Critique #288 – Draft Report: Scientific Findings of the Alcohol Intake & Health Study for Public Comment

Authors

Shield, K., Keyes, K.M., Martinez, P., Milam, A.J., Naimi, T.S., Rehm, J.

Citation

https://www.stopalcoholabuse.gov/media/pdf/Report-on-Alcohol-Intake-and-Health.pdf

Author’s Abstract

  • The Alcohol Intake and Health Study assessed relationships between alcohol use and health in the U.S., with a focus on morbidity and mortality (i.e., deaths and reduced life expectancy) from health conditions causally related to alcohol use.
  • Most of the evidence evaluated in this report is from systematic reviews and meta-analyses of “observational” cohort studies. It did not include data from quasi-experimental (e.g., Mendelian randomization) studies or randomized controlled trials.
  • We assessed relationships between different levels of alcohol consumption and the risk of dying from health conditions that are causally related to alcohol consumption in the U.S. We also modelled the total alcohol-specific mortality at different levels of consumption based on all alcohol-related conditions.
  • Among the U.S. population, the risk of dying from alcohol use begins at low levels of average use. Higher levels of alcohol consumption are linked with progressively higher mortality risk. Depending on the level of use, men are at a similar risk of health harms from alcohol use compared to women.
  • In the United States, males and females have a 1 in 1000 risk of dying from alcohol use if they consume more than 7 drinks per week. This risk increases to 1 in 100 if they consume more than 9 drinks per week.
  • Males and females who consumed 1 drink per day had an increased risk of liver cirrhosis, oesophageal cancer, oral cancer, and injuries, but a lower risk for ischemic stroke. In addition, females had a higher risk for liver cancer and a lower risk for diabetes mellitus when they drank 1 drink per day. However, drinking patterns shape risk. Specifically, even infrequent high per-occasion drinking may eliminate the lower levels of risk for ischemic stroke.
  • Alcohol use is associated with increased mortality for seven types of cancer (colorectal, female breast, liver, oral cavity, pharynx, larynx, oesophagus [squamous cell type]). Increased risk for these cancers begins with any alcohol use and increases with higher levels of use. Women experience a much greater risk of an alcohol-attributable cancer per drink consumed.
  • ‘Per occasion’ alcohol use refers to how much alcohol is consumed within a short time period, as opposed to how much is consumed on average. Higher levels of per occasion use result in higher blood alcohol levels and higher risks for injuries. Risks increase starting at one drink per occasion and are particularly pronounced for women consuming more than three drinks and men consuming more than four drinks per occasion.
  • For individuals who start consuming alcohol at age 15, the risk of an alcohol-attributable death between the ages of 15 and 20 varies by consumption level. The risk of an alcohol-attributable death increases linearly with alcohol consumption. For males, the risk of an alcohol-attributable death ranges from 0.07 (95% CI: 0.05, 0.08) per 1000 for those who consume 1 drink per week, to 1.76 (95% CI: 1.42, 2.28) per 1000 for those who consume 3 drinks per day (i.e., 21 drinks per week). Similarly, for females, the risk increases from 0.03 (95% CI: 0.02, 0.04) per 1000 for those who consume 1 drink per week, to 0.75 (95% CI: 0.57, 1.06) per 1000 for those who consume 3 drinks per day (i.e., 21 drinks per week). Most of these alcohol-attributable deaths are caused by road traffic crashes, unintentional injuries, and intentional injuries. These deaths represent a substantial proportion of all deaths for individuals 15 to 20 years of age.

Forum Summary

ICCPUD initiated analyses published as a draft report entitled: “Scientific Findings of the Alcohol Intake & Health Study for Public Comment”, to suggest that the recommendations for alcohol consumption should be lowered compared to the current Drinking Guidelines for Americans 2020-2025.

A main criticism of the Draft Report is that it catalogues all possible negative risks associated with alcohol consumption. It then models these negative risks in a non-transparent and methodologically criticized way without comparing their model-generated overall outcomes with real observations on overall mortality. 

Further, with the reason to use lifetime abstainers as the reference group to limit abstainer bias, the authors ignore most of the overwhelming data showing risk reductions for several diseases associated with light to moderate alcohol consumption. This is in contrast to research showing that former drinker bias is not a main driving factor in the alcohol–disease associations.

Another major issue in this Draft Report is a lack of common scientific checks for internal and external consistency, simply refuting most high-quality studies that disagree with the authors’ opinions.

Forum comments

Background to the Draft Report

The Dietary Guidelines for Americans 2020-2025 (DGA) are being updated this year. Certain U.S. institutions have pre-emptively initiated analyses to suggest that the recommendations for alcohol consumption should be adjusted and, more specifically, lowered compared to the current guidelines, which say that women should drink no more than seven drinks per week and men should drink no more than 14 drinks per week.

One of those initiatives is the “Scientific Findings of the Alcohol Intake & Health Study for Public Comment” (Draft Report), published by the Interagency Coordinating Committee on the Prevention of Underaged Drinking (ICCPUD), which is a US Department of Health and Human Services Committee (USHHSC) (Shield, K., Keyes, K.M., Martinez, P., Milam, A.J., Naimi, T.S., Rehm, J.R. 2025).

Although the publication concerns alcohol consumption and health in the USA, two of the Draft Report authors (Naimi, T.S., Rehm, J.R.) are affiliated with the Center of Addiction and Mental Health in Canada and the Canadian Centre on Substance Use and Addiction. They were both Scientific Expert Panel members of Canada’s Guidance on Alcohol and Health (2023), evaluated in ISFAR critique #261[1]. In addition, at least two authors have received funding from Movendi International – a recognised temperance organisation. They promote that “the lifestyle of the 21st century is about living free from alcohol and other drugs[2]”.[3]

The Draft Report also follows a similar approach as Canada’s Guidance on Alcohol and Health (2023) in that all potential risks of alcohol consumption are considered and taken into account, while the potential benefits for health are underrepresented. The main criticism of ISFAR members in critique #261, however, was the important omission of not considering the association of alcohol consumption with overall mortality. Rather they based their advice on a model using evidence of low to very low quality. Furthermore, one of the Draft Report authors (Niami, T.S.) co-authored the meta-analysis review solely selected to evaluate cardiovascular disease (CVD) risk and alcohol.

