Risk factors tracked how sharp people were, but not how fast they declined

Large expert reviews estimate that a big share of dementia cases might be preventable by tackling risk factors such as little schooling, diabetes, hearing loss or physical inactivity. But most of that evidence comes from older people or from patients in clinics, and it is about getting a diagnosis. Whether the same factors shape the ordinary, slow change in thinking that almost everyone goes through with age is much less clear.

A new study, posted as a preprint (not yet peer-reviewed), separates two questions that are often mixed up: how well someone does on thinking tests at a given age (their level), and how fast their scores fall over the following years (their rate of decline). The answer was that different things were linked to each.

What they did

• Who: 1,396 adults aged 36 to 108 from four US sites, all living at home and free of diagnosed neurological or psychiatric disease when they joined. They were very well educated: on average 17.3 years of schooling.
• Follow-up: 915 people were tested more than once, over an average of five years (one to eight).
• Tests: the main outcome combined five tests of memory, attention and speed into one score; the team also looked at separate memory and "fluid" (reasoning and processing speed) scores.
• Risk factors: 18 in total. Twelve were modifiable ones taken from a 2024 expert list, such as education, blood pressure, LDL cholesterol, diabetes, obesity, alcohol, loneliness and head injury. Motor function (grip strength and walking speed), hearing and vision were measured with tests and compared with people of the same age, rather than asked about.
• Blood markers: in 928 people, the team measured three proteins in the blood that rise with brain changes: pTau217, a marker of Alzheimer's-type tau changes, and GFAP and NfL, signs of stress or damage in brain cells. A pTau217 level above a set cut-off was counted as "Alzheimer's-biomarker positive".

What they found

• Age mattered most. Scores were lower in each older age group, and the drop sped up after 70. Within people, yearly decline on the main score grew from −0.013 standard deviations (a measure of how spread out scores are) at ages 36–49 to −0.140 at 80 and over.
• Five factors were linked to level. More education went with higher scores; diabetes, weaker motor function, poorer hearing and poorer vision went with lower ones. Together, the modifiable factors explained 8 to 13 percentage points of the differences in scores, beyond age and sex.
• Others were not. High blood pressure, high LDL cholesterol, obesity (as a yes/no category), heavy drinking, a tendency to negative emotions, loneliness and head injury were not linked to scores. Neither was carrying APOE4, the best-known gene variant for Alzheimer's risk.
• No factor was linked to the rate of decline. Using four different statistical approaches, none of the modifiable factors was linked to how fast scores fell over about five years.
• The blood marker was. 12.4% of those tested were pTau217-positive, rising to 32.6% at 80 and over. They declined faster on all three scores, by an extra 0.058 standard deviations a year on the main score. Blood markers were linked to test scores only from age 65 on, not in younger people.
• Not explained by hidden Alzheimer's. In the 813 people who were pTau217-negative, the links between the five factors and test level looked almost the same.

What to keep in mind

• The authors' own caveats. The group was highly educated, so the education comparison is mostly college versus postgraduate degrees. Most findings are cross-sectional (one point in time), so they show associations, and reverse causation is possible: poorer thinking could lead to some of these factors rather than the other way round.
• Short follow-up. About five years is short for cognitive aging, and the participants were healthy at the start. The team notes that people getting used to the tests and the dropout of older, less healthy participants could hide real links with decline. Longer follow-up may find some.
• Blood markers came late. For most people the blood sample was taken later, on average 3.1 years into the study, so the marker results are not true predictions. Alzheimer's changes were not confirmed with brain scans or spinal fluid, and other brain diseases were not measured.
• Missing pieces. Smoking and air pollution could not be analysed, race and ethnicity were not examined, and many tests were run, so borderline results need repeating.
• Interests. Two of the authors are named on a patent application related to blood biomarker tests.

Why it matters

A risk factor that goes with lower scores at one point in time is often read as a sign that changing it would slow decline. This study suggests the two can come apart: in healthy adults, factors such as education, hearing, vision and motor function were linked to how sharp people were, while the speed of decline was linked to early Alzheimer's-type changes. The team stresses that this does not mean the modifiable factors are unimportant, only that showing an effect on decline needs longer studies that measure change directly. The cohort's follow-up is continuing.