits, depending on the alcoholic strength of the drink. The guideline consumption threshold is also usually accompanied by textual information that may include, among other things, the source of the guideline (e.g. the CMO recommends . . .), a reason to listen to the advice (e.g. to reduce your risk of alcohol-related disease . . .) and supplementary recommendations (e.g. do not drink every day).
Figure 1 shows that drinking guidelines often vary markedly between countries. There may be a number of reasons for this. First, the purpose of guidelines often differs between countries. For example, Denmark provided guidelines to help people avoid high-risk levels of alcohol consumption until 2010, whereas the UK has tended to issue guidelines to promote low-risk consumption.8 Other countries use guidelines primarily to inform government policy (e.g. to assess levels of excessive drinking) or clinical practice, rather than as health promotion information for the public. Second, the riskiness of drinking may vary between countries depending on the prevalence of competing or compounding risk factors, such as smoking and obesity. Third, guidelines are usually expressed in units or standard drinks, but these measures vary widely across countries. For example, 1 unit is 8 g of pure ethanol in the UK, 10 g in Ireland, 12 g in Finland, 14 g in the USA and 20 g in Austria.9 If guidelines are to be easily understood by the public and not expressed in decimals, variation in the definition of 1 unit necessitates variation in drinking guidelines.
FIGURE 1 Weekly drinking guidelines for women and men in selected countries in 2016. Shaded countries at the bottom of the figure are those with the same guideline for men and women. (Figure not reproduced here; see the original article.)
Guidelines also vary over time. The earliest quantified guideline dates back to Anstie’s limit of 28–43 g per day, which was developed in the 1860s and used in life assurance in the UK and USA.10 More recently, UK health authorities published ‘safe limits’ for men of 64 g per day in 1979 and then 144 g per week in 1984, followed by ‘sensible limits’ of 168 g per week in 1987 and then 24–32 g per day in 1995.11 These changes reflect improvements in epidemiological evidence, societal standards regarding the acceptability of health risks and the purpose of guidelines, as described above. They also reflect changes in how guidelines are set. As described by Holmes et al.,12 the evidence used by guideline development committees is increasingly likely to include bespoke epidemiological modelling exercises in addition to reviews of scientific literature.
Despite their widespread use, drinking guidelines do not enjoy universal support from public health actors. As discussed below, their effectiveness as a public health intervention is questioned,13,14 and some researchers believe that they are counterproductive. For example, Casswell15 argued in 1996 that guidelines are unlikely to affect behaviour when presented in the context of the pro-alcohol messages that suffuse contemporary media, a view shared by the authors of the leading summary of alcohol policy effectiveness.13 In 2012, Casswell16 reiterated this point and added that promoting guidelines draws government away from more effective interventions. Similarly, Bellis17 argues that, when implemented in practice, drinking guidelines serve the interests of the alcohol industry as, ‘rather than explaining risks, [they] can read more like an alcohol promotion slogan’ that presents recommended consumption levels.
Other public health actors view guidelines more positively. Heather18 argues that they are essential as the public has a right to know about alcohol-related health risks and they facilitate discussions about alcohol consumption in clinical practice. Marteau19 further suggests that, although guidelines may not lead directly to reductions in alcohol consumption, they increase public concern about the risks of alcohol consumption and may make it easier to introduce effective alcohol control policies.
Drinking guidelines in the UK, 1995–2016
Prior to 2016, the UK drinking guidelines had not changed since 1995, when the Sensible Drinking report20 recommended that men should not regularly consume more than 3–4 units a day and women should not regularly consume more than 2–3 units a day, where regularly means drinking this amount every day or on most days.
The House of Commons Science and Technology Committee launched an inquiry in 2011 to assess whether or not the 1995 guidelines were ‘robust’ as a foundation for alcohol policy.11 It sought submissions on the evidence underpinning the guidelines, whether or not that evidence could be improved, how well the government communicates the guidelines and how the guidelines compare with other countries. The report concluded that the passage of time, greater evidence on the causal relationship between alcohol and cancer and increased scientific scepticism regarding the health benefits of moderate drinking meant that the guidelines should be reviewed by an expert group.