Critique of the Draft Report

  1. General methodology criticisms

1.1 The Draft Report primarily relies on observational data to assess the relationship between alcohol and health outcomes.

A main criticism of the Draft Report is that it catalogues all possible negative risks associated with alcohol consumption. It then models these negative risks in a complex and non-transparent way without comparing their model-generate overall outcomes with real observations on overall mortality.  The main biases in observational studies are selection bias, information bias and measurement errors, confounding and Simpson’s paradox. All the biases compromise validity, that is, selection bias compromises external validity while confounding compromises both internal and external validity.

1.2 Reference group

The authors state in the methods section that lifetime abstainers were selected as the reference group to limit abstainer bias, which can arise when studies combine lifetime abstainers with former drinkers who may have quit after becoming sick or experiencing some health effects related to their alcohol use. This way of selecting data may seem logical, but previous research has shown that former drinker bias is not a main driving factor in the alcohol–disease associations (Klatsky & Udaltsova, 2013). This choice, however, allows the authors to ignore most of the overwhelming data showing risk reductions for several diseases associated with light to moderate alcohol consumption. Indeed, even when Wood et al., (2018) in their extensive combined analysis in The Lancet eliminated non-drinkers from their analysis, their data still supported those individuals consuming more often than twice per week or engaging in binge drinking, consumption of up to 200 g alcohol/week does not seem to increase mortality compared to non-drinkers.

Furthermore, as recognised by Bondy and Rehm in 1998, “interpreting risk differences for current drinkers compared with lifetime abstainers may be difficult, because lifetime abstainers in many societies differ from the general population in several other ways, such as diet, religion, or socioeconomic status”.

1.3 Selection bias

Selection bias seems probable when looking at Figure 1 and Table 2 on the selection process, which is the basis for their report. The authors found 13,998 potentially relevant reports in unspecified databases (and one additional report added by the experts, it is unclear which report this is and why it was added). Of these 13,998 records, 6,705 duplicates were removed. Then another 7003 were excluded without providing insight into the reason for such exclusion. There is a double asterisk to this statement, but the text belonging to the double asterisk is not shown. The text states: “based on title and abstract screening, the Annex indicates 2 criteria for exclusion, namely beverage-specific studies and studies focusing on patterns of alcohol use”. Being aware of the much of the literature on alcohol and health it seems highly unlikely that 7003 papers needed to be excluded based on these two criteria, simply because very few studies focused on beverage-specific effects or drinking patterns only.

This left 291 reports for retrieval and full-text assessment. It is unclear what retrieval means in this context, but all reports were retrieved. Of these 291 reports another 235 were excluded for various specified reasons leaving 58 studies to be included.

1.4 Lack of checking for internal and external consistency

Another major issue in this Draft Report is a lack of the common scientific checks for internal and external consistency. Internal consistency means checking whether the results obtained correspond with what was found with other parameters and outcomes evaluated in the same study, while external consistency means checking whether the results obtained correspond to what is found in other studies published in the area or other related areas. This is to ensure that the findings align with what is already established while also presenting new insights that are not contradicted by previous research; this involves examining whether the data within the study itself is consistent (internal) and if it aligns with broader scientific understanding (external). Key points to consider on internal consistency include: a literature review – thoroughly reviewing existing research on the topic to ensure the study aligns with established knowledge and does not contradict major findings; generalizability – considering whether the study sample and conditions are representative enough to apply the results to a wider population or context; and comparison with other studies – comparing findings to similar studies conducted by other researchers. The authors of the Draft Report, unfortunately, do not explain why their results show that alcohol consumption at the lowest levels is associated with considerable risk. The vast majority of other authors, however, find a J-shaped association between light and moderate alcohol consumption and all-cause mortality.

Similar to the J-shaped association between alcohol consumption and coronary heart disease, a J-shaped association between alcohol consumption and overall mortality has been established in many studies over the past five decades. A recent study again confirmed such an association (Bryazka et al., 2022), while a similar conclusion was reached by one of the Draft Report authors (Rehm, J.R.) in yet another paper. In that meta-analysis of the alcohol consumption and all-cause mortality association, it was concluded that: “The overall beneficial effect of light to moderate drinking remained under all scenarios, indicating a high validity of the overall shape despite the heterogeneity between studies.” (Gmel et al., 2003). In the discussion section of the Draft Report, the authors indicate, however, that the use of all-cause mortality studies to examine population estimates about alcohol and health is problematic since the cause-and-effect relation between alcohol consumption and health outcomes is not established. They used the epidemiological definitions of causality, where alcohol had to be necessary, either alone or in combination with other antecedent conditions as a component cause (Rothman et al. 2008), rather than diseases and injuries fully attributed to alcohol[4]. Interestingly, one author of the Draft Report (Rehm, J.R.), was a coauthor of at least two publications related to alcohol and infectious diseases informing the WHO (2017). Correlation implies associations but not causality and, conversely, causation implies association but not correlation (Altman and Krywinski 2015). As the cause-and-effect relation for alcohol has not been definitively established by the WHO for all the listed diseases and injuries used for this modelling study either, this argument is invalid[5].

1.5 Generalising observational studies to broader populations

The Draft Report generalises or extrapolates observational studies to model harm across the broader US population. Successful extrapolation hinges on similarities and differences between populations at different levels. No differences present at any of these levels such as demographics are taken into account (Khosrow, 2019) and any errors are compounded because the results of observational studies risk containing confounding biases. For example, the Draft Report states in the Discussion section, that cohort studies predominantly “consisted of middle-class, middle-aged participants from similar cultural backgrounds”. As correctly stated in this Discussion section, “the RR derived from these studies may have limited generalizability to other groups [in the US population] defined by factors such as sex, age, socio-economic status, and other modifiers (112, 113)”.

