Alcohol and Health: The Surprising Reason Moderate Is No Longer Safe

If you grew up hearing that a glass of red wine was good for your heart, you were not alone. For much of the late 20th century, moderate drinking carried a halo of legitimacy in public health messaging. Guidelines in the United States and elsewhere suggested that one to two drinks per day was not just acceptable but possibly protective. That picture has changed significantly.

In January 2023, the World Health Organization published a report stating that no safe level of alcohol consumption exists for health. The agency then followed up with a formal fact sheet reinforcing the position that alcohol causes cancer through direct biological mechanisms and that no threshold has been identified below which consumption is considered safe. Shortly after, in January 2025, the U.S. Surgeon General issued an advisory calling alcohol the third leading preventable cause of cancer in the United States, behind only tobacco and obesity, and recommending updated warning labels on alcoholic beverages.

This updated position reflects a genuine shift in how public health bodies interpret decades of accumulated evidence. Understanding why that shift happened requires looking at what the old evidence actually showed, where it fell short, and what newer research methods reveals.

Where the Safe Message Came From

The idea that moderate drinking was harmless or beneficial came from observational epidemiology studies conducted throughout the 1970s, 1980s, and 1990s. Those studies found what's called a J-shaped curve: when researchers plotted alcohol consumption against mortality or cardiovascular disease risk, the line dipped before it rose back up. Non-drinkers appeared to fare worse than light or moderate drinkers, who seemed to have lower rates of heart disease than both pabstainers and heavy drinkers.

This pattern was replicated across dozens of studies and across multiple countries. A meta-analysis of 44 observational studies found a J-shaped association between alcohol use and coronary heart disease risk. Another analysis including over one million participants observed that low to moderate alcohol consumption was inversely associated with total mortality in both men and women, again in a J-shaped pattern. The consistency of the finding gave it considerable credibility. Public health bodies translated it into population-level guidance, and "drink in moderation" became embedded in dietary and health recommendations.

The Problem With Previous Research

The observational studies that produced the J-shaped curve had a critical methodological flaw that took years to fully appreciate: the comparison group was contaminated.

In most of those studies, the abstainer category, the people who reportedly did not drink, was not composed exclusively of lifelong non-drinkers. It also included former drinkers, or people who had reduced or stopped drinking because they were already sick. This is now called the "sick quitter" effect. When someone quits drinking because of cardiovascular disease, cancer, liver problems, or other serious illness, they shift from the moderate drinker category into the abstainer category. The result is that the abstainer group looks sicker, not because abstaining is harmful, but because many abstainers had already been harmed by years of drinking before they stopped.

A systematic review and meta-analysis examined 87 studies and found that those with higher quality study designs, specifically those that excluded former drinkers from the abstainer reference group, showed little to no protective effect of moderate drinking on all-cause mortality. When the comparison was made against true lifetime abstainers, the apparent benefit largely disappeared.

Additional confounding variables made the problem worse. Moderate drinkers, as a group, tend to have higher socioeconomic status, better access to healthcare, more consistent exercise habits, and healthier diets than non-drinkers. These lifestyle differences were not always controlled for in early research. The apparent health advantage of moderate drinkers may have reflected who they were, rather than what they drank.

The Mendelian Randomization Studies

A methodological advance called Mendelian randomization helped clarify the picture further. Instead of asking people how much they drink and then tracking their health outcomes, Mendelian randomization studies use naturally occurring genetic variants associated with how much alcohol a person tends to consume as a proxy for exposure. Because these genetic variants are assigned at conception and cannot be changed by lifestyle choices or illness, they sidestep many of the causation problems that plagued earlier observational work.

Across multiple Mendelian randomization analyses, the apparent cardioprotective signal from moderate drinking did not hold up. Genetically predicted higher alcohol consumption was not associated with reduced cardiovascular risk and was instead linked to adverse outcomes, including elevated blood pressure (Holmes et al., 2014; Millwood et al., 2019). A large study of cardiovascular risk factors using this method concluded that the J-shaped curve observed in conventional epidemiology does not appear to reflect a causal cardioprotective effect of alcohol.

