A single blood test for tau predicted dementia 11 years later better than lifestyle checklists

A blood test measuring a form of the protein tau picked out people likely to have dementia more than a decade later, and did so better than any of eight questionnaire-style risk scores built from lifestyle and health factors, a Norwegian study reports. Combining the blood test with a risk score helped a little more. It is a preprint, not yet peer-reviewed, and like all such studies it predicts risk in groups; it does not diagnose individuals.

In Alzheimer's disease, two proteins pile up in the brain: amyloid in plaques between nerve cells and tau in tangles inside them. A modified form called pTau217 leaks into the blood in small amounts and rises with these changes, often years before symptoms. A second blood marker, NfL (neurofilament light), is released when nerve fibres are damaged, whatever the cause.

The other approach is a dementia risk score: a tally of known risk factors from someone's behaviour, heart and metabolic health, mental health and background. Several such scores exist and are used in prevention research. Nobody knew how they compare with the new blood tests, or whether the two tell you the same thing.

What they did

• The team used the HUNT Study, which has followed the population of a Norwegian county for decades. Blood samples stored in 2006–2008 were measured for pTau217 and NfL in 3,935 people, aged 58 to 91 at the time.
• About 10.7 years later, all of them had a full assessment at age 70 or older. A panel of doctors diagnosed dementia without knowing the blood results.
• The samples were picked to include many people who went on to develop dementia: 28% had it at follow-up.
• From the 2006–2008 data, they calculated eight risk scores, including CAIDE, LIBRA, the UK Biobank score and one based on the Lancet Commission's list of risk factors.
• They then checked how well age, sex and education alone, or together with each marker or score, separated those who later had dementia from those who didn't. This is measured by the AUC, a number between 0.5 (a coin toss) and 1 (perfect).

What they found

• Age, sex and education alone gave an AUC of 0.747. Age was the strongest single predictor.
• Adding pTau217 raised it to 0.773, the biggest gain of anything tested.
• Adding a risk score on top raised it a little more, to between 0.775 and 0.783. The gains were small, though statistically significant for every score except CAIDE.
• NfL added nothing once pTau217 was known.
• Together, they sorted people into risk groups. People with both a high risk score and high pTau217 had roughly 5 to 11 times the odds of dementia of those low on both, depending on the score and cut-off used. In the highest-risk groups, more than half had dementia at follow-up; in the lowest, about one in six.
• It worked better in some groups than others: better in women and in carriers of APOE4, the best-known genetic risk variant for Alzheimer's, and much worse in people under 65 at the start (AUC around 0.62–0.66).

What to keep in mind

• Not yet peer-reviewed.
• A selected group. Only people who survived and took part in the follow-up at 70+ were included, and the sample was deliberately enriched with dementia cases. The team corrected for this statistically and got similar results, but say some bias can't be ruled out, including from people who died before follow-up.
• Older adults only. The youngest were about 58 at the start, so the results say little about midlife.
• Imperfect scores. Not every risk score could be rebuilt exactly; some parts were approximated or left out.
• Dementia type uncertain. Diagnoses were clinical, without brain scans or spinal fluid tests, so Alzheimer's and other dementias may be mixed up.
• One blood draw. The blood wasn't taken fasting, and only one baseline NfL value was used; the team suggests changes in NfL over time might predict better.
• Small differences between scores. The authors warn against reading much into which risk score came out on top.

Why it matters

Blood tests for Alzheimer's-related proteins are becoming practical, and this study shows that a single pTau217 measurement carries warning signs more than ten years ahead. The risk scores matter in a different way: they barely improved prediction overall, but among people with high pTau217, those with lower risk-factor scores were less likely to have dementia. The authors read this as a sign that blood markers and risk factors capture different parts of the picture: the disease process on one side, a lifetime of exposures on the other. Whether lowering those risk factors changes the outcome for people with high pTau217 is a question this kind of study can't answer.