Critique 240: Are there health risks from amounts of alcohol referred to as being within “low-risk drinking” guidelines? — 8 July 2020

Sherk A, Thomas G, Churchill S, Stockwell T.  Does Drinking Within Low-Risk Guidelines Prevent Harm? Implications for High-Income Countries Using the International Model of Alcohol Harms and Policies.  J Stud Alcohol Drugs 2020;81:352–361.

 Authors’ Abstract

Objective: Many countries propose low-risk drinking guidelines (LRDGs) to mitigate alcohol-related harms. North American LRDGs are high by international standards. We applied the International Model of Alcohol Harms and Policies (InterMAHP) to quantify the alcohol-caused harms experienced by those drinking within and above these guidelines. We customized a recent Global Burden of Disease (GBD) analysis to inform guidelines in high-income countries.

Method: Record-level death and hospital stay data for Canada were accessed. Alcohol exposure data were from the Canadian Substance Use Exposure Database. InterMAHP was used to estimate alcohol-attributable deaths and hospital stays experienced by people drinking within LRDGs, people drinking above LRDGs, and former drinkers. GBD relative risk functions were acquired and weighted by the distribution of Canadian mortality.

Results: More men (18%) than women (7%) drank above weekly guidelines. Adherence to guidelines did not eliminate alcohol-caused harm: those drinking within guidelines nonetheless experienced 140 more deaths and 3,663 more hospital stays than if they had chosen to abstain from alcohol. A weighted relative risk analysis found that, for both women and men, the risk was lowest at a consumption level of 10 g per day. For all levels of consumption, men were found to experience a higher weighted relative risk than women.

Conclusions: Drinkers following weekly LRDGs are not insulated from harm.  Greater than 50% of alcohol-caused cancer deaths are experienced by those drinking within weekly limits. Findings suggest that guidelines of around one drink per day may be appropriate for high-income countries.

Forum Comments

 This study uses global data from different cultures in an attempt to judge whether there are health risks for subjects who are consuming alcohol considered to be within “low-risk drinking guidelines”.  Unfortunately, their arguments and conclusions are weakened from their using an unusual method for estimating alcohol intake and very little information on the pattern of drinking (regular and moderate versus only on one or two days/week, with or without food, type of beverage, etc.).  Further, these analyses utilize a combination of data from a variety of high-income countries, many of which have very different drinking cultures; this gives values of risk of certain health outcomes that actually apply to no single culture (as described in previous Forum critiques 199, 206, and 219).

Approach for estimating “under-reporting” of alcohol intake:  The authors make a lot of assumptions in their analyses, including a questionable way of adjusting for under-reporting of alcohol.  To “adjust” for differences between sales figures for alcohol and the reported amounts consumed by subjects in epidemiologic studies, they increase the reported amounts of alcohol consumed for all individuals in a country (by 80% in the present analyses).  Reviewer Ellison noted: “It has been shown that when such adjustments are based on individual characteristics, subjects who may be classified as ‘likely under-reporters’ and those classified as ‘likely not under-reporters’ have many differences in health characteristics and health outcomes (Klatsky et al, 2006; Klatsky et al, 2014).  The method of the authors of the present paper, increasing the reported intake of alcohol by 80% for all subjects to adjust for under-reporting, yields data that cannot be interpreted in terms of judging the health effects of a given amount of alcohol for evaluating ‘low-risk drinking’.”

Reviewer Skovenborg agreed: “The questionable theoretical rationale for adjustment by 80% of local recorded plus unrecorded per capita alcohol consumption is inadequate.  As stated, mixing data from likely and unlikely under-reporters, using a blanket adjustment of alcohol intake for all participants, would be expected to give flawed results.”

Combining former heavy drinkers with current moderate drinkers:  In almost all previous studies of western countries, former drinkers tend to be mainly former heavy drinkers, and are at higher risk for most conditions.  For many analyses, the authors combined their data from former drinkers (who were apparently former heavy drinkers) with data from drinkers consuming “within guidelines”.  The focus of their arguments tends to be based on combining these two groups when stating that “More than 50% of alcohol-attributable cancer deaths were experienced by the population that is currently drinking within weekly guidelines, when including former drinkers”.

According to the data presented in the paper, the former drinkers in this analysis had the highest risks of the typically “alcohol-related” cancers, higher than even the group classified as current “above-guidelines” group.  For an outcome that is usually considered to be fully attributable to alcohol, liver cancer, their data show among women that 29 of the instances were in the “former drinker” category, and only 3 in the within guidelines and 3 in the above guidelines current drinker groups.  For men, of 64 deaths from liver cancer, 23 were former drinkers, 11 were drinkers in the within guidelines group, and 30 in the above guidelines group.  Hence, the within guidelines drinkers made up only a small percentage (14/99) of the liver cancer subjects.  It thus suggests that results in former drinkers should have been combined with those of the above-guidelines drinkers; in doing so, the conclusions of the authors suggesting increases in risk among the lighter drinkers would have been quite different . . . much, much lower.