This does not mean every observational study on the topic was wrong. The J-shaped pattern is real. What the newer methods suggest is that the pattern was an artifact of the way those studies were designed, not evidence that moderate drinking protects the heart.

Why Alcohol is Labeled as a Carcinogen

While the cardiovascular evidence was being reexamined, the cancer evidence was accumulating in a different direction. The International Agency for Research on Cancer, which is part of the WHO, classified alcoholic beverages as Group 1 carcinogens in 1988. Group 1 is the highest risk category and includes asbestos, tobacco, and ionizing radiation. Ethanol and acetaldehyde, the primary metabolites produced when the body breaks down ethanol, are both classified as carcinogenic to humans.

When ethanol is metabolized, it's converted to acetaldehyde. Acetaldehyde is genotoxic because it reacts directly with DNA, forming adducts and causing mutations. Additional damage occurs through oxidative stress, as alcohol metabolism generates reactive oxygen species that can damage DNA, proteins, and lipids. Alcohol also interferes with folate absorption and one-carbon metabolism, impairs immune function, disrupts gut microbiome composition, and raises estrogen levels, which is relevant to breast cancer risk (NCI, 2024).

Alcohol causes at least seven types of cancer: oral cavity, pharynx, larynx, esophagus, liver, colorectum, and female breast. Globally, an estimated 741,300 new cancer cases in 2020 were attributable to alcohol consumption, representing approximately 4.1% of all new cases that year.

The WHO's 2023 statement specifically noted that half of all alcohol-attributable cancers in the European Region are caused by light and moderate consumption, not heavy drinking. This was a direct challenge to the assumption that cancer risk was only meaningful at high intake levels. Breast cancer risk in particular increases at drinking levels as low as one drink per day.

Why No Threshold Has Been Established

The absence of a safe threshold is not a rhetorical move. Unlike some toxicological exposures where the body can fully neutralize a low dose before harm accumulates, acetaldehyde interacts with DNA even at low concentrations. As the WHO's 2023 statement puts it, currently available evidence cannot indicate the existence of a threshold at which the carcinogenic effects of alcohol switch on. That means that the risk does not begin above a certain intake level. It begins with any intake.

Individual susceptibility varies. People with genetic variants affecting alcohol metabolism, particularly variants in the ALDH2 enzyme common in East Asian populations, accumulate acetaldehyde more rapidly and face substantially elevated cancer risk when they drink (Brooks et al., 2009). But the underlying mechanism, DNA damage through acetaldehyde exposure, operates across populations regardless of these differences.

What This Means for People Who Train

The shift in public health messaging does not mean that anyone who has a drink with dinner is in immediate danger. What it means is that the old framework, under which moderate drinking was characterized as essentially neutral or possibly beneficial, is no longer supported by the best available evidence.

For people focused on fitness and long-term health, the cancer risk from alcohol exists on a continuum and compounds over time. It does not switch on at some clear intake threshold. Reduction lowers risk. Cessation lowers it further. The IARC evidence review confirms that reducing or stopping alcohol consumption does reduce cancer risk for several alcohol-attributable cancer types, though the degree of risk reversal varies by cancer site and duration of exposure (IARC, 2024).

None of this requires a binary choice. But it does require that conversations about alcohol and health be grounded in what the current evidence actually shows rather than what the evidence appeared to show thirty years ago. The J-shaped curve was never a license to drink. It turned out to be a measurement artifact. The carcinogenic mechanism of ethanol, however, has held up across decades of biological and epidemiological investigations.

Where the Evidence Stands Now

The core findings are as follows. Alcohol is a Group 1 carcinogen. It causes at least seven types of cancer through well-characterized biological mechanisms. The risk increases with consumption and lacks a clearly established lower threshold. The cardiovascular protection once attributed to moderate drinking has not been confirmed by more rigorous study designs. And the majority of alcohol-attributable cancers occur not in heavy drinkers, but in people who drink at levels once considered moderate and safe.

That is the evidence base these updated guidelines are built on. The question of what any individual chooses to do with that information is a personal one. But the information itself is no longer ambiguous.

Copyright 2026 The Arena Group, Inc. All Rights Reserved

This story was originally published July 8, 2026 at 7:34 AM.