Does Alcohol Cause Cancer? Why the Genetic Evidence Disagrees

Alcohol is classified as a carcinogen, governments are moving towards cancer warnings on bottles, and fewer than half the public know the link exists. Then in December 2025 a study using genetic data across four biobanks failed to find the association for two of the cancers the policy case rests on most heavily. Both of those things are true, and the gap between them is the interesting part.

TL;DR

• The observational case is well established. Alcohol is linked to at least seven cancers, and around 17,000 UK cancer cases in 2020 — about 4.1% of the total — were attributed to drinking.
• Policy is moving. The US Surgeon General called for cancer warning labels in January 2025. Ireland legislated for them, then deferred the requirement to September 2028. England has promised mandatory health warnings without confirming whether cancer will be among them.
• A December 2025 genetic study found no association with cancer overall across four biobanks and around 1.5 million people.
• It did support three of the seven: head and neck cancer most clearly, with weaker signals for oesophageal and colorectal.
• It found no evidence for breast cancer and an inconsistent picture for liver — which matters, because breast and colorectal account for over half the UK's alcohol-attributable cases.
• Whether stopping drinking lowers risk is a separate question. IARC found sufficient evidence that cutting down lowers oral and oesophageal cancer, but only limited evidence for breast and colorectum — which reflects how hard cessation is to study, not doubt about cause.
• A 2026 review of 88 genetic studies disagreed, classing the breast cancer evidence as robust — though it could not pool its studies, because they draw on overlapping samples.
• Neither method settles it. The genetic study also produced unexplained inverse findings for several cancers, and its authors list five substantial limitations.

Where the policy stands

On 3 January 2025 the US Surgeon General issued an advisory on alcohol and cancer risk, naming at least seven cancers causally linked to drinking. It attributed roughly 100,000 cancer cases and about 20,000 cancer deaths a year in the United States to alcohol, and called for cancer warnings to be added to alcohol labels and for daily drinking limits to be reassessed.

Ireland legislated for exactly that, with labels stating that alcohol is directly linked to cancer due to commence on 22 May 2026 — then deferred the requirement to 3 September 2028, citing economic circumstances. England has committed to mandatory health warnings on alcoholic drinks, currently voluntary, but has not confirmed whether cancer will be among them.

What the observational evidence says

The case for those warnings rests on a large body of population research, and on the share of cancer cases the International Agency for Research on Cancer attributes to drinking. For the UK, House of Commons Library figures put alcohol-attributable cancers at around 17,000 new cases in 2020, roughly 4.1% of all cancers. The distribution matters for what follows: colorectal cancer accounts for about 6,100 of those and breast cancer about 4,100, together more than half the total.

This is the evidence that has driven forty years of public health policy, and nothing below overturns it. It is also the kind of evidence that carries known weaknesses: it compares people who drink with people who don't, and the two groups differ in many other ways — in smoking above all, but also in diet and deprivation.

What the genetic evidence adds

Mendelian randomisation is an attempt to get round that problem. Because the gene variants that influence how much people drink are allocated at conception and aren't affected by lifestyle, comparing cancer rates across people with different genetic propensities to drink should be less vulnerable to confounding than comparing drinkers with non-drinkers.

A study published in BMC Medicine in December 2025 applied that method at scale, using 95 genetic variants associated with alcohol consumption and testing them against 20 cancers. The primary analyses drew on UK Biobank (367,643 people), FinnGen (500,348), All of Us (169,312) and the Million Veteran Program (451,206), with additional cancer-specific consortium data. The headline result was null: genetically-predicted alcohol consumption showed no association with cancer overall, at an odds ratio of 0.96 (p = 0.45). Within the individual cancers:

• Head and neck cancer showed a clear positive association, significant after correction for multiple testing (OR 1.51, p = 0.001).
• Colorectal (OR 1.21, p = 0.035) and oesophageal cancer (OR 1.42, p = 0.033) showed positive but only nominally significant estimates.
• Liver cancer was null overall (OR 1.40, p = 0.10), though positive in one cohort.
• Breast cancer showed nothing, in either the biobank data (OR 1.09, p = 0.25) or the consortium data (OR 0.98, p = 0.84), the latter drawn from 133,384 cases.
• Kidney (OR 0.64) and endometrial cancer (OR 0.56) showed apparently protective associations, both significant after correction, with nominally inverse estimates for non-Hodgkin's lymphoma, myeloma and some ovarian subtypes.