Reviewer Ellison commented on this problem.  “Overall, the authors had data from 3,942,691 subjects, 2,556,729 of whom were listed as drinking within-guidelines; among these, there were a total of 140 alcohol-attributable deaths (0.005%).  There were 488,438 subjects who consumed above the weekly guidelines, associated with 1,283 alcohol-attributable deaths (0.263%).  There were 351,091 reportedly lifetime abstainers and 546,433 former drinkers; among the latter there were 630 alcohol-attributable deaths (0.115%).  Hence, the death rate for the within-guidelines subjects was many times lower (about 50-fold lower!) than for those consuming above-guidelines, and even in comparison with former drinkers it was more than 20 times lower.  Yet, throughout the text, the authors tend to combine data from former drinkers and within-guidelines drinkers and state that for hospitalizations and deaths, events were greater than those in the above-guidelines group.

“The authors also focus on cancer deaths, where they state specifically: ‘More than 50% of alcohol-attributable cancer deaths were experienced by the population that is currently drinking within weekly guidelines, when including former drinkers.’  The actual numbers of alcohol-attributable cancer deaths among the within-guidelines group is 123 of 2,556,729 (0.005%), while it is 206 of 546,443 former drinkers (0.038%) and 312 of 488,438 above-guidelines subjects (0.064%).  (The authors tend to report numbers of attributable cases but not occurrence rates, even though the denominators for these groups are vastly different.)  Further, combining data from former drinkers and current moderate drinkers is absurd, yet is used as the most-touted result of this study.”

Forum member Stockley agreed with other reviewers, adding that “Our Forum has previously reviewed the GBD paper from 2018 pointing out that combining data from a large number of diverse cultures to determine a single association between alcohol consumption and health creates real problems.  Many lifestyle and cultural factors strongly modify the relation of a given amount of alcohol to health and disease.  When such modifying factors are not taken into consideration, the estimated intake of a given amount of alcohol (even if accurate), provides an incomplete assessment of the effects of alcohol on health, even among high-income countries.”

Alcohol effects relate to more than just estimated number of drinks per week:  Reviewer de Gaetano was concerned that these authors considered alcohol as a drug, a separate entity, and did not evaluate drinking in the context of life habits, including smoking on one side and nutrition on the other.  And, even more importantly, they do not include details on the pattern of drinking (e.g., with or without food, type of beverage), or socio-economic factors (which markedly modify the risk of adverse effects of a given amount of alcohol).  Further, using high-income country mortality weights instead of global weights, the three RR functions were shown to present a J-shaped curve at low daily levels of consumption (they state for less than 12 g/day for men and less than 17 g/day for women).  We note again, however, that these estimates were based on the majority of meta-analyses that indicate a degree of cardioprotection associated with low-level alcohol use, an observation that continues to be contested by some, such as Chikritzhs et al; Ortolá et al; Zhao et al (the highly selected references included in this paper).”

Forum member de Gaetano added: “After presenting their results based on global data, showing only adverse health effects of alcohol, the authors then show results based on high-income countries, more appropriate for Canada.  The results are quite different, and show, at least for the distribution of the burden of disease in Canada — which may be broadly representative of other high-income countries — that there may exist a safe, and even slightly (statistically significant, anyway) health-protective level of alcohol intake, corresponding with low levels of daily consumption.”

Unusual differences between men and women:  Extensive studies have demonstrated that, for a given amount of alcohol, women tend to exhibit more adverse effects than men.  However, this study finds, for example, much larger protective effects against ischemic heart disease and ischemic stroke among women than among men.  They conclude that “If national drinking guidelines are based on aligning risks between drinkers and abstainers, our study suggests limits of approximately 12 g/day for men and 17 g/day for women.”  Such results are confusing and inconsistent with most previous research, and may relate to the many assumptions made by the authors in the present analyses.

Presentation of results:  The figure in the paper showing the relative risk of all “alcohol-attributable causes” for alcohol intake (in grams of ethanol/day) shows result from 0 to about 150 g/day of intake.  The curve extends well beyond usual intakes in the population (the actual reported consumption as well as the “adjusted” levels of intake are not stated in this paper); presumably, much of this curve is based on extrapolation rather than actual data.  The authors do include an expanded version of the curve extending only up to 30 grams/day (an upper level that includes the vast majority of values of intake in most studies).  In the expanded version, there is a strong suggestion of a “J-shaped curve” (similar to what is almost always seen in epidemiologic studies).  Because of the many assumptions made by the authors (regarding adjustments to reported intake, combination of data from many cultures, etc.), the data do not permit a valid estimation of the level of alcohol intake that exceeds the calculated risk of never drinkers and that would be of value when setting limits for “low-risk” drinking.