In 2012, the UK Government published its alcohol strategy5 and announced that the CMO would lead a review of the guidelines, which ran from 2013 to 2016. Although the review was led jointly by the four UK CMOs, it was conducted primarily by a guideline development group (GDG) that mainly comprises civil servants, clinicians and academics with expertise in health and behavioural science. The GDG considered published studies, commissioned new systematic reviews, epidemiological modelling and focus group research, and heard presentations from expert advisors. In 2014, its interim report concluded that that there was sufficient new evidence to justify changes to the guidelines. The CMOs accepted this conclusion.
In 2016, the GDG published its final report21 along with the new draft guidelines,22 which recommended that the guidelines should be revised to state that men and women could keep their risks from alcohol consumption low by consuming no more than 14 units per week. Additional accompanying guidance recommended that drinkers consuming ≥ 14 units per week spread their alcohol consumption across several days and have at least 2 drink-free days per week. These recommendations differed from the 1995 guidelines in three main ways. First, they switched from a daily to a weekly guideline, citing evidence that the previous guideline was often misinterpreted as a limit on consumption on any single occasion and seen as less useful as a result.23 Second, they set the same guideline for men and women, a decision that triggered substantial comment and criticism as only two other countries take this approach (see Figure 1). Third, they reduced the previous guideline for men. That guideline was often interpreted as ≤ 14 units per week for women and ≤ 21 units for men, a legacy of pre-1995 guidance, and the revisions were, therefore, viewed as reducing the guideline for men by one-third.
The GDG also revised the guidance in two other key areas. First, it updated the guidance on drinking alcohol during pregnancy. Previously, the guidance in Scotland was that women should not drink during pregnancy, whereas the guidance in England was that women should not drink more than 1 or 2 units once or twice per week. The revised guidance removed this inconsistency and took a precautionary approach that recommended:
. . . the safest approach is not to drink alcohol at all, to keep risks to your baby to a minimum.
Second, the GDG provided updated guidance on single occasions of drinking and elected not to provide a specific guideline but instead recommended that drinkers keep their risk to a low level by:
Limiting the total amount of alcohol you drink on any single occasion, drinking more slowly, drinking with food and alternating with water, and planning ahead to avoid problems.
Notably, this guidance did not incorporate the widely used definition of binge drinking as the consumption on a single occasion of more than 6 units (women) or more than 8 units (men). The GDG elected not to set a guideline consumption threshold for single drinking occasions as risks strongly vary substantially between individuals depending on their ability to reduce such risks by other means.21
The new guidelines were announced publicly by the CMOs in January 2016 and were formally adopted by the UK Government in August 2016 following a public consultation.25
The effectiveness of drinking guidelines
Little evidence is available on the effects of drinking guidelines, and reviews note that there have been no rigorous evaluations of the impact of producing, revising or promoting drinking guidelines on alcohol consumption or on other factors related to behaviour change.13,26–28 This is somewhat surprising as public guidance on a wide range of health-related behaviours, including smoking, physical activity, drink-driving and nutrition, has produced small to moderate effects on knowledge, attitudes and behaviour.28–32
A number of factors may explain the lack of evidence on the effectiveness of drinking guidelines. Opportunities to evaluate the impact of guidelines are rare as revisions take place at intervals of ≥ 10 years and, when they do happen, the new guidelines are often not promoted via large-scale campaigns after the initial announcement. Furthermore, few countries routinely collect data that would facilitate robust evaluation. For example, data on awareness, knowledge or exposure to guidelines are not a common feature of public health surveys, and alcohol consumption data are typically collected on an annual basis only, which makes robust analysis of intervention effects difficult as they are likely to be modest given the scale of promotional activity. For these reasons, the small literature that does attempt to examine effects of promoting drinking guidelines relies on relatively weak research designs.
Studies in several countries have used cross-sectional analyses to examine public knowledge of guidelines. These analyses generally find that large minorities of the population are aware of drinking guidelines and can correctly identify guideline consumption levels, but there is little evidence that this influences drinking behaviours.33–35 For example, studies in Australia, Denmark and the UK have examined changes in guideline-related knowledge and perceptions over time using annual cross-sectional data.36–39 In each case, the results suggest that knowledge of guidelines improved over time. In Denmark, this appeared to be linked to annual mass media campaigns to promote the guidelines, but, in Australia, beliefs about what constitutes safe drinking also changed.
