Constructing and responding to low‐risk drinking guidelines: Conceptualisation, evidence and reception

Despite significant scientific and communication challenges, government and non-government agencies in numerous countries continue to issue advice to the populace on how to reduce health and safety risks when alcohol is consumed. Often referred to as ‘low-risk drinking guidelines’ (LRDG), the advice typically recommends both average and upper daily levels for consumption for men and women, as well as when to abstain completely. The scientific and conceptual bases for these guidelines have only rarely been precisely spelt out and subjected to independent scrutiny. Having both been engaged in the development and revision of LRDGs in Australia and Canada at different times over the past 15 years—and having come to different conclusions [1] , [2] —we hope the various perspectives and empirical analyses presented here will contribute to informed debate, clearer thinking and perhaps eventually an improved conceptual and empirical basis for future guidelines. The challenges in developing simple upper limits to consumption on a particular occasion and in terms of volume over time are immense despite—or maybe because of—the voluminous literature linking levels of alcohol consumption with risks of various types of social and health harm. In the first place, there is enormous variation in the riskiness of a particular dose or volume of alcohol. Individual responses to a standard dose of alcohol vary substantially. Whether or not food has been taken beforehand can affect obtained blood alcohol levels by two or threefold [3] ; body weight varies across individuals to an even greater extent; both metabolic and functional tolerance to the effects of alcohol vary substantially across individual drinkers, but also over time for one person as a function of their current pattern of drinking and age [4] ; risk of injury from a given dose is influenced to a large degree by drinking context [5] ; for longer-term or chronic risks of serious illnesses, a whole range of other lifestyle, genetic and other risk factors interact with the effects of alcohol [6] . In the second place, there are substantial problems in measuring and communicating the dosage of alcohol. It is extremely difficult outside of a controlled experimental environment to determine the typical doses of alcohol people administer themselves in their daily lives [7] . In most countries which permit the sale of alcohol, there are typically thousands, even tens of thousands, of different types of alcoholic products varying in terms of taste, price, alcohol content and volume. Beers can vary in strength from 0.5% to 25% or more as just one example—and of course serve sizes can go from a sip to a barrel. Epidemiological studies of patterns of alcohol use by necessity simplify this variation using such devices as the concept of a ‘standard drink’, even though in practice there is wide divergence in actual serve sizes and understandings of this term [7] , [8] . In the third place, most epidemiological studies of alcohol, health and safety use self-report measures of consumption which typically underestimate actual consumption by large margins [7] , [9] . Even when one thinks one is on firm ground by using ‘abstainers’ as a comparison group, a close examination of this assumption reveals it also to be ephemeral: estimates of risk can shift substantially depending on whether former drinkers (whether moderate or heavy) are included in this category [10] . If there are difficulties with the quantification of exposure to risk from alcohol, assessment of the harms can be equally fraught. There are at least 60 major categories of serious health consequences and over 200 specific diagnostic codes in the International Classification of Diseases-10 (ICD-10) to which drinking contributes causally, with a wide range of different risk functions of the relation between level of consumption and the disease—the shape can be linear, U-shaped, J-shaped, exponential or other [11] . Against such odds, is it scientifically defensible to offer simple ‘ballpark’ upper limit advice to the general public? Two commentators in this issue present opposing views: Nick Heather suggests there is a moral imperative to provide consumers with advice to protect their health [12] ; Sally Casswell argues there is wide scope for their misunderstanding, weak evidence for the effectiveness of alcohol education and the worrying possibility of diverting attention from more effective public policies [13] . Many of the public submissions in the commentary period on the draft Australian guidelines in 2007 opposed specification of a low-risk level, feeling that a call for no drinking would be the best message [14] . And Rehm and Patra argue that there is substantial empirical justification for guidelines varying between different societies [15] . Anyone who has been involved in the process of developing LRDGs will agree that it can never be a simple matter of asking the question: ‘What does the science say?’ Faced with identical sets of evidence, different expert panels have come up with divergent advice. A clear example centres around interpretation of the typically J-shaped curve representing risk of death from all causes as a function of average daily alcohol consumption. Earlier Canadian guidelines [16] selected the point on the curve of lowest risk (approximately one drink per day for women, two drinks per day for men on average); the new Canadian guidelines [1] opt instead for a ‘breakeven point’ where the relative risk of all-cause mortality is no more than that of a lifetime abstainer, that is, where putative health benefits and health risks cancel each other out. But basing the guideline on the total mortality curve also builds in presumptions which are controversial. The 2009 Australian guidelines take into account only the adverse effects of drinking in terms of particular causes of death for which the causal connection of alcohol has been most firmly established, noting that the extent of health benefit from drinking is a subject of controversy, and may well have been overestimated, and adding that ‘the potential cardiovascular benefits from alcohol can also be gained from other means, such as exercise or modifying the diet’[17] . Both sides of the issue of the causal status of putative health benefits from drinking are further argued in two commentaries in this issue [18] , [19] . The new guidelines for Canada and Australia also differ in that the former take a relative risk approach, while the latter use an absolute risk approach. A relative risk considers the proportionate change in risk for an individual or social category from a given consumption of alcohol, while an absolute risk approach is concerned with the category's absolute increment in risk from the alcohol consumption, regardless of the level of risk from other factors. Since men behave more riskily than women even when not drinking, the main difference in results between the two approaches is that with relative risks the limit per occasion for men is set higher than for women. Guidelines such as the new Canadian and Australian ones are primarily based on data on mortality, the adverse consequence for which the most detailed analyses exist. It has been taken largely for granted that guidance thus based will be applicable also to the much more common adverse consequences of illness, injury and dependence. This assumption is put to the test in a detailed analysis of a major national US survey of alcohol consumption by Dawson and colleagues [20] in which 27 alternative LRDGs are compared on their ability to predict the greatest proportion of self-reported health and social consequences. In some instances, different methodologies converge in the advice they suggest; in others they differ, but usually by no more than one drink per day. So far, we have considered the issues in estimating and setting appropriate upper limits for daily consumption for otherwise healthy adults. However, LRDGs are typically intended for the whole population, and guidelines committees have typically also taken on guidance for a number of subpopulations and circumstances, for instance, concerning consumption under the legal drinking age, by young adults, by older adults, during pregnancy, when breast-feeding, with mental health problems, as part of a balanced diet, before driving and in combination with other psychoactive substances. In this issue, we focus on guidelines for the two subpopulations which have received the most attention: drinking during pregnancy [21] and drinking by youth [22] . The review of the evidence on the former by O'Leary and Bower [21] concludes that ‘there is no strong research evidence to implicate low levels of prenatal exposure with fetal harm’, but the new Canadian and Australian guidelines both nevertheless advise abstention during pregnancy. This advice is clearly based on a very different social calculus of what level of risk is acceptable in connection with pregnancy than for other life situations. In the opposite direction, in terms of the relation with empirical findings, there has been hesitation about following up the implications of analyses of differential risk among young adults. Both the Canadian and Australian guidelines make differentiations for those below the legal age, recommending lower levels than for adults. The Australian guidelines do not differentiate younger adults from older. In the full version of the Canadian guidelines [23] , adults under age 25 are given an upper daily limit of three drinks for men and two for women, one drink lower than for older adults, but this nuance is dropped in summary versions [24] . Thompson and colleagues [22] find ‘significant risks for young people at moderate consumption levels’ below the levels specified in the Canadian guidelines, and increased risk among young adults has also been found in Australia [25] . The primary focus in the papers discussed so far is on the evidence on and conclusions from the empirical relationships between alcohol consumption and harm to the drinker's health. The other side of a consideration of LRDGs is the question of their reception: how widely are they recognised and to what extent do they affect norms and behaviour? The existing literature on this side is slim indeed, and the papers here make a significant contribution. An analysis of the responses to the formal consultation process on draft guidelines which preceded the final revision of the current Australian guidelines lays bare the fundamental philosophical differences in how different individuals and agencies approach the ethical considerations around communicating risk to the whole population. At one end of the spectrum is the application of the precautionary principle that wherever there is any doubt or possibility of risk, the advice should be abstinence; at the other end of the spectrum is the view that the advice should not be oriented to evidence of level of risk so much as tailored to what it is presumed that the audience is willing to hear [14] . Several contributions in this issue consider issues in the extent of awareness and comprehension of LRDGs in different populations and countries. In Anglophone countries, LRDGs are expressed in terms of standard drinks or units (although with a different unit in each country), and the guidance will not be easily and correctly understood if the audience does not know the size of the unit. In a review, Kerr and Stockwell [7] report the clear finding that units are widely misunderstood, most often with the result of underestimation of the amount actually consumed. They also report evidence from studies showing that the addition of standard drink labels on alcohol containers