Other studies estimating attributable deaths due to alcohol:  The authors cite a small number of studies to support their thesis and criticize most previous research.  Reviewer Skovenborg wrote: “The authors brought up the tired old horse of abstainer misclassification and residual confounding used in the meta-analysis of Zhao et al to discredit the reduction of coronary heart disease associated with light to moderate alcohol consumption in men.  However, as shown many decades ago by Boffetto et al and in almost all prospective studies since, when adjusting for former drinking and all potential confounders, the vast majority of studies show a significant reduction in the risk of cardiovascular disease, especially ischemic heart disease, to be associated with light-to-moderate drinking.  Among such studies is one from Denmark (Eliasen et al) of alcohol-attributable deaths; those authors found that 73 and 81% of deaths for women and men, respectively, occurred within the high alcohol consumption group (defined as > 14 drinks/week for women or > 21 drinks/week for men).  Furthermore, they reported that for heavy drinkers, total compliance with sensible drinking guidelines with a low risk limit of <7 drinks/week for women and 14 drinks/week for men would dramatically lower alcohol-attributable deaths.”

Deaths from cancer attributed to alcohol consumption:  The authors calculated that 32% of cancer deaths in women and 15% of cancer deaths in men were attributable to subjects stating that they consumed alcohol within the drinking guidelines.  However, these percentages are not based on the risk of disease within each group as the numbers of subjects in each group (the denominators) differed markedly.  For example, of the 53 cases of pancreatic cancer reported in this paper, 35 were in former drinkers, 12 in above-guideline drinkers, and 6 in within-guidelines drinkers.  However, the risk ratios for the subjects in these groups were not given: there were a total of 546,443 former drinkers and 488,438 subjects above-guideline drinkers, compared with 2,556,729 within-guidelines drinkers.  So, unlike what the authors insinuate, the rates of such cancers is extremely low in the true within-guidelines drinkers.

Skovenborg reports on a key paper by Tramacere et al on pancreatic cancer: “The results of a meta-analysis of the dose-risk relation of alcohol drinking and pancreatic cancer risk provides ‘strong evidence for the absence of a role of moderate drinking in pancreatic carcinogenesis, coupled to an increased risk for heavy alcohol drinking.  Given the moderate increase in risk and the low prevalence of heavy drinkers in most populations, alcohol appears to be responsible for only a small fraction of all pancreatic cancers’ (Tramacere et al).  In a recent Lancet review by Mizrahi et al, risk factors for developing pancreatic cancer included ‘family history, obesity, type 2 diabetes and tobacco use’.”

Potential use of the present study for setting drinking guidelines:  Forum members considered that the potential application of the results of this study for setting guidelines is quite limited, and do not agree with the authors that these data could support a lowering of alcohol guidelines.  Stated Forum member Finkel: “We have here a meta-study interpreted by possibly prejudiced authors with a very limited selection of references, based on very indirect data (both the quantitative drinking prevalence and the attributions of death and disease to alcohol), and inferences just as weakly based.  One could argue endlessly over the accuracy of ‘attributable’ in many cases.  And without detailed data on such vital matters as type of beverage and drinking pattern, I can’t begin to make sense, much less draw conclusions about and execute plans based upon this paper.”

Reviewer Van Velden noted: “This is a very emotive article that singles alcohol out as a risk factor without taking into consideration socio-economic-, cultural-, political-, and educational-factors.  Alcohol consumption should not be seen in isolation from other health-affecting factors such as smoking, exercise, socioeconomic status, balanced nutrition, and effective stress management techniques.  We do not talk here about a health elixir, but about a responsible and balanced lifestyle where alcohol consumption with meals is the norm.  Refined carbohydrates (sugar) consumption contribute much more than alcohol consumption to the epidemic of the chronic and degenerative lifestyle-related diseases of the 20th century.  At least the sugar in grape juice is fermented to alcohol in wine . . . and wine does not cause insulin resistance.”

References from Forum review:

Boffetta P, Garfinkel L. Alcohol drinking and mortality among men enrolled in an American Cancer Society prospective study. Epidemiology 1990;1:342-348.