No study design or statistical method can account for confounders and bias as can randomised clinical trials. They are considered the ‘gold standard’ for producing reliable evidence as they ensure that potential confounding variables are evenly distributed between groups[6]. According to the widely-accepted hierarchy of evidence, the most reliable evidence comes from systematic reviews, followed by evidence from randomized controlled trials, cohort studies and then case-control studies. The greatest value of observational studies is that they provide preliminary evidence that can be used as the basis for hypotheses in randomised clinical trials.

1.6 Validity of correlation of burden of disease studies and low-risk drinking guidelines

A statistical modelling project, such as that described in the Draft Report, provides population-level estimates of risk attributed to alcohol consumption. It is useful for exploring cause-specific indicators and trends over time or across countries. The Global Burden of Disease (GBD) study is another example of this approach used to inform our understanding of country, regional, or global-level attributable risk and burden of disease.

In 2015, Rehm et al. pondered “how alcohol should figure in burden-of-disease studies, because for good reason these focus on aggregating risks across specific causes of death and disability. The same average drinking level will have markedly different effects in different countries based on their different composition of causes of death, often markedly different from the composition in the cohort studies.” This is ignored in the Draft Report.

In general, however, modelling the burden of disease based on a population-risk approach may not be relevant for each individual. Each individual has their own risk profile depending on their characteristics, circumstances and lifestyle. As such, the modelling approach leads to very general information that may or may not be relevant for all individuals. Also, the approach chosen by ICCPUD focuses on and emphasizes all negative outcomes that may be associated with alcohol in a causal or noncausal relationship. Consequently, this general information may make individuals better aware of the possible risks but has little relevance as a basis for guidelines or legislation. More so, since the outcomes of this Draft Report do not correspond to observations on overall outcomes such as all-cause mortality.

1.7 Correction for presumed under-reporting of alcohol consumption

The Draft Report challenges the accuracy of information about alcohol exposure, since a lot of earlier evidence is based on potentially inaccurate self-reporting of alcohol consumption (Gilligan et al. 2019). Accordingly, the authors themselves selected relative risk (RR) estimates for the modelling based on publication volume, which potentially biases toward specific perspectives or their own papers. This method of selecting RR is not transparent given the pre-registered systematic scoping review used by the authors was not publicly available, nor is the rationale for their choice of RR. Further, this method is not scientifically validated.

While it is true that in observational studies epidemiologists have to rely on their human participants to tell them what they actually drink, there is always the chance that there will be mistakes in their reporting, especially under-reporting of alcohol. Consequently, self-reported alcohol consumption data are prone to bias and are challenging to harmonise across studies conducted over different time periods that used varying instruments and methods to record such data.  In The Svalbard Study, for example, which was a unique setting for validation of self-reported alcohol consumption, the self-reported volume accounted for approximately 40% of the sales volume (Høyer et al. 1995), where coverage of sales estimates by surveys varied between 39% for Germany and 56% for France.

Underestimation of alcohol consumption in observational studies, however, is less than in typical population surveys (Boniface et al. 2014, Stockwell et al. 2018), where increasingly epidemiologists have become better able to identify and adjust for such potential bias, and are increasingly able to recognize under-reporting of alcohol (Klatsky et al. 2006, Klatsky and Udlaltsova 2007, Klatsky et al. 2014).  The Kaiser Permanente Study, for example, found the underreporting of alcohol consumption actually partially explained the increased prevalence of hypertension among individuals reporting one to two drinks per day (Klatsky et al. 2006). It also partially explained the apparent increased risk of cancer among light-moderate drinkers (Klatsky et al., 2014). In addition, it also partially explained the apparent magnitude of the benefit of lighter drinking (Klatsky and Udaltsova 2007). Furthermore, in an observational study from Italy of men aged 45-64 years who were followed for total mortality from 1965 to 1995, men reporting drinking approximately five drinks/day had a longer life expectancy than occasional drinkers and heavy drinkers. Underreporting was not an issue in this population with almost no non-drinkers and total acceptance of drinking wine with meals (Farchi et al. 2000).

However, in addition to average volume of alcohol consumption in determining the risk relationships between alcohol consumption and alcohol-related diseases and injuries, Rehm et al. (2003) found that pattern of drinking was an additional influencing factor for alcohol-related diseases such as coronary heart disease. Consequently, the influence of self-reported patterns of drinking rather than self-reported volume in alcohol exposure may be underestimated because pattern measures have not been consistently included in epidemiologic studies (Martin et al. 2014).

Consequently, while underreporting and its correction are important, over-estimation is also important, but neither can be accurately corrected using the method described in the Draft Report. Indeed, as shown by Gilligan et al. (2019) who support previous findings of the inaccuracy of alcohol consumption volume in surveys but also demonstrate that an assumption of underestimation cannot be applied to all individual reports of consumption.

2 Criticisms specific to the relationship between alcohol and CVD

2.1 Of the 58 selected studies by the authors, 19 concerned cardiovascular diseases (CVDs), the main cause of death in the USA. In Table 2, five CVDs were mentioned including ischaemic heart disease (IHD), the leading cardiovascular disease; for IHD four systematic reviews were selected. However, of these four systematic reviews only the report by Zhao et al. (2017) was selected for inclusion in their modelling without motivation.

Equally for other diseases evaluated, not all meta-analyses initially selected were used. For every disease outcome, one study was selected, which narrows down the basis for their evaluation from 58 reviews in Figure 2 to 17 reports for 17 disease outcomes. Of these 17, some reports were used for multiple disease outcomes leaving only 15 reports underlying the author’s analysis.

2.2 The analysis by Zhao et al. (2017) builds on, or updates, that of Fillmore et al. (2006) and Stockwell et al. (2016). The methodology and hence findings of Fillmore et al. (2006) were discredited by reanalysis of the studies, which showed it to be biased, as repeat studies adjusting for the claimed bias still showed a cardioprotective effect for regular light to moderate alcohol consumption. This is described in a complete supplement of the Annals of Epidemiology in 2007 (Ellison et a. 2007). Also, repeat studies adjusting for the claimed bias still showed a cardioprotective effect for regular light to moderate alcohol consumption (Di Castelnuovo et al., 2006, Ronksley et al., 2011).

