How Safe Is Safe Enough? International Frameworks for Constructing Alcohol Drinking Guidelines

1. Introduction

Drinking alcohol, like any human behaviour, carries risk. This relationship has been well understood since early times. One of the earliest drinking guidelines is reflected in the writings of the Greek poet Eubulus, who wrote in the 4th century BCE, “I mix three kraters only for those who are wise. One is for good health, which they drink first. The second is for love and pleasure. The third is for sleep, and when they have drink it those who are wise wander homewards. The fourth is no longer ours. But belongs to arrogance. The fifth leads to shouting. The sixth to a drunken revel. The seventh to black eyes. The eighth to a summons. The ninth to bile. The tenth to madness, in that it makes people throw things” (Hunter, 1983).[1]

Over time, the concept of safety in relation to drinking alcohol has given way to a focus on alcohol’s contribution to health risks. The World Health Organization (WHO) includes the harmful use of alcohol among the leading modifiable global risk factors for death and disease [2,3,4,5] and harmful drinking is a significant contributor to social cost in terms of premature death and disability, with a financial burden from healthcare, enforcement, and other downstream costs (WHO, 2010). This shift in emphasis reflects the evolution of scientific insight into the relationship between alcohol and health, and the importance of both drinking level and drinking pattern. It also reflects a more general and aspirational shift towards the lowest possible level of risk in all things.

Risk mitigation involves various approaches, with a global emphasis on regulating both the physical and economic availability of alcohol. [6,7,8] However, proper risk mitigation also requires raising public understanding about health relationships and ensuring transparent access to complete information about drinking and its various outcomes, both negative and positive. The purpose of drinking guidelines is to support this process and to provide an accessible, scientific, evidence-based message that can be applied to individual choices about drinking.

The risk equation for drinking reflects biological, epidemiological and toxicological evidence. Yet the same body of evidence has led to widely varied assessments and recommendations across national drinking guidelines. These reflect differences in the selection of evidence, methodological approaches, risk endpoints, and the framing of the advice provided. Each of these matters a great deal, with implications for how the science is interpreted and the conclusions on which recommendations are based. Equally important is the public health message the guidelines are intended to convey, whether to indicate the lowest risk levels or to provide guidance on moderate drinking. It is also worth acknowledging at the outset that guidelines do not recommend drinking for any potential health or other reasons. Drinking guidelines are a public health information tool with a cautionary purpose.

The methodology used to develop drinking guidelines has implications that extend well beyond the recommended number of drinks. Guideline development shapes dietary advice, chronic disease prevention strategies, cancer prevention policies, public health messaging, and consumer understanding of alcohol-related risk. It is also central to ongoing debates about moderate alcohol consumption, for example, wine consumed with meals as part of traditional dietary patterns such as the Mediterranean diet. Divergent conclusions about health effects often reflect differences in the conceptual frameworks used to assess and communicate alcohol-related risk rather than disagreements about the underlying epidemiological evidence. Understanding how drinking guidelines are constructed is therefore essential for interpreting their recommendations and appreciating why national guidance can differ despite drawing on much of the same scientific literature.

In this review, we examine the principal methodological frameworks used to develop drinking guidelines over the past two decades. Rather than simply comparing national recommendations, we explore how differences in risk assessment, quantitative modelling, policy judgement and risk communication have shaped guideline development across countries, often yielding markedly different conclusions despite drawing on much of the same epidemiological evidence. We identify five broad conceptual approaches to guideline development and examine their respective strengths and limitations. Drawing on these approaches, we propose an integrated framework that combines their complementary attributes to provide a transparent, scientifically robust and accessible basis for future drinking guidelines and to improve public understanding of alcohol-related risk. This framework also provides a clearer context for interpreting evidence on moderate alcohol consumption, including wine consumed as part of dietary patterns, by distinguishing between the assessment of health risks and their communication in public health guidance.

2. International Drinking Guidelines

On the whole, individual decision-making about whether to drink alcohol or engage in other voluntary activities does not rely on science-based risk calculations. Individual decisions are based on a subjective trade-off between perceived potential harms and potential benefits that depend on several factors that are personally determined – the degree of control over exposure and outcomes, how familiar the risk is, and views on the relative importance of immediate and long-term consequences [9]. This calculation is highly subjective, but drinking guidelines can help ground this decision accessibly in what science tells us.