The UK data are particularly relevant to the present study. The Office for National Statistics (ONS) Opinion Survey collected data on knowledge of units and guidelines between 1997 and 2009 and summarised the findings in a 2010 report.40 The report shows that the proportion of drinkers who had heard of units rose from 79% in 1997 to 90% in 2009. Similarly, the proportion who knew how much beer, wine and spirits constituted 1 unit also rose by up to 10–15 percentage points. In 2009, 63% of respondents knew that 1 unit of beer was approximately half a pint, 69% knew that 1 unit of spirits was a single measure, but only 27% knew that 1 unit of wine was less than a 125-ml glass. The proportion of respondents who had heard of guidelines increased from 54% in 1997 to 75% in 2009. However, only 44% of men and 55% of women in 2009 correctly identified the guideline for their sex, an improvement of only 6 percentage points among men and of 14 percentage points among women over a period of 12 years. Moreover, only 13% of respondents kept a check on their units, with no increase in this proportion between 1997 and 2009. There was little variation in these measures by sex, but heavier drinkers generally showed greater awareness and knowledge of the guidelines and of units.
Prior to the present study, the most detailed evaluation of promoting drinking guidelines focused on an Australian campaign aiming to raise awareness of guidelines and alcohol-related cancer risks.41 The study used repeat cross-sectional surveys of 150–200 women at three time points over a 15-month period and concluded that multiple waves of advertising led to good recognition and awareness of the campaign content. Promoting drinking guidelines also led to increased cancer risk awareness, knowledge of guidelines and intentions to reduce alcohol consumption among heavier alcohol drinkers in particular. However, there was no evidence of impacts on drinking behaviour.
In addition to these quantitative analyses, a number of qualitative studies have sought to understand how drinkers respond to and use drinking guidelines. These studies report that drinkers typically disregard guidelines as they do not align well with their current drinking practices and rely on drinkers understanding units, which is often not the case.23,42,43 Instead, drinkers monitor their consumption through embodied means; that is, they make judgements on the riskiness of their drinking based on how intoxicated they feel and their previous experiences of alcohol and its after-effects.
Effectiveness of alcohol health warning labels
Although there is little robust research on the effectiveness of promoting drinking guidelines, a separate body of research exists examining the effectiveness, potential effectiveness and appropriate design of alcohol health warning labels. This literature is important, as approximately 80% of products sold in the UK in 2014 included the CMO’s guidelines on their label,44,45 and this is a key location where drinkers are likely to encounter the guidelines when they are not being promoted via more active means.46 More generally, information on labels is argued to be a potential influence on social norms around alcohol consumption and increase support for other alcohol policies.47
The USA introduced mandatory health warning labels on alcoholic products in 1989, which read ‘Government Warning: (1) According to the Surgeon General, women should not drink alcoholic beverages during pregnancy because of the risk of birth defects. (2) Consumption of alcohol impairs your ability to drive a car or operate machinery, and may cause health problems’.48 Evaluations of the impact of these labels constitute a large proportion of the alcohol labelling literature and the highest-quality evidence on the effectiveness of alcohol health warning labels as a public health intervention.49 Overall, the evaluations suggest that awareness and recall of USA alcohol warning labels increased initially,48,50,51 but plateaued after around 3–4 years in both adults and adolescents.52,53 As with drinking guidelines, there was little evidence that these gains in awareness and knowledge translated into changes in alcohol-related behaviours, such as reduced drink-driving or alcohol consumption.49,53,54
More recent research has used survey, focus group and experimental methods to examine several themes relating to how consumers may use and respond to information and warnings provided on alcohol product labels. For example, focus group research in Australia found that participants favoured labels providing messages related to short-term risks and also found messages with positive rather than fear-based or numerical framings, general health messages rather than ones relating to cancer or specific cancers, and risk-based rather than deterministic messages more believable, convincing and personally relevant.55 A small experimental study of students at a UK university56 found no significant differences between text-based messages and pictorial messages in terms of fear arousal and intentions to reduce or stop alcohol consumption, and Hobin et al.57 obtained a similar finding in an experimental study with Canadian adults. This is in contrast to a general view that pictorial messages are more effective,58 although Hobin et al.57 obtained a similar finding in an experimental study.
