Researchers sorted people by the foods they said they liked, not the foods they ate. The sweet-tooth group went on to have 79% more dementia

Most diet research asks what you ate. It is a hard question to answer honestly. People forget, round down the biscuits and round up the vegetables, and a week of food diaries may not look much like the other fifty-one. A paper published on 7 October 2026 in the journal GeroScience tries a different question, one that is easier to answer and harder to fudge: what do you like?

The data come from UK Biobank, the long-running British study that has followed hundreds of thousands of volunteers through their health records. Part of that group filled in a food preferences questionnaire, rating how much they liked a long list of foods and drinks. The researchers, based in Hong Kong, Beijing, Weifang, Nanjing and Notre Dame in the United States, ran those answers through a statistical method called latent profile analysis. It looks for clusters of people who answer in similar ways, without being told in advance what the clusters should be.

Three profiles came out. The authors call them health-aligned, broad-preference and sweet-tooth. The names describe a pattern of liking, and nobody in the study was labelled by how much sugar they actually ate. An earlier analysis of the same questionnaire by a team at the University of Surrey, published in 2024, found three very similar groups among more than 180,000 volunteers: one that liked fruit and vegetables and was cool on sweet and animal-based foods, one that liked nearly everything, and one with a strong liking for sweet foods and sweetened drinks.

The new paper then followed people forward through their medical records. Compared with the health-aligned group, the sweet-tooth group had a 79% higher risk of dementia from any cause, a 62% higher risk of Alzheimer's disease and a 202% higher risk of vascular dementia, which works out at roughly three times the rate. Vascular dementia is the kind caused by damage to the blood supply of the brain, so it is the type you would expect to track most closely with diet, blood sugar and blood pressure. That the largest figure lands there fits the rest of what is known, although it does not prove anything on its own.

The researchers also looked at three measures of biological ageing. Two are scores built from routine blood tests and body measurements that estimate whether a body looks older or younger than its birth certificate. The third is the length of telomeres, the protective caps on the ends of chromosomes, measured in white blood cells. On all three the sweet-tooth group came out worse: older on both scores and with shorter telomeres.

The obvious next step is to ask whether faster ageing is the route from a sweet tooth to dementia. The answer in this study is: barely. The ageing markers statistically accounted for between 2.28% and 4.82% of the association. More than 95% of the link runs through something these markers do not capture. That could be blood sugar, weight, blood vessels or sleep, or it could be something that has nothing to do with food.

Genes mattered too. People with a high inherited risk of dementia who also had a sweet-tooth profile did worse than you would predict by adding the two risks together. The authors read this as a reason to aim dietary advice at people with a family or genetic risk.

Now the cautions, and there are several. This is an observational study. People with a sweet-tooth profile may differ from the others in income, education, smoking, exercise, weight and mood, and no statistical adjustment removes all of that. The earlier Surrey analysis found that the sweet-tooth group also had more depression, diabetes and stroke, each of which is linked with dementia in its own right. There is also a question of which came first. Changes in taste and a growing preference for sweet food are recognised in some forms of dementia, sometimes years before a diagnosis, so some of the liking could be an early sign of the illness and not a cause of it.

A note on what we could check. The full paper sits behind the journal's paywall, and the published abstract gives the percentages above but not the number of participants, the number of dementia cases, the length of follow-up or the confidence intervals. We have left those out instead of guessing. A relative figure such as 79% also says nothing about how common dementia was in either group; in studies of middle-aged and older volunteers the absolute numbers are usually small.

So what can a reader take from this? A liking for sweet things is not a diagnosis, and nobody chooses their palate. Still, what you like is a fair guide to what you will eat over forty years, and this study adds dementia to a list that already includes diabetes and stroke. Tastes do shift with habit. People who cut back on sugar in tea or coffee often find, a few weeks later, that the old amount tastes too sweet. If dementia runs in your family, that small retraining may be worth the effort.