Drinking guidelines are issued by governments and quasi-governmental organisations in countries around the world. At last count, some form of guideline was available in 56 countries [10]. Most offer specific advice on daily or weekly drinking thresholds, expressed in terms of grams of pure alcohol. A smaller proportion provides guidance about the number of ‘glasses’, ‘servings’, or ‘drinks’. In a handful of countries, advice is simply confined to cautioning about drinking less, drinking in moderation, or avoiding harmful use. While presumably based on the same science, guidelines arrive at vastly different conclusions and provide disparate recommendations. For example, daily recommendations range from 10 g to 40 g, and weekly recommendations range from 70 g to 220 g. Furthermore, while most guidelines differentiate between thresholds for men and women, others set a single one for both (Table 1).

The disparity across guidelines reflects complex choices about approach. This includes decisions about whether to rely on relative or absolute risk, and where (and whether) to set a threshold for an ‘acceptable’ risk level. Other choices concern how to use the available scientific evidence to develop risk curves and which evidence to include in formulating guidelines. Related to this is the choice of risk endpoints, with some guidelines based on combined overall risk and others on the risk of individual diseases.

3. Conceptual Frameworks Underlying Drinking Guideline Development

Despite substantial differences between national drinking guidelines, most can be grouped into five broad methodological frameworks: (1) synthesis of observational evidence; (2) quantitative disease-risk modelling; (3) precautionary minimum-risk approaches; (4) continuum-of-risk communication; and (5) integrated absolute-risk frameworks. These frameworks differ not only in the scientific evidence they draw upon but also in the assumptions they make about how alcohol-related risk should be assessed, communicated and translated into public health advice.

3.1. Defining Risk and an Acceptable Threshold

The first consideration in developing drinking guidelines is how alcohol-related risk should be measured. Traditionally, recommendations have been based on observational epidemiology, comparing health outcomes among drinkers with those of non-drinkers or lifetime abstainers using relative risk estimates. This approach commonly uses all-cause mortality as the principal endpoint because it captures the net relationship between alcohol consumption and premature death across multiple diseases and causes, rather than focusing on individual conditions alone [11]. It also reflects the long-standing observation that, compared with people who do not drink alcohol, light and moderate drinkers have often been found to experience lower overall mortality and lower risks of several major chronic diseases, including cardiovascular disease and type 2 diabetes [12,13,14,15,16,17,18,19,20,21,22]. These findings underpin the widely described J-shaped relationship between alcohol consumption and health outcomes. However, the existence and interpretation of the J-shaped curve remain the subject of ongoing scientific debate because of concerns about residual confounding and other methodological limitations inherent in observational studies [23,24,25].

An alternative approach derives recommendations from explicit estimates of absolute lifetime risk. Rather than comparing drinkers with non-drinkers, this methodology begins by selecting an acceptable level of alcohol-attributable risk and then estimating the level of alcohol consumption associated with that threshold. Because alcohol consumption is a voluntary behaviour, the selection of an acceptable level of risk inevitably involves a policy judgement as well as a scientific one. Room and Rehm (2012) argued that these two processes should be explicitly distinguished, thereby making transparent both the estimation of risk and the value judgment underlying the chosen threshold [26]. Building on this conceptual framework, Rehm, Lachenmeier and Room (2014) proposed that alcohol should be evaluated against a more stringent lifetime voluntary risk threshold of 1:1,000, illustrating how the choice of threshold can substantially influence the resulting recommendations [27].

3.2. Choosing the Endpoint

Guidelines also differ according to the health outcomes used to derive recommendations. Some are anchored in all-cause mortality, providing an overall assessment of the balance between increased and decreased risks across multiple diseases. Others focus on specific outcomes, particularly cancers such as breast cancer, for which risk increases even at relatively low levels of alcohol consumption [28,29]. More recent frameworks also incorporate acute alcohol-related harms, including injuries and violence, recognising that chronic disease and immediate harms arise through different drinking patterns.