The authors' conclusion was that the genetic data support causal effects for head and neck, oesophageal and colorectal cancer at moderate-to-weak strength, are inconsistent for liver, and provide no evidence for breast. Cancer mortality showed a nominally positive association (OR 1.44, p = 0.003) that attenuated once smoking heaviness was adjusted for.

That study is not the whole genetic literature, and the rest reads differently. A 2026 systematic review in Drug and Alcohol Dependence collected 88 Mendelian randomisation studies covering 630 alcohol–disease investigations, and classed the evidence linking alcohol to breast, colorectal, oropharyngeal and hepatocellular cancer as robust. Two things temper it. The reviewers could not pool their studies, because those studies draw on overlapping samples, so a robust rating can reflect several papers re-analysing one biobank rather than independent confirmations. And 77% of the investigations produced evidence they called insufficient or non-evaluable.

The awkward part

Line the evidence up and the difficulty is obvious. The UK estimate is dominated by colorectal and breast cancer — the two where the December 2025 genetic evidence is, respectively, only nominally significant and absent altogether — while the cancer that data supports most strongly, head and neck, contributes far fewer cases.

IARC has examined a related but different question, and the two shouldn't be run together. Its Handbooks working group assessed what happens when people cut down or stop, finding sufficient evidence that doing so lowers the incidence of oral cavity and oesophageal cancer, limited evidence for larynx, colorectum and breast, and inadequate evidence for pharynx and liver. That group did not reassess whether alcohol causes cancer in the first place, and the same document restates that alcohol is an established cause of breast cancer.

The distinction matters, because evidence that drinking causes cancer and evidence that stopping lowers incidence are not interchangeable. The second is simply harder to produce: it depends on how accurately researchers capture changes in consumption, how long people are followed, and whether other factors obscure any later shift in risk. IARC's own reasons for the thinner ratings are narrower still — for breast cancer, few studies broke results down by hormone receptor subtype; for liver, people with advanced cirrhosis who keep drinking may die before any cancer is diagnosed. Limited evidence of reversibility does not weaken an established causal classification.

Why the methods disagree

Start with the inverse findings, because they show how hard a single genetic estimate is to read on its own. The apparent protection against kidney and endometrial cancer might reflect a real difference between cancer types, residual bias, or a broken assumption in the method — the authors go as far as saying their data suggest there may be inverse associations for certain cancers. What none of it supports is drinking to avoid cancer; these are results needing replication before they mean anything.

They list five limitations: the genetic variants may not be valid instruments, the analyses had limited power, testing twenty cancers invites false positives in both directions, cancer detection rates vary between the datasets, and the datasets aren't representative of the general population. Mendelian randomisation also estimates the effect of a lifetime genetic tendency to drink somewhat more or less, which is not the same question as the effect of the drinking patterns people really have.

The observational evidence has its own weaknesses, which is why the genetic work was done at all — not least that people who have stopped drinking because they are ill get counted as non-drinkers, which flatters the comparison group. So these are two imperfect methods pointing in overlapping directions. For head and neck, oesophageal and probably colorectal cancer, both support a causal link. For breast cancer, a large body of observational evidence says yes and one large genetic analysis cannot detect it.

What This Actually Means

For an individual deciding what to drink this weekend, this changes very little: none of it offers a reason to drink more. The study found no association with cancer overall; it did not find alcohol harmless. And cancer was never the only reason to moderate — alcohol's effects on the liver, blood pressure, sleep and dependence aren't in dispute and don't depend on any of this.

What it should change is how you read the next headline. "No safe level" is accurate in its own terms — no amount has been shown to carry no cancer risk at all — but it says nothing about size, and the risk carried by an occasional drink is not the risk carried by heavy drinking. "Alcohol doesn't cause cancer" is the opposite distortion, and the same study that produced the breast cancer null produced a clear signal for head and neck cancer.

