Background

Proving whether anti-aging compounds or lifestyle interventions increase lifespan and healthspan typically takes decades. To shorten this period, epigenetic clocks that estimate biological age from DNA methylation (DNAm) patterns have gained attention as alternatives. However, previous studies used different clocks and analytical methods, making direct comparisons difficult. It remains unestablished whether changes in clock values truly reflect changes in aging rate or mortality risk.

Early clocks such as the Horvath and Hannum models focused on predicting chronological age. Subsequent models like PhenoAge and GrimAge were designed to reflect disease and mortality risks, while DunedinPACE was developed to capture the pace of aging. Principal component-based clocks to reduce measurement errors and interpretable clocks to show organ-specific changes, such as in the liver, lungs, and kidneys, have also emerged. The issue lies in the lack of a comparative benchmark to determine which clock is both sensitive and consistently responsive to interventions.

Key Findings

Researchers built the TranslAGE database by standardizing whole-blood DNAm data from 51 public and private longitudinal intervention studies. They calculated 16 representative epigenetic clocks and 94 DNAm indicators for smoking, immune cell composition, and proteins and metabolites using a unified computational framework. After removing the effect of chronological age via regression analysis and normalizing values by standard deviation in reference groups, they compared pre- and post-intervention samples using paired t-tests. According to the Nature Medicine paper, aging-promoting events had an average effect size of 0.14, while longevity-targeting interventions recorded −0.06 (P<0.0001).

Among the 51 interventions, 19 significantly reduced epigenetic age when all 16 clocks were considered, and 13 remained significant after multiple testing correction. Five interventions increased it, and 26 showed no significant change. The average effect size for drug interventions was −0.09307, and for lifestyle interventions, it was −0.0393. No significant reductions were observed in the supplement and medical procedure categories.

Differences among clocks were also notable. DunedinPACE decreased in 16 interventions and increased in only one, showing the broadest responsiveness. The average effect size across all interventions was −0.0891 (P=0.0012). PCGrimAge showed a slightly smaller decrease of −0.07843 but had the strongest statistical support (P=0.0003). Among 15 lifestyle interventions, DunedinPACE decreased in 8, while among 14 drug interventions, GrimAgeV2 responded in 8. This suggests that each clock captures different biological signals.

Implications and Outlook

This analysis demonstrates that treating all epigenetic clocks as the same 'aging meter' can lead to misinterpretation of intervention effects. Second-generation clocks that learn mortality risk or aging pace showed greater sensitivity and consistency compared to first-generation clocks that predict chronological age. Anti-TNF therapy and the Mediterranean diet showed reproducible directions across multiple studies and clocks, while senolytic studies yielded conflicting results depending on the clock and study.

The health status of study participants also influenced outcomes. In diseased populations, the average effect sizes for PCPhenoAge, PCGrimAge, and SystemsAge were −0.32, −0.22, and −0.213, respectively, and remained significant after multiple testing. In healthy populations, the significance was weaker. It cannot be ruled out that improvements in inflammation or metabolic status due to disease treatment may be overinterpreted as 'systemic aging reversal'.

This study is based on reanalysis of existing data and does not provide clinical evidence that specific interventions improve lifespan or disease incidence. Limitations remain, including differences in sample size, study duration, and target diseases across studies, as well as the restriction to whole-blood DNAm. To use epigenetic clocks as surrogate endpoints in clinical trials, prospective trials are needed to confirm whether clock changes are linked to long-term healthspan and mortality improvements.