The choice of endpoint is therefore not merely a technical decision. Whether recommendations are based on all-cause mortality, disease-specific risks, acute harms, or lifetime alcohol-attributable mortality has a profound influence on the resulting guideline. National differences reflect not only the interpretation of scientific evidence but also differing judgements about which health outcomes should receive the greatest emphasis in public health advice.

3.3. Implications for Guideline Development

The methodological choices outlined above underpin the diverse approaches adopted internationally. The following sections examine five representative frameworks—observational evidence (U.S.), disease-risk modelling (U.K.), precautionary minimum-risk (Netherlands), continuum-of-risk communication (Canada), and integrated absolute-risk (Australia)—to illustrate how different approaches to risk assessment and communication can yield markedly different drinking recommendations, despite drawing on much of the same epidemiological evidence.

4. Observational Evidence Synthesis: United States

Since their introduction in 1980, U.S. alcohol recommendations have been incorporated within the Dietary Guidelines for Americans, reflecting the view that alcohol consumption should be considered within the broader context of diet and health [30]. The basis for the guidelines was a conventional systematic approach that reviewed observational studies of risk endpoints across different drinking levels, compared with that for non-drinkers. The original guideline advised, “One or two drinks daily appear to cause no harm in adults. If you drink, you should do so in moderation” [30]. Nuance was added in 1995, differentiating between risk for women and men and advising consumption of no more than one standard drink (the equivalent of 14g) and two standard drinks a day, respectively [31].

Tasked with informing the 2025 update of the guidelines, a scientific panel selected by the U.S. National Academies of Sciences, Engineering, and Medicine (NASEM) once again relied on the conventional approach of reviewing observational evidence relating to several risk endpoints including all-cause mortality, cardiovascular disease, neurocognitive outcomes, several cancers, and maternal and infant outcomes related to alcohol consumption during lactation [32]. Based on a comprehensive review, the panel concluded that, compared with never drinking alcohol, moderate alcohol consumption was associated with lower all-cause mortality, lower risk of nonfatal myocardial infarction and nonfatal stroke, and increased risk of breast cancer [32]. No conclusions could be drawn about the relative risk of other cancers, including colorectal, oral cavity, pharyngeal, oesophageal, or laryngeal cancers. Similarly, evidence was deemed insufficient to draw conclusions about dementia, cognitive decline, and Alzheimer’s disease risk. The panel concluded that the available evidence did not justify revising the existing recommendations of 1 standard drink per day for women and 2 for men and considered these limits consistent with the balance of the available evidence.

In light of concerns about confounding and other limitations of observational studies, a second review was conducted in parallel and outside of the standard 5-year reviews by the Departments of Agriculture and Health and Human Services. The assessment, the Alcohol Intake and Health Study (AIAS), was undertaken under the auspices of a different government agency, the Interagency Coordinating Committee on the Prevention of Underage Drinking (ICCPUD). Unlike the NASEM review, which synthesised observational evidence across a broad range of health outcomes [32], the AIAS adopted a modelling approach that began by specifying an acceptable lifetime risk threshold and then estimating the level of alcohol consumption associated with it [33]. Instead, its starting point was to define an ‘acceptable’ risk threshold at 1:1,000, as proposed by Rehm, Lachenmeier and Room (2014), and work backwards using a mathematical model [27,33]. The model incorporated diseases and injuries considered causally attributable to alcohol consumption, according to established comparative risk assessment methodologies [34]. Causality was attributed not only to acute injury and noncommunicable diseases but also to some infectious diseases, such as tuberculosis and HIV/AIDS. While the model also assessed the risk of all-cause mortality, the approach was unconventional: it did not measure all-cause mortality directly but instead constructed individual risk curves for each outcome and then combined them.

Based on the modelling approach, the report concluded that the level of drinking corresponding to an ‘acceptable’ risk of 1:1,000 was no more than 7 drinks per week for both men and women and recommended revising the advice in the U.S. Dietary Guidelines by lowering the weekly recommendation for men from 14 to 7 drinks [33]. Importantly, the estimated drinking threshold proved highly sensitive to the selected acceptable-risk criterion. If the threshold were set at 1:100 rather than 1:1,000, the corresponding drinking limit would increase from 7 to approximately 8.5 drinks per week [33]. This illustrates how policy judgements regarding acceptable risk can materially influence the resulting recommendations.