Similarly, ISFAR has previously criticised certain researchers at the Canadian Institute for Substance Use Research for biased meta-analysis. For example, Stockwell et al. (2016) ‘cherrypicked’ a small number of studies for their meta-analysis – they discarded 2,575 studies and analysed only 87. The studies that they analysed, related reported intake to disease but they carefully avoided hundreds of validated studies that showed reduced disease among moderate drinkers. Inappropriately, the Draft Report continues to rely on discredited studies using demonstrably flawed techniques.

Indeed, Zhao et al. (2017) failed to address these criticisms. For example, the authors include in their analyses numerous older epidemiologic studies and do not acknowledge that when the “errors” that they have commented on in the past (such as including heavy ex-drinkers in the no-alcohol referent group) have been dealt with in the majority of studies over the past decade.  The authors still excluded the vast majority of these well-done studies in their new meta-analysis. Results of essentially all studies that adjust for their concerns continue to show a significant and meaningful reduction in the risk of CVD and total mortality from the moderate intake of wine and alcohol.

They showed that in fully adjusted models all current drinkers had a decreased risk (a relative risk of about 0.81-0.86) for IHD for those consuming 1.3 – 24.99 g of alcohol per day, indicating cardioprotection from low to medium volume alcohol consumption. In that study, however, the authors used a similarly limited number of ‘higher quality[7]’ studies to reach their conclusion that “neither partially nor fully adjusted estimates indicated significantly decreased risk of CHD for low- and medium-volume drinkers, although the estimated RRs for all current drinkers were less than unity”. According to the authors, higher quality studies (not specified in the paper) were mainly free from abstainer bias.

2.3 The updated analysis by Zhao et al. (2017) also markedly distorted the accumulated scientific evidence on alcohol and CVD and mortality. This biased selection of studies selected by Zhao et al. (2017) undermines the value of the paper, but more importantly, promulgates misinformation in the name of the appropriate scientific method. Failure to acknowledge the robust body of knowledge supporting the opposite conclusion and disqualifying extensive studies that offer plausible biological explanations of observed benefits is unconscionable.

Indeed, Zhao et al. (2017) have been criticized in multiple peer-reviewed publications since its publication. One of the criticisms was that only using lifelong abstainers as a reference group will introduce bias as well. Such a reference group will represent a small unrepresentative fraction of the population with an overrepresentation of religious minorities as Mormons and Adventists (Ding & Mukamal, 2017). In their commentary, Ding and Mukamal also mention that meta-analyses suggest that former drinkers are at higher risk for CHD mortality but not CHD incidence (Roerecke and Rehm, 2011), which suggests that the choice of the reference group is at least irrelevant for CHD incidence.

2.4 Of course, there are problems with assessing alcohol consumption in a population and of course, former drinkers will be less healthy than abstainers once these former drinkers decide to stop drinking due to health problems. However, one can try on the one hand to assess the contribution of that former drinker’s group and correct for it, or on the other hand discard most of the studies performed in this area. Choosing the first option (assessing the contribution) has been done and has shown that such a contribution does not always affect the overall outcome of protection.  Choosing the second option will introduce another form of bias, namely selection bias, which will exclude most of the studies performed in this area and dramatically limit the use of available data. Data collected in studies performed according to the prevailing scientific rules and quality guarantees approval by various committees and funding parties.

2.5 The Draft Report also considers the “Impact of drinking patterns on cardiovascular disease risk”. The authors use the study by Mostofsky et al. for the immediate risk after consuming alcohol rather than long-term risk. That paper states:  Moderate alcohol consumption was associated with an immediately higher cardiovascular risk that was attenuated after 24 hours, and even protective for myocardial infarction and hemorrhagic stroke (≈2-4 drinks: relative risk=30% lower risk) and protective against ischemic stroke within 1 week (≈6 drinks: 19% lower risk). In contrast, heavy alcohol drinking was associated with higher cardiovascular risk in the following day (≈6-9 drinks: relative risk=1.3-2.3) and week (≈19-30 drinks: relative risk=2.25-6.2).

2.6 The authors of the Draft Report only used the last information on heavy drinking (9 drinks per day) and omitted the following: When accounting for the amount of alcohol consumed in the 24 hours prior to MI onset, there was a U-shaped association (p curve<0.001) between alcohol intake and MI risk, with the greatest benefit following approximately 28 grams of alcohol (≈2 drinks) in one day (RR=0.67) and a higher risk following approximately 108 grams (≈9 drinks) in one day (RR=1.59).

So, by selective referencing and selective use of data focusing on heavy drinking rather than on light or moderate drinking, data on the association between alcohol consumption and ischaemic heart disease are distorted, which results in an underestimation of the beneficial effect of light and moderate alcohol consumption on cardiovascular disease.

2.7 The Draft Report also does not include the conclusions of Roerecke and Rehm (2014) from their meta-analysis on the role of confounders in explaining the observed association of alcohol with health outcomes. Roerecke and Rehm (2014) stated that the “Results from our quantitative meta-analysis showed that drinkers with average intake of < 30 g/day and no episodic heavy drinking had the lowest IHD (ischaemic heart disease) risk (relative risk = 0.64, 95% confidence interval 0.53 to 0.71). Drinkers with episodic heavy drinking occasions had a risk similar to that of lifetime abstainers (relative risk = 1.12, 95% confidence interval 0.91 to 1.37).”  Further, these two authors who have traditionally been concerned that confounding and errors weaken the purported relation between alcohol and a lower risk of CVD, concluded: “For drinkers having one to two drinks per drinking day without episodic heavy drinking, there is substantial and consistent evidence from epidemiological and short-term experimental studies for a beneficial association with IHD risk when compared to lifetime abstainers. The alcohol-IHD relationship fulfils all criteria for a causal association proposed by Hill.”