The UK’s Royal Society for Public Health surveyed drinkers to determine what information they find most useful.47 The results revealed that drinkers view alcoholic strength and units as the most important information to include on labels and consider warnings against drink-driving and drinking during pregnancy as moderately important; however, drinking guidelines and links to online information are viewed as substantially less important.47 Thomson et al.59 found that drinkers in Australia supported the provision of health information and warnings on product labels, whereas other studies found that drinkers in Australia report poor recall or use of industry-backed logos and weblinks on product labels60 but good awareness of standard drink labels.61 In Canada, drinkers demonstrate a high level of ability to use standard drink information on product labels, and a focus group study of drinkers in Yukon, Canada, also found that detailed product labels including a health warning, guidelines and standard drink information were seen as of intrinsic value.57,62 The information provided in this study was viewed by participants as new, useful, important and something they had a right to know. Participants also suggested that it could prompt new conversations and potentially influence behaviour. However, other studies suggest that, particularly among some younger drinkers, standard drink information may be used to maximise the volume of alcohol purchased rather than to moderate consumption.47,61,63 More broadly, a criticism common across research on alcohol labels is that the information provided on real-world labels is too small, discreet, vague or ambiguous, or otherwise lacking in visual impact, to inform or influence consumer behaviour.44,55,60
The breadth of findings here and the reliance on small-scale, cross-sectional, focus group or experimental research are perhaps the greatest weaknesses of the literature. Hassan and Shiu64 reviewed studies of the efficacy of alcohol warning labels published between 2000 and 2015 and concluded that the literature is ‘a weak and fragmented evidence base to guide policy-makers or practitioners, which requires further research on most key questions’.64 They particularly criticise the lack of consistent research approaches and measures, which prevents comparison across studies, and the lack of long-term research on the effects of labels. As with the literature on drinking guidelines, we further note that few of these more recent studies following the US warning label evaluations provide evidence on whether such labels can or do affect drinking behaviour. One exception is Maynard et al.,63 who found that providing calorie or unit information did not affect alcohol consumption by UK undergraduates in a laboratory study. Maynard et al.’s63 additional qualitative analyses suggest that this is because such information fits poorly with drinking to get drunk or socialise, echoing findings from previous studies.23,43
Despite their apparent lack of direct impact on alcohol-related behaviours, alcohol health warning labels may, like drinking guidelines, be important in other ways. The US labels are often viewed by the authors of evaluation studies as important because they inform consumers about alcohol-related risks and reinforce this information, even if they do not affect drinking behaviour.52 Others note that such labels help to stimulate conversations about alcohol and its risks as well as changing social and cultural norms, potentially leading to behavioural changes in future generations,47,54,62 or greater acceptance of alcohol control measures.19,47 More simply, warning labels may fulfil consumers’ right to know about the risks of a product they are purchasing.47,62
The literature on the evaluation of alcohol labels has three main implications for our evaluation of drinking guidelines. First, it supports the evidence from the limited literature on the effectiveness of promoting drinking guidelines in showing that health-related information pertaining to alcohol can be recalled and used by drinkers. Second, it also supports the evidence that such information does not generally lead to changes in alcohol consumption behaviour. As suggested by behaviour change theory, such behaviour changes may require well-designed and implemented messaging.65 In this context, the text around the numerical guideline consumption level may matter as much as the guideline consumption level itself and the content of promotional activity may be as important as its presence. Third, despite the importance of such design and implementation, there is insufficient evidence available to date to identify clear recommendations regarding what is effective in the design of drinking guidelines or guideline-related messaging. Literature from health promotion in other areas, particularly tobacco warning labels, can provide some guidance, but it is unclear whether or not this can be translated unproblematically to alcohol, the risks of which are smaller in scale, different in nature and interact with more positive public attitudes.64 This means that it is difficult to assess if we would expect the guidelines as written to be effective or what effects they may have.
Rationale for the present study
Despite concerns about their effectiveness, drinking guidelines still play an important role in alcohol policy debate in the UK and internationally. Successive UK governments have been reluctant to increase regulation of the alcohol market and have focused instead on promoting ‘responsible drinking’ with the support of alcohol producers and retailers.66 The 2012 alcohol strategy5 initially departed from this approach by promising the introduction of minimum unit pricing (MUP) for alcohol, a greater role for public health within alcohol licensing and possible restrictions on price-based promotional deals for alcohol. However, all of these measures were removed from the strategy following a public consultation,67 leaving the review of the drinking guidelines as one of the headline measures alongside increased attempts to identify and deliver brief interventions to drinkers within primary care. The 2016 revisions to the UK drinking guidelines therefore represented both a central plank of government alcohol policy and a rare opportunity to undertake a prospective evaluation of a widely used but under-researched intervention.