The U.S. drinking guideline process illustrates how different methodological frameworks can produce different recommendations despite drawing on much of the same scientific evidence. The NASEM review synthesised observational evidence to evaluate the health effects of moderate drinking [32], whereas the AIAS adopted an explicit absolute-risk modelling approach based on a predefined acceptable-risk threshold [33]. These contrasting approaches demonstrate that differences in drinking guidelines arise not only from the evidence itself but also from the conceptual framework used to interpret and translate that evidence into public health recommendations.

The guideline revision process also attracted considerable public and scientific debate regarding the evidence considered and the methodologies employed, although discussion of these issues lies beyond the scope of this review. Ultimately, neither the NASEM nor the AIAS recommendations were adopted in the 2025–2030 Dietary Guidelines for Americans, which instead advise adults who choose to drink alcohol to drink less. At the time of writing, however, the U.S. Centers for Disease Control and Prevention (CDC) continue to describe moderate drinking as up to 1 standard drink per day for women and up to 2 for men [35].

5. Disease-Risk Modelling: United Kingdom

An alternative approach rests on assessing drinking levels that may be considered ‘low-risk’. Like the U.S. AIAS, it sets an ‘acceptable’ threshold and works backwards to determine corresponding drinking levels. The difference, however, is in the selection of endpoints. The ‘low-risk’ approach sets the thresholds for men and women at the lowest risk level (and thus consumption level) for any individual health outcome.

The revision of the U.K. Chief Medical Officers' (CMO) drinking guidelines marked a fundamentally different approach to developing drinking guidelines compared with that adopted in Australia and discussed later in this review. Rather than beginning with an explicitly defined acceptable level of lifetime risk, in 2008 the U.K. Department of Health commissioned the Sheffield Alcohol Research Group to develop a comprehensive epidemiological model estimating the relationship between alcohol consumption and a broad range of alcohol-related diseases, injuries and causes of death. The objective was not to identify a 'safe' level of drinking, but to estimate the level of alcohol consumption at which the aggregate risk across multiple alcohol-related diseases and injuries remained comparatively low.

The resulting Sheffield Alcohol Policy Model integrated evidence from systematic reviews and meta-analyses with U.K.-specific epidemiological, mortality, and drinking-pattern data. Dose-response relationships were modelled across numerous chronic diseases, including cancers and cardiovascular disease, as well as acute outcomes such as injuries [36]. This comprehensive modelling framework was among the first attempts to synthesise all alcohol-related health risks into a single quantitative model to inform national drinking guidelines.

A notable outcome of this modelling was the recommendation that both men and women should limit alcohol consumption to 14 U.K. units per week (approximately 112 g ethanol). The U.K.’s previous guidelines had recommended higher limits for men, but revised evidence indicated that the overall risk difference between men and women had narrowed substantially when multiple health outcomes were considered together. Breast cancer, which contributes to risk at relatively low levels of alcohol consumption, was recognised as an important factor in the overall balance of harms for women, while acute injury risks continued to rise with heavier episodic drinking, particularly for men.

Conceptually, the U.K.’s approach differs from Australia's integrated absolute-risk framework in an important respect. Australia first defines an acceptable level of lifetime risk and then estimates the alcohol consumption associated with that risk. The Sheffield Alcohol Policy Model, by contrast, integrates dose–response relationships for multiple chronic diseases and acute injuries with the U.K.’s population drinking data to estimate the level of alcohol consumption associated with a relatively low overall population risk [37,38,39]. Thus, the U.K.’s recommendations arise from disease-risk optimisation rather than from an explicit acceptable lifetime risk threshold.

6. Precautionary Minimum-Risk Approach: The Netherlands

The Dutch Health Council (the Council) adopted perhaps the most precautionary interpretation of the current evidence on alcohol and health [40]. Rather than seeking an acceptable level of lifetime risk or modelling the balance of competing health outcomes, the Council based its recommendations on the level of alcohol consumption that minimises the overall disease burden [42].