Thus, Zhao et al. (2017) purportedly greatly reduce some of the challenges of earlier evidence, but instead markedly distort the accumulated scientific evidence on alcohol and CVD, where the biased selection of studies that are included undermines the value of the papers, but more importantly promulgates misinformation in the name of appropriate scientific method.  Failure to acknowledge the robust body of knowledge supporting the opposite conclusion and disqualifying extensive animal and cell culture studies that offer plausible biological explanations of observed benefits is unconscionable. There is an immense amount of experimental data, not only animal experiments but also trials in humans, that have described the mechanisms by which moderate alcohol consumption have been shown to decrease essentially all the risk factors for CVD. These mechanisms include low HDL-cholesterol, elevated LDL-cholesterol, endothelial dysfunction, coagulopathies, inflammation, abnormal glucose metabolism, and many others (Brien et al., 2011, Ronksley et al., 2011, Hendriks, 2020). The consistent finding of lower CVD risk among moderate drinkers in all well-done cohort studies is strongly supported by experimental evidence of the mechanisms.

Indeed, the overwhelming body of observational scientific data, as well as an immense number of experimental studies, support the contention that, for most middle-aged and older men and women who choose to do so, the regular consumption of small amounts of an alcoholic beverage can be considered as one component of a ‘healthy lifestyle’, associated with a lower risk of CVD and total mortality.

Of note is that in Figure 6 of the Draft Report, their data actually show a J-shaped relationship and hence a reduction in IHD and ischaemic stroke for two drinks per day. This is neither emphasized nor discussed in the Draft Report.

Specific Comments from Forum Members

Forum Member McIntosh states “The authors claim on page 2 that increasing alcohol consumption from 7 drinks per week to 9 will increase the risks of getting cancer by a factor of 10. I don’t believe this.

While I don’t understand what they are doing and have not read all the papers my understanding of what most other researchers have done is that the statement is false.

Here is the argument. Looking at Bergmann et al. (2103), Table 4 for cancer deaths and both genders shows hazard ratios for two categories: below recommended levels at 0.0.85 and 0.87 for light to moderate drinking when the hazard ratios have been renormalized to make never drinking the reference category. So there is a 2% increase in risk as respondents move up one category, an increase of about 2 drinks per day, nowhere near what the authors claim. It is important to note that both levels yield much lower risk levels than not drinking. This result can be shown in a large number of studies that have been carried out in the last 20 years. I chose Bergmann et al. (2013) because I believe it is the best of all studies that I have seen. First, it is a competing risk model which is important and second, former drinkers are included as a category avoiding Stockwell-type objections. Unfortunately, some of the regressors are correlated with the alcohol variables so the adjustments do not deal with the net vs gross effect of alcohol use. As we saw in the BRFSS data drinking 7 drinks a week does not minimize the risk of getting heart disease or diabetes.

To write a response to this Draft Report, I think it has to be to understand what they have actually done. It has something to do with adjusting the number of drinks so that they are consistent with alcohol sales. At a minimum, there should be a table in the Draft Report giving the distribution of respondent drinks per day that they think is appropriate. This is not done. They also say that what Bergmann et al. (2013), Mi et al. (2107), and Woods et al. (2019) suggest is wrong without explaining why. They don’t ever refer to these papers. Knowing what the CCSA guideline study does is not helpful. There are large numbers of factual errors in the paper such as the number of deaths due to cancer in 2018 is around 3200 and not 17000.”

Forum Member Ellison states “other Forum members have described in detail many inaccuracies, false assumptions, and unusual measures taken by the authors to support their long-stated premise, which seems to be to urge zero consumption of alcohol.  A few additional points can be made as follows:

1.   First, all of the figures showing ‘alcohol-attributable deaths’ obviously cannot show deaths unrelated to alcohol.  It would be important to also show total mortality risk to clarify the relationship of alcohol consumption to the risk of death. 

2.   Studying alcohol and health assumes people drink (or are injected with) pure alcohol, whereas the relevant relation is between the consumption of different beverages that contain alcohol and health outcomes.  Here the health effects tend to relate primarily to the amount, the type of beverage, and the pattern of drinking (frequency of consumption, in binges or not, with or without food, for example).  In this Draft Report, the authors do not separate effects according to these important characteristics. 

3.   The analyses presented are based on the authors’ combination of data from:

(a)  young people (where there are generally no health benefits but increased deaths predominantly from accidents and violence related to alcohol abuse);

(b)  middle age (when decreases in a few cardiovascular deaths are usually exceeded by increases in the risk of death from liver disease and some cancers, mainly among heavier drinkers); and

(c)  the elderly, when deaths from cardiovascular diseases and dementia are especially common.

Mixing data from all ages in the same statistical calculations does not permit a clear appraisal of the health effects of alcoholic beverage consumption.

4.   The authors have adjusted for under-reporting of intake, which is appropriate. They have, however, applied the same formula to all subjects, whereas it has been demonstrated that alcohol abusers tend to understate their consumption more than moderate drinkers (Klatsky et al., 2006, Klatsky et al., 2014).

5.   The exclusions of certain publications by these authors are always interesting.  They report that of 7,294 publications screened, 7,003 were excluded, leaving 291 for review.  Of these 291 studies, 235 were excluded primarily because they had a “wrong outcome”, “wrong exposure”, or “wrong study design”.  For the final group of 56 studies considered in their very extensive analyses, there were only 24 providing data on neoplasms, 19 for cardiovascular diseases (but only four for IHD or ischaemic stroke), and fewer than 10 for other outcomes.  Especially noted is that there were only two studies for assessing the effects of alcohol on the risk of diabetes, a rapidly increasing and important disease that is almost always found to be reduced by moderate drinking.

Overall, for the results given in this report, deficiencies have been noted indicating that little can be used to assist our attempts to judge the net health effects of moderate alcohol consumption.  We require reliable and unbiased data upon which to base recommendations to the public.”