Designing an evaluation is not straightforward as promoting drinking guidelines is a complex intervention. Promotional activity is delivery by many different organisations and actors in varying forms, with an extended causal chain from implementation to outcome. Organisations involved include governmental and non-governmental bodies, such as Public Health England (PHE), the NHS, Drinkaware (London, UK) (an industry-funded charity), alcohol producers, retailers and social aspect organisations, public health charities, and news media. Individual actors include social media users, public commentators, individual politicians and the general public to the extent that each of these discusses the guidelines. Likewise, promotional activity ranges across mass media campaigns, interactive social media, consultations with health professionals, product labels, point-of-sale advertising and general public debate. Behaviour change theory suggests that any changes in alcohol consumption that arise from this network of promoters and promotion will be mediated by changes in behavioural influences, including individuals’ knowledge of the guidelines and alcohol-related risks, their motivations and their social context.65,68 Therefore, a robust and comprehensive evaluation requires an understanding of not just the effects of promoting drinking guidelines on key outcomes, such as alcohol consumption, but also how and by whom guidelines are promoted and the effects of this promotion on proximal outcomes, such as behavioural influences. Appendix 1 presents a causal logic model indicating our understanding of the connections between these processes and informs the design of the evaluation.
Aims and objectives
The aim of this study was therefore to conduct a prospective evaluation of promoting revised UK drinking guidelines on alcohol consumption behaviour. To do so, it sought to meet the following objectives:
- to document the timing, audience and content of major promotional activity following the publication of revised drinking guidelines
- to use interrupted time series analysis of monthly survey data to assess whether or not trends in alcohol consumption behaviour, as a primary outcome, changed following publication and promotion of the revised drinking guidelines
- to undertake subgroup analyses to examine whether or not there are variations in intervention effects across groups of the population defined by age, sex and socioeconomic status
- to use difference-in-difference methods to examine whether or not direct and frequent exposure to promotion of drinking guidelines increases their effectiveness
- to undertake pathway analyses to validate theorised capability, opportunity and motivation to change behaviour and behaviour itself
- to assess cost-effectiveness of any identified effects on alcohol consumption using the Sheffield Alcohol Policy Model framework.69
During the project, the objectives were amended as follows in consultation with the Project Steering Committee. First, as we identified no large-scale promotional activity relating to the guidelines beyond the announcement of the revisions, the first objective was abandoned. This was replaced with a new objective: to conduct a review of the scale and content of news media coverage relating to the guidelines. Second, preliminary analyses showed no changes in the primary outcome measure and, in the absence of any promotional activity, we decided not to conduct a cost-effectiveness analysis. Third, the lack of any substantial change in the outcome measures meant that we did not pursue the difference-in-difference analysis or pathway analyses.
At the same time, a decision was also taken to not conduct a cost-effectiveness analysis as preliminary analyses indicated that, in addition to there being no promotional activity that attracted a cost, there was also no change in the primary outcome (alcohol consumption). We judged, in consultation with the Project Steering Committee, that conducting the news media review was a better use of resources. The cost-effectiveness analysis is therefore not discussed further in this report.
Patient and public involvement
The research team conducted two patient and public involvement (PPI) sessions during the project to support the interpretation of the early and final results of the evaluation analyses. In the absence of a pre-existing and appropriate PPI panel in England, we worked with an established PPI panel based in Scotland. The panel comprised approximately 30 adult drinkers from the general population and living in or around Stirling, where the panel meets. The UK Centre for Tobacco and Alcohol Studies recruited the initial panel members in 2014 via social media, with additional members added over time to maintain the group’s size. The panel meets two or three times per year and provides members with a £25 voucher plus an offer to cover travel costs for each meeting attended.
One member of this PPI panel also joined the Project Advisory Group and attended the group’s meetings.
Structure of the report
The remainder of the report comprises a further four chapters. Chapter 2 describes the initial attempt to construct a timeline of promotional activity and the early findings prior to the abandonment of this exercise. Chapter 3 describes the review of print news media coverage of the guidelines. Chapter 4 describes the interrupted time series analyses. Chapter 5 brings together the main findings and discusses the strengths and limitations of the research and its implications for future research, policy and practice before presenting our conclusions.