The 2015 Dutch Dietary Guidelines concluded that observational evidence supporting health benefits from moderate drinking had become increasingly uncertain, whereas evidence linking alcohol consumption to several cancers, particularly breast cancer, had strengthened, despite also being derived from observational studies [28,40,41,42,43,44]. Consequently, the Council recommended that adults either abstain from drinking alcohol or limit intake to no more than 1 standard drink per day [40]. Unlike previous Dutch guidance, this recommendation no longer sought to identify a level of drinking that might confer health benefits, instead reflecting a precautionary interpretation of the available epidemiological evidence.

The Dutch approach, therefore, differs conceptually from both the Australian and U.K. frameworks. It does not define an explicit acceptable lifetime risk threshold, nor does it seek to optimise the balance between competing disease risks through quantitative modelling. Rather, it adopts a minimum-risk philosophy, in which the recommended level of consumption corresponds to the lowest observed overall disease risk 40,42,42]. This approach places greater emphasis on precaution than on explicit risk trade-offs and is particularly influenced by evidence on alcohol-related cancers, for which no convincing threshold has been identified [28,43,44].

Although highly precautionary, the Dutch guideline highlights another important dimension of drinking guideline development. Public health authorities may choose to prioritise avoiding even small increases in disease risk over communicating acceptable risk thresholds or relative differences in risk. Consequently, the Dutch recommendation reflects a policy judgement about the desirable level of protection for the population, rather than a quantitative determination of an acceptable lifetime risk.

Viewed alongside the approaches adopted in Australia, Canada, the U.K. and the U.S., the Dutch guidelines show that differences between national recommendations arise not only from varying interpretations of epidemiological evidence but also from differing philosophies about how risk should be evaluated, communicated and translated into public health advice.

7. Continuum-of-Risk Communication: Canada

Recent Canadian guidance emphasised a graded continuum of risk and shifted public messaging towards lower thresholds and risk reduction rather than a single moderation limit [45,46]. Official drinking guidelines in Canada, first issued in 2011, are billed as ‘low-risk’ recommendations [47], clarifying that ‘(a)ny amount of alcohol can have risks to your health.’ In terms of advice, these ‘low-risk’ guidelines differ markedly from those issued in both the U.K. and the Netherlands. Recommended drinking limits exceed even those supported by the NASEM review in the U.S. where women are advised to consume no more than 2 drinks per day or 10 per week, while the advice for men is not to exceed 3 per day or 15 per week [47]. Given the comparable amount of alcohol in a standard drink in Canada and the U.S. (13.45 g vs 14 g), the Canadian recommendation, despite its title, is considerably higher, making it, in practice, a ‘moderate’ drinking guideline. The choice of terminology, while confusing, once again underscores the arbitrariness inherent in crafting drinking guidelines.

In 2023, the Canadian Centre on Substance Use and Addiction (CCSA) reviewed the evidence underpinning the 2011 guidelines and issued its own drinking recommendations [45]. The result is a hybrid framework that, according to the CCSA, "is based on the principle of autonomy and harm reduction and the fundamental idea behind it that people living in Canada have a right to know that all alcohol use comes with risk" [45]. Unlike previous drinking guidelines, which sought to identify a single low-risk consumption limit, the CCSA deliberately reframed alcohol consumption as a continuum of progressively increasing risk. Rather than defining a threshold below which drinking could be considered ‘safe’, the guidance was designed to communicate how risk increases incrementally with increasing consumption, thereby supporting informed decision-making and harm reduction [45,46]. This continuum-of-risk framework operationalises drinking recommendations accordingly: at zero drinks per week, there is no alcohol-attributable risk, while increasing consumption is associated with progressively higher levels of risk, categorised as low, moderate and increasingly high. The threshold for the latter category was set at 7 standard drinks per week, closely aligning with the recommendation emerging from the U.S. ICCPUD review [45,48]. The convergence between the two recommendations is noteworthy. Several members of the expert panels participated in both evidence reviews, although each review was conducted independently within its own policy context [33,46].