helps overcome such misunderstandings. de Visser and Birch [8] demonstrate the misunderstandings experimentally and quite dramatically in samples of English teenagers and young adults. This study also shows that the UK LRDGs are not well known to those in these samples; thus, only one-third of university students and less than one-quarter of school students knew the UK guideline on the weekly limit. Similarly, Livingston [26] finds that almost two-thirds of men and one-third of women in the Australian general population give an estimate of the amount which they can safely drink on an occasion which is higher than the relevant LRDG. On the other hand, in a convenience sample of relatively affluent young Australians, Bowring et al. [27] found over 70% of the sample gave accounts of ‘safe’ amounts of drinking to avoid long-term harms or injury from alcohol which were in agreement with the 2009 guidelines, although approximately one-half reported actually drinking more than six drinks at least once a month, clearly over the limit for injury. Both these studies thus find a tendency for respondents to estimate a ‘safe’ level for short-term harms higher than the 2009 guidelines, but a ‘safe’ level for long-term harms that is within the guidelines. The actual guidelines were not well known to Bowring et al.'s samples; only 32% reported being aware of them, and there was not a strong relationship between awareness and giving ‘safe’ amounts in agreement with the guidelines. Wettlaufer and colleagues [28] fill in part of the picture of the social context of competing messages which LRDG messages face, behaviour advice to social hosts in entertainment magazines and from vendors on the appropriate dosage of alcohol to have on hand for guests at parties, luncheons and other events. The advice they garnered from these on-the-record sources was somewhat higher than the LRDGs; the paper's epigraph from The Office gives a hint of the much higher quantities which may be recommended informally. Interestingly, Livingston's analysis gives the first evidence of which we are aware that can be interpreted as an effect of LRDGs on public opinion about safe levels of drinking [26] . Between the 2001 and the 2009 Australian guidelines, there was little change in the recommended guidelines for women, while the change for men (signalled already in the 2007 draft revision put out for public comment) was to reduce their levels to those for women. Livingston finds that men, more than women, reduced their estimate of the daily drinking level which avoided long-term risks, although there was no such gender difference in change in levels for avoiding short-term risks. This evidence of an effect on risk perception is equivocal. Given the paucity of publicity about the 2009 Australian guidelines after their publication, any effect is not likely to have been through direct health promotion effects. Rather, particularly in light of the substantial public discussion when the draft revision of the guidelines was put out for comment in 2007, any effect is likely to have been through more diffuse processes of public discourse and debate. Such processes may turn out, more generally, to be the main justification for the existence and periodic review of LRDGs. Although LRDGs are framed in a discourse of individual risk and governance, their main point may be as a contribution to the public debate on what are appropriate collective norms on drinking and levels of consumption in the society. Evidence on effects of guidelines on actual drinking behaviour remains scanty. But a recent Australian study [29] measured the drinking behaviour of Australian pregnant women in 2000, when the official guideline was for abstinence during pregnancy, and in 2003 and 2006, when the guideline was to have ‘less than 7 standard drinks [in a week], and, on any one day, no more than 2 standard drinks’. Between the two times, the proportion drinking moderately or more heavily dropped a little (25% to 20%; χ2(1) = 5.22, P = 0.022). But there was a non-significant trend for a smaller proportion, among those drinking a low or zero amount, to be drinking zero in the second period (27% vs. 23%; χ2(1) = 2.21, P = 0.137). Another study also found a reduction in abstinence between cohorts of mothers of babies born before and after the 2001 guidelines change [30] (p. 129). There may thus have been some change in behaviour in the direction of change in the guideline, although this might also be interpreted as respondents being more prepared to admit to light drinking during pregnancy rather than an actual behaviour change. In any case, the results of a study by Thomson and colleagues [31] suggest that LRDGs have a bright future. In their study of public opinion on health information and warning labels, support was nearly unanimous among Australian adults for ‘recommended guidelines for low-risk alcohol consumption’ and for warning labels keyed to the guidelines. The appetite for communicating such advice continues to grow, and with increasing global concern of the considerable public health burden from hazardous consumption [32] , there is likely to be pressure eventually for some kind of international consensus. In Canada until late in 2011, there had not even been a national consensus on low-risk levels [28] . We hope the interesting and lively contributions in this special issue will contribute to a better understanding of the issues raised by LRDGs, and to moving towards an evidence-based understanding of their functions and effects. Stockwell's work on this issue was supported by the Endowment Fund of the Centre for Addictions Research of British Columbia at the University of Victoria, British Columbia, Canada. Room's work was supported by the Victorian Department of Health and the Foundation for Alcohol Research and Education, Canberra.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.