A study commissioned by the Biden administration and published this week in the journal Studies on Alcohol and Drugs bears the main conclusion its authors wanted to write into the U.S. Dietary Guidelines: Alcohol has no protective effect on deaths at any level, and the health risks start at one drink a day. The Trump administration declined to include the findings in the 2025-2030 Dietary Guidelines for Americans, advising instead that people drink “less alcohol for better overall health.”
According to the analysis, men who consume 14 standard drinks per week – the upper limit allowed in previous guidelines – have a risk of dying from alcohol of 1 in 25. The risk starts at about seven drinks per week for both sexes. The researchers found that even levels considered “moderate” increased the risk of premature death and more than 200 diseases, including heart disease and cancer.
But a closer look at the graphs and tables shows that the conclusions are not as clear as they seem.
At one drink per week, the study’s central estimate—the model’s single most likely value—is not a positive number representing harm. This is less than 16.30 alcohol-related deaths per 1,000 men. That’s three drinks a week, minus 10.56. Negative numbers here indicate protection: drinking at these levels is associated with fewer deaths than not drinking at all.
The confidence interval — the statistical range within which the true value reasonably lies — exceeds zero at these doses, making the result uncertain, the researchers note. A confidence interval of -47 to plus 11 deaths per thousand, as for men who drink one drink per week, corresponds to strong protection, mild harm, and everything in between—all at once.
But this is a simplistic reading of what is a statistical expression of poor study design.
Behind him is the structure of the study. It is a modeling exercise based on relative risks derived from observational research – studies that record health outcomes in people who drink versus those who do not drink, without random assignment. Drinkers and non-drinkers differ in ways that resist full statistical correction: income, diet, physical activity, and access to health care.
The researchers were not aware of these traps. They chose lifelong abstainers – people who had never drunk – as their reference group rather than all non-drinkers, a deliberate attempt to exclude former drinkers who might have given up drinking because they were already sick, and whose presence might make abstinence appear artificially unhealthy. They also adjusted self-reported consumption upwards against national sales data, to correct for participants’ tendency to underestimate their drinking.
These are real methodological choices, made in good faith and documented transparently.
In their limitations section, however, the authors acknowledge that the definition of lifetime abstinence varies across the source studies from which the relative risks are extracted—which means that the correction, no matter how carefully designed, depends on inputs that cannot be fully standardized.
This shows in the numbers, which ultimately undermines the very point she was trying to make.
Vinay Prasad, an American epidemiologist who until recently served as a top official at the US Food and Drug Administration under the Trump administration, has chosen some unusual consequences. For example, among women with modest consumption levels, calculations of disability-adjusted life years—a measure of healthy living after accounting for time lost to illness and premature death—show a marginal gain rather than a loss. This is not a finding on which any public health policy can be based.
Alcohol is so socially embedded and commercially powerful that it is not possible for this debate to resolve itself on scientific grounds alone, let alone that the scientific foundations appear to be fragile.
After the study’s draft findings were published, trade associations described it as “irretrievably flawed”. The House Oversight Committee called it “fraught with bias” and accused the report’s authors of making predetermined conclusions. Robert Vincent, the government official who led the study, told the Associated Press that he was “asked to kill her” within the Trump administration, but he refused and was later fired.
Two of the five officials charged with recommending stricter guidelines — including leading alcohol policy at the Centers for Disease Control and Prevention — were fired in the spring staff cuts. A competing $1.3 million, congressionally commissioned study was itself the product of industry pressure, according to two former alcohol lobbyists who disclosed it to the media.
The industry campaign has taken aim at the study’s methodology — but not because of flaws that lead, inadvertently, to show alcohol as a protective factor at low doses; But because its main recommendation appears to be, by all indications, an editorial choice.
This does not diminish the documented harm of alcohol at higher doses determined by more robust epidemiological studies. The scientific consensus is: Ethanol is metabolized to acetaldehyde, a compound that forms chemical bonds with DNA – adducts – that disrupt normal reproduction and can become permanent mutations.
Alcohol also impairs folate metabolism and raises estrogen levels, both of which have been linked to hormone-sensitive cancers.
For liver disease, the progression from fatty liver to hepatitis to fibrosis to cirrhosis to hepatocellular carcinoma is dose-responsive and mechanistically determined.
For injuries, the relationship with blood alcohol concentration is sharp: every 0.02% increase in blood alcohol content increases the odds of a fatal road accident by approximately 74%.
With respect to cardiovascular disease, Mendelian randomized studies — which use naturally occurring genetic variants that alter alcohol metabolism as a proxy for lifelong drinking, bypassing confounders that distort observational data — have generally found no protective effect.
The most powerful of these studies is the China Kadoorie Biobank study published in The Lancet in 2019, which followed 512,715 people for ten years. While conventional analysis indicated cardiovascular protection, genetic analysis found none.
In some ways, the new study harkens back to the early days of the pandemic, a time when trust in science eroded, rather than improved.
At the time, in the spring of 2020, as lockdowns spread, some public health guidelines framed masks primarily as a way to protect the wearer. The strongest evidence was source control: masks reduce what an infected person spreads more reliably than what an uninfected person picks up. When this gap became visible, it was used to discredit all mask guidelines, including source control guides, which had never been seriously questioned.
The results for cancer and the liver are not in dispute. The evidence is established, the mechanisms are understood, and no serious scientific body disputes them. The issue of low-dose mortality is different: the data are uncertain, the methodology is disputed, and this study does not resolve it. For the average person who looks to science and public health messages to understand what they are putting in their bodies, this distinction is important. Presenting both outcomes under the same heading, as if the uncertainty of one is as stable as the solidity of the other, links the credible and the contested. It gives anyone interested in the outcome — industry, ideology, or otherwise — a way in.