What is curious, however, is that, using the same methodology and a ‘tolerable’ risk threshold of 1:1,000, the same drinking level was framed as the cutoff for ‘high-risk’ in one instance (Canada) but not in the other (U.S.) [26,45]. While the CCSA recommendations were intended to replace the original ‘low-risk’ advice issued by Health Canada, both guidelines are currently in use, with the latter still appearing as the official government drinking advice [49]. Canada, therefore, represents a distinct methodological framework in which the principal innovation lies not in estimating alcohol-related risk, but in communicating that risk transparently as a continuum to support informed consumer choice and harm reduction [45,46].

8. Integrated Absolute-Risk Framework: Australia

Australia occupies a distinctive place in the evolution of drinking guidelines. While many countries have revised their drinking recommendations in response to emerging epidemiological evidence, Australia was among the first to explicitly recognise that alcohol poses two fundamentally distinct forms of risk: the cumulative lifetime risk associated with regular alcohol consumption and the short-term risk associated with heavier drinking on individual occasions. Rather than attempting to capture these risks in a single recommendation, the Australian guidelines addressed them separately, acknowledging that they arise from different biological mechanisms, drinking patterns, and public health considerations [50,51,52].

This integrated approach was first introduced in the 2001 National Health and Medical Research Council (NHMRC) Australian Alcohol Guidelines and refined in the 2009 revision [50,51], making Australia one of the earliest countries to systematically distinguish between chronic and acute alcohol-related harms within a single national guideline. The 2009 guidelines recommended limiting average daily consumption to reduce the lifetime risk of alcohol-related disease and premature death, while separately recommending limits on the amount consumed on any one occasion to reduce acute harms such as injury, violence and road traffic crashes. In doing so, Australia moved beyond the traditional concept of ‘moderate drinking’ and recognised that no single drinking limit adequately captures the multiple dimensions of alcohol-related risk.

Although this framework pre-dated the work of Room and Rehm (2012), it aligns closely with the conceptual rationale they subsequently articulated. They argued that drinking guidelines should distinguish explicitly between the scientific estimation of alcohol-related risk and the policy judgement concerning what level of voluntary risk society considers acceptable, thereby making both the evidence and the underlying value judgement transparent [26]. Building on this framework, Rehm, Lachenmeier and Room (2014) proposed that alcohol should be evaluated against an explicit lifetime voluntary risk threshold, illustrating how the choice of acceptable risk directly influences the resulting recommendations [27].

The 2020 NHMRC Guidelines to Reduce Health Risks from Drinking Alcohol retained Australia's integrated lifetime- and short-term-risk framework while updating the recommendations using contemporary epidemiological evidence and formal quantitative risk modelling [52,53,54]. Rather than identifying a threshold below which drinking could be considered ‘safe’, the guidelines identified a level of consumption associated with a lifetime risk of alcohol-attributable death of less than 1% (1:100) [54]. This led to revised advice that healthy adults should consume no more than 10 standard drinks per week and no more than four standard drinks on any one day. The accompanying public health message—’the less you drink, the lower your risk of harm’—further reinforced that alcohol-related risk exists on a continuum rather than at a single threshold [52].

Australia's experience also illustrates an important distinction between constructing and communicating risk. The underlying methodology is based on an explicit absolute lifetime risk criterion, yet the advice communicated to the public integrates this long-term perspective with practical guidance on avoiding acute harms associated with heavier drinking occasions. In contrast with approaches that focus exclusively on a single disease endpoint, a single absolute-risk threshold, or a continuum of increasing risk, the Australian guidelines recognise that different drinking patterns produce different types of harm and that meaningful public health advice should address both [52,53,54].

From an international perspective, Australia may be the first mature example of an integrated drinking guideline, one that brings together epidemiological evidence, explicit lifetime risk modelling, and practical advice to avoid acute harms, all within a single coherent framework. This distinction between lifetime and acute alcohol-related risk has also contributed to broader international discussion of drinking guideline methodology and remains one of the clearest examples of integrating multiple dimensions of alcohol-related risk into a single public health framework [26,52]. Table 2 compares these frameworks.

