News · Nutrition & Diet
Cutting alcohol in midlife showed no breast cancer benefit yet
EPIC researchers tracked drinking changes in 123,679 European women. Four years after the change, postmenopausal breast cancer risk had not moved either way.
Based on a peer-reviewed prospective cohort study in the European Journal of Nutrition
- Researchers compared drinking at recruitment with drinking a median 9.8 years later in 123,679 women in the European Prospective Investigation into Cancer and Nutrition, publishing in the European Journal of Nutrition.
- Over a median 4.0 years after that second measurement, 2,173 postmenopausal breast cancers were diagnosed.
- Each 10 g/day change in alcohol intake showed no association with risk (HR, 0.97; 95% CI, 0.93-1.01).
- Nothing appeared by tumor subtype either, across estrogen receptor, progesterone receptor and HER2 categories.
- This does not contradict the established finding that alcohol causes breast cancer. The authors state that outright in their opening line.
- The follow-up after the change was short, a median of four years, which is brief for a cancer that takes years to develop.
- Observational, with alcohol intake self-reported at two points and missing data filled in statistically.
That alcohol causes breast cancer is settled. What nobody had properly tested is whether drinking less, later in life, buys anything back. A study in the European Journal of Nutrition ran that test on 123,679 women and came up empty, which is a more useful result than it first sounds.
The authors set the scene in a single line: alcohol consumption is a cause of breast cancer, yet the association between changes in alcohol consumption during adulthood and the risk of BC has been examined little. BC is the paper’s shorthand for breast cancer.
Measuring a change, not a level
Nearly every study on this topic measures how much someone drinks once and follows them. This one measured twice.
Within the European Prospective Investigation into Cancer and Nutrition cohort including 123,679 women, changes in alcohol intake were obtained by comparing middle-aged participants’ alcohol intake assessed at recruitment and during follow-up, 9.8 years (median) later.
Ten years apart is a long enough gap for real change. Some women drank more, some less, some the same. The question was whether the direction of travel predicted what happened next.
What the cancers showed
Then the counting started. During a median follow-up time of 4.0 years after the follow-up assessment, 2,173 cases of postmenopausal BC were diagnosed.
The primary comparison was per 10 grams of alcohol a day, roughly one standard drink, added or removed.
No associations were observed between alcohol changes and BC risk. The estimate landed at 0.97 with a range from 0.93 to 1.01, which sits on top of no effect rather than near it.
The researchers also split cancers by their hormone receptors, since alcohol’s suspected mechanism runs partly through estrogen. If the effect were real and fast, receptor-positive tumors should have moved first. Nothing appeared in any subtype.
Why a null result here is not a green light
The obvious misreading is that alcohol turns out not to matter. The study does not support that, and the authors take care to say so before presenting anything.
The far more likely explanation is time. Breast cancer takes years to develop from the first cellular changes to a diagnosis. Cancers found within four years of a woman changing her drinking were, in many cases, already in progress before she changed anything.
Put differently, this study measured whether a late change alters a process that mostly ran earlier. Finding no effect over four years says something about the timescale, not about whether alcohol matters.
The other limits
Alcohol was self-reported at both points, and self-reported drinking is consistently under-stated. If misreporting shifted between the two questionnaires, some real change would be smoothed away.
Missing information about follow-up alcohol intake was filled in statistically rather than measured. That is standard practice in cohorts this size, and it is still a modeling assumption.
And this is observational. Women who cut down may differ from women who did not in ways no questionnaire captures, including reasons for cutting down that are themselves health-related.
What to take from it
The conclusion is stated plainly: changes in alcohol consumption during midlife were not associated with the risk of postmenopausal BC, either overall or by hormonal receptor status.
That is worth knowing, and it is worth not over-reading. Cutting back in your fifties does not appear to visibly shift breast cancer risk in the following few years. Whether it shifts it over 15 or 20 years, this study cannot say, because nobody has followed anyone that long after a measured change.
The advice from cancer agencies has not moved, and it rests on more than one endpoint. What this paper adds is a realistic expectation about how quickly the benefit of drinking less should be expected to show up in the statistics, which is: not soon.
People also ask
Does this mean alcohol does not cause breast cancer?
No, and the paper says the opposite in its first sentence: alcohol consumption is a cause of breast cancer. That conclusion rests on decades of consistent evidence and a plausible biological mechanism. This study asked a narrower question, which is whether changing how much you drink in midlife shifts risk within the following few years. It found no detectable shift. Those are different questions with different answers.
Why would cutting down not show a benefit?
Timing is the leading explanation. Breast cancer develops over many years, and the cancers diagnosed in this window were plausibly already underway before anyone changed their drinking. With a median of four years between the second measurement and the end of follow-up, there may simply not have been enough time for a change in exposure to alter what gets diagnosed.
What does a hazard ratio of 0.97 mean here?
It means the estimate sat essentially on top of no effect, and the range of uncertainty around it, 0.93 to 1.01, comfortably includes no effect. This is a genuine null result rather than a hint of benefit that missed a threshold. With 123,679 women and 2,173 cases, the study had enough size to detect a modest effect if one had been there over this timeframe.
Is this study large enough to trust?
The size is a real strength. EPIC is one of the largest nutrition cohorts in the world, spanning multiple European countries, and 123,679 women is far beyond what most dietary studies manage. The weaknesses are elsewhere: alcohol was self-reported, which people tend to underestimate, and missing values were filled in statistically rather than measured.
So should anyone bother cutting back?
This is general information rather than medical advice, and one study on one timeframe does not overturn the wider evidence. Major cancer agencies still advise limiting alcohol, and their reasoning covers more than breast cancer, including liver disease, several other cancers and cardiovascular risk. What this paper suggests is that the breast cancer benefit specifically may take longer than four years to appear, not that it never does.
References
- Antoniussen CS, Ibsen DB, Olsen A, et al. Changes in alcohol consumption and the risk of postmenopausal breast cancer in the European Prospective Investigation into Cancer and Nutrition cohort. European Journal of Nutrition (2026).
- National Cancer Institute. Alcohol and Cancer Risk.
- Centers for Disease Control and Prevention. Breast Cancer Risk Factors.