Forum Member Harding makes the following observations:

  1. The entire Draft Report is based on associations observed from epidemiological studies, expressed as relative risks, without even beginning to address whether the association is causal, in whole or in part.  This means that throughout, the ‘relative risks’ are not real risks, they are a measure of the statistical association between two variables.  Indeed, studies that might help to inform the issue of causation (e.g., randomised control trials) have been deliberately excluded (see second bullet point of the summary, page 3).  It follows that the use of the word ‘risk’ in all the remaining bullet points of the summary, indeed throughout the Draft Report, in the way the authors intend, is unjustified.
  2. It is clear from the section on ‘Methods’ (page 8) that the epidemiological data has been subjected to mathematical modelling, without providing any information on the nature of these models.  All models need a central criterion against which the data is compared (as discussed on pages 42 and 43), e.g., ‘no more than a 1 in 100 chance of an alcohol-attributable lifetime mortality risk’.  Whatever the criterion, it is by its nature totally arbitrary, completely made-up.  If another figure were chosen (e.g., 1 in 1000), then the acceptable lifetime level of consumption, e.g., x g/day, would change.   Given that this whole exercise is intended to inform the population what is sensible consumption for them as individuals, this public health message actually depends on an arbitrary number in the model, and not, as the public will assume and have the right to expect, derived from the actual evidence of the effect of alcohol on their disease risk.  If a different criterion were chosen, then different public health advice would emerge from the model and would be completely unrelated to the actual evidence.  That cannot be right.
  3. Whatever level of consumption that emerges from the models will be passed off as appropriate consumption for all individuals in the population, when it is nothing of the sort.  It is simply the level of consumption that everyone must adhere to in order to meet the made-up criteria in the mathematical model.  It means that everyone is required to behave in a particular way to achieve some arbitrary national target.
  4. It would appear that deaths from all alcohol-related causes have been included in the models, both chronic and acute, i.e., both those as a result of long-term regular consumption and those as a result of being intoxicated on single occasions.  This means that everyone in the population will be advised to drink less because, e.g., some people drive their cars into trees or fall off ladders when under the influence of alcohol.  This is completely unreasonable and unacceptable.  Instead, the population should simply be advised not to get drunk, and public health advice should focus on what is sensible consumption.

Forum Member Goldfinger states “as a cardiologist, I have stressed the consistent reduction of cardiovascular disease events and reduction in total mortality with moderate alcohol consumption, despite a small increase in certain cancers with alcohol. Note the meta-analysis by Lucerón-Lucas-Torres et al. (2023) where wine has not been associated with any cancers and, in fact, may be protected.”

Forum Member Waterhouse highlights that the guidelines call for recommendations based on “the preponderance of scientific and medical knowledge which is current.”  While it is not explicit, it would appear that a key criterion is that “knowledge” be peer-reviewed scientific papers, since the bibliographies of both the NASEM and ICCPUD reports cite such papers. 

Thus, the Draft Report is flawed in the following ways:

  1. As many have pointed out, the authors demonstrate a blatant bias in excluding the vast majority of the scientific literature. This includes papers that underwent rigorous peer review in the top-tier medical journals, to cherry-pick the papers that help them achieve their political goal to claim that moderate alcohol consumption is unhealthy. Many of these papers are found in third-rate journals with sloppy peer review;
  2. The authors were not satisfied with just a biased selection of papers but created flawed models of morbidity and mortality to ‘substantiate’ their claims. These models have NOT been scientifically peer-reviewed and scientifically validated; and
  3. In the section on the purpose of this Draft Report, the authors do not even claim to be creating advice for a healthy diet but instead are setting out only to define the harms of alcohol use.  Of course, this is not surprising, since the stated purpose of the ICCPUD is to reduce alcohol consumption in the youth population, and their modus operandi is to show how much harm occurs to that population because of alcohol consumption. 

Furthermore, this Draft Report has not been peer-reviewed and is authored by an ad-hoc subset of the panel or their unnamed advisors and experts.” 

Forum Member Skovenborg considers that “this ISFAR critique takes on many issues and has many good arguments, the problem is, however, that for the media, the population, and the politicians it is a very complicated and hardly understandable discussion. The very purpose of the flood of biased reviews produced by Stockwell et al. since 2006 is to create a complex narrative like professor A says that moderate consumption of alcohol is fine for individuals aged 40+ while professor B says that no intake is without risk and any intake is associated with cancer ─ a narrative that is not really about science, it’s about creating affective polarization against alcoholic drinks at the individual level in general and the political level in particular. 

The change from science to politics started in 1994, when a WHO publication published this conclusion about the association between light drinking and reduced risk of coronary heart disease: “Evidence on the protective effect of drinking on coronary heart disease is conclusive at the level of association, highly suggestive at the level of causation, but not, on present analysis, significant at the policy level. Most of the achievable benefit is likely to be obtained at an alcohol intake between one drink every 2 days, and two drinks a day.” (Edwards et al., 1994). The same year a report of a WHO Scientific Group on cardiovascular disease risk factors (World Health Organization, 1994a.)  concluded that “moderate drinking (10-30 g of ethanol daily, i.e., 1-3 drinks) provides a moderate protective effect against CVD, as compared with abstention and heavy drinking. The protective effect of moderate alcohol consumption is related to the modulation of several well-recognized pathogenic mechanisms leading to atheroma.”

The sound WHO science reports were soon contradicted by a WHO political statement from Hans W. Emblad, then director of the WHO’s Division of Substance Abuse (World Health Organization, 1994b). He said that “there is no minimum threshold below which alcohol can be consumed without any risk, and assertions that moderate alcohol consumption can be good for health, are wrong. Scientific studies show that only very low consumption, of the order of one drink every other day, is liable to reduce the risk of cardiovascular diseases compared with total abstinence.” He went on to say that “there are other ways of reducing the risk of cardiovascular disease: avoid smoking, engage in physical activity, eat less fat. Those already taking these precautions are unlikely to reduce their risk still further with light drinking. We strongly believe that less is better. If you reach the point of zero, from the health point of view that is optimal.”