These frameworks should not be regarded as mutually exclusive. Rather, they reflect different emphases in guideline development. Most contemporary drinking guidelines draw on several frameworks, although one usually predominates. The challenge for future guideline development is therefore not to identify a single universally correct methodology, but to integrate the complementary strengths of observational evidence, quantitative modelling, explicit risk assessment and effective risk communication into a coherent and transparent process.

9. Toward a Coherent Risk Framework for Drinking Guidelines

Each decision about how best to craft drinking guidelines is, at least in principle, scientifically defensible, and each has its strengths and shortcomings. Given their purpose as a tool for informing individual decisions about drinking, however, disparate and contradictory recommendations with inconsistent terminology and framing are likely to create confusion and perhaps raise more questions than they answer.

An important consideration regarding drinking guidelines is whether they are likely to have a significant and meaningful impact on the public, drinking behaviours, and harms. A more coherent approach includes making the risk-constructing process and the rationale for risk thresholds explicit, explaining how existing evidence was used, which endpoints are prioritised and why, and how uncertainty is handled. While this may not be widely comprehensible to the general public, transparency is paramount. A subsequent step is to create drinking advice that is accessible to the lay public, along with an explanation of how it relates to individual circumstances. For the average person deciding whether to have that glass of wine, this is what counts.

The following proposition includes reasonable elements for inclusion in drinking guidelines.

Net population health. A sensible starting point is to use all-cause mortality risk as the ‘anchor’ for population-level thresholds, with transparent discussions of disease-specific risks (such as cancers) at different consumption levels. This approach was used in the U.S. NASEM review and offers specificity.

Chronic vs. acute harms. The Australian and Canadian guidelines provide recommendations for average weekly, daily, and even occasion-based intake limits (Australia). This is important public health advice that reflects the short-term risks of drinking, related to intentional and unintentional injury, and the long-term risk of developing non-communicable diseases or increasing the chance of dying prematurely. Such information also requires explanation of specific situational risk, for example, as relates to driving, pregnancy, or medications, and individual-level risks like being in recovery from alcohol use disorders or a family history of cancer or cardiovascular disease. Understanding how these factors may interact with drinking, particularly at lower levels of consumption, can help relate abstract risks to personal circumstances. It also offers individuals an indication of whether drinking alcohol can be part of a balanced lifestyle.

Voluntary risk context. To be credible, thresholds and recommendations should acknowledge that a ‘zero-risk’ scenario is hypothetical when it comes to virtually every human activity, and that drinking alcohol is no different from anything else. Risks from drinking alcohol are not much different from those from other voluntary risks. Therefore, an appropriate calibration of thresholds is important. Voluntary risk thresholds are variously set at 1:100 or 1,000. Where on this range alcohol risk should sit is subject to debate, but it matters a great deal. It is important to acknowledge that the use of the 1:1,000 benchmark, although defended by its proponents, is fundamentally arbitrary.

Consistency in endpoints. While drinking guidelines in most countries distinguish between men and women in terms of their recommendations, not all do. There is no disagreement that drinking alcohol affects men and women differently and at different levels, yet some countries, for example, the U.K. and the proposed AIAS limits in the U.S., set the same threshold for men and women. The reason for this is the use of two different endpoints – for men, the recommendation is guided by risk of acute harm, while for women, the driver is breast cancer risk. Without a proper explanation, this inconsistency sends a misleading message that risk is the same for both. Even with a proper explanation, the question remains: why is acute harm from road traffic crashes, for example, addressed through guidelines rather than with other, more targeted interventions?

Risk stratification. Risk thresholds and exposures vary across populations and cohorts, so advice needs to be tailored to their specific needs, depending on age, gender, health status, and other factors. Differentiating between men and women has been discussed in the context of appropriate endpoints, and specific recommendations are also needed for older adults that consider age-related changes in responses to alcohol. Currently, few countries offer specific advice for older adults, an area in need of attention given the world’s ageing population.

