Why Do Only Some Cohort Studies Find Health Benefits From Low-Volume Alcohol Use? A Systematic Review and Meta-Analysis of Study Characteristics That May Bias Mortality Risk Estimates
Systematic review
Study highlights
Previous studies demonstrating the health benefits of moderate alcohol consumption were of low quality, according to a new systematic review.
In many cases, these inferior studies would label older participants who had quit alcohol use due to ill health as alcohol consumers “who consumed less than weekly”.
By comparing such “alcohol users” to younger, healthier people who consumed alcohol more regularly, many of these flawed studies came to the erroneous conclusion that low-dose alcohol intake had health benefits.
But when researchers identified and discounted these flawed studies and focused only on thorough research in their systematic review, the researchers from the Canadian Institute for Substance Use Research (CISUR) found no evidence that light-to-moderate alcohol users lived longer.
Exposing flawed research, correcting decades-old mistakes
Over decades studies seemed to suggest that any alcoholic beverage, when consumed in “moderation”, could offer health benefit. One study published in 1997, for instance, found that men and women who consumed one or two alcoholic drinks per day had 30-40% lower total cardiovascular disease mortality than people who did not consume alcohol.
To assess and investigate the validity of these kinds of findings, a research group from CISUR and the Canadian Centre on Substance Use and Addiction (CCSA) reevaluated 107 longitudinal alcohol-health studies, which accounted for 4,838,825 participants and 425,564 alcohol deaths.
The researchers divided the studies into either “lower” or “higher” quality categories based on whether the comparison group of teetotalers/irregular alcohol users included participants with ill health.
The present study, an extension of Zhao et al. (2023), is essentially a study of the methodological limitations and resulting biases that affect mortality risk estimates in observational, longitudinal research.
Researchers found that estimates for all-cause mortality risk for low-volume alcohol users were significantly higher in the studies that:
- used younger cohorts,
- excluded potential participants with current or prior ill health,
- assessed alcohol use over less than 30 days,
- didn’t suffer from “abstainer bias” in the reference group, and
- didn’t control for smoking status.
The researchers concluded that participant selection biases had warped the findings of dozens of alcohol mortality studies over the years, creating “the false appearance of health benefits from moderate [alcohol consumption].”
If you look at the weakest studies, that’s where you see health benefits,” said Tim Stockwell, former director of the CISUR, as per Technology Networks Applied Sciences reporting.
Prof. Tim Stockwell, former director, Canadian Institute for Substance Use Research (CISUR)
The health toll of alcohol
Beyond these flawed studies, high-quality alcohol research has illuminated alcohol’s real health risks, such as the risk of developing bowel and female breast cancer – risks that increase from the first drop of alcohol intake and that increase substantially the more alcohol a person consumes.
In early 2023, the World Health Organization (WHO) issued a statement based on a review of world-class science:
The risks and harms associated with drinking alcohol have been systematically evaluated over the years and are well documented. The World Health Organization has now published a statement in The Lancet Public Health: when it comes to alcohol consumption, there is no safe amount that does not affect health.
WHO Statement “No level of alcohol consumption is safe for our health”, January 2023
WHO is very clear: It is the alcohol that causes harm, not the beverage. Ethanol (alcohol) causes cancer through biological mechanisms as the compound breaks down in the body, which means that any beverage containing alcohol, regardless of its price and quality, poses a risk of developing cancer.
We cannot talk about a so-called safe level of alcohol use – the risk to the [alcohol consumer’s] health starts from the first drop of any alcoholic beverage,” explained Dr Carina Ferreira-Borges, acting Unit Lead for Noncommunicable Disease Management and Regional Advisor for Alcohol and Illicit Drugs in the WHO Regional Office for Europe.
Dr Carina Ferreira-Borges, acting Unit Lead for Noncommunicable Disease Management and Regional Advisor for Alcohol and Illicit Drugs, WHO Europe
Despite this, the question of beneficial effects of alcohol has been a contentious issue in research for years.
In the WHO press release, Dr Jürgen Rehm, member of the WHO Regional Director for Europe’s Advisory Council for Noncommunicable Diseases and Senior Scientist at the Institute for Mental Health Policy Research and the Campbell Family Mental Health Research Institute at the Centre for Addiction and Mental Health, Toronto, Canada, explained the science clearly:
Potential protective effects of alcohol consumption, suggested by some studies, are tightly connected with the comparison groups chosen and the statistical methods used, and may not consider other relevant factors.”
Dr Jürgen Rehm, Senior Scientist, Institute for Mental Health Policy Research and the Campbell Family Mental Health Research Institute, Centre for Addiction and Mental Health, Toronto, Canada
Study background
An increasing number of studies have challenged prior findings suggesting that alcohol consumed at low levels can protect against serious diseases, notably ischemic heart disease (IHD), ischemic stroke, and type 2 diabetes.
Although many observational studies suggest that people consuming alcohol at “moderate” (i.e., low-volume) levels live longer and have fewer illnesses than people assessed as “abstainers,” many potential sources of selection bias contaminate such simple comparisons.
Assumptions about the validity of the alcohol and health benefit hypothesis have major implications for estimates of the global burden of disease and the formulation of national guidelines on low-risk alcohol use.
A recent estimate of alcohol’s global burden of disease by the International Health Modelling and Evaluation (IHME) group used an assumption that consuming up to nine alcoholic drinks per day protected against IHD. This contributed to an estimate of 1.8 million deaths globally in 2020 attributable to alcohol use, 1 million less than the estimate for 2016 that assumed more modest protective effects against IHD.
