Older Germans with early memory problems drank a third of a drink less a day after a year of online diet classes that barely mentioned alcohol. There was no group without classes to compare them with

Mild cognitive impairment is the in-between stage: memory and thinking are measurably worse than expected for a person's age, but daily life still runs on its own. Some people with it go on to develop dementia, some stay the same and some improve. Heavy drinking is one of the 14 risk factors for dementia that the Lancet Commission counts as changeable, which makes people at this stage an obvious group to help cut down. A team in Erlangen and Berlin asked whether general nutrition advice, not an alcohol programme, would do that on its own.

The people came from BrainFit-Nutrition, a German trial that screened 1,111 adults aged 60 and over in 2022 and randomised 271 with mild cognitive impairment. Everyone did a computer brain-training programme and one of two online nutrition courses: twelve 90-minute group video sessions, one every two weeks. One course taught a whole-food, plant-based diet; the other taught the standard recommendations of the German Nutrition Society, which allow meat and fish. This new analysis, published in BMC Public Health on 3 October 2026, looks only at the 189 who drank at the start and went to at least one session. Their average age was 70.6, just over half were women, and 60% had a high level of education.

Alcohol was not the subject of either course. It came up in a single session on oils, fats, nuts, seeds and drinks. Both courses told people that alcohol harms the brain and that the best amount is none; the plant-based course also gave grams of alcohol and calories for common drinks and limits of 10 g a day for women and 20 g for men. Oddly for 2022, both also said that anyone who did drink should drink in moderation and preferably red wine.

On the numbers people gave in telephone interviews, average drinking fell from 1.61 standard drinks a day (a standard drink here is 10 g of alcohol, roughly a small glass of wine) to 1.28 a year later. Both courses saw a fall: from 1.52 to 1.27 in the plant-based group and from 1.70 to 1.29 in the standard-advice group. The difference between the two courses was small enough to be chance. Men cut down more than women, and women in the plant-based course barely changed (1.46 to 1.35, not statistically significant).

Here is the problem. Because the trial was designed to compare two diets, every person in this analysis got nutrition classes. There was no group that got nothing, so there is no way to know how much drinking would have fallen anyway. Several ordinary forces push the same direction. People often join a health study at a moment when they are already worried and already changing. Drinking tends to fall with age. Some people drink less as their memory slips, which the authors themselves mention. And people asked twice about a habit they have just been warned about may report less the second time, especially with no blood test to check.

One finding points straight at that. Among the 84 people who did cut down (fewer than half of the 189), the strongest predictor of how much they cut was how much they had drunk at the start. That fits the authors' explanation, that heavier drinkers had more to gain from the advice. It is also exactly what regression to the mean looks like: people who report an unusually high figure once tend to report a lower one next time, with or without help. Attending more sessions went with a bigger cut, which is the most encouraging sign for the classes themselves, but keener attenders may simply be keener to change.

The authors are open about most of this. They call the study an exploratory secondary analysis, note that the brain-training arm could have muddied the results, and say a future study needs an untreated group. Their conclusion that the German Nutrition Society advice 'should be followed to ensure maximum benefit to both genders' leans on differences between the courses that their own tests could not separate from chance.

Our reading: a year after general healthy-eating classes, a group of older adults with early memory problems said they were drinking about a third of a drink a day less, and nobody can yet say how much of that was the classes. The size of the fall, about two to three standard drinks a week, is in the same range as digital programmes built specifically to cut drinking, which is a reason to test this properly rather than a result in itself. Even afterwards the average was still more than one drink a day. If you or someone close to you has memory concerns and drinks most days, cutting down is reasonable advice on its own merits, and a doctor is a better place to start than a red wine.