The change of focus from science to politics has also happened to Timothy R. Stockwell who in 2000 began his Medical Journal of Australia editorial with these two statements: “The evidence amassed to date on the link between moderate alcohol intake and reduced risk of dying of cardiovascular disease might be thought already sufficient to bracket sceptics of alcohol’s protective effect with doubters of manned lunar missions and members of the Flat Earth Society. Published studies demonstrating this link can now be counted in the hundreds, and no fewer than six plausible underlying biological mechanisms have been identified”; and “Alternative explanations for the protective effect of moderate alcohol intake, relying on ever more tenuous confounding effects, have been discredited one by one. The apparently protective effect of moderate alcohol consumption has so far survived the use of controls for sociodemographic status; for the ‘sick quitter hypothesis’! (i.e., the suggestion that many abstainers have stopped drinking because of serious illness), for the amount of cholesterol in the diet, and even for the degree of social isolation.” (Stockwell, 2000.)

The best way to maintain focus on sound science and resist the determined attempts to create confusion with abstainer bias and confounding factors like socioeconomic status and lifestyle is to focus on results from studies such as that described in the peer-reviewed publication Impact of healthy lifestyle factors on life expectancies in the US population (Li et al., 2018). Key points are:

  • Data from Nurses’ Health Study (n=78 865) and the Health Professionals Follow-up Study (n=44 354) = no confounding by socioeconomic status;
  • Alcohol consumption assessed every four years during follow-up (NHS 34 years, HPFS 28 years) = valid updated drinker status and no abstainer bias;
  • Information about five low-risk lifestyle factors (never smoking, BMI 18.5 to 24.9 kg/m2, ≥30 min/d of moderate to vigorous physical activity, moderate alcohol intake (5-15 g/d for women, 5 to 30 g/d for men), and a high diet quality score = no confounding by lifestyle;
  • The projected life expectancy at age 50 years was on average 14.0 years longer among women with five low-risk factors compared with those with zero low-risk factors; for men, the difference was 12.2 years; and
  • To address the concern about the potential adverse effects of moderate alcohol intake, a healthy lifestyle score based on the other four low-risk factors without alcohol was created. The projected life expectancy at age 50 years was on average 11.4 years longer among women with four low-risk factors compared with those with zero low-risk factors; for men, the difference was 10.0 years = moderate alcohol intake was associated with 2.6 years longer life expectancy for women and 2.2 years for men.

References

Altman, N., & Krzywinski, M. (2015) Association, correlation and causation. Nat Methods 12, 899–900 (2015). https://doi.org/10.1038/nmeth.3587

Bondy, S., & Rehm, J. (1998) The interplay of drinking patterns and other determinants of health. Drug Alcohol Rev, 17:399–412. https://doi.org/10.1080/09595239800187241.

Bryazka, D., Reitsma, M. B., Griswold, M. G., Abate, K. H., Abbafati, C., Abbasi-Kangevari, M., ET AL. (2022). Population-level risks of alcohol consumption by amount, geography, age, sex, and year: a systematic analysis for the Global Burden of Disease Study 2020. Lancet (London, England), 400(10347), 185–235. https://doi.org/10.1016/S0140-6736(22)00847-9

Di Castelnuovo, A., Costanzo, S., Bagnardi, V., Donati, M.B., Iacoviello, L., & de, Gaetano. G. (2006). Alcohol dosing and total mortality in men and women: an updated meta-analysis of 34 prospective studies. Arch Internal Med, 166(22), 2437–2445. https://doi.org/10.1001/ARCHINTE.166.22.2437

Ding, E. L., & Mukamal, K. J. (2017). Robustness of the J-shaped association of alcohol with coronary heart disease risk. J Studies Alcohol Drugs, 78(3), 389–391. https://doi.org/10.15288/jsad.2017.78.389

Edwards et al. (1994) Alcohol Policy and the Public Good. Oxford, Oxford University Press, 226 p.

Farchi, G., Fidanza, F., Giampaoli, S., Mariotti, S., & Menotti, A. (2000) Alcohol and survival in the Italian rural cohorts of the Seven Countries Study.  Int J Epidemiol, 29, 667-671

Gilligan, C., Anderson, K.G., Ladd, B.O., Yong, Y.M., & David, M. (2019) Inaccuracies in survey reporting of alcohol consumption. BMC Public Health, 19, 1639. https://doi.org/10.1186/s12889-019-7987-3

Gmel, G., Gutjahr, E., & Rehm, J. (2003). How stable is the risk curve between alcohol and all-cause mortality and what factors influence the shape? A precision-weighted hierarchical meta-analysis. Eur J Epidemiology, 18(7), 631–642. https://doi.org/10.1023/a:1024805021504

Høyer, G., Nilssen, O., Brenn, T., & Schirmer, H. (1995) The Svalbard study 1988-89: a unique setting for validation of self-reported alcohol consumption.  Addiction, 90, 539-544.

Khosrowi, D. (2019): Extrapolation of causal effects – hopes, assumptions, and the extrapolator’s circle. J Econ Method. https://doi.org/10.1080/1350178X.2018.1561078

Klatsky, A.L., Udaltsova, N., Li, Y., Baer, D., Tran, N.H., Friedman, G.D. (2014) Moderate alcohol intake and cancer: the role of underreporting.  Cancer Causes Control, 25, 693-699. https://doi.org/10.1007/s10552-014-0372-8

Klatsky, A. L., & Udaltsova, N. (2013). Abounding confounding: sick quitters and healthy drinkers. Addiction (Abingdon, England), 108(9), 1549–1552. https://doi.org/10.1111/add.12157

Klatsky, A.L., Gunderson, E.P., Kipp, H., Udaltsova, N., Friedman, G.D. (2006) Higher prevalence of systemic hypertension among moderate alcohol drinkers: an exploration of the role of underreporting.  J Stud Alcohol, 67, 421-428.

Klatsky, A.L., & Udaltsova, N. (2007) Alcohol Drinking and Total Mortality Risk.  Ann Epidemiol, 17, S63–S67.