Media coverage of this high-profile study supposedly showing light alcohol use brought better health, especially among older people, was swift and wide spread, and often accompanied by sensational headlines.
But how did several hundred authors of the IHME group publish a study in the pre-eminent medical journal The Lancet lose a million global alcohol deaths and nobody is raising concerns?
Prof. Tim Stockwell shared his detailed analysis of the infamous study. Prof. Stockwell dug deep into the appendix of the study and what he found is not pretty. He uncovers flaws, shortcomings, and methodological problems that cast doubt about the foundations of the study and its conclusions.
Previous systematic reviews and meta-analyses of alcohol use and all-cause mortality have highlighted the crucial role of study design, particularly the criteria applied to define the reference group (usually abstainers) against which alcohol consumers are compared.
Naimi et al. (2017)provided a theoretical analysis of how selection biases accumulate across the life course resulting in the false appearance of health benefits for low-volume alcohol use, especially in studies of older people. Empirical evidence supports this theory, showing how the usual reference group of abstainers fills up among older cohorts with people who have stopped or cut down on their alcohol intake for health reasons.
Thus, studies of younger cohorts followed up to older age should be less prone to such selection biases, a prediction confirmed in a meta-analysis of IHD studies. Similarly, a study of an older Spanish cohort found that the appearance of protection against all-cause mortality for low-volume alcohol users vanished when a measure of lifetime rather than recent alcohol use was incorporated.
A recent systematic review and meta-analysis of alcohol use and all-cause mortality confirmed that estimates of mortality risk linked with low-volume alcohol use become smaller and nonsignificant when adjustment is made for key study characteristics (e.g., if lifetime abstainers were the reference group, study cohort age).
However, statistical study-level controls for these methodological characteristics may not eliminate their effects; thus, lower-quality studies would still have influenced this pooled meta-analysis. This point was also made repeatedly in a critique of Zhao et al. (2023) by the International Scientific Forum for Alcohol Research (ISFAR) (2023), a group with significant ties to the alcohol industry.
The present study is a response to the criticism that the researchers should have excluded studies from our meta-analysis that did not meet pre-registered quality criteria. To illustrate the difference these criteria make, the researchers provide separate meta-analyses of both lower- and higher-quality studies using up to three key quality criteria. Specifically, we group studies by key characteristics predicted to be important based on the model of lifetime selection bias outlined above.
Study conclusions
We suggest that our main findings present further support for the existence of lifetime selection biases in cohort studies of alcohol and mortality creating the false appearance of health benefits from moderate drinking. As individuals age, it has been repeatedly shown that those who cut down or completely stop alcohol use are prone to ill health, thus rendering those who continue drinking to look healthy by comparison. We found that subgroups of studies least likely to be biased had no significant reduction of mortality risk among lower-volume drinkers. Conversely, subgroups of studies more likely to be biased showed substantial apparent health benefits.”
Stockwell T, Zhao J, Clay J, Levesque C, Sanger N, Sherk A, Naimi T. Why Do Only Some Cohort Studies Find Health Benefits From Low-Volume Alcohol Use? A Systematic Review and Meta-Analysis of Study Characteristics That May Bias Mortality Risk Estimates. J Stud Alcohol Drugs. 2024 Jul;85(4):441-452. doi: 10.15288/jsad.23-00283. Epub 2024 Jan 30. PMID: 38289182.
Apparent health benefits can be consistently created through studies that bias the abstainer reference group toward ill health. The relatively few published studies meeting minimal quality criteria to avoid this problem do not show significantly lower mortality risk for low-volume alcohol users.
The identified sources of bias here pervade the field of alcohol epidemiology and have confused public communications and alcohol and health risks.
In addition, studies of non-smokers find no J-shaped curve.
Future research should investigate whether adjustment for smoking status and/or SES is always appropriate or whether it may bias low-volume alcohol users toward appearing healthier than current abstainers.
Abstract
Objective
Assumptions about alcohol’s health benefits profoundly influence global disease burden estimates and alcohol use guidelines. Using theory and evidence, the researchers identify and test study characteristics that may bias estimates of all-cause mortality risk associated with low-volume alcohol consumption.
Method
The researchers identified 107 longitudinal studies by systematic review with 724 estimates of the association between alcohol consumption and all-cause mortality for 4,838,825 participants with 425,564 recorded deaths.
“Higher-quality” studies had a mean cohort age of 55 years or younger, followed up beyond 55 years, and excluded former and occasional alcohol users from abstainer reference groups.
“Low-volume” alcohol use was defined as between one alcoholic drink per week (>1.30 g ethanol/day) and two alcoholic drinks per day (<25 g ethanol/ day).
Mixed linear regression was used to model relative risks (RRs) of mortality for subgroups of higher- versus lower-quality studies.
Results
As predicted, studies with younger cohorts and separating former and occasional alcohol users from abstainers estimated similar mortality risk for low-volume alcohol users as abstainers.
Studies not meeting these quality criteria estimated significantly lower risk for low-volume alcohol users.
In exploratory analyses, studies controlling for smoking and/or socioeconomic status had significantly reduced mortality risks for low-volume alcohol users. However, mean RR estimates for low-volume alcohol users in nonsmoking cohorts were above 1.0.
Conclusions
Studies with life-time selection biases may create misleading positive health associations. These biases pervade the field of alcohol epidemiology and can confuse communications about health risks.
Future research should investigate whether smoking status mediates, moderates, or confounds alcohol-mortality risk relationships.