The practical reading sits where it did. UK guidance is no more than 14 units a week, spread across three or more days. If you want a liver-specific reason to drink less, our Liver Reset sets out what the evidence supports. And if you've been told you have liver disease, the advice is firmer than any of this: drink nothing.

Frequently Asked Questions

Is alcohol a carcinogen?

Yes. Alcohol is classified as a carcinogen and is linked to at least seven cancers: breast in women, colorectal, oesophageal, liver, mouth, throat and larynx. Around 17,000 UK cancer cases in 2020 were attributed to drinking.

Does alcohol cause breast cancer?

Alcohol is an established cause of breast cancer in IARC's assessment, and it accounts for roughly 4,100 of the UK's alcohol-attributable cases. A large genetic study in December 2025 did not detect that association, in either its biobank or its consortium data, which raises questions about the size of the effect and the limits of the method. A 2026 review of 88 genetic studies reported robust evidence for it. The individual studies disagree; the broader causal assessment stands.

How much alcohol increases cancer risk?

There's no agreed threshold, which is why the US Surgeon General called for daily limits to be reassessed. UK guidance remains no more than 14 units a week, spread over three or more days, as a level of low rather than no risk.

Are cancer warnings coming to alcohol labels?

In some places. Ireland legislated for them but deferred the requirement to September 2028. England has committed to mandatory health warnings without confirming cancer will be included. The US Surgeon General recommended them in January 2025.

Does the genetic study mean alcohol is safe?

No. It supported a causal link for head and neck cancer, its authors list five substantial limitations, and it produced inverse findings for several cancers that need replication before they mean anything.

The Bottom Line

Alcohol's link to cancer is established enough to justify warning labels, and an estimated 17,000 UK cancer cases in 2020 were attributable to alcohol. The genetic evidence published in December 2025 supports that for head and neck, oesophageal and probably colorectal cancer, and cannot detect it for breast cancer — one of the two largest contributors to the total. A 2026 review of the wider genetic literature reaches the opposite conclusion on breast, and IARC's causal assessment is unchanged.

That is a real tension rather than a reason to dismiss either side. What survives the argument is unchanged: drinking less reduces several risks that have nothing to do with cancer, and the honest answer to "how much is safe" is still that nobody has established one.

This article is for general information and is not medical advice. If you are concerned about your drinking or your cancer risk, speak to your GP.

Sources

• Larsson SC, Mason AM, Cronjé HT, et al. Alcohol consumption and risk of cancer: a Mendelian randomization analysis of four biobanks and consortium data. BMC Medicine, 2025;23:676. https://doi.org/10.1186/s12916-025-04543-8
• Kassaw NA, et al. A systematic review of Mendelian randomization evidence on alcohol-related phenotypes and multiple human diseases. Drug and Alcohol Dependence, 2026. https://doi.org/10.1016/j.drugalcdep.2026.113330
• IARC Handbooks of Cancer Prevention Volume 20A: Reduction or Cessation of Alcohol Consumption. International Agency for Research on Cancer. https://www.iarc.who.int/faq/iarc-handbooks-of-cancer-prevention-volume-20a-reduction-or-cessation-of-alcohol-consumption/
• Alcohol and Cancer Risk: the U.S. Surgeon General's Advisory, 3 January 2025. https://ascopost.com/news/january-2025/surgeon-general-s-report-calls-for-cancer-warning-labels-on-alcoholic-beverages/
• Alcohol and cancer. House of Commons Library research briefing, 2025. https://commonslibrary.parliament.uk/research-briefings/cdp-2025-0157/
• Alcohol labelling in Ireland: timeline for compliance deferred. Arthur Cox, 2025. https://www.arthurcox.com/knowledge/alcohol-labelling-in-ireland-timeline-for-compliance-deferred/
• UK Chief Medical Officers' low risk drinking guidelines. Department of Health and Social Care, 2016. https://www.gov.uk/government/publications/alcohol-consumption-advice-on-low-risk-drinking