Which Aging Biomarkers Best Predict Who Dies Early? A 16-Year Study Answers
A 16-year mortality study pits proteomic organ clocks against epigenetic, brain, and physical biomarkers — revealing what actually predicts death.
Summary
Scientists compared many ways of measuring biological aging to see which best predicts who dies sooner. They followed 861 Scottish adults from age 73 for 16 years, during which 444 died. Blood-protein clocks estimating the age of organs like the liver, immune system and heart did predict death. But an epigenetic clock (GrimAge2), smaller total brain and gray matter volume, weaker lung function and poorer thinking skills predicted it more strongly. The team also screened nearly 9,700 blood proteins and found 368 linked to mortality. The strongest were GDF15, WFDC2 and TIMP1. Proteins tied to higher risk were mostly immune-related, while protective ones were involved in DNA stability and cell maintenance.
Detailed Summary
Many tools now claim to measure biological age, but there is no consensus on which best predicts health outcomes. Recent plasma proteomic clocks estimate the age of 11 organs from blood proteins and have been linked to higher mortality. However, nobody had tested them head-to-head against established aging biomarkers in the same people. This matters for deciding which measures are worth using for risk stratification.
Researchers analyzed the Lothian Birth Cohort 1936: 861 participants (445 men, 416 women) with a mean age of 72.5 at baseline. Over about 16 years of follow-up, 444 died. They compared the 11 proteomic organ clocks with the GrimAge2 epigenetic clock, leukocyte telomere length, grip strength, 6-meter walk time, lung function, brain MRI measures and general cognitive function. Cox regression models were adjusted stepwise for age and sex, then smoking, alcohol and BMI, then seven chronic conditions. P-values were FDR-corrected.
Among organ clocks, accelerated liver aging had the strongest link to mortality (hazard ratio per standard deviation 1.43, 95% CI 1.30–1.58), followed by immune (1.42, 1.29–1.57) and heart (1.38, 1.25–1.53). Several other markers did better. Older epigenetic age, smaller total brain and gray matter volumes, reduced respiratory function and poorer cognition had hazard ratios of 1.44–1.62 per standard deviation. So the organ-specific proteomic clocks were informative but not superior to existing measures.
The second aim was to screen 9,703 plasma protein targets (after quality control from 11,083 on the SomaScan 11K platform) individually for mortality associations. 368 proteins were associated with mortality. The top three were GDF15 (HR 1.56, 1.42–1.72), WFDC2 (1.47, 1.33–1.62) and TIMP1 (1.45, 1.29–1.62). Enrichment analysis showed that higher-risk proteins were concentrated in immune functions. Protective proteins were involved in genomic stability and cellular maintenance. The authors also ran multivariable models with 21 biomarkers and an elastic net-penalized Cox regression across all proteins.
For practice, the results suggest that cheap functional measures such as lung function and cognition, plus brain imaging and epigenetic clocks, carry substantial prognostic information. Organ-specific proteomic clocks may add mechanistic detail about which systems are aging fastest. The protein hits point to inflammation, immune aging and weakened cellular maintenance as targets for further study.
Caveats: this is an observational study in a single cohort of people born in Scotland in 1936. Baseline was about age 73, so the findings may not apply to younger adults or other populations. Hazard ratios show association, not causation. Organ clocks were trained in a different cohort, and disease status was self-reported. Some aptamers were removed because of batch differences between waves.
Key Findings
- Cohort of 861 adults (mean age 72.5 at baseline) followed ~16 years, with 444 deaths
- Accelerated liver aging was the strongest proteomic organ clock for mortality (HR per SD 1.43, 95% CI 1.30–1.58)
- Immune (HR 1.42) and heart (HR 1.38) organ-age acceleration were next-strongest among proteomic clocks
- Epigenetic age, smaller total brain and gray matter volumes, reduced respiratory function and poorer cognition outperformed organ clocks (HR per SD 1.44–1.62)
- Of 9,703 plasma protein targets, 368 were associated with all-cause mortality (FDR < 0.05)
- Top mortality proteins were GDF15 (HR 1.56), WFDC2 (HR 1.47) and TIMP1 (HR 1.45)
- High-risk proteins were enriched for immune functions; low-risk proteins for genomic stability and cellular maintenance
Methodology
Prospective cohort analysis of the Lothian Birth Cohort 1936 (861 participants, 444 deaths, ~16-year follow-up from wave 2, mean age 73). Plasma proteins were measured with SomaScan 11K (9,703 aptamers after quality control), and organ clocks were derived from published Oh et al. models. Cox proportional hazards models gave hazard ratios per SD, with three levels of adjustment (demographics; lifestyle; chronic disease), a 21-biomarker multivariable model, and Nagelkerke pseudo-R2 comparison. Proteins were also analyzed with elastic net Cox regression (alpha 0.5, lambda 0.16) and pre-ranked gene set enrichment, with Benjamini-Hochberg FDR correction.
Study Limitations
The study is observational and based on one cohort of Scottish-born adults followed from about age 73, so generalizability to younger or more diverse populations is uncertain. Organ clocks were trained in a separate cohort, some aptamers were removed because of wave-to-wave variation, and disease status was self-reported. Conflicts of interest were not stated in the text provided.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe:
