Senescence Protein Score Linked to Higher Mortality

Evidence

Higher scores on a protein-based measure linked to cellular senescence—a state in which cells stop dividing—were associated with higher mortality among 7,639 UK Biobank participants.

The senescence-protein score study used deep learning to combine biologically selected senescence-associated secretory phenotype proteins into a single score. These proteins are associated with substances released by senescent cells. Researchers developed the score using UK Biobank Pharma Proteomics Project data.

Over a mean follow-up of 13.3 years, each one-standard-deviation higher score—a standard unit describing variation between participants—was associated with a 41% higher adjusted hazard of death. Hazard describes the rate of death during follow-up, not the absolute probability of dying. The hazard ratio was 1.41, with a 95% confidence interval of 1.25–1.60, after adjustment for baseline characteristics.

The score also correlated negatively with Healthy Physical Status, a measure scored from 0 to 1: the correlation was −0.63, meaning higher protein scores tended to accompany lower physical-status scores. Higher scores were also significantly associated with the incidence of chronic kidney disease, heart failure, lung cancer, delirium, dementia and Parkinson’s disease.

In a separate setting, the hazelnut-extract trial randomly assigned 60 hospitalized female Holstein calves with neonatal diarrhea requiring intravenous therapy to two groups of 30. Both received standardized care, including fluids and antimicrobials. One group additionally received 10 grams of hazelnut skin extract daily with milk for five days.

By day 6, four of 30 calves had died in each group, leaving 26 survivors per group. Median hospitalization was eight days with standard care and nine days with added extract. The middle half of hospital stays ranged from 7–11 days and 8–12 days, respectively. The hospitalization difference was not statistically significant.

Analysis

Interpretation: The protein findings suggest that the combined score captures a pattern associated with future mortality and several health outcomes. They do not show that the measured proteins cause those outcomes, that reducing the score improves health, or that the score is clinically useful.

The calf trial addresses a different question: whether adding an extract improves outcomes under a specific treatment protocol. Its reported results support no detected short-term added benefit in that setting. These are separate single-study findings involving different species, measurements and outcomes; they do not confirm one another or support a shared biological conclusion.

Limitations

The protein associations remain observational despite baseline adjustment. Dependence on the original protein-measurement platform may limit transfer to other settings. The protein-score briefing does not provide individual disease effect estimates or sample sizes for cohorts beyond the test sample, limiting assessment of those additional findings and the reported external evaluation.

For the calf trial, 30 animals per group and five days of extract administration leave uncertainty about undetected differences and longer-term effects. Reddish fecal discoloration prevented adequate blinding, potentially introducing observer bias. Equal observed death counts do not establish equivalence, and the findings should not be extended beyond the selected hospitalized calves and treatment protocol.

What to watch

For the protein score, mortality-prediction performance in a clearly sized external cohort using a different protein-measurement platform would test transferability. For the calf intervention, confidence would change with a larger trial using blinded outcome assessment and confidence intervals showing whether mortality or hospitalization benefits are present—or excluding a prespecified, clinically meaningful benefit.