Li, Y., Pan, A., Wang, D.D., Liu, X., Dhana, K., Franco, O.H., et al. (2018) Impact of healthy lifestyle factors on life expectancies in the US population. Circulation, 138(4), 345-355. doi: 10.1161/CIRCULATIONAHA.117.032047. Erratum in: Circulation. 2018 Jul 24;138(4):e75. https://doi.org/10.1161/CIR.0000000000000587

Lucerón-Lucas-Torres, M., Cavero-Redondo, I., Martínez-Vizcaíno, V., Bizzozero-Peroni, B., Pascual-Morena, C. & Álvarez-Bueno, C. (2023) Association between wine consumption and cancer: a systematic review and meta-analysis. Front Nutr, 10, 1197745. https://doi.org/10.3389/fnut.2023.1197745

Rehm, J., Gmel, G.E. Sr., Gmel, G., Hasan, O.S.M., Imtia, S., Popova, S., Probst, C. et al. (2017) The relationship between different dimensions of alcohol use and the burden of disease-an update. Addiction. 112(6), 968-1001. https://doi.org/10.1111/add.13757

Rehm, J., Room, R., Graham, K., Monteiro, M., Gmel, G., & Sempos, C.T. (2003) The relationship of average volume of alcohol consumption and patterns of drinking to burden of disease: an overview. Addiction, 98(9), 1209-28.

Roerecke, M., & Rehm, J. (2011). Ischemic heart disease mortality and morbidity rates in former drinkers: a meta-analysis. Am J Epidemiol, 173(3), 245–258. https://doi.org/10.1093/aje/kwq364

Ronksley, P. E., Brien, S. E., Turner, B. J., et al. (2011). Association of alcohol consumption with selected cardiovascular disease outcomes: a systematic review and meta-analysis, Br Med J, 342, d671. https://doi.org/10.1136/bmj.d671

Rothman K. J., Greenland S., & Lash T. L. Modern Epidemiology. Philadelphia, PA: Lippincott Williams & Wilkins; 2008.

Shield, K, Keyes, KM, Martinez, P, Milam, AJ, Naimi, TS, & Rehm, J. (2025). Draft Report: Scientific Findings of the Alcohol Intake & Health Study for Public Comment. 1–81.

Stockwell, T.R. (2000) Alcohol and cardiovascular disease: still a research priority? Med J Aust, 173(3), 116-7. https://doi.org/10.5694/j.1326-5377.2000.tb125559.x

Stockwell, T., Zhao, J., Sherk, A., Rehm, J., Shield, K., & Naimi, T. (2018) Underestimation of alcohol consumption in cohort studies and implications for alcohol’s contribution to the global burden of disease. Addiction, 113(12), 2245-2249. https://doi.org/10.1111/add.14392

Valencia Martín, J.L., González, M.J., & Galán I. (2014) Methodological issues in the measurement of alcohol consumption: the importance of drinking patterns. Rev Esp Salud Publica, 88(4), 433-46. https://doi.org/10.4321/S1135-57272014000400002

Wood, A. M., Kaptoge, S., Butterworth, A., et al. (2018). Risk thresholds for alcohol consumption: combined analysis of individual participant data for 599 912 current drinkers in 83 prospective studies. Lancet (London, England), 391(10129), 1513–1523. https://doi.org/10.1016/S0140-6736(18)30134-X

World Health Organization (1994a) Technical Report – Scientific Group on cardiovascular disease risk factors, 841, 1-53.

World Health Organization (1994b) https://www.upi.com/Archives/1994/11/01/WHO-Even-moderate-drinking-is-harmful/7072783666000/

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

Henk Hendriks, PhD, Zeist, The Netherlands

Creina Stockley, PhD, MBA, Independent consultant and Adjunct Senior Lecturer in the School of Agriculture, Food and Wine at the University of Adelaide, Australia

James McIntosh, PhD, Professor of Economics, Concordia University, Montreal, Canada

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

Richard Harding, PhD, Formerly Head of Consumer Choice, Food Standards and Special Projects Division, Food Standards Agency, UK

Erik Skovenborg, MD, Former Family Medicine specialist, member of the Scandinavian Medical Alcohol Board, Aarhus, Denmark

Fulvio Mattivi, Fulvio Mattivi, MSc, Professor and Scientific Advisor, Research and Innovation Centre, Fondazione Edmund Mach, in San Michele all’Adige, Italy

Andrew L. Waterhouse, PhD, Professor Emeritus of Enology, Department of Viticulture and Enology, University of California, Davis, CA, USA

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

Tedd Goldfinger, DO, FACC, Desert Cardiology of Tucson Heart Center, University of Arizona School of Medicine, Tucson, AZ, USA

Lynda H. Powell, MEd, PhD, Chair, Dept. of Preventive Medicine, Rush University Medical School, Chicago, IL, USA

Matilda Parente, MD, Consultant in Molecular Pathology/Genetics and Emerging Technologies, San Diego, CA, USA

Fulvio Ursini MD, Emeritus Professor of Biochemistry, University of Padova, Padova, Italy

Ramon Estruch, Ramon Estruch, MD, PhD.  Associate Professor of Medicine, University of Barcelona, Spain

Luc Djoussé, MD, DSc, Dept. of Medicine, Division of Aging, Brigham & Women’s Hospital and Harvard Medical School, Boston, MA, USA

Linda McEvoy, PhD, Department of Radiology, University of California at San Diego (UCSD), La Jolla, CA, USA

Mladen Boban, MD, PhD, Professor and Head of the Department of Pharmacology, University of Split School of Medicine, Croatia

Dominique Lanzmann-Petithory, MD, PhD, Nutrition/Cardiology, Praticien Hospitalier Hôpital Emile Roux, Paris, France Monika Christmann, PhD, Head of Institute, Department of Enology and Professorship for Enology, Hochschule Geisenheim University, Germany

[1] https://www.alcoholresearchforum.org/critique-261/

[2] https://movendi.ngo/who-we-are/the-movendi-way/what-we-do/

[3] https://www.telegraph.co.uk/health-fitness/diet/alcohol/abstemious-scientists-accused-flawed-research/

[4] https://onlinelibrary.wiley.com/doi/full/10.1111/add.13757

[5] https://icd.who.int/browse10/Content/statichtml/ICD10Volume2_en_2019.pdf

[6] https://doi.org/10.1016/j.lansea.2024.100415

[7] Only 45 studies were selected from a total of 2768 studies identified

Proudly powered by WordPress. Theme by Infigo Software.