A blood-protein clock followed the same people for 14 years, and its age moved with their health

Two people born in the same year do not age at the same pace. Researchers try to capture this with “aging clocks”: computer models that read some molecular pattern in the body and estimate an age from it. If a 60-year-old’s clock says 64, that person shows age acceleration. Many clocks read chemical tags on DNA; others read the mix of proteins circulating in the blood, a field called proteomics.

Most such clocks are built from a single blood sample per person. That shows who looks older than whom, but not whether a person’s clock can speed up or slow down as their life and health change. This study set out to measure that.

What they did

The team used KORA, a long-running health study of residents of the Augsburg region in southern Germany. Participants were examined up to three times, about seven years apart, between 1999 and 2014. Each time they gave blood and answered questions about their health and habits. In total, 9,105 blood samples from 4,100 people aged 25 to 88 were analysed with mass spectrometry, a technique that weighs fragments of proteins to identify and count them. This measured 842 proteins.

A type of machine-learning model (a small neural network) learned to predict each person’s age from their proteins. To keep it honest, all samples from one person were held out together when the model was tested on them. The clock’s estimates were within about 6.4 years of true age on average.

The researchers then asked two kinds of question. Between people: do those with a given risk factor have older-looking proteins? Within people: when the same person’s risk factor changes between visits, does their protein age change too? The second question is useful because each person is compared with themselves, so stable differences such as genes or upbringing cannot explain the result. They also linked the clock to deaths over up to 22 years.

What they found

• Risk factors went with an older protein age. People with diabetes looked about 2.7 years older than those without it. Higher body weight (BMI), high LDL cholesterol and smoking were also linked to older readings. Better kidney function and being physically active were linked to younger ones.
• Changes tracked changes. Within the same person, a rise in BMI, LDL cholesterol moving above the cut-off, or developing diabetes went with a higher protein age at that visit. The strongest link was for kidney function: a one-step improvement (one standard deviation) went with a reading about 1.8 years younger. Kidney function was calculated from blood tests that were not part of the clock.
• A fast clock was linked to earlier death. People whose clock ran furthest ahead for their age died sooner than those whose clock ran furthest behind. Adding the clock to true age slightly improved the prediction of who would die.
• Different habits, different proteins. Weight, blood pressure and kidney function shared many of the same proteins in the clock; smoking and physical activity each had their own distinct pattern. Many important proteins come from the liver or are involved in blood clotting and inflammation.
• Some proteins may play a causal role. Using genetic data (Mendelian randomization), 28 of the 842 proteins showed evidence consistent with an effect on aging-related traits such as frailty or longevity. Half of these were among the proteins the clock relied on most.

What to keep in mind

• Not yet peer-reviewed. This is a preprint.
• Associations, not proof. The within-person design removes some confounding, but the authors note it cannot rule out factors that change over time alongside the risk factors. Their statistical estimate that part of the extra death risk runs through faster protein aging rests on assumptions that “cannot be verified” in observational data.
• No replication yet. No other cohort with repeated measurements of this kind was available to check the within-person results.
• Less precise than some clocks. The authors say the clock is less accurate at estimating age than some larger protein or DNA clocks.
• One population. All participants were from one region of Western Europe.
• Blood proteins capture only part of what happens as the body ages.

Why it matters

A biological-age test is only useful for tracking health if it responds to change. This study suggests that a blood-protein clock does: when people’s weight, cholesterol, blood sugar or kidney function got better or worse, their protein age tended to follow. That is consistent with the idea that biological aging is not fixed and moves with risk factors people and doctors can work on. It does not show that improving them will slow aging; that would take trials. The strong link with kidney function also fits the authors’ view that the kidneys and the rest of the body age closely together. For another approach to protein-based clocks, see sex-specific organ clocks.