Absolute risk, relative risk, and uncertainty. The concepts of absolute and relative risk are not easily comprehensible, and relative risk, often used in public discussions and by the media, is difficult to understand and may present a disproportionate picture of the true magnitude of risk. While an increase in risk for drinkers relative to abstainers may seem large, in real or absolute terms, considering the prevalence of particular diseases, this difference may in fact be small. Information about absolute risk, reflecting the baseline incidence of diseases, would provide the necessary context for health outcomes within a particular national setting. Comparative risks for other activities would help show how alcohol-associated risk fits into overall lifestyle and behaviours.

In practice, this approach could be envisioned as follows.

Weekly or daily guidance for healthy adults based on ACM, that is complemented by information on: (a) acute-harm occasions and drinking patterns to avoid; (b) situations in which drinking may be inadvisable; (c) groups for whom reduced intake may be advisable or those who should consider not drinking at all, with an explanation of why this is the case; (d) when to seek clinical advice; and (e) contextualizing risk within other lifestyle choices and activities.

Studies have examined the impact of drinking guidelines and their uptake, showing that when unaccompanied by proper explanation and context, they are likely to be perceived as out of sync with realistic lifestyles and may be ignored. A review of the response to the U.K.’s CMO guidelines found little change in drinking behaviour, partly because understanding of the guidelines and their crafting was low [55,56]. A common misperception was that the guidelines were limiting personal freedom of choice, and media reports highlighted inconsistencies between the U.K.’s guidelines and international best practice, without a clear understanding of either the approach or the intent [57]. A survey conducted in the U.K. and Sweden found that subjective perceptions of drinking and the perceived level of risk it poses determine how guidelines are viewed and whether they are likely to be followed [58].

Finally, the discussion of drinking guidelines and their construction is missing a key element, likely the most important in practice. Most people drink simply because they enjoy it. While the pleasurable dimension of drinking alcohol, optimal when drinking occurs in moderation, cannot be quantified and included in drinking guidelines, it needs at least to be acknowledged. Across cultures, alcohol consumption is a well-integrated aspect of daily life and has been for millennia. It should be recognised as such. Efforts to impose draconian limits are likely to be perceived as killjoy public health messages intended to de-normalise drinking and strip drinkers of agency. Regardless of how they are crafted, a fundamental requirement for meaningful guidelines and advice is to recognise that people drink for many different reasons and that the point of guidelines is to equip them to do so safely, or at least without inordinate risk.

10. Conclusions

This review has shown that differences between national drinking guidelines arise not only from the interpretation of epidemiological evidence but also from the conceptual frameworks used to assess, interpret and communicate alcohol-related risk. Whether recommendations are derived from observational evidence, quantitative disease-risk modelling, precautionary public health principles, continuum-of-risk communication or explicit absolute-risk thresholds has a profound influence on the advice ultimately provided to consumers.

The development of drinking guidelines is therefore both a scientific and a policy process. Science provides estimates of risk, but decisions about acceptable risk levels, the weighting of health outcomes and the communication of uncertainty inevitably involve value judgements. Greater transparency in distinguishing scientific evidence from policy decisions would improve the consistency, credibility and public understanding of future drinking guidelines.

Drawing on the strengths of the principal international approaches, we propose that future guideline development should adopt an integrated framework that combines robust epidemiological evidence with explicit risk thresholds, distinguishes between lifetime and short-term harms, and communicates alcohol-related risk within the broader context of dietary patterns, lifestyle behaviours and individual circumstances. Such an approach would better reflect the complexity of alcohol's relationship with health than reliance on a single disease endpoint or a single numerical threshold.

Finally, drinking guidelines should recognise that alcohol is consumed within social and dietary contexts rather than in isolation. Drinking patterns, food consumption, overall dietary quality, physical activity, smoking and other lifestyle behaviours all influence health outcomes and should be considered when translating scientific evidence into public health advice. This is particularly relevant to moderate wine consumption, which is often consumed with meals as part of dietary patterns such as the Mediterranean diet. While no guideline can eliminate uncertainty, those that acknowledge this broader context are more likely to support informed decision-making, improve health literacy and help individuals make choices that reflect both scientific evidence and their own values and preferences.

Future drinking guidelines should be judged not only by the scientific evidence they rely on, but also by the transparency of the framework through which that evidence is translated into meaningful public health advice.