Chikritzhs T, Fillmore K, Stockwell T.  A healthy dose of scepticism: Four good reasons to think again about protective effects of alcohol on coronary heart disease. Drug and Alcohol Review 2009;28, 441–444.  doi:10.1111/j.1465-3362.2009.00052.x

Eliasen M, Becker U, Grønbæk M, Juel K, Tolstrup JS.  Alcohol-attributable and alcohol-preventable mortality in Denmark: an analysis of which intake levels contribute most to alcohol’s harmful and beneficial effects.  Eur J Epidemiol 2014;29:15-26.

Forum critique 199.  Potential mechanisms for the greater risk and fewer health advantages from alcohol consumption for subjects with low socio-economic status – 25 May 2017.

Forum critique 206.  Importance of considering cultural factors in determining effects on health of alcohol consumption – 3 October 2017.

Forum critique 219.  A global overview of alcohol consumption and health – 3 September 2018.

GBD 2016 Alcohol Collaborators. Alcohol use and burden for 195 countries and territories, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016.  Published online August 23, 2018.

Klatsky AL, Gunderson EP, Kipp H, Udaltsova N, Friedman GD.  Higher Prevalence of Systemic Hypertension Among Moderate Alcohol Drinkers: An Exploration of the Role of Underreporting. J Stud Alcohol 2006;67:421-428.

Klatsky AL, Udaltsova N, Yan L, Gaer D, Tran HN, Friedman GD.  Moderate alcohol intake and cancer: the role of underreporting.  Cancer Causes Control 2014;25:693–699

Mizrahi JD, Surana R, Valle JW, Shroff RT.  Pancreatic cancer. Lancet 2020;395:2008-2020.

Ortolá R, García-Esquinas E, López-García E, León-Muñoz LM, Banegas JR, Rodríguez-Artalejo F. (2019). Alcohol consumption and all-cause mortality in older adults in Spain: An analysis accounting for the main methodological issues. Addiction 2009;114,59–68.  doi:10.1111/add.14402

Tramacere I, Scotti L, Jenab M, Bagnardi V, Bellocco R, Rota M, Corrao G, Bravi F, Boffetta P, La Vecchia C.  Alcohol drinking and pancreatic cancer risk: a meta-analysis of the dose-risk relation.  Int J Cancer 2010;126:1474-1486.

Zhao J, Stockwell T, Roemer A, Naimi T, Chikritzhs T.  Alcohol consumption and mortality from coronary heart disease: An updated meta-analysis of cohort studies.  Journal of Studies on Alcohol and Drugs 2017;78:375–386.  doi:10.15288/jsad.2017.78.375

Forum Summary

The authors of this paper use very questionable data to suggest that drinking guidelines for all countries should be lowered to no more than 1 drink/day.  In their analyses, they make a very large number of assumptions, including adding 80% to reported amounts of alcohol for all subjects to make up for alcohol disappearance data, and do not modify this according to their individual likelihood of being an under-reporter of intake.  Further, they are not able to adjust for the pattern of drinking or type of beverage consumed, and it is considered bizarre that they grouped ex-drinkers (many of whom were apparently heavy drinkers) with current drinkers who report consuming within the guidelines to state that such subjects have a high risk of adverse effects.  From the numbers presented in the tables in this paper, the risk of alcohol-attributable death for consumers of alcohol within-guidelines was fifty times lower than for those consuming above the guidelines.  Such deficiencies in methodology, and especially possibly deliberate combinations of groups to support their premise, make it impossible to utilize their conclusions for setting guidelines for “low-risk” drinking.

Further, their conclusions do not reflect what the vast majority of well-done epidemiologic studies have shown for decades that light-to-moderate alcohol consumption (when defined as up to 1 drink/day for women and up to 2 drinks/day for men and not consumed in binges) is associated with a significant decrease in the risk of cardiovascular mortality as well as total mortality.  The results of the current paper are not sufficient to warrant changing drinking guidelines.

*                  *                    *                      *                      *                      *

Contributions to this critique by the International Scientific Forum on Alcohol Research were provided by the following members:

Erik Skovenborg, MD, specialized in family medicine, member of the Scandinavian Medical Alcohol Board, Aarhus, Denmark

Pierre-Louis Teissedre, PhD, Faculty of Oenology–ISVV, University Victor Segalen Bordeaux 2, Bordeaux, France

R. Curtis Ellison, MD, Professor of Medicine, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA

Giovanni de Gaetano, MD, PhD, Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy

David van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa

Harvey Finkel, MD, Hematology/Oncology, Retired (Formerly, Clinical Professor of Medicine, Boston University Medical Center, Boston, MA, USA)

Creina Stockley, PhD, MSc Clinical Pharmacology, MBA; Adjunct Senior Lecturer at the University